Graphs and Combinatorics (1999) 15 : 137±142
Graphs and Combinatorics ( Springer-Verlag 1999
2; k-Factor-Critical Gr...
12 downloads
402 Views
81KB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Graphs and Combinatorics (1999) 15 : 137±142
Graphs and Combinatorics ( Springer-Verlag 1999
2; k-Factor-Critical Graphs and Toughness Mao-Cheng Cai1*, Odile Favaron2, and Hao Li2 1 Institute 2 LRI,
of Systems Science, Academia Sinica, Beijing 100080, China Bat. 490, Universite Paris-Sud, 91405 Orsay cedex, France
Abstract. A graph is
r; k-factor-critical if the removal of any set of k vertices results in a graph with an r-factor (i.e. with an r-regular spanning subgraph). We show that every ttough graph of order n with t V 2 is
2; k-factor-critical for every non-negative integer k U minf2t ÿ 2; n ÿ 3g, thus proving a conjecture as well as generalizing the main result of Liu and Yu in [4].
1. Introduction All graphs G
V ; E under consideration are simple and ®nite of order jV j n. If A J V , we denote by G ÿ A the subgraph obtained from G by deleting the vertices in A together with the edges incident with vertices in A. For a vertex x of G; NA
x is the set of neighbors of x in A and dA
x jNA
xj. If the two subsets A and B of V are disjoint, E
A; B is the set of edges between A and B and e
A; B jE
A; Bj. We denote by d
G and k
G the minimum degree and the vertex connectivity of G. If S is a cutset of G, o
G ÿ S is the number of connected components of G ÿ S. When G is not complete, its toughness is de®ned by jSj S is a cutset of G . It is clear from the de®nition that t
G : min o
G ÿ S
d
G V k
G V 2t
G. The graph G is said to be t-tough for every positive t U t
G. For a clique Kn , we usually put k
Kn n ÿ 1 and t
Kn y. An rfactor of G, where r is a positive integer, is an r-regular spanning subgraph of G. In particular, a 1-factor is a perfect matching. When r is odd, only graphs of even order can admit an r-factor. Tutte gave necessary and su½cient conditions for a graph to have an r-factor ([6] for r 1 and [7] for r V 2). Let us recall these conditions for r V 2. For a given positive integer r and a pair S; T of disjoint subsets of V, a component C of G ÿ
S U T is said to be odd if rjV
Cj e
T; V
C is odd. We denote by o1
S; T the number of odd components of G ÿ
S U T and let * Research partially supported by National Natural Science Foundation of China and by the CoopeÂration Franco-Chinoise PRA 93-M10
138
M.-C. Cai et al.
qG
S; T : rjSj ÿ rjTj
X
dGÿS
v ÿ o1
S; T:
vAT
Note that when r is even, the odd components C are de®ned by e
T; V
C odd and, inP particular, e
T; V
C V 1 for an odd component C. This implies that for r dGÿS
v ÿ o1
S; T V 0 and thus qG
S; T V rjSj ÿ rjTj. even, vAT
Theorem A. (Tutte [7]): (i) qG
S; T has always the same parity as nr. (ii) G has an r-factor if and only if qG
S; T V 0 for every pair S; T of disjoint subsets of V. The graph G is said
r; k-factor-critical if G ÿ X admits an r-factor for every subset X of k elements of V (when r 1, we usually simply say k-factor-critical). For r V 2, the study of these graphs has been initialized by Liu and Yu [4] under the name of
r; k-extendable graphs. We prefer to keep here the term factorcritical because usually, in the term extendable, X is not any subset of k vertices but must satisfy some given properties. Using Tutte's Theorem, Liu and Yu found the following characterization of
r; k-factor-critical graphs. Theorem B. (Liu and Yu [4]): Let r; k be integers with r V 2 and k V 0, and G a graph of order n V r k 1. Then G is
r; k-factor-critical if and only if qG
S; T V rk for any pair S; T of disjoint subsets of V with jSj V k. They also gave a su½cient condition for a graph to be
2; k-factor-critical in terms of its toughness. Theorem C. (Liu and Yu [4]): Let G be a graph of order n and toughness t
G V 3. Then G is
2; k-factor-critical for every integer k such that 3 U k U t
G and k U n ÿ 3. However they think Theorem C is not best possible and propose Conjecture D. (Liu and Yu [4]): Let G be a graph of order n and toughness t
G. If t
G V q and n V 2q 1 for some integer q V 1, then G is
2; 2q ÿ 2factor-critical. Our purpose is to prove this conjecture with the necessary restriction t
G V 2 since it is proved in [1] that for any positive real number e there exists an
r ÿ etough graph which has no r-factor. 2. The main result Theorem 1. Let G be a t-tough graph of order n with t V 2. Then G is
2; k-factorcritical for every non-negative integer k U minf2t ÿ 2; n ÿ 3g and the bound 2t ÿ 2 on k is sharp. Proof: First we may assume G to be not complete for otherwise the result is obvious. Now suppose, to the contrary, that G is not
2; k-factor-critical. Then, by
2; k-Factor-Critical Graphs and Toughness
139
Theorem B, there exist a pair S; T of disjoint subsets of V with jSj V k such that qG
S; T < 2k, implying by Theorem A(i) qG
S; T U 2k ÿ 2: Among all such pairs, we choose a pair S; T with additional properties: (i) qG
S; T is minimum and (ii) subject to (i), T is minimal under inclusion. P dGÿS
v 0; o1
S; q 0, and thus qG
S; T Then T 0 q. For otherwise vAT
2jSj V 2k, a contradiction. We write U : V
G ÿ
S U T;
jSj s;
jTj t:
Thus s V k and t V 1. In order to prove the theorem we need the following claims, whose ideas already appeared in [4]. Claim 1: (a) T is independent. (b) For all y A T, each edge of E
y; U joins y to an odd component of G ÿ
S U T, and, moreover, di¨erent edges of E
y; U join y to di¨erent odd components of G ÿ
S U T. Proof of Claim 1: For any y A T, let L
y be the component of G ÿ
S U
T n fyg containing y, and put h
y 1 if L
y is odd for the pair S; T n fyg, 0 otherwise. Let also u
y be the number of odd components of GP ÿ
S U T joined to y by some edges. Then qG
S; T n fyg 2s ÿ 2
t ÿ 1 dGÿS
v ÿ
o1
S; T ÿ u
y h
y. On the other hand, qG
S; T n fyg V v A Tnfyg
qG
S; T 2 by the choice of
S; T and by parity. Hence 2 UqG
S; T n fyg ÿ qG
S; T 2 ÿ dGÿS
y u
y ÿ h
y 2 ÿ dT
y ÿ e
y; U u
y ÿ h
y. But dT
y V 0, e
y; U ÿ u
y V 0 and h
y V 0. Therefore, dT
y 0, implying (a), h
y 0 and e
y; U u
y, implying (b). Note that by (b), there is no edge between T and the even components of G ÿ
S U T. Claim 2: For t V 2, s e
T; U ÿ o1
S; T V tt. Proof of Claim 2: Let us label the odd components Ci of G ÿ
S U T in such a way that for i U o2 , each component Ci contains at least one vertex with exactly one neighbor in T, and for i > o2 , no vertex of Ci has exactly one neighbor in T. For short we put o1
S; T o1 and note that 0 U o2 U o1 . For each i U o2 , we choose in Ci one vertex ui such that dT
ui 1 and put L fu1 ; u2 ; ; uo2 g. If o2 0 then L q. We denote by W the set NU
T (by Claim 1(b), W J 6 V
Ci ) and put Z W nL. Each vertex of W has at least one 1UiUo1
neighbor in T. Moreover by the de®nition of o2 and the fact that e
T; Ci is odd for all i, each component Ci with i > o2 contains at least one vertex having at least three neighbors in T. Therefore e
T; U V jW j 2
o1 ÿ o2 jZj o1
140
M.-C. Cai et al.
o1 ÿ o2 V jZj o1 . The equality e
T; U jZj o1 occurs if and only if o2 o1 and each vertex of W has exactly one neighbor in T. On the other hand, by the construction of Z, the t vertices of the set T, which is independent by Claim 1, belong to t di¨erent components of G ÿ
S U Z. Since t V 2, S U Z is a cutset of G and thus jS U Zj V tt. Hence s e
T; U ÿ o1
S; T V tt, as required. Claim 3: For t V 2 and s 2; s e
T; U ÿ o1
S; T > tt. Proof of Claim 3: Suppose s e
T; U ÿ o1
S; T tt. Then by the proof of Claim 2, o2 o1 , every vertex of W NU
T has exactly one neighbor in T, and tt jS U Zj, that is, the number of components of G ÿ
S U Z is exactly t (in particular G ÿ
S U T has no even component). Now, since d
G V 2t
G V 4 and s 2, and by Claim 1(b), some vertex y of T has one neighbor in at least two di¨erent components Ci , say y1 in C1 and y2 in C2 . In the choice of L, let us take u1 y1 and u2 y2 . For the set R S U Z U fyg of cardinality jZj 3, the number of components of G ÿ R is at least
t ÿ 1 2 t 1, and thus t
t 1 U jZj 3. This contradicts tt jS U Zj jZj 2 since t V 2. We distinguish three cases according to the values of t. Case t 1. Then s V 2t ÿ 1. Indeed, say T f yg. If e
y; U U 1, clearly s V 2t ÿ 1. And if e
y; U V 2, then, by Claim 1(b), o1
S; T e
y; U and S U fyg is a cutset, implying s V to1
S; T ÿ 1 V 2t ÿ 1. Hence s V k 1 and qG
S; T V 2s ÿ 2t V 2k, a contradiction. Case t 2. By Claim 1, either E
T; U is empty or there exists an odd number of edges, and thus exactly one edge, between T and each odd component of G ÿ
S U T. Therefore the two vertices of T belong to two di¨erent components of G ÿ S and S is a cutset. Hence s V 2t V k 2 and thus qG
S; T V 2s ÿ 2t V 2k, a contradiction, Case t V 3. Then qG
S; T s ÿ 2t
s e
T; U ÿ o1
S; T V s ÿ 2t tt by Claim 2, with a strict inequality when s 2 by Claim 3. As t V 2, then k U s U qG
S; T U 2k ÿ 2
implying k V 2. Let us show k V 3. Indeed, if k 2, the equality occurs everywhere in
and s 2, contradicting Claim 3. k2 , we have As s V k; t V 3 and t V 2 k2 5k ÿ2 ÿ 3 > 2k ÿ 2; qG
S; T V k 3 2 2 a contradiction. Since the assumption qG
S; T < 2k leads to a contradiction in all cases, qG
S; T V 2k and thus G is
2; k-factor-critical by Theorem B. The proof is complete. To show the sharpness of our result, consider the graph G consisting of a clique of vertex set A S U fyg with jSj k V 3, a second clique C of order at least 2,
2; k-Factor-Critical Graphs and Toughness
141
all the edges between C and S, and one edge yz for some vertex z A C. The graph k1 V 2 and is not
2; k-factor-critical since G ÿ S has G has toughness t
G 2 no 2-factor. Hence a t-tough graph with t V 2 and n V 2t 2 is not necessarily
2; 2t ÿ 1-factor-critical. r An obvious consequence of Theorem 1 is the following. Corollary 3. (conjecture D): Let q be an integer V 1 and G a graph of order n V 2q 1 and toughness t
G V maxfq; 2g. Then G is
2; 2q ÿ 2-factor-critical. Note that for the values of t
G belonging to intervals of the form 2q 1 ; q 1 , Theorem 1 shows that G is
2; 2q ÿ 1-factor-critical and thus is 2 slightly stronger than Conjecture D. 3. Open problem For r 1, a theorem similar to Theorem 1 already exists: Theorem E. (Favaron [3]). Let G be a t-tough graph of order n with t > 1. Then G is
1; k-factor-critical for every non-negative integer k such that n k is even, k < 2t and k U n ÿ 2. It would be interesting to generalize Theorems E and 1 to larger values of r and to determine functions t0
f and k0
t; f such that any t-tough graph with t > t0
f is
r; k-factor-critical for every non-negative integer k with
n k f even, k U k0
t; f and k U n ÿ
f 1. Added, as a partial answer to the open problem, the following results have been obtained [5, 8]: (1) Every t-tough graph of order n V 12 with t V 4 is
3; k-factor-critical for every non-negative integer k such that n k even and k U minf2t ÿ 3; n ÿ 7g. (2) Every t-tough graph of order n V 14 with t V 5 is
4; k-factor-critical for every non-negative integer k U minf2t ÿ 4; n ÿ 8g. Acknowledgments. This work was done while the ®rst author was visiting LRI, Universite Paris-Sud, he wishes to thank LRI and Professor Hao Li for their hospitality. The authors are grateful to Professor H. Enomoto for his stimulating discussions, which led to Theorem 1 extended to case t 2.
References 1. Enomoto, H., Jackson, B., Katerinis P., and Saito, A.: Toughness and the existence of k-factors, J. Graph Theory 9, 87±95 (1985) 2. Enomoto, H.: Toughness and the existence of k-factors III, Discrete Math. 189, 277± 282 (1998)
142
M.-C. Cai et al.
3. Favaron, O.: On k-factor-critical graphs, Discussiones Mathematicae-Graph Theory 16, 41±51, (1996) 4. Liu G., Yu, Q.: k-factors and extendability with prescribed components, submitted 5. Shi, M., Yuan, X., and Cai, M.:
3; k-Factor-critical graphs and toughness, submitted 6. Tutte, W.T.: The factorization of linear graphs, J. London Math. Soc. 22, 107±111, (1947) 7. Tutte, W.T.: The factors of graphs, Canad. J. Math. 4, 314±328, (1952) 8. Yuan, X., Shi, M., and Cai, M.:
4; k-Factor-critical graphs and toughness, in preparation
Received: December 16, 1996 Revised: September 17, 1997