See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/266704065

On the total domination number of graphs

Article in Utilitas Mathematica · March 2007

CITATIONS READS
29 3,409

2 authors:

Peter Che Bor Lam Bing Wei

Caritas Institute of Higher Education University of Mississippi
90 PUBLICATIONS 1,071 CITATIONS 99 PUBLICATIONS 1,027 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Bing Wei on 26 February 2016.

The user has requested enhancement of the downloaded file.

 On the Total Domination Number of Graphs1
Peter Che Bor LAM
Department of Mathematics, Hong Kong Baptist University, Hong Kong
Bing WEI
Department of Mathematics, University of Mississippi, University, MS 38677, USA
and
Institute of Systems Science, Academy of Mathematics and System Sciences
Chinese Academy of Sciences, Beijing 100080, China

Abstract

Let G be a graph of order n with minimum degree at least two and S2 be a

vertex set consisting of all vertices of degree two in G. We call a vertex set T a
total dominating set if for any vertex u ∈ V (G) there exists some vertex v ∈ T
such that u ∈ N (v). The total domination number denoted by γt (G) is the
minimum cardinality of the total dominating sets. In this paper, we will prove
that γt (G) ≤ n2 if the length of the longest paths in the subgraph induced by
S2 is at most one. As a consequence, we have that γt (G) ≤ n2 if the minimum
degree of G is at least three.

1 Preliminaries and Main Results

We consider only finite undirected graphs without loops or multiple edges. The set of
vertices of a graph G is denoted by V (G) or just by V ; the set of edges by E(G) or just by
E. We use |G| (the order of G) as a symbol for the cardinality of V (G). If H and S are
subsets of V (G) or subgraphs of G, we denote by NH (S) the set of vertices in H which
are adjacent to some vertex in S, and set dH (S) = |NH (S)|. In particular, when H = G,
S = {u}, then let NG (u) = N (u) and set dG (u) = d(u). Set S2 = {x : x ∈ V (G), d(x) = 2}.
Paths and cycles in a graph G are considered as subgraphs of G. We use δ(G) to denote
the minimum degree of G, use G[S] to denote the subgraph induced by the vertex set S,
use dG (x1 , x2 ) to denote the distance of two vertices x1 , x2 in G and use G − E 0 to denote
the graph obtained by deleting an edge set E 0 of G. For an edge set E 0 of the complement
graph of G, we use G + E 0 to denote the graph by adding E 0 to G. In particular, we write
G + xy for G + E 0 when E 0 = {xy}.
The total domination number is defined by Cockayne et al. in [2]. We call a vertex
set T a total dominating set if for any vertex u ∈ V (G) we have u ∈ N (T ). The total
domination number denoted by γt (G) is the minimum cardinality of the total dominating
sets. There are several papers discussed the total domination number of graphs. We refer
1
supported by National Natural Science Foundation of China and the Croucher Foundation of Hong
Kong

1
the reader who is interested in this problem to [1], [2], [3] and [4]. In the rest of the
paper, we will write MTDS for minimum total dominating set and write TDS for total
dominating set.
The following theorem was proved in [2]:
Theorem 1. If G is a connected graph of order n ≥ 3, then γt (G) ≤ 2n 3 .

Sun [4] improved Theorem 2 as follows:

Theorem 2. Let G be a connected graph of order n. If δ(G) ≥ 2, then γt (G) ≤ b 4(n+1)
7 c.
Although the upper bounds of the above two theorems are best possible, the graphs
which show the sharpness of the theorems contain some vertices of degree two. This leads
Favaron et al. to make the following conjecture in [3].
Conjecture [3]. If G is a graph of order n with δ(G) ≥ 3, then γt (G) ≤ n2 .
Recently, the above Conjecture was proved in [1]. In this paper, we will consider the
total domination number of a graph G under the following conditions:
(1) G is a graph of order n with δ(G) ≥ 2 and the length of the longest path of G[S2 ]
is at most one.
We shall prove the following main theorem.
Theorem 3. If G satisfies (1), then γt (G) ≤ n2 .
Since for any graph with δ(G) ≥ 3, we have S2 = ∅. Thus, the Conjecture is a
consequence of Theorem 3. Notice that the proof in [1] is much shorter. We tried to
simplify the proof of Theorem 3. But it seems to be not so easy. We hope that the proof
techniques and methods developed in this paper will be useful for further research on
investigating total domination number of graphs.
To show the sharpness of Theorem 3, we construct the following graphs. For an
integer k ≥ 3, let Gk be the graph obtained from two disjoint cycles C1 = x1 x2 · · · xk x1
and C2 = y1 y2 · · · yk y1 by adding the edges xi yi (1 ≤ i ≤ k. Let Gjk be the graph
obtained from Gk by dividing each edge xi yi (1 ≤ i ≤ k) with zip (1 ≤ p ≤ j) vertices.
Then |V (G2k )| = 4k and γt (G2k ) = 2k as |T ∩ {xi , zi1 , zi2 , xi }| = 2 for any MTDS T of
G2k (1 ≤ i ≤ k), which implies that the upper bound n2 of Theorem 3 is sharp. Since
|V (G33 )| = 15 but γt (G33 ) = 8, the condition that the length of the longest path of G[S2 ] is
at most one cannot be relaxed.

2 Some Lemmas

In order to prove our main theorem, we first give a definition and show some lemmas.
Definition: An edge e of G is called removable, if G − e still satisfies (1).
By the definition of the total domination number, we have
Lemma 1. (i) γt (G) ≤ γt (G0 ) for any spanning subgraph G0 of G;

2
 (ii) γt (G) ≤ γt (G1 ) + γt (G2 ), where G1 and G2 are two disjoint subgraphs of G and
|G1 | + |G2 | = |G|.
Lemma 2. Let x1 , x2 be two vertices of G with dG (x1 , x2 ) = 2 and G0 = G + x1 x2 .If one
of the following conditions is satisfied, then γt (G0 ) = γt (G):
(i) there exists an MTDS T of G0 such that T ∩ NG (x1 ) ∩ NG (x2 ) 6= ∅;
(ii) there exists an MTDS T of G0 such that {x1 , x2 } ∩ T = ∅;
(iii) dG0 (xi ) = 2 and xi ∈ T for some 1 ≤ i ≤ 2 and some MTDS T of G0 ;
(iv) dG0 (x1 ) = dG0 (x2 ) = 2.
Proof. Notice that γt (G0 ) ≤ γt (G) by Lemma 1. Since any MTDS T of G0 satisfying (i)
or (ii) is a TDS of G, we have γt (G) ≤ γt (G0 ), which implies γt (G0 ) = γt (G).
(iii) Since dG (x1 , x2 ) = 2, there is some vertex, say v in NG (x1 ) ∩ NG (x2 ). Without
loss of generality, let dG0 (x1 ) = 2 and x1 ∈ T . If x2 ∈ T , then (T − {x1 }) ∪ {v} is a TDS
of G. Thus γt (G0 ) = γt (G). If x2 ∈ / T , then v ∈ T as dG0 (x1 ) = 2 and γt (G0 ) = γt (G) by
(i).
(iv) Since dG0 (x1 ) = dG0 (x2 ) = 2, then for any MTDS T of G0 we have either T ∩
NG (x1 ) ∩ NG (x2 ) 6= ∅ or T ∩ {x1 , x2 } =6 ∅. Thus by (i) or (iii), γt (G0 ) = γt (G). 2
Lemma 3. If G is a graph with δ(G) ≥ 3, then there is a spanning subgraph G0 of G such
that δ(G0 ) = 2 and the set of all vertices of degree two in G0 is an independent set of G0 .
Proof. Let d(v) = δ(G). If δ(G) = 3, then by deleting an edge which is incident
with v we can get the required spanning subgraph G0 . Now, by induction, assume that
Lemma 3 holds for all graphs with minimum degree t ≥ 3 and consider the graph G with
δ(G) = t + 1. We first delete an edge incident with v to get a spanning subgraph G∗ of G
such that δ(G∗ ) = t and then use the induction hypothesis to get the required spanning
subgraph G0 of G∗ . Thus G0 is also a required spanning subgraph of G. 2
Lemma 4. Let P = x1 x2 x3 x4 be a path in G.
/ E and G0 = G[V (G−P )]+{u1 u2 : u1 ∈ NG−P (x1 ), u2 ∈ NG−P (x4 ), u1 u2 ∈
(i) If x1 x4 ∈ /
E(G)}, then γt (G) ≤ γt (G0 ) + 2.
(ii) If x1 x4 ∈ E and G0 = G[V (G−P )]+{u1 u2 : u1 , u2 ∈ NG−P (x1 )∪NG−P (x4 ), u1 u2 ∈
/
0
E(G)} then γt (G) ≤ γt (G ) + 2.
Proof. (i) Let T be an MTDS of G0 . If (NG−P (x1 )∪NG−P (x4 ))∩T = ∅, then T ∪{x2 , x3 }
is a TDS of G and γt (G) ≤ γt (G0 ) + 2. If NG−P (x1 ) ∩ T 6= ∅ but NG−P (x4 ) ∩ T = ∅ or
NG−P (x1 ) ∩ T = ∅ but NG−P (x4 ) ∩ T 6= ∅, then T ∪ {x3 , x4 } or T ∪ {x1 , x2 } is a TDS of
G and γt (G) ≤ γt (G0 ) + 2. If NG−P (x1 ) ∩ T 6= ∅ and NG−P (x4 ) ∩ T 6= ∅, then T ∪ {x1 , x4 }
is a total dominating set of G and again γt (G) ≤ γt (G0 ) + 2.
(ii) For any MTDS T of G0 , it is easy to check that T ∪ {x1 , x4 } is a TDS of G since
x1 x4 ∈ E. Thus, γt (G) ≤ γt (G0 ) + 2. 2
Remark: In Lemmas 1–4, we do not require that G satisfies (1).
Lemma 5. Let G be a graph satisfying (1) and P = u1 x1 x2 u2 be a path in G with

3
{x1 , x2 } ⊆ S2 . If |NG−P (u1 ) ∪ NG−P (u2 )| ≥ 3, then there exist a vertex set Q and a graph
G0 with vertex set V (G) − Q such that G0 satisfies (1) and γt (G) ≤ γt (G0 ) + |Q| 2 .
Proof. Let M = V (P ) ∪ NG−P (u1 ) ∪ NG−P (u2 ). If for any vertex y in NG−P (u1 ) ∪
NG−P (u2 ) we have N (y) ∩ (V (G) − M ) = ∅, then take Q = M and {u1 , u2 , v} is an MTDS
of G[Q] for some vertex v in NG−P (u1 ) ∩ NG−P (u2 ) when NG−P (u1 ) ∩ NG−P (u2 ) 6= ∅ or
{u1 , u2 , v1 , v2 } is a MTDS of G[Q], where vi ∈ NG−P (ui ) (i = 1, 2), when NG−P (u1 ) ∩
NG−P (u2 ) = ∅. In either case, we can check that Lemma 5 holds. Thus in the rest
of the proof, we asume that there exists a vertex y in NG−P (u1 ) ∪ NG−P (u2 ) such that
N (y) ∩ (V (G) − M ) 6= ∅.
If u1 = u2 , let Q = {x1 , x2 } and G0 = G[V (G) − Q]. Since dG0 (u1 ) ≥ 3, G0 satisfies
(1). Let T be an MTDS of G0 . When u1 ∈ T , then T is a TDS of G and Lemma 5 holds.
When u1 ∈ / T , then T ∩ N (u1 ) 6= ∅ and T ∪ {u1 } is a TDS of G. Thus γt (G) ≤ γt (G0 ) + 1
and Lemma 5 holds.
If u1 6= u2 , since the length of the longest path in G[S2 ] is at most one, we have
N (u1 ) − {x1 , u2 } = 6 ∅. When u1 u2 ∈ E, then G0 = G[V (G −
6 ∅ and N (u2 ) − {x2 , u1 } =
P )] + {vw : v, w ∈ NG−P (u1 ) ∪ NG−P (u2 ), vw ∈ / E} satisfies (1). By Lemma 4(ii), we
0 |P |
have γt (G) ≤ γ(G ) + 2 . When u1 u2 ∈ / E, then |NG−P (ui )| ≥ 2 (i = 1, 2).
Let Q = V (P ) and let G0 = G[V (G) − Q] + {wz : w ∈ NG−P (u1 ), z ∈ NG−P (u2 ), wz ∈ /
0
E}. If G satisfies (1), we can easily check that Lemma 5 holds by using Lemma 4(i). If
G0 does not satisfy (1), then |NG−P (u1 ) ∪ NG−P (u2 )| = 4, NG−P (u1 ) ∩ NG−P (u2 ) = ∅,
G[NG−P (u1 ) ∩ NG−P (u2 )] is a K2,2 and there exists only one vertex, say y, in NG−P (u1 ) ∪
NG−P (u2 ) such that N (y) ∩ (V (G) − M ) 6= ∅. Without loss of generality, let y ∈ N (u1 )
and set NG−P (u1 ) = {y, u01 }.
Take Q0 = N (u2 ) ∪ {u2 } and G1 = G[V (G) − Q0 ] + u01 x1 . Then G1 satisfies (1) with
dG1 (u01 ) = dG1 (x1 ) = 2. and γt (G[Q]) = 2, by Lemma 2(iv), we can derive that Lemma 5
holds.
If G1 does not satisfies (1), then |NG−P (y) ∩ (V (G) − M )| = 1, d(z) = 2 for z ∈
NG−P (y) ∩ (V (G) − M ) and d(z1 ) = 2 for z1 ∈ N (z) − {y}. Let Q” = M and G01 =
G[V (G) − Q”] + zz2 for z2 ∈ N (z1 ) − {z}. Then G01 satisfies (1). Since {u1 , y, u2 , x2 } is
a TDS of G[Q”] and dG01 (z) = dG01 (z1 ) = 2, we can easily check that Lemma 5 holds by
Lemma 2. 2
Lemma 6. Let G be a graph satisfying (1) and containing no removable edges. If G[S2 ]
contains a path of length one, then there exist a vertex set Q and a graph G0 with vertex
set V (G) − Q such that G0 satisfies (1) and γt (G) ≤ γt (G0 ) + |Q|
2 .

Proof. Choose x1 , x2 in S2 such that x1 x2 ∈ E(G) and |N (u1 ) ∪ N (u2 )| as large as

possible, where u1 ∈ N (x1 ) − {x2 } and u2 ∈ N (x2 ) − {x1 }. Let P = u1 x1 x2 u2 (possibly,
u1 = u2 ). Since the length of the longest path in G[S2 ] is one, N (u1 ) − {x1 , u2 } 6= ∅ and
N (u2 ) − {x2 , u1 } =
6 ∅. Let M = (V (P ) ∪ NG−P (u1 ) ∪ NG−P (u2 )). By Lemma 5, we may
assume that |NG−P (u1 ) ∪ NG−P (u2 )| ≤ 2.
Case 1. u1 6= u2 .

4
 Then P is a path of four vertices of G.
Case 1.1. u1 u2 ∈ E.
Then there exists a vertex y in NG−P (u1 )∪NG−P (u2 ) such that N (y)∩(V (G)−M ) 6= ∅,
since otherwise, take Q = M and we can easily check that Lemma 6 holds as γt (G[Q]) = 2
({u1 , u2 } is an MTDS of G[Q]).
Case 1.1.1. |NG−P (u1 ) ∪ NG−P (u2 )| = 2.
Let u0i ∈ NG−P (ui ) (i = 1, 2) and without loss of generality, we may assume that
|N (u01 ) ∩ (V (G) − M )| ≥ |N (u02 ) ∩ (V (G) − M )| and N (u01 ) ∩ (V (G) − M ) 6= ∅.
If |N (u01 ) ∩ (V (G) − M )| ≥ 2, then take Q = {u01 , u1 , x1 , x2 } and let G0 = G[V (G) −
Q] + {wu2 : w ∈ N (u01 ) ∩ (V (G) − M )}. By Lemma 4, we can easily check that Lemma
6 holds. Thus |N (u01 ) ∩ (V (G) − M )| = 1 and |N (u02 ) ∩ (V (G) − M )| ≤ 1. Let v1 ∈
N (u01 ) ∩ (V (G) − M ). When N (u02 ) ∩ (V (G) − M )) = ∅, then N (v1 ) ∩ (V (G) − M ) 6= ∅ as
δ(G) ≥ 2. Let Q = {x1 , x2 , u2 , u02 } and G0 = G[V (G) − Q] + u1 v1 . Then G0 satisfies (1).
Let T be an MTDS of G0 . Then T ∪ {u01 , u2 } whenever u1 ∈ T or T ∪ {u1 , u2 } whenever
u1 ∈ / T is a TDS of G. Thus Lemma 6 holds. Hence we only need to deal with the case
when N (u02 ) ∩ (V (G) − M ) 6= ∅. Let v2 ∈ N (u02 ) ∩ (V (G) − M ).
If v1 = v2 , then by taking Q = M ∪ {v1 } when d(v1 ) = 2 or Q = V (P ) and G0 =
G[V (G) − Q] when d(v1 ) ≥ 3, we can derive that Lemma 6 holds by Lemma 1 or Lemma
4.
If v1 6= v2 , since |N (u01 ) ∩ (V (G) − M )| = |N (u02 ) ∩ (V (G) − M )| = 1, we have
v1 u02 ∈
/ E and v2 u01 ∈ / E. By the choice of x1 , x2 , |{u0i , vi } ∩ S2 | ≤ 1 (i = 1, 2)(otherwise,
P 0 = ui u0i vi wi for some wi ∈ N (vi ) − {u0i } is a path with |NG−P 0 N (ui ) ∪ NG−P 0 (wi )| ≥ 3).
When u01 u02 ∈ / E, then we may assume that {v1 , v2 } ∩ S2 6= ∅ by Lemma 4 and without
loss of generality, say v1 ∈ S2 . Then d(u01 ) ≥ 3, which implies u01 u2 ∈ E. By the choice of
x1 , x2 again, d(w1 ) ≥ 3 for w1 ∈ N (v1 ) − {u01 }. Thus u01 u2 is removable, a contradiction.
When u01 u02 ∈ E, then {v1 , v2 } ∩ S2 6= ∅, since u01 u02 is not removable. Without loss of
generality, let v1 ∈ S2 and w1 ∈ N (v1 ) − {u01 }. Thus by the choice of x1 , x2 , we have
either d(w1 ) = 2 and w1 = v2 or d(w1 ) ≥ 3. In the former case, Lemma 6 holds by taking
Q = M ∪ {v1 , v2 } and applying Lemma 1. In the latter case, we have d(v2 ) = 2 and
d(w2 ) = 2 for w2 ∈ N (v2 ) − {u02 }, since u01 u02 is not removable. Thus P 0 = u02 v2 w2 w20 for
w20 ∈ N (w2 ) − {v2 } is a path with |NG−P 0 (u02 ) ∪ NG−P 0 (w20 )| ≥ 3, contrary to the choice
of x1 , x2 .
Case 1.1.2. |NG−P (u1 ) ∪ NG−P (u2 )| = 1.
Let v ∈ NG−P (u1 ) ∪ NG−P (u2 ). If dG−P (v) ≥ 3, let G0 = G[V (G − P )]. Then G0
satisfies (1) and we can easily check that Lemma 6 holds.
If dG−P (v) = 2, let v1 , v2 ∈ NG−P (v). Without loss of generality, assume that d(v1 ) ≥
d(v2 ). When d(v2 ) ≥ 3, take Q = V (P ) and G0 = G[V (G) − Q] and we can easily check
that Lemma 6 holds. When d(v2 ) = 2 but d(v1 ) ≥ 3, then by the choice of {x1 , x2 }, we
have d(v20 ) ≥ 3 for v20 ∈ N (v2 ) − {v} (otherwise P 0 = vv2 v20 w will be a path of G for
some w ∈ N (v20 ) with |NG−P 0 (v) ∪ NG−P 0 (w)| ≥ 3). Thus, by taking Q = V (P ) and
G0 = G[V (G) − Q], we can easily check that Lemma 6 holds.

5
 Hence d(v1 ) = 2 and d(v2 ) = 2. Let vi0 ∈ N (vi ) − {v} (i = 1, 2). By the choice
of {x1 , x2 }, we have d(vi0 ) ≥ 3 (i = 1, 2) (otherwise,either n = 7 and we can directly
check that Lemma 6 holds or P 0 = vvi vi0 w will be a path of G for some w ∈ N (vi0 ) with
|NG−P 0 (v) ∪ NG−P 0 (w)| ≥ 3). When v10 = v20 , then take Q = M and G0 = G[V (G) −
Q] + v1 v2 which satisfies the (1). Notice that dG0 (v1 ) = dG0 (v2 ) = 2 and dG (v1 , v2 ) = 2.
By Lemma 2, γt (G0 ) = γt (G[V (G) − Q]). Thus Lemma 6 holds. When v10 6= v20 and
v10 v20 ∈ E, then take Q = M ∪ {v2 } and G0 = G[V (G) − Q] + v1 v20 which satisfies (1).
For any MTDS T of G0 , T ∪ {v10 , u1 , u2 } whenever v20 ∈ T or T ∪ {v20 , u1 , u2 } whenever
v20 ∈/ T implying v10 ∈ T will be a TDS of G. Thus Lemma 6 holds. When v10 6= v20 and
v10 v20 ∈/ E, then take Q = M ∪ {v1 , v2 , v20 } whenever N (v10 ) ∩ N (v20 ) = ∅ and we can check
that G0 = G[V (G) − Q] + {wv10 : w ∈ N (v20 )} satisfies (1). By Lemma 4, we can get that
Lemma 6 holds. Take Q = M ∪ {v2 } whenever N (v10 ) ∩ N (v20 ) 6= ∅, say w ∈ N (v10 ) ∩ N (v20 ).
Then G0 = G[V (G) − Q] + v1 v20 will satisfy (1). For any MTDS T of G0 , T ∪ {v, u1 , u2 } as
long as v1 ∈ / T or T − {v1 } ∪ {w, v, u1 , u2 } as long as v1 ∈ T implying {v10 , v20 } ∩ T 6= ∅ is
a TDS of G. Thus Lemma 6 holds.
If dG−P (v) = 1, let w ∈ NG−P (v). Take Q = V (P ) and G0 = G[(V (G) − Q)] + vw0
for some w0 ∈ NG−P (w). Then G0 satisfies (1) and dG0 (v) = 2. For any MTDS T of G0 ,
when T ∩ {v, w} 6= ∅, then by Lemma 2, γt (G) ≤ γt (G0 ) + 2. When T ∩ {v, w} = ∅, then
T ∪ {u1 , u2 , } is a TDS of G. Thus Lemma 6 holds.
Case 1.2. u1 u2 ∈
/ E.
Then |NG−P (u1 )| = |NG−P (u2 )| = 2 and NG−P (u1 ) = NG−P (u2 ). Let {u01 , u02 } ⊆
NG−P (u1 ). Without loss of generality, let dG−P (u01 ) ≥ dG−P (u02 ).
Case 1.2.1. u01 u02 ∈ / E.
If dG−P (u2 ) ≥ 2 or dG−P (u02 ) = 1 but dG−P (u01 ) ≥ 2, let Q = V (P ) and G0 =
0

G[V (G) − Q] + u01 u02 . Then G0 satisfies (1) by the choice of x1 , x2 , and Lemma 6 holds by
Lemma 4(i).
If dG−P (u02 ) = 1 and dG−P (u01 ) = 1, let vi ∈ NG−P (u0i ) (i = 1, 2). When v1 = v2 or
d(vi ) ≥ 3 (i = 1, 2), by taking Q = V (P ) and G0 = G[V (G) − Q)] + u01 u02 , we can derive
that Lemma 6 holds as before. Thus, we may assume that v1 6= v2 and without loss of
generality, assume that d(v2 ) = 2. Let wi ∈ N (vi ) ∩ (V (G) − M ) (i = 1, 2). By the choice
of {x1 , x2 }, we have d(w2 ) ≥ 3 and |{v1 , w1 } ∩ S2 | ≤ 1.
If w2 = v1 , take Q = V (P ) ∪ {u02 , v2 } and G0 = G[V (G) − Q] + w1 u01 which satisfies
(1). For any MTDS T of G0 , by Lemma 2, we can get that Lemma 6 holds when u01 ∈ T
and T ∪ {u02 , u1 , u2 } is a TDS of G when u01 ∈ / T.
If w2 6= v1 , then v1 v2 ∈ / E as d(v2 ) = 2. When w1 = w2 , let Q = M and G0 =
G[V (G) − Q] + v1 v2 which satisfies (1). Notice that dG0 (v2 ) = 2 and γt (G[Q]) = 3. For
any MTDS T of G0 , we can get that Lemma 6 holds by Lemma 2(iii) whenever v2 ∈ T or
by the fact that T ∪ {u02 , u1 , u2 } is a TDS of G whenever v2 ∈ / T . Thus we may assume
that w1 6= w2 .
When N (v1 ) ∩ N (w2 ) = ∅, let Q = M ∪ {v1 , v2 } and G0 = G[V (G) − Q] + {ww2 : w ∈
N (v1 ) ∩ (V (G) − Q)}. Then G0 satisfies (1) and for any MTDS T of G0 , T ∪ {v1 , v2 , x1 , x2 }

6
whenever {w, w2 } ⊆ T for some w ∈ N (v1 ) ∩ (V (G) − Q) or T ∪ {v2 , u2 , u02 , x2 } whenever
N (v1 ) ∩ (V (G) − Q) ∩ T 6= ∅ but w2 ∈ / T or T ∪ {v1 , u1 , u01 , x1 } whenever N (v1 ) ∩ (V (G) −
Q) ∩ T = ∅ but w2 ∈ T or T ∪ {u1 , u01 , u2 , u02 } whenever N (v1 ) ∩ (V (G) − Q) ∩ T = ∅
and w2 ∈ / E will be a TDS of G Thus Lemma 6 holds. When N (v1 ) ∩ N (w2 ) 6= ∅, say
w1 w2 ∈ E. Then let Q = V (P ) ∪ {v2 , u02 } and G0 = G[V (G) − Q] + u01 w2 which satisfies
(1), since |{v1 , w1 } ∩ S2 | ≤ 1. For any MTDS T of G0 , T ∪ {u1 , u2 , u02 } whenever u01 ∈ / T or
0 0 0 0
(T − {u1 }) ∪ {w1 , u1 , u2 , u2 } whenever u1 ∈ T (implying {v1 , w2 } ∩ T 6= ∅ as dG (u1 ) = 2)
0

will be a TDS of G. Thus Lemma 6 holds.

If dG−P (u02 ) = 0, let Q = {x1 , x2 } and G0 = G[V (G) − Q] + u1 u2 . Since u01 u02 ∈ / E,
0
T ∩ {u1 , u2 } 6= ∅ for any total dominating set T of G . Thus we can easily check that
Lemma 6 holds.
Case 1.2.2. u01 u02 ∈ E.
Since u01 u02 is not removable, we have |N (u01 ) ∩ (V (G) − M )| = 0, and |N (u02 ) ∩ (V (G) −
M )| = 0. Thus by taking Q = M and G0 = G[V (G) − Q], we can derive that Lemma 6
holds.
Case 2. u1 = u2 .
Case 2.1. d(u1 ) = 4.
Let v1 , v2 ∈ N (u1 ) − {x1 , x2 }. If N (vi ) ∩ (V (G) − M ) = ∅ (i = 1, 2), then by taking
Q = M and G0 = G[V (G) − Q], we can check that Lemma 6 holds by Lemma 1. Thus,
without loss of generality, we may assume that N (v1 ) ∩ (V (G) − M ) 6= ∅.
Since δ(G) ≥ 2, N (vi ) − {u1 } 6= ∅ and let wi ∈ N (vi ) − {u1 } (i = 1, 2). By the choice
of {x1 , x2 }, |{vi , wi } ∩ S2 | ≤ 1. Thus we may assume that d(vi ) = 2 (i = 1, 2), otherwise,
by taking Q = {x1 , x2 } and G0 = G[V (G) − Q] which satisfies (1), we can derive that
Lemma 6 holds as T ∩ ({u1 } ∪ (N (u1 ) − Q)) 6= ∅. Thus v1 v2 ∈ / E and d(wi ) ≥ 3 (i = 1, 2).
When w1 = w2 , then take Q = {x1 , x2 , u1 , v2 } and G0 = G[V (G) − Q] + v1 w for some
w ∈ N (w1 ) − {v1 , v2 }. Then G0 satisfies (1) and for any MTDS T of G0 , T ∪ {w1 , u1 }
whenever v1 ∈ T or T ∪ {v2 , u1 } whenever v1 ∈ / T will be a TDS of G. When w1 6= w2 ,
then take Q = {x1 , x2 , u1 , v2 } and G = G[V (G) − Q] + v1 w2 . Then G0 satisfies (1) and
0

T ∪ {u1 , v2 } is a TDS of G for any MTDS T of G0 . Thus Lemma 6 holds.

Case 2.2. d(u1 ) = 3.
Let v ∈ N (u1 )−{x1 , x2 }. Then we may assume that N (w)∩(V (G)−(M ∪N (v)) 6= ∅ for
some vertex w ∈ N (v) − {u1 }, otherwise, by taking Q = M ∪ N (v) and G0 = G[V (G) − Q],
we can get that Lemma 6 holds by Lemma 1, as γt (G[Q]) = 2.
Case 2.2.1. d(v) ≥ 4.
Let Q = M and G0 = G[V (G) − Q] + {xy : x, y ∈ N (v) ∩ (V (G) − Q), xy ∈
/ E(G)}.
0 0
Then G satisfies (1) and T ∪ {u1 , v} is a TDS of G for any MTDS T of G .
Case 2.2.2. d(v) = 3.
Let w1 , w2 ∈ N (v) − {u1 } and assume that d(w1 ) ≥ d(w2 ). Notice that |{w, wi } ∩
S2 | ≤ 1 for any w ∈ N (wi ) ∩ (V (G) − M ) (i = 1, 2), otherwise, P 0 = u1 wi ww0 for
w0 ∈ N (w) ∩ (V (G) − M ) will be a path which contradicts the choice of P . We may

7
assume that d(w2 ) ≤ 3 and d(w1 ) = 3 when d(w2 ) = 3 (otherwise, w1 v is removable, a
contradiction).
If w1 w2 ∈/ E, then let Q = M and G0 = G[V (G) − Q] + w1 w2 . G0 satisfies (1) and
T ∪ {u1 , v} is a TDS of G for any MTDS T of G0 .
If w1 w2 ∈ E, then d(w1 ) ≥ 3. When d(w2 ) = 2, then take Q = M and G0 =
G[V (G) − Q] + ww2 for some w ∈ N (w1 ) ∩ (V (G) − Q). Then G0 satisfies (1). Noting
that dG0 (w2 ) = 2, for any MTDS T of G0 , we have T ∪ {u1 , v} is a TDS of G whenever
|T ∩ {w, w2 }| ≤ 1 or (T − {w2 }) ∪ {w1 , v, u1 } whenever {w, w2 } ⊆ T is a TDS of G. Thus
Lemma 6 holds. When d(w2 ) = 3, then d(w1 ) = 3. Let wi0 ∈ N (wi ) ∩ (V (G) − {v, w1 , w2 })
(i = 1, 2). Then d(wi0 ) = 2 (i = 1, 2), since both w1 v and w2 v are not removable. By
the choice of x1 , x2 , we have either w10 = w20 ∈ E or d(w) ≥ 3 for w ∈ N (wi0 ) − {w1 , w2 }
(i = 1, 2). In the former case, Lemma 6 holds by taking Q = M ∪ {w1 , w2 , w10 } and
applying Lemma 1. In the latter case, we have w1 w2 is removable, a contradiction.
Case 2.2.3. d(v) = 2.
Let w ∈ N (v) − {u1 }. Then by the choice of {x1 , x2 }, we have d(w) ≥ 3. If d(w) ≥
4, by taking Q = M and G0 = G[V (G) − Q], we can get Lemma 6 holds as before.
Thus d(w) = 3. Let w1 , w2 ∈ N (w) − {v} and assume that d(w1 ) ≥ d(w2 ). Then
(N (w1 ) ∪ N (w2 )) − {v, w1 , w2 } = 6 ∅, otherwise, by taking Q = M ∪ {w, w1 , w2 } and
0
G = G[V (G) − Q], we can derive that Lemma 6 holds.
Since |{w0 , wi } ∩ S2 | ≤ 1 (i = 1, 2) for any w0 ∈ N (wi ) by the choice of x1 , x2 , we may
assume that d(w1 ) = d(w2 ) = 2, otherwise, by taking Q = M and G0 = G[V (G) − Q], we
can derive that Lemma 6 holds. Thus w1 w2 ∈ / E.
0
Let wi ∈ N (wi ) − {w} (i = 1, 2). Then by the choice of {x1 , x2 } and Case 1, we have
d(wi0 ) ≥ 3 (i = 1, 2). If w10 = w20 , then take Q = M ∪ {w} and G0 = G[V (G) − Q] + w1 w2 .
Notice that dG0 (w1 ) = dG0 (w2 ) = 2 and γt (G[Q]) = 2. By Lemmas 1 and 2, we can derive
that Lemma 6 holds. If w10 6= w20 , then we may assume that w10 w20 ∈ / E, otherwise, take
Q = M ∪ {w, w2 } and G = G[V (G) − Q] + w2 w1 . For any MTDS T of G0 , we can get
0 0

Lemma 6 holds when w1 ∈ T by Lemmas 1 and 2, since dG0 (w1 ) = 2 or T ∪ {u1 , v, w} is a

TDS of G when w1 ∈ / T.
Now, when N (w10 ) ∩ N (w20 ) = ∅, then let Q = M ∪ {w, w1 , w2 , w20 } and let G0 =
G[V (G) − Q] + {w10 w : w ∈ N (w20 ) ∩ (V (G) − Q), ww10 ∈ / E(G)}. Then G0 satisfies
0
(1). Notice that w1 ww2 w2 is a path of four vertices in G[V (G) − M ] and γt (G[M ]) = 2.
By Lemmas 1 and 4, we can get γt (G) ≤ γt (G0 ) + 4, which implies Lemma 6 holds.
When N (w10 ) ∩ N (w20 ) 6= ∅, let z ∈ N (w10 ) ∩ N (w20 ) and take Q = M ∪ {w, w2 }. Then
G0 = G[V (G) − Q] + w1 w20 satisfies (1). Noting that dG0 (w1 ) = 2, for any MTDS T of G0 ,
T ∪ {u1 , v, w} whenever w1 ∈ / T or (T − {w1 }) ∪ {u1 , v, w, z} whenever w1 ∈ T implying
0 0
T ∩ {w1 , w2 } =
6 ∅ will be a TDS of G. Thus Lemma 6 holds.
The proof of Lemma 6 is complete. 2

8
3 The Proof of Theorem 3
Let G be a graph which satisfies (1). By Lemma 1, we may assume that G contains
no removable edges, which implies that S2 6= ∅ by Lemma 3.
We will prove Theorem 3 by induction. If n is small, we can easily check that Theorem
3 holds. Thus we may assume that for any graph G with less than n vertices which satisfies
(1), γt (G) ≤ |G|
2 . Now we consider the graph G with n vertices. If S2 is not independent,
then by Lemma 6 and the induction hypothesis, we can get γt (G) ≤ n2 . Hence we only need
to consider the case when S2 is independent. By Lemma 1 and the induction hypothesis,
we may further assume that G is connected.
Claim 1. Let x1 , x2 ∈ V (G) and x1 x2 ∈ E. Then
(i) min{d(x1 ), d(x2 } ≤ 3;
(ii) d(x2 ) = 2 or d(x2 ) = 3 and d(x) = 2 for any x ∈ N (x2 ) − {x1 } if d(x1 ) ≥ 4.
(iii) (N (x1 ) − {x2 }) ⊆ S2 or (N (x2 ) − {x1 }) ⊆ S2 , if {x1 , x2 } ∩ S2 = ∅ and N (x1 ) ∩
N (x2 ) = ∅, that is, V (P ) ∩ S2 6= ∅ for any path P of G with |P | ≥ 4, if the length of the
shortest cycle of G is at least 4 .
Proof. If neither (i) nor (ii) holds, then G contains a removable edge, a contradiction.
(iii) If there exist y1 ∈ N (x1 )−{x2 } and y2 ∈ N (x2 )−{x1 }) such that {y1 , y2 }∩S2 = ∅,
then P = y1 x1 x2 y2 is a path with V (P ) ∩ S2 = ∅, since N (x1 ) ∩ N (x2 ) = ∅ and S2 is
independent, we have that x1 x2 is removable, a contradiction. 2
Claim 2. Let C = x1 x2 · · · xt x1 (t ≥ 3) be a shortest cycle of G, then
(i) V (C) ∩ S2 6= ∅;
(ii) |V (C) ∩ S2 | ≥ |C|
4 ≥ 1 when |C| ≥ 4.
Proof. (i) By contradiction, assume that V (C) ∩ S2 = ∅. Then d(xi ) ≥ 3. If there exists
some 1 ≤ i ≤ t − 1 such that N (xi ) ∩ N (xi+1 ) = ∅, then xi xi+1 is removable as S2 is
independent in G. If N (xi ) ∩ N (xi+1 ) 6= ∅ for any 1 ≤ i ≤ t − 1, let y1 ∈ N (x1 ) ∩ N (x2 )
and y2 ∈ N (x2 ) ∩ N (x3 ). When y1 6= y2 , then d(x2 ) ≥ 4 which contradicts to Claim 1(ii)
as xt ∈ N (x1 ) and d(xt ) ≥ 3. When y1 = y2 , then d(y1 ) ≥ 3 and x1 x2 is removable, a
contradiction.
(ii) By Claim 1(iii), for any 1 ≤ i ≤ t, we have {xi , xi+1 , xi+2 , xi+3 } ∩ S2 6= ∅. Thus
(ii) holds. 2
Claim 3. Let v1 , v2 be two distinct vertices in S2 with d(v1 , v2 ) = 2. Suppose w ∈
N (v1 ) ∩ N (v2 ) and yi ∈ N (vi ) − {w} (i = 1, 2). If one of the following conditions holds,
then γt (G) ≤ n2 :
(i) d(yi ) ≥ 4 (i = 1, 2);
(ii) d(yi ) ≥ 4 and d(yj ) + Σx∈N (yj ) d(x) ≥ 10 (i 6= j, i, j ∈ {1, 2});
(iii) d(yi ) + Σx∈N (yi ) d(x) ≥ 10 (i = 1, 2), when N (y1 ) ∩ N (y2 ) ∩ S2 = ∅.
Proof. Let G1 = G − {v1 y1 , v2 y2 } and G0 = G1 + v1 v2 . If one of the above conditions
holds, then it is easy to check that G0 satisfies (1), as S2 is independent. Notice that
dG0 (v1 ) = dG0 (v2 ) = 2. By the induction hypothesis and Lemmas 2 and 6, γt (G1 ) =

9
γt (G0 ) ≤ n
2. Thus γt (G) ≤ γt (G1 ) ≤ n
2 by Lemma 1, as G1 is a spanning subgraph of G.
2
Choose a shortest cycle C of G such that |V (C) ∩ S2 | as large as possible. By Claim
2, |V (C) ∩ S2 | ≥ 1. Let C = x1 x2 · · · xt x1 (t = |C|).
Claim 4. If |C| = 3, then γt (G) ≤ n2 .
Proof. By Claim 2, we may assume that d(x1 ) = 2. Since S2 is independent, d(x2 ) ≥ 3 and
d(x3 ) ≥ 3. Without loss of generality, let d(x2 ) ≥ d(x3 ). If d(x2 ) ≥ 4, then d(x3 ) = 3 by
Claim 1(i). When N (x2 ) ∩ N (x3 ) − {x1 } = 6 ∅, let v ∈ N (x2 ) ∩ N (x3 ) − {x1 }. Then d(v) = 2
0
by Claim 1(ii). Thus G = G[V (G) − {x1 }] satisfies (1). By the induction hypothesis,
γt (G0 ) ≤ n−1 0
2 . Since for any MTDS T of G , {x2 , x3 }∩T 6= ∅, T is a TDS of G. Thus Claim
4 holds. When N (x2 ) ∩ N (w3 ) − {x1 } = ∅, then the graph G0 = G[V (G) − {x1 , x3 }] + yx2
for some y ∈ N (x3 ) − V (C) satisfies (1). Thus γt (G0 ) ≤ n2 − 1. Since for any MTDS T of
G0 , {x3 } ∪ T whenever {x2 , y} ∩ T 6= ∅ or {x2 } ∪ T whenever {x2 , y} ∩ T = ∅ (implying
(N (x2 ) − V (C)) ∩ T 6= ∅) is a TDS of G, we obtain γt (G) ≤ n2 .
If d(x2 ) = 3, then d(x3 ) = 3. Let ui ∈ N (xi ) − V (C) (i = 2, 3). When u2 = u3 ,
then d(u2 ) ≥ 3 since G is connected. Thus u2 x2 is removable, since S2 is independent,
a contradiction. Thus u2 6= u3 . Assume that d(u2 ) ≥ d(u3 ). When d(u2 ) ≥ 3, then
G0 = G[V (G) − {x1 , x3 }] + u3 x2 satisfies (1). For the same reason as above, we can get
Claim 4 holds. Thus d(u2 ) = 2 and d(u3 ) = 2. Since S2 is independent, d(y) ≥ 3 for any
y ∈ N (u2 ) ∪ N (u3 ). Thus when N (u2 ) ∩ N (u3 ) = ∅, G0 = G[V (G) − (V (C) ∪ {u3 })] +
yu2 for some y ∈ N (u3 ) − V (C) satisfies (1) and Claim 4 holds by Lemma 4. When
N (u2 ) ∩ N (u3 ) 6= ∅, let v ∈ N (u2 ) ∩ N (u3 ). Then d(v) ≥ 3. Taking Q = V (C) ∪ {u3 } and
G0 = G[V (G) − Q] + yu2 for some y ∈ N (v) − {u2 , u3 }, we can check that G0 satisfies (1)
and for any MTDS T of G0 , T ∪ {v, x2 } whenever u2 ∈ T or T ∪ {x2 , x3 } whenever u2 ∈ /T
will be a TDS of G. Thus Claim 4 holds. 2
Claim 5. If |C| = 4, then γt (G) ≤ n2 .
Proof. Since S2 is independent, 1 ≤ |V (C) ∩ S2 | ≤ 2 by Claim 2.
Case 1. |V (C) ∩ S2 | = 2.
Let {x1 , x3 } ⊆ S2 . Since C is a shortest cycle, we have x2 x4 ∈
/ E. Since G is connected,
we may assume that there exists some vertex y ∈ NG−C (x2 ) ∪ NG−C (x4 ) such that N (y) −
(N (x2 ) ∪ N (x4 )) 6= ∅ (otherwise, {x1 , x2 } when n = 5 or {x1 , x2 , x4 } when n ≥ 6 will
be a TDS of G). If G0 = G[V (G) − {x1 , x3 }] + x2 x4 satisfies (1), then for any MTDS
T of G0 , T ∪ {x1 } when {x2 , x4 } ∩ T 6= ∅ or T ∪ {x2 } when {x2 , x4 } ∩ T = ∅ (implying
NG (x2 ) ∩ T 6= ∅) is a TDS of G. Thus by the induction hypothesis, Claim 5 holds.
If G0 = G[V (G) − {x1 , x3 }] + x2 x4 does not satisfy (1), then d(x2 ) = d(x4 ) = 3,
NG−C (x2 ) ∩ NG−C (x4 ) = ∅ and min{d(v2 ), d(v4 )} = 2, where vi ∈ NG−C (xi ) (i = 2, 4).
Without loss of generality, let d(v4 ) = 2. When v2 v4 ∈ E, then d(v2 ) ≥ 3 and G0 =
G[V (G − C)] + wv4 for some vertex w ∈ N (v2 ) − (V (C) ∪ {v4 }) satisfies (1), as S2 is
independent. For any MTDS T of G0 , T ∪ {x2 , v2 } whenever v4 ∈ T or T ∪ {x1 , x4 }
whenever v4 ∈ / T is a TDS of G. Thus Claim 5 holds by the induction hypothesis. When

10
 / E, then G0 = G[V (G−C)]+v2 v4 satisfies (1), as S2 is independent. For any MTDS
v2 v4 ∈
T of G0 , T ∪ {x2 , x4 } whenever {v2 , v4 } ⊆ T or T ∪ {xi , xi−1 } whenever vi ∈
/ T (i ∈ {2, 4})
is a TDS of G. Thus Claim 5 holds.
Case 2. |V (C) ∩ S2 | = 1.
Let d(x4 ) = 2. Then by Claim 1(ii), d(x1 ) = d(x3 ) = 3. Let ui ∈ N (xi ) − V (C)
(i = 1, 3). Since S2 is independent in G and d(x2 ) ≥ 3, we have d(u1 ) = d(u3 ) = 2 by
Claim 1(iii). By the choice of C, u1 6= u3 , x2 u1 ∈
/ E and x2 u3 ∈
/ E. Let wi ∈ N (ui ) − {xi },
then d(wi ) ≥ 3 (i = 1, 3) and x1 w3 ∈
/ E, x3 w1 ∈ / E by Claim 1(iii).
Case 2.1. w1 6= w3 .
Then N (x3 ) ∩ N (w1 ) = ∅. Let Q = {w1 , u1 , x1 , x4 }. Then G0 = G[V (G) − Q] +
{wx3 : w ∈ N (w1 )} satisfies (1). Thus by Lemma 4 and the induction hypothesis, we can
derive that Claim 5 holds.
Case 2.2. w1 = w3 .
Since dG (u1 , x4 ) = 2 and d(x3 ) + Σx∈N (x3 ) d(x) ≥ 10, we may assume that d(w1 ) = 3
by Claim 3(ii). Let v ∈ N (w1 ) − {u1 , u2 }.
If v = x2 , then n = 7, when d(x2 ) = 3, as G is connected, and {x1 , x2 , x3 } is a TDS of G
which implies Claim 5 holds. When d(x2 ) ≥ 4, then G0 = G[V (G)−{u1 , u3 , w1 , x4 }]+x1 x3
satisfies (1) and for any MTDS T of G0 , T ∪{x1 , x3 } whenever x2 ∈ T or T ∪{x2 } whenever
x2 ∈/ T (implying {x1 , x3 } ⊆ T as dG0 (x1 ) = dG0 (x3 ) = 2) is a TDS of G. Thus Claim 5
holds by the induction hypothesis. Hence in the rest of the proof, we assume that v 6= x2 .
If N (x2 ) ∩ N (v) = ∅, let G1 = G − {u1 w1 , x3 x4 } and G0 = G1 + u1 x4 and G01 =
G [V (G0 ) − {x3 , u3 , w1 , v}] + {x2 v 0 : v 0 ∈ N (v) − {w1 }}. Then G01 satisfies (1). Notice that
0

x3 u3 w1 v is a path in G0 . By the induction hypothesis and Lemma 4, γt (G0 ) ≤ n2 . Since

dG0 (x4 ) = dG0 (u1 ) = 2 and G1 is a spanning subgraph of G, we can derive that Claim 5
holds by Lemmas 1 and 2.
If N (x2 ) ∩ N (v) 6= ∅, let v1 ∈ N (x2 ) ∩ N (v). When d(v) = 2, then d(v1 ) ≥ 3. Since
dG (v, u3 ) = 2, d(v1 ) + Σy∈N (v1 ) d(y) ≥ 10 and d(x3 ) + Σy∈N (x3 ) d(y) ≥ 10, we have Claim 5
holds by Claim 3(iii). When d(v) ≥ 3, then d(v1 ) = 2 by Claim 1(iii) and vx2 ∈ / E by the
choice of C. Let G0 = G[V (G)−{w1 , u1 , x1 , x4 , x3 , u3 }]+x2 v. Then G0 satisfies (1) and for
any MTDS T of G0 , T ∪ {w1 , u1 , x1 } whenever x2 ∈ T or T ∪ {x1 , x2 , x3 } whenever x2 ∈ /T
(implying v ∈ T as dG0 (v1 ) = 2) is a TDS of G. Thus Claim 5 holds by the induction
hypothesis. 2
Claim 6. If |C| = 5, the γt (G) ≤ n2 .
Proof. Since S2 is independent, we have |V (C) ∩ S2 | = 2. Let {x1 , x3 } ⊆ S2 . Noting that
dG (x1 , x3 ) = 2, we may assume that d(x4 ) = d(x5 ) = 3 and (N (x4 ) ∪ N (x5 ) − {x1 , x3 }) ∩
S2 6= ∅ by Claim 3.
Let vi ∈ N (xi )−V (C) (i = 4, 5) and assume, without loss of generality, that d(v4 ) = 2.
We distinguish the following two cases.
Case 1. d(v5 ) = 2.
Let wi ∈ N (vi ) − V (C) (i = 4, 5). Then d(wi ) ≥ 3, since S2 is independent. If

11
w4 = w5 , by Claim 3, we may assume, without loss of generality that d(x2 ) = 3 and
d(w2 ) = 2 for w2 ∈ N (x2 ) − V (C), as dG (x3 , v4 ) = 2. When N (w2 ) ∩ N (w4 ) 6= ∅, then
d(w) ≥ 3 for w ∈ N (w2 )∩N (w4 ), since S2 is independent in G. Notice that dG (w2 , x3 ) = 2
and d(w) + Σx∈N (w) d(x) ≥ 10 and d(v4 ) + Σy∈N (v4 ) d(y) ≥ 10. By Claim 3(iii), Claim 6
holds. When N (w2 ) ∩ N (w4 ) = ∅, let G0 = G[V (G) − (V (C) ∪ {v4 , v5 , w4 }] + {w2 w : w ∈
N (w4 ) − {v4 , v5 }}. Then G0 satisfies (1) and for any MTDS T of G0 , T ∪ {x2 , x3 , w4 , v5 }
is a TDS of G. Thus Claim 6 holds by the induction hypothesis.
If w4 6= w5 and w4 w5 ∈ / E, then let G1 = G − {x1 x5 , x3 x4 } and set G0 = G1 + x1 x3 ,
G01 = G0 [V (G) − {v4 , x4 , x5 , v5 }] + w4 w5 . Notice that v4 x4 x5 v5 is a path in G0 , dG0 (x1 ) =
dG0 (x3 ) = 2 and G1 is a spanning subgraph of G, we can get γt (G) ≤ γt (G1 ) = γt (G01 ) ≤ n2
by Lemmas 1, 2, 6 and the induction hypothesis. Thus Claim 6 holds.
If w4 6= w5 and w4 w5 ∈ E, then we may assume that d(wi ) = 3 (i = 4, 5), otherwise,
let G1 = G − {x1 x5 , x3 x4 }, G0 = G1 + x1 x3 , and G01 = G0 [V (G) − {v4 , x4 , x5 , v5 }]. Then
G01 satisfies (1). Using the same argument as above, we can derive that Claim 6 holds.
When N (wi ) ∩ N (x2 ) = ∅ (i = 4, 5), let G0 = G[V (G) − {x3 , x4 , v4 , w4 }] + x2 w40 for
some w40 ∈ N (w4 ) − {v4 , w5 } and G01 = G0 [V (G0 ) − {x1 , x5 , v5 , w5 }] + x2 w50 for some w50 ∈
N (w50 )−{v5 , w4 }. Then G01 satisfies (1). Notice that x3 x4 v4 w4 is a path of G and x1 x5 v5 w5
is a path of G0 . Thus by Lemma 4 and the induction hypothesis, we can derive that Claim
6 holds. When there exists some i ∈ {4, 5}, say i = 4, such that N (w4 ) ∩ N (x2 ) 6= ∅, let
w40 ∈ N (w4 ) ∩ N (x2 ). Then w40 ∈ S2 by Claim 1(iii) and we may assume that d(x2 ) = 3 by
Claim 3(ii), as dG (v4 , x3 ) = 2. Let G0 = G[V (G) − (V (C) ∪ {v4 , w4 , w40 })] + v5 w0 for some
w0 ∈ N (w5 )−{v5 , w4 }. Then G0 satisfies (1) and for any MTDS T of G0 , T ∪{w5 , x2 , x3 , x4 }
whenever v5 ∈ T or T ∪ {x2 , x4 , x5 , w40 } whenever v5 ∈ / T is a TDS of G. Thus Claim 6
holds.
Case 2. d(v5 ) ≥ 3.
Since d(x2 ) ≥ 3 and x5 v5 is not removable, we have d(v5 ) = 3, (N (v5 ) − {x5 }) ⊆ S2
and N (v4 ) ∩ N (v5 ) = ∅ as S2 is independent. Let w4 ∈ N (v4 ) − V (C) and w5 , w6 ∈
N (v5 ) − V (C). If N (w4 ) ∩ {w5 , w6 } = ∅, then let G1 = G − {x3 x4 , x1 x5 }, G0 = G1 + x1 x3
and G01 = G0 [V (G) − {v4 , x4 , x5 , v5 }] + {w4 w5 , w4 w6 }. Then G01 satisfies (1). Notice that
v4 x4 x5 v5 is a path of G0 and G1 is a spanning subgraph of G. We can derive that Claim 6
holds as before. If N (w4 ) ∩ {w5 , w6 } = 6 ∅, without loss of generality, say w4 w5 ∈ E. Then
dG (v4 w5 ) = 2. Since d(x4 ) + Σy∈N (x4 ) d(y) ≥ 10 and d(v5 ) + Σy∈N (v5 ) d(y) ≥ 10, we can
derive that Claim 6 holds by Claim 3(iii). 2
Claim 7. If |C|=6, then γt (G) ≤ n2 .
Proof. Since S2 is independent, by Claim 2, we have 2 ≤ |V (C) ∩ S2 | ≤ 3.
Case 1. |V (C) ∩ S2 | = 3.
Without loss of generality, let d(xi ) = 2, i = 1, 3, 5. If max{d(x2 ), d(x4 ), d(x6 )} ≥ 4,
say d(x2 ) ≥ 4, then d(x4 ) = d(x6 ) = 3 by Claim 3(i) and d(ui ) = 2 for ui ∈ N (xi ) −
V (C) (i = 4, 6) by Claim 3(ii). Let wi ∈ N (ui ) − V (C) (i = 4, 6). When w6 6= w4
or N (w6 ) ∩ N (x2 ) = ∅, then let G1 = G − {x1 x6 , x3 x4 } (or G1 = G − {x5 x6 , x2 x3 }),
G0 = G1 + x1 x3 (or G0 = G1 + x3 x5 ) and G01 = G0 [V (G0 ) − {x4 , x5 , x6 , u6 }] + u4 w6 (or

12
G01 = G0 [V (G0 )−{x2 , x1 , x6 , u6 }]+{w6 w : w ∈ N (x2 )−V (C)}). Then G01 satisfies (1) and
G1 is a spanning subgraph of G in either case. Notice that both x4 x5 x6 u6 and x2 x1 x6 u6 are
paths of G. By the induction hypothesis and Lemmas 2, 6, we have γt (G1 ) = γt (G0 ) ≤ n2
and consequently, γt (G) ≤ n2 by Lemma 1. When w6 = w4 and N (w6 ) ∩ N (x2 ) 6= ∅, then
d(w6 ) = 3 by Claim 3(i), as dG (u4 , x3 ) = 2. Let w ∈ N (x2 ) ∩ (N (w6 ) − {u4 , u6 }). Then
d(w) = 2 by Claim 1(ii). Let G0 = G[V (G)−{x1 , x3 , x4 , x5 , x6 , u4 }]+wu6 . Then G0 satisfies
(1). Notice that dG0 (u6 ) = dG0 (w6 ) = 2. Thus for any MTDS T of G0 , {w, u6 } ∩ T 6= ∅
and w6 ∈ T when w ∈ / T , which imply that T ∪ {x4 , x5 , x6 } whenever |T ∩ {w, u6 }| = 1
or (T − {u6 }) ∪ {w6 , x4 , x5 , x6 } whenever |T ∩ {w, u6 }| = 2 is a TDS of G. Thus by the
induction hypothesis, we can derive that Claim 7 holds.
Hence, in the rest of the proof, we need only to deal with the case when d(x2 ) =
d(x4 ) = d(x6 ) = 3, let ui ∈ N (xi ) − V (C) (i = 2, 4, 6). If N (u2 ) ∩ N (u4 ) = ∅ or
N (u2 ) ∩ N (u6 ) = ∅, let G0 = G[V (G) − {u2 , x2 , x3 , x4 }] + {u4 z : z ∈ N (u2 ) − V (C)} or
G0 = G[V (G) − {u2 , x2 , x1 , x6 }] + {u6 z : z ∈ N (u2 ) − V (C)} and for the same reason as
above, we can derive that Claim 7 holds. Hence we may assume that N (u2 ) ∩ N (u4 ) 6= ∅
and N (u2 ) ∩ N (u6 ) 6= ∅. Let w4 ∈ N (u2 ) ∩ N (u4 ) and w6 ∈ N (u2 ) ∩ N (u6 ).
If w4 = w6 , then min{d(u4 ), d(u6 )} = 2 by Claim 1(iii). Without loss of generality,
let d(u4 ) = 2. Notice that dG (x3 , u4 ) = 2 and N (x2 ) ∩ N (w4 ) = {u2 }. By Claim 3(iii),
we may assume that d(u2 ) = 2 and furthermore, d(u6 ) = 2, as dG (x5 , u4 ) = 2 and
N (x6 ) ∩ N (w4 ) = {u6 }. When n = 10, then {x2 , u2 , x4 , x5 , x6 } is a TDS of G and Claim 7
holds. When n ≥ 11, then d(w4 ) ≥ 4. Let G0 = G[V (G) − {x3 , x4 , x5 , x6 , u4 , u6 }] + u2 x1 .
Then G0 satisfies (1). Notice that dG0 (x1 ) = dG0 (x2 ) = 2. For any MTDS T of G0 ,
T ∪ {x4 , x5 , x6 } whenever |T ∩ {u2 , x1 }| = 1 or (T − {x1 }) ∪ {x2 , x4 , x5 , x6 } whenever
|T ∩ {u2 , x1 }| = 2 is a TDS of G. Thus by the induction hypothesis, we can derive that
Claim 7 holds.
If w4 6= w6 , then d(u2 ) ≥ 3 and |{w4 , u4 } ∩ S2 | = |{u6 , w6 } ∩ S2 | = 1 by Claim
1(iii), as S2 is independent. since dG (u4 , x3 ) = dG (u6 , x1 ) = 2, we may assume that
d(u4 ) = d(u6 ) ≥ 3 by Claim 3(iii). Thus d(w4 ) = d(w6 ) = 2. When N (u4 ) ∩ N (u6 ) = ∅,
let G0 = G[V (G) − {u4 , x4 , x5 , x6 }] + x1 x3 . Then G0 satisfies (1) and we can derive that
Claim 7 holds as before. When N (u4 ) ∩ N (u6 ) 6= ∅, let w ∈ N (u4 ) ∩ N (u6 ). By Claim
1(iii), we have d(w) = 2. Since dG (w, w6 ) = 2 and dG (w4 , w6 ) = 2 and xi ∈ N (ui )
(i = 2, 4, 6), we may assume that d(u2 ) = d(u4 ) = d(u6 ) = 3 by Claim 3. Hence, n = 12
and {x2 , u2 , x4 , u4 , x6 , u6 } is a TDS of G which implies that Claim 7 holds.
Case 2. |V (C) ∩ S2 | = 2.
Let x1 , xi ∈ S2 . If dG (x1 , xi ) = 2, say x3 ∈ S2 , then {x4 , x5 , x6 } ∩ S2 = ∅. By Claim 3,
we can derive that Claim 7 holds as before. Thus dG (x1 , xi ) ≥ 3 and i = 4 which implies
d(xi ) ≥ 3 (i = 2, 3, 5, 6).
If d(xi ) ≥ 4 or d(xi ) = 3 but d(u) ≥ 3, where u ∈ N (xi ) − V (C) and i ∈ {2, 3, 5, 6},
without loss of generality, say i = 2, then d(x3 ) = 3 and d(u3 ) = 2 for u3 ∈ N (x3 )−V (C) by
Claim 1(ii). Let w3 ∈ N (u3 )−{x3 }. Then d(w3 ) ≥ 3, w3 x5 ∈ / E and N (w3 )∩N (x5 ) = ∅ by
the choice of C and the previuos cases. Thus G0 = G[V (G) − {w3 , u3 , x3 , x4 }] + {x5 y : y ∈

13
N (w3 ) − {u3 }} satisfies (1). Hence by the induction hypothesis and Lemma 4, Claim 7
holds.
If d(xi ) = 3 and d(ui ) = 2, where ui ∈ N (xi ) − V (C) (i = 2, 3, 5, 6), then u2 u5 ∈ / E
and u3 u6 ∈ / E as S2 is independent. By the choice of C and Case 1, N (u3 ) ∩ N (u5 ) = ∅
and N (u2 ) ∩ N (u6 ) = ∅. Let G0 = G[V (G) − {u2 , x2 , x1 , x6 }] + w2 u6 and G01 = G0 [V (G0 ) −
{u3 , x3 , x4 , x5 }] + w3 u5 , where wi ∈ N (ui ) − {xi } (i = 2, 3). Then G01 satisfies (1) and
|G0 |
γt (G01 ) ≤ 21 by the induction hypothesis. Notice that u3 x3 x4 x5 is a path in G0 and
u2 x2 x1 x6 is a path in G. By Lemma 4, we can derive that Claim 7 holds. 2
Now, we turn to prove Theorem 3. By Claims 4–7, we may assume that |C| ≥ 7. If
there exits some 1 ≤ i ≤ t such that{xi , xi+1 , xi+2 } ∩ S2 = ∅(the subscripts are taken
modulo t), then d(xi+2 ) = 3 by Claim 1(ii) and {xi+3 , u} 6⊆ S2 for u ∈ N (xi+2 ) − V (C) by
Claim 1(iii). Let w ∈ N (u) − {xi+2 }. Then N (w) ∩ N (xi+4 ) = ∅ by the choice of C. Thus
G0 = G[V (G) − {w, u, xi+2 , xi+3 }] + {xi+4 y : y ∈ N (w) − {u}} satisfies (1) and Theorem
3 holds by the induction hypothesis and Lemma 4.
If there exits some 1 ≤ i ≤ t such that {xi , xi+1 } ∩ S2 = ∅, then we may assume
that d(xi+1 ) = 3 by Claim 1(i), d(xi+2 ) = 2 and d(xi−1 ) = 2. When d(xj ) ≥ 4 or
d(xj ) = 3 but d(uj ) ≥ 3 for uj ∈ N (xj ) − V (C), where j ∈ {i, i + 1}, say j = i, then
d(ui+1 ) = 2 for ui+1 ∈ N (xi+1 ) − V (C) by Claim 1(iii). Since N (wi+1 ) ∩ N (xi+3 ) = ∅ for
wi+1 ∈ N (ui+1 ) − V (C) by the choice of C, we can derive that Theorem 3 holds as before.
When d(xj ) = 3 and d(uj ) = 2 for uj ∈ N (xj ) − V (C) (i = i, i + 1), let wj ∈ N (uj ) − {xj }
(j = i, i + 1), then wi wi+1 ∈ / E, N (wi ) ∩ N (xi−2 ) = ∅ and N (wi+1 ) ∩ N (xi+3 ) = ∅ by
0
the choice of C. Let G = G[V (G) − {wi , ui , xi , xi−1 }] + {xi−2 y : y ∈ N (wi ) − {ui }}
and G01 = G0 [V (G0 ) − {wi+1 , ui+1 , xi+1 , xi+2 }] + {xi+3 y : y ∈ N (wi+1 ) − {ui+1 }}. Then
|G0 |
G01 satisfies the conditions of Theorem 3 and γt (G01 ) ≤ 21 by the induction hypothesis.
Notice that wi+1 ui+1 xi+1 xi+2 is a path in G0 and wi ui xi xi−1 is a path in G. By Lemma
4, we can derive that Theorem 3 holds.
Hence, in the rest of the proof, we may assume that {xi , xi+1 } ∩ S2 6= ∅ for any
1 ≤ i ≤ t (the subscripts are taken modulo t). Since S2 is independent, we have |C| is even
and |C| ≥ 8 by Claims 4–7. Without loss of generality, let {x1 , x3 , · · · , xt−1 } ⊆ S2 , then
d(x2i ) ≥ 3 (i = 1, 2, · · · , 2t ). By Claim 3(iii), we may assume that d(x2 ) = 3 and d(u2 ) = 2
for u2 ∈ N (x2 ) − V (C). Let w2 ∈ N (u2 ) − {x2 }. By the choice of C, N (w2 ) ∩ N (xt ) = ∅.
If d(x6 ) + Σy∈N (x6 ) d(y) ≥ 10, let G1 = G − {x2 x3 , x5 x6 }, G0 = G1 + x3 x5 and G01 =
G0 [V (G0 ) − {w2 , u2 , x2 , x1 }] + {xt y : y ∈ N (w2 ) − {u2 }}. Then G01 satisfies (1) and
|G0 |
γt (G01 ) ≤ 21 by the induction hypothesis. Notice that w2 u2 x2 x1 is a path in G0 and G1
is a spanning subgraph of G. By Lemmas 1, 2 and 4, we can derive that Theorem 3 holds.
If d(x6 ) + Σy∈N (x6 ) d(y) ≤ 9, then d(x6 ) = 3 and d(u6 ) = 2 for u6 ∈ N (x6 ) − V (C).
Let w6 ∈ N (u6 ) − {x6 }, then N (w6 ) ∩ N (x8 ) = ∅ by the choice of C. When w2 6= w6 , let
G1 = G − {x2 x3 , x5 x6 }, G0 = G1 + x3 x5 , G01 = G0 [V (G0 ) − {w2 , u2 , x2 , x1 }] + {xt y : y ∈
N (w2 ) − {u2 }} and G02 = G01 [V (G01 ) − {w6 , u6 , x6 , x7 }] + {x8 y : y ∈ N (w6 ) − {u6 }}. Then
|G0 |
G02 satisfies (1) and γt (G02 ) ≤ 22 by the induction hypothesis. Notice that w6 u6 x6 x7 is a
path in G01 , w2 u2 x2 x1 is a path in G0 and G1 is a spanning subgraph of G. By Lemmas

14
 1, 2 and 4, we can derive that Theorem 3 holds. When w2 = w6 , then |C| = 8 and
N (u) ∩ N (w2 ) = ∅ for any u ∈ N (x8 ) − V (C). Let G1 = G − {x2 x3 , x5 x6 }, G0 = G1 + x3 x5 ,
and G01 = G0 [V (G0 ) − {w2 , u2 , x2 , x1 , x8 , x7 , x6 , u6 }] + {uy : u8 ∈ N (x8 ) − V (C), y ∈
|G0 |
N (w2 ) − {u2 }}. Then G01 satisfies (1) and γt (G01 ) ≤ 21 by the induction hypothesis.
Since for any MTDS T of G01 , we have T ∪ {w2 , u2 , x8 , x7 } is a TDS of G0 , γt (G0 ) ≤ n2 .
Since G1 is a spanning subgraph of G and dG0 (x3 ) = dG0 (x5 ) = 2, by Lemmas 1 and 2, we
can derive that Theorem 3 holds.
Therefore, the proof of Theorem 3 is complete.

References

[1] D. Archdeacon, J. Ellis-Monaghan, D. Froncek, P. Lam, B. Wei and R. Yuster,

Some Remarks on Domination, Journal of Graph Theory
[2] E. J. Cockayne, R. M. Dawe and S. T. Hedetniemi, Total domination graphs,
Networks 10 (1980), 211–219.
[3] O. Favaron, M. Henning, C. M. Mynhart and J. Puech, Total domination in graphs
with minimum degree three, Journal of Graph Theory 34 (2000), 9–19.
[4] L. Sun, An upper bound for total domination number, J. Beijing Inst. Technol.,
Engl. Ed. 4 (1995), 111–114.

View publication stats

15