Discrete Mathematics 340 (2017) 140–144
Contents lists available at ScienceDirect
Discrete Mathematics journal homepage: www.elsevier.com/locate/disc
Intersecting k-uniform families containing a given family Jun Wang a , Huajun Zhang b, * a b
Department of Mathematics, Shanghai Normal University, Shanghai 200234, PR China Department of Mathematics, Zhejiang Normal University, Jinhua 321004, PR China
article
a b s t r a c t
info
A family A of sets is said to be intersecting if A ∩ B ̸ = ∅ for any A, B ∈ A. Let m, n, k and r be positive integers with m ≥ 2k ≥ 2r > n ≥ k. A family F of sets is called
Article history: Received 19 September 2014 Received in revised form 29 July 2016 Accepted 31 July 2016
([m]) (m, n, k, r)-intersecting family if F is an intersecting subfamily of k containing {A ∈ (an [m]) : |A ∩ [n]| ≥ r }. Maximum (m, k, k, k)- and (m, k + 1, k, k)-intersecting families k
were determined by the well-known Erdős–Ko–Rado theorem and Hilton–Milner theorem, respectively. Recently, Li et al. determined maximum (m, n, k, k)-intersecting family when n = 2k − 1, 2k − 2, 2k − 3 or m is sufficiently large. In this paper, we determine all the maximum (m, n, k, r)-intersecting families. © 2016 Elsevier B.V. All rights reserved.
Keywords: Family of sets Intersecting family Erdős–Ko–Rado theorem Hilton–Milner theorem
1. Introduction A finite set X is called an n-set if its cardinality |X | equals n. In particular, let [n] denote the n-set {1, . . . , n} and for a positive integer k ≤ n, define [k, n] = {k, k + 1, . . . , n}. A(family of subsets of X is said to be intersecting if no two members ) (|X |) X are disjoint. The family of all k-subsets of X is denoted by k , whose cardinality is the well-known binomial coefficient k . ([n]) We are interested in the intersecting subfamilies of k . Here, a maximum intersecting family is an intersecting family of ([n]) ([n]) the largest size. Note that k is intersecting if n < 2k. While for n = 2k, an intersecting subfamily F of k contains at most one of A ∈
([n]) k
and its complement A = [n] \ A. So a maximum intersecting subfamily of
([n])
([2k]) k
has size
1 2k 2 k
( )
=
(2k−1) k−1
.
If all members of a family F ⊂ k contain a fixed element, say t ∈ [n], then F is obviously an intersecting family and is said to be a star, or a t-star ([n]if) the element t needs (n−1to ) be indicated. A star is said to be a trivial intersecting family. A trivial intersecting subfamily of k can have at most k−1 members. The general case is solved by Erdős, Ko, and Rado [3], whose
([n])
(n−1)
pioneering result states that if F is an intersecting subfamily of k , then |F | ≤ k−1 subject to n ≥ 2k, and equality holds if and only if F is a star, except for n = 2k. This theorem was reproved many times, and initiated a new field of combinatorial set theory. The reader is referred to Deza and Frankl [2], Frankl [4], Borg [1] for( surveys on relevant results. Hilton and ([nand ) (n− ) ]) n−1 k−1 Milner [6] proved that if F is a nontrivial intersecting subfamily of k , then |F | ≤ k−1 − k−1 + 1, and equality implies
([n])
that F consists of a fixed A ∈ k together with a t-star S such that t ̸ ∈ A and every member of S intersects A, except for ([k+1]) k = 3. By symmetry we may assume that t = 1 and A = [2, k + 1]. Then F contains . k In [7], Li et al. introduced a generalization of the above structure, which we state in a more general form as follows. Let m, n, k and r be positive integers with m ≥ 2k > n ≥ k ≥ r > 0, and set A(m, n, k, r) =
*
{ A∈
( ) [m ] k
} : |A ∩ [n]| ≥ r .
Corresponding author. E-mail addresses:
[email protected] (J. Wang),
[email protected] (H. Zhang).
http://dx.doi.org/10.1016/j.disc.2016.07.027 0012-365X/© 2016 Elsevier B.V. All rights reserved.
J. Wang, H. Zhang / Discrete Mathematics 340 (2017) 140–144
Clearly, A(m, n, k, k) =
, A(m, n, k, r) =
([n])
([m])
k
k
∑k (n)(m−n)
141
if m = n + k − r, and the size of A(m, n, k, r), written as f (m, n, k, r),
equals i=r i k−i . 2r > n, A(m, n, k, r) is an intersecting family. We call a family F (m, n, k, r)-intersecting if A(m, n, k, r) ⊆ F ⊆ ([m]When ) . Let α (m, n, k, r) denote the largest size of (m, n, k, r)-intersecting families. k Let F be an (m, n, k, r)-intersecting family. For every A ∈ F \ A(m, n, k, r), if |A ∩ [n]| ≤ n − r, then A would not intersect some B ∈ A(m, n, k, r). Therefore, n − r + 1 ≤ |A ∩[n]| ≤ r − 1. From this it is seen that if n = 2r − 1, then F = A(m, n, k, r), which is a trivial case. Suppose n ≤ 2r − 2 and set B(m, n, k, r) =
{ A∈
( ) [m ]
}
: n − r + 1 ≤ |A ∩ [n]| ≤ r − 1 k = A(m, n, k, n − r + 1) \ A(m, n, k, r).
(1)
Then F ⊆ A(m, n, k, r) ∪ B(m, n, k, r). Let β (m, n, k, r) denote the size of maximum intersecting subfamilies of B(m, n, k, r). Then it is easily seen that
α (m, n, k, r) = f (m, n, k, r) + β (m, n, k, r). Our tasks are indeed to determine β (m, n, k, r) and the intersecting families of size β (m, n, k, r). For any t ∈ [n], set Bt (m, n, k, r) = {A ∈ B(m, n, k, r) : t ∈ A}, and define Ht (m, n, k, r) = A(m, n, k, r) ∪ Bt (m, n, k, r).
Then Ht (m, n, k, r) is an (m, n, k, r)-intersecting family of size h(m, n, k, r) =
) k ( )( ∑ n m−n k−i
i
i=r
+
( )( ) r −1 ∑ n−1 m−n . i−1 k−i i=n−r +1
Consider the special case when n = 2r − 2:
) ( )} [2r − 2] [2r − 1, m] ,B ∈ . r −1 k−r +1 { ([2r −2]) ([2r −1,m])} For any maximum intersecting subfamily T of r −1 , set BT (m, 2r − 2, k, r) = A ∪ B : A ∈ T , B ∈ k−r +1 , and define B(m, 2r − 2, k, r) =
{
(
A∪B:A∈
HT (m, 2r − 2, k, r) = A(m, 2r − 2, k, r) ∪ BT (m, 2r − 2, k, r).
Then HT (m, 2r − 2, k, r) is an (m, 2r − 2, k, r)-intersecting family of size h(m, 2r − 2, k, r). Theorem 1.1. Let m, n, k and r be positive integers with m ≥ 2k ≥ 2r ≥ n + 2 ≥ k + 2. If F is an (m, n, k, r)-intersecting family, then |F | ≤ h(m, n, k, r), and equality holds ([2r −if2]and ) only if F = Ht (m, n, k, r) for some t ∈ [n], or F = HT (m, 2r − 2, k, r) for some maximum intersecting family T in r −1 , except for m = 2k. Note that, in [7], Li et al. dealt with the case r = k. They proved that the result of Theorem 1.1 holds for n = 2k − 1, 2k − 2, 2k − 3 or m sufficiently large, and asked whether or not it is always the case. Theorem 1.1 has answered their question in a more general setting. Our tool is the shift operation, which was introduced by Erdős, Ko and Rado in the original paper [3] (see also [4]). We shall review it in the next section, and present some preliminary results in Section 3. The proof of Theorem 1.1 is given in Section 4. 2. Shift operation For a family F ⊆
{ sij (F ) =
([m]) k
and i, j ∈ [m] with i < j, define
(F \ {j}) ∪ {i}, F,
if j ∈ F , i ̸ ∈ F , (F \ {j}) ∪ {i} ̸ ∈ F ; otherwise
for each F ∈ F , and sij (F ) = {sij (F ) : F ∈ F }. Clearly, |sij (F )| = |F |, and it was proved in [3] that sij (F ) is an intersecting family if F is so. Apply this procedure to F until we obtain a family H such that sij (H) = H for ([many ) 1 ≤ i < j ≤ m. Such an H is called compressed. ] Now, let F be an intersecting subfamily of k satisfying sim (F ) = F for each i < m. Then F has a canonical decomposition F0 ∪ F1 , where F0 = {F ∈ F : m ̸ ∈ F } and F1 = {F ∈ F : m ∈ F }. Then F0 is an intersecting ([mF−1= ]) subfamily of . A key point that was proved in [3] is that the family F1′ = {F \ {m} : F ∈ F1 } is an intersecting k subfamily of
([m−1]) k−1
if m > 2k.
142
J. Wang, H. Zhang / Discrete Mathematics 340 (2017) 140–144
3. Preliminary results Recall that
{
A(m, n, k, r) =
( ) [m ]
A∈
k
} : |A ∩ [n]| ≥ r ,
which is of size f (m, n, k, r) =
) k ( )( ∑ n m−n k−i
i
i =r
.
For any t ∈ [n], define At (m, n, k, r) = {A ∈ A(m, n, k, r) : t ∈ A}. Then At (m, n, k, r) is an intersecting subfamily of A(m, n, k, r), which is of size f (m, n, k, r) = ∗
)( ) k ( ∑ n−1 m−n i−1
i =r
k−i
.
Lemma 3.1. Let m, n, k (nand ) r be positive integers with m (m≥ ) n ≥ k ≥ r > 0, n ≥ k + r and m − n ≥ k − r. Then (1) f (m, n, k, k) = k and f (m, m − k + r , k, r) = k ; and for m > n, (2) f ∗ (m, n, k, r) = f ∗ (m − 1, n, k, r) + f ∗ (m − 1, n, k − 1, r); (3) h(m, n, k, k) = h(m − 1, n, k, k) + f ∗ (m − 1, n, k − 1, n − k + 1); (4) h(m, n, k, r) = h(m − 1, n, k, r) + h(m − 1, n, k − 1, r) where k > r. Proof. (1) is obtained by the following simple observation: A(m, n, k, r) = k if r = k and A(m, n, k, r) = if k m = n + k − r. (2)–(4) can be directly verified in a similar way. Here we show only (3) and leave the others to the reader.
([n])
([m])
( )( ) k−1 ∑ n−1 m−n k i−1 k−i i=n−k+1 ( ) ( ) [( ) ( )] k−1 ∑ n n−1 m−1−n m−1−n = + + k i−1 k−i k−i−1 i=n−k+1 ( ) ( )( ) ( )( ) k−1 k−1 ∑ ∑ n n−1 m−1−n n−1 m−1−n = + + k i−1 k−i i−1 k−i−1 i=n−k+1 i=n−k+1
h(m, n, k, k) =
( ) n
+
= h(m − 1, n, k, k) + f ∗ (m − 1, n, k − 1, n − k + 1).
□
Lemma 3.2. Let m, n, k and r be positive integers with m ≥ n ≥ k ≥ r > 0, n ≥ k + r and m > n + k − r. If F is an intersecting subfamily of A(m, n, k, r) with sim (F ) = At (m; n; k; r) for some t ∈ [n] and i < m, then F = At (m; n; k; r). Proof. Note that m > n ≥ t. Suppose i ̸ = t. By definition, for each A ∈ A(m; n; k; r), A ∈ At (m; n; k; r) if and only if t ∈ A, or equivalently, t ∈ sim (A), which implies the statement of the lemma. Suppose now i = t and write G = stm (F ) = At (m; n; k; r). For j ≥ r, set Fj = {A ∈ F : |A ∩ [n]| = j} and Gj = {A ∈ G : |A ∩ [n]| = j}. Clearly, for each A ∈ Fj , stm (A) ∈ Fj+1 subject to stm (A) ̸ = A, hence Gr ⊆ Fr . Note that
{ Gj =
α∪β :α ∈
(
[n]
)
j
with t ∈ α, β ∈
(
[n + 1, m] k−j
)}
,
j = r , . . . , k,
and n ≥ k + r and m − k > k − r. If Fk−1 ̸ ⊆ Gk−1 , that is, there is an A ∈ Fk−1 with t ̸ ∈ A, then we may find a B ∈ Gr with A ∩ B = ∅. Hence Fk−1 ⊆ Gk−1 , that is, Fk−1 is invariant under the action of stm , which yields Fk = Gk . From this it follows Fr = Gr , which implies Gr +1 ⊆ Fr +1 . This process is briefly expressed as Gr ⊆ Fr ⇒ Fk−1 ⊆ Gk−1 ⇒ Fk = Gk ⇒ Fr = Gr ⇒ Gr +1 ⊆ Fr +1 . Repeating this process we obtain that Fj = Gj for all r ≤ j ≤ k, that is, Fr = Gr . This completes the proof. □ Lemma 3.3. Let m, n, k and r be positive integers with m ≥ n ≥ k ≥ r > 0, n ≥ k + r and m − n ≥ k − r. If F is an intersecting subfamily of A(m, n, k, r), then
|F | ≤ f ∗ (m, n, k, r) =
)( ) k ( ∑ n−1 m−n i=r
i−1
k−i
,
and equality implies that F = At (m, n, k, r) for some t ∈ [n] unless n = 2k = 2r or m − n = k − r > 0.
(2)
J. Wang, H. Zhang / Discrete Mathematics 340 (2017) 140–144
143
Proof. Note that A(m, n, k, r) = k for k = r, and A(m, n, k, r) = k for m = n + k − r. In both cases, the classical Erdős–Ko–Rado theorem implies inequality (2), and if equality holds, then F is a t-star for some t ∈ [n] unless n = 2k = 2r or m − n = k − r > 0. Suppose k > r. We apply induction on m to prove inequality (2). We have seen that it holds for m = n + k − r. Assume that m > n + k − r. It is easily seen that A(m, n, k, r) is compressed. By Lemma 3.2 we may assume that sim (F ) = F for each i < m. We now consider the canonical decomposition of F = F0 ∪ F1 , where F0 = {A ∈ F : m ̸ ∈ A} and F1 = {A ∈ F : m ∈ A}. Clearly, F0 is an intersecting subfamily of A(m − 1, n, k, r) with m − 1 ≥ n + k − r, and F1′ = {A \ {m} : A ∈ F1 } is an intersecting subfamily of A(m − 1, n, k − 1, r) with m − 1 > n + k − 1 − r. Then, by the induction hypothesis we have
([n])
([m])
|F | = |F0 | + |F1 | ≤ f ∗ (m − 1, n, k, r) + f ∗ (m − 1, n, k − 1, r) = f ∗ (m, n, k, r). Suppose that the equality holds. Then |F0 | = f ∗ (m − 1, n, k, r) and |F1 | = f ∗ (m − 1, n, k − 1, r). Keeping m > n + k − r in mind, we apply induction on k to show F = At (m, n, k, r) for some t ∈ [n]. It is clearly true for k = r unless n = 2k = 2r. Assume that the result holds for k − 1 ≥ r. By the induction hypothesis we have that F1′ = At (m − 1, n, k − 1, r) for some t ∈ [n] because n ≥ k + r > k − 1 + r ≥ 2r. Note that F0 ∪ F1′ is intersecting as F0 and F1′ are intersecting. Clearly, for each set A in A(m; n; k; r) \ At (m; n; k; r), we can choose a set B in At (m − 1; n; k − 1; r) such that A ∩ B = ∅. Since F1′ = At (m − 1; n; k − 1; r), we therefore have F ⊆ At (m; n; k; r), and the equality follows from |F | = |At (m; n; k; r)|. □ 4. Proof of Theorem 1.1
([m])
If m = 2k, then for each A ∈ k , either A ∈ Ht (m; n; k; r) or [m] \ A ∈ Ht (m; n; k; r) for t ∈ [m], so Ht (m; n; k; r) is ([m]) a maximum intersecting subfamily of k . Assume now m > 2k. Let F be a maximum (m, n, k, r)-intersecting family. We ∗ have seen that F = A(m, n, k, r) ∪ F , where F ∗ is a maximum intersecting subfamily of B(m, n, k, r). Let us consider the case n = 2r − 2:
) ( )} [2r − 2] [2r − 1, m] B(m, 2r − 2, k, r) = A ∪ B : A ∈ ,B ∈ . r −1 k−r +1 ([2r −1,m]) Since m > 2k, |[2r − 1, m]| > 2(k − r + 1) so that k−r +1 is not an intersecting family. From this it follows that F ∗ = ([2r −2]) BT (m, 2r − 2, k, r), that is, F = HT (m, 2r − 2, k, r), for some maximum intersecting family T in r −1 (cf. [5,8,9], or [7, {
(
Proof of Theorem 10]). Assume n < 2r − 2. By Lemma 3.2 we may assume that F is invariant under the action of sim for each i < m, and consider the canonical decomposition F0 = {A ∈ F : m ̸ ∈ A}, F1 = {A ∈ F : m ∈ A} and F1′ = {A \ {m} : A ∈ F1 }. Then F0 is an (m − 1, n, k, r)-intersecting family, and F1′ is an (m − 1, n, k − 1, r)-intersecting family if k > r. The proof will be completed by induction on m. We have seen that the result is true for m = 2k. Assume now m > 2k. There are two cases to be considered. First, suppose k = r. In this case, by the induction hypothesis, |F0 | ≤ h(m − 1, n, k, k) and equality holds if and only if F0 = Ht1 (m − 1, n, k, k) for some t1 ∈ [n], except for m − 1 = 2k; and F1′ ⊆ A(m − 1, n, k − 1, n − k + 1). It is easy to check that the parameters satisfy the conditions in Lemma 3.3, so |F1′ | ≤ f ∗ (m − 1, n, k − 1, n − k + 1), and equality implies F1′ = At2 (m − 1, n, k − 1, n − k + 1) for some t2 ∈ [n]. By Lemmas 3.3 and 3.1(3) we have that
|F | = |F0 | + |F1 | ≤ h(m − 1, n, k, k) + f ∗ (m − 1, n, k − 1, n − k + 1) = h(m, n, k, k). Then maximality of F implies the equality. In this case it is easy to show t1 = t2 , hence F = Ht (m, n, k, k) for some t ∈ [n]. Next, suppose k > r. In this case, we have seen that F0 is an (m − 1, n, k, r)-intersecting family, and F1′ is an (m − 1, n, k − 1, r)-intersecting family. By the induction hypothesis and Lemma 3.1(4) we have that
|F | = |F0 | + |F1 | ≤ h(m − 1, n, k, r) + h(m − 1, n, k − 1, r) = h(m, n, k, r). Similarly to the first case, the maximality of F implies F = Ht (m, n, k, r) for some t ∈ [n]. □ Acknowledgments The first author is supported by the National Natural Science Foundation of China (Nos. 11171224 and 11231004); the second author is supported by the National Natural Science Foundation of China (No. 11371327). References [1] P. Borg, Intersecting families of sets and permutations: a survey, in: A.R. Baswell (Ed.), in: Advances in Mathematics Research, vol. 16, Nova Science Publishers, 2011, pp. 283–299. [2] M. Deza, P. Frankl, Erdős–Ko–Rado theorem–22 years later, SIAM J. Algebr. Discrete Methods 4 (1983) 419–431.
144 [3] [4] [5] [6] [7] [8] [9]
J. Wang, H. Zhang / Discrete Mathematics 340 (2017) 140–144 P. Erdős, C. Ko, R. Rado, Intersection theorems for systems of finite sets, Quart. J. Math. Oxford Ser. 2 (12) (1961) 313–318. P. Frankl, The shifting technique in extremal set theory, in: C. Whitehead (Ed.), Combinatorial Surveys, Cambridge Univ. Press, 1987, pp. 81–110. P. Frankl, An Erdős–Ko–Rado Theorem for direct products, European J. Combin. 17 (1996) 727–730. A.J.W. Hilton, E.C. Milner, Some intersection theorems for systems of finite sets, Q. J. Math. 2 (1967) 369–384. W. Li, B. Chen, K. Huang, K. Lih, Intersecting k-uniform families containing all the k-subsets of a given set, Electron. J. Combin. 20 (2013) #P38. H.J. Zhang, Primitivity and independent sets in direct products of vertex-transitive graphs, J. Graph Theory 67 (2011) 218–225. H.J. Zhang, Independent sets in direct products of vertex-transitive graphs, J. Combin. Theory Ser. B 102 (2012) 832–838.