536
Views
2
CrossRef citations to date
0
Altmetric
Articles

Specified holes with pairwise disjoint interiors in planar point sets

&

Abstract

A k-hole of a planar point set in general position is a convex k-gon whose vertices are elements of the set and whose interior contains no elements of the set. We discuss the minimum size of a point set that contains specified holes with disjoint interiors.

1 Introduction

In 1935, Erdős and Szekeres [Citation1] stated that for every integer t3 there is a smallest number f(t) such that every set of at least f(t) points in general position in the plane, contains a subset of points that are the vertices of a convex t-gon. The exact value of f(t) is a long standing open problem. A construction due to Erdős and Szekeres [Citation2] shows that f(t)2t2+1, which is also conjectured to be sharp. It is known that f(4)=5,f(5)=9 [Citation3] and f(6)=17 [Citation4]. The best known upper bound is due to Tóth and Valtr [Citation5], f(t)2t5t3+1. For a more detailed description of the Erdős and Szekeres theorem and its many ramifications, see the surveys by Bárány and Károlyi [Citation6] and Morris and Soltan [Citation7].

Erdős [Citation8] also asked the following combinatorial geometry problem in 1979: Find the smallest integer n(k) such that any set of n(k) points in general position in the plane, contains the vertices of a convex k-gon, whose interior contains no points of the set. Such a subset is called an empty convex k-gon or a k-hole of the set. Klein [Citation1] found n(4)=5, and n(5)=10 was determined by Harborth [Citation9]. Horton [Citation10] constructed arbitrarily large point sets which do not contain any 7-holes, so n(k) does not exist for k7. For the remaining case of n(6), Overmars exhibited a set of 29 points, the largest known, with no empty convex hexagons [Citation11]. About 10 years ago, the existence of n(6) was proved by Gerken [Citation12] and independently by Nicolás [Citation13]. Later, Valtr [Citation14] gave a similar version of Gerken’s proof. And recently, Koshelev improved the upper bound to n(6)463 [Citation15]. Therefore the current record of n(6) is 30n(6)463.

Fig. 1(a) n(4, 4) ≥ 9.

Fig. 1(a) n(4, 4) ≥ 9.

Fig. 1(b) m(4, 4) ≥ 7.

Fig. 1(b) m(4, 4) ≥ 7.

A pair of holes is said to be disjoint if their convex hulls do not intersect. We denote n(k,l) for kl by the smallest integer such that any set of n(k,l) points in general position in the plane, contains both a k-hole and an l-hole that are disjoint. Clearly, n(3,3)=6 and Horton’s result implies that n(k,l) does not exist for all l7. For this function, we showed n(3,4)=7 [Citation16] and n(4,4)=9 in [Citation17], and also determined n(3,5)=10,12n(4,5)13 and 17n(5,5)20 in [Citation20] later tightened to n(4,5)=12 and also improved the upper bound of n(5,5) to 19 [Citation21].

In [Citation19], we considered several problems for disjoint holes. Let n(k1,k2,,kl) be the smallest integer such that any set of n(k1,,kl) points contains a ki-hole for each i, 1il, where the holes are pairwise disjoint. We showed that n(2,3,4)=9,n(2,3,5)=11,n(3,4,4)=12,n(4,4,4)=14,15n(4,4,5)17 and more. In particular, any set of 15 points in general position in the plane is partitioned into a 1-hole, a 2-hole, a 3-hole, a 4-hole and a 5-hole which are pairwise disjoint, that is n(1,2,3,4,5)=15.

In this paper, the related problem is considered as follows. A family of holes is with disjoint interiors if their interiors are pairwise disjoint. We define m(k,l) for kl by the smallest integer such that any set of m(k,l) points in general position in the plane contains both a k-hole and an l-hole with disjoint interiors. Clearly, m(k,l)n(k,l) holds for any k,l, and also m(k,l) does not exist for all l7 by Horton’s result.

For example, an 8-point set in does not contain two disjoint 4-hole, implying that n(4,4)9. However, it contains two holes with disjoint interiors, formed by {v1,p1,2,p3,4,p4,1} and {v2,p2,3,p3,4,p1,2}. And shows m(4,4)7, that is, this 6-point set has no two 4-holes with disjoint interiors. We discuss two specified holes in Section 3 and three specified holes in Section 4.

2 Preliminaries

2.1 Notations and definitions

We first give notations and definitions used in the proofs. Throughout this work, we consider only planar point sets in general position. For such a point set P, we distinguish the vertices V(P) on the convex hull boundary from the remaining interior points I(P). Let V(P)={v1,v2,,vt} in clockwise order. We remark that when indexing a set of t points, we identify indices modulo t. If I(P)=, the convex t-gon formed by V(P) is called empty. A line segment vivi+1¯ is simply called an edge of V(P). Let R be a closed region in the plane. A point of P in the interior of R is generally said to be an interior point of R, and R is empty if it contains no interior points.

We denote the closed convex cone by γ(a;b,c) or γ(a;c,b) such that a is the apex and b and c lie on the boundary. If γ(a;b,c) is not empty, we define an attack point α(a;b,c) as the interior point of γ(a;b,c) such that γ(a;b,α(a;b,c)) is empty. A quasi-attack point α̃(a;b,c) in γ(a;b,c) is the point c or α(a;b,c) if γ(a;b,c) is empty or not, respectively. For δ=b or c of γ(a;b,c), let δ be a point collinear with a and δ so that a lies on the line segment δδ¯. Then we can consider a convex cone of γ(a;b,c) or γ(a;b,c).

Let l(a,b) be the line through a and b. Denote the closed half-plane bounded by the line l(a,b) that contains c or does not contain c by H(ab;c) or H(ab;c¯), respectively. For any elements a,b,c of P, we let P1 or P2 be a subset of P on H(ab;c) or H(ab;c¯), respectively, where P1P2={a,b}. Then we say that the cutting line l(a,b) divides P into P1 and P2.

An interior point pi,i+1 of P is said to be a friend to the edge vivi+1¯ of V(P) if γ(vi;vi+1,pi,i+1)γ(vi+1;vi,pi,i+1) is empty, e.g. . We represent a k-hole H by H=(v1vk)k if V(H)={v1,,vk} is in clockwise order.

2.2 Lemmas

We now present two lemmas used throughout the paper. Let P be a set of n points for n4, and V(P)={v1,,vt} in clockwise order.

Lemma 1

If there exists an edge of V(P) with no friend, then we have a cutting line which divides P into a 4-hole and the remaining n2 points.

Proof

We consider any edge of V(P), say v1v2¯. First, if v1v2v3 is empty, then v1v2¯ has no friend. For the quasi-attack point a1=α̃(v1;v3,v4), there exists a cutting line l(v1,a1) which divides P into a 4-hole of (v1v2v3a1)4 and the remaining n2 points, see (a).

If v1v2v3 is not empty, there is an attack point a2=α(v1;v2,v3), see (b). We remark that if the convex cone γ(a2;v1,v2) is empty, then a2 is the friend to the edge v1v2¯. Otherwise, for a3=α(a2;v1,v2), there exists a cutting line l(a2,a3) which divides P into a 4-hole of (v1v2a2a3)4 and the n2 points.

Fig. 2 Shaded areas are empty.

Fig. 2 Shaded areas are empty.

We remark that for any i, a friend pi,i+1 must lie in γ(vi;vi+1,vi+2)γ(vi+1;vi,vi1). Thus, any pair of consecutive edges does not have a common friend except for the case in which |P|=4 and |V(P)|=3. If |V(P)|>|I(P)|, then there exists an edge of V(P) having no friend. Therefore we give the next lemma.

Lemma 2

If |V(P)|>|I(P)|, then we have a cutting line which divides P into a 4-hole and the remaining n2 points.

3 Two holes with disjoint interiors

In this section, we discuss values of m(k,l), that is we consider two holes with disjoint interiors. If k=3, then the values are easily shown. For example, any set of four points has a 3-hole of T=(abc)3 and the remaining point p. Since p can see some edge of T, say ab¯, we obtain another 3-hole of (abp)3 such that these two holes are with disjoint interiors. Thus,

Proposition 1

m(3,3)=4.

Using n(4)=5, any set of five points has a 4-hole. The remaining point can also see some edge of the 4-hole. Thus,

Proposition 2

m(3,4)=5.

The next result is clearly shown by n(5)=10.

Proposition 3

m(3,5)=10.

The next result shows that a set of seven points has two 4-holes with disjoint interiors, and the value is tight.

Theorem 1

m(4,4)=7.

Proof

shows the lower bound of m(4,4)7. To prove the upper bound, let P be a set of seven points. If |V(P)|4, there exists a cutting line dividing P into a 4-hole and the remaining 5-point set S by Lemma 2. Then we can find another 4-hole of S using n(4)=5.

The remaining case is for |V(P)|=3. If there is an edge having no friend, we have a desired cutting line dividing into a 4-hole and the 5-point set containing another 4-hole by Lemma 1. Otherwise, there are three friends pi,i+1 to each edge vi,vi+1¯ of V(P)={v1,v2,v3}. Denote Ti=pi1,ivipi,i+1 for i=1,2,3. If the remaining point p lies in some Ti, say T2, we obtain two 4-holes of (pp3,1v1p1,2)4 and (pp2,3v3p3,1)4 with disjoint interiors. If p lies in p1,2p2,3p3,1, we also obtain (pp3,1v1p1,2)4 and (pp2,3v3p3,1)4.

The next result is a set of ten points has a 4-hole and a 5-hole with disjoint interiors, and the value is tight. Since a 10-point set has a 5-hole, we consider configurations of the remaining five points to prove the upper bound.

Theorem 2

m(4,5)=10.

Proof

Any set of ten points has a 5-hole by n(5)=10, so m(4,5)10. To prove m(4,5)10, let F=(v1v2v3v4v5)5 be a 5-hole of a given 10-point set and consider the closed convex cones γi=γ(vi;vi1,vi+1) for 1i5. Without loss of generality, we assume that γ1 contains the largest number of interior points of all the γi’s. Let I(γi) be a set of interior points of γi for any i, and we have three cases according to the number of I(γ1).

Case 1: |I(γ1)|3. Since there are at least five points on γ1, we have F and a 4-hole on γ1 by n(4)=5.

Case 2: |I(γ1)|=2. If γ(v1;v5,v2) is not empty, we have F and a 4-hole on H(v1v2;v5¯). And more if γ5 is not empty, there exist (v1v3v4v5α(v5;v1,v4))5 and a 4-hole on γ(v1;v3,v5). Thus, we consider the case in which γ5 is empty. By the same way, γ3γ(v4;v3,v5) is also empty, see . We have three subcases.

(a) |I(γ2)|=0: We obtain F and a 4-hole on H(v4v5;v1¯).

(b) |I(γ2)|=1: If I(γ1) lies in γ(v2;v3,v1), we have F and a 4-hole on H(v2v3;v1¯). Otherwise, we have a 6-hole of (v1wv2v3v4v5)6 for some point w of I(γ1). Then if γ(v2;v3,v4) is empty, we obtain (v1wv2v3v4)5 and a 4-hole on γ(v4;v2,v1). If not so, we obtain (v1wv2v4v5)5 and a 4-hole on γ(v2;v1,v4).

(c) |I(γ2)|=2: If γ(v2;v3,v1) is not empty, we obtain F and a 4-hole on H(v2v3;v1¯). Otherwise, we have a 5-hole of (v1wv2v4v5)5 for some point w of I(γ1) and a 4-hole on γ(v2;v1,v4).

Case 3: |I(γi)|=1 for each i. Let wi be precisely one interior point of γi.

(a) w1 lies on H(v2v3;v1): Clearly, we have a 6-hole (v1w1v2v3v4v5)6. If w3 lies in γ(v2;v3,v4), we have (v1w1v2v4v5)5 and (w3v4v2v3)4. Otherwise, we have (v1w1v2v3v4)5 and a 4-hole on γ(v4;v2,v1).

(b) w1 lies on H(v2v3;v1¯): If w2 lies on H(v3v4;v2), we have a 6-hole (v1v2w2v3v4v5)6 and we are done by the same way as in (a). Hence, w2 lies on H(v3v4;v2¯). If w2 is not contained in v2w1v3, we have F and (v3v2w1w2)4. Otherwise, we obtain F and (v3w2w1w3)4.

Fig. 3 Illustration of Case 2.

Fig. 3 Illustration of Case 2.

We next consider the case of two 5-holes with disjoint interiors. The upper bound is showed by the simple way using n(5)=10.

Theorem 3

14m(5,5)18.

Proof

A 13-point set as shown in gives m(5,5)14. To prove the upper bound, we consider an 18-point set, and let v1,v2 and v3 be three consecutive vertices of the set. Then there exists an interior point p such that each of γ(v2;v1,p) and γ(v2;p,v3) contains exactly ten points and it has a 5-hole by n(5)=10.

Fig. 4 m(5,5)14.

Fig. 4 m(5,5)≥14.

4 Three holes with disjoint interiors

In this section, we discuss the cases of three holes with disjoint interiors. Let m(k1,k2,k3) denote the smallest integer such that any set of m(k1,k2,k3) points contains a k1-hole, a k2-hole and a k3-hole with disjoint interiors. We first consider some cases of m(3,k,l) for 3kl5.

Proposition 4

m(3,3,3)=5,m(3,3,4)=6,m(3,3,5)=10,m(3,4,4)=7,m(3,4,5)=10.

Proof

Let P be a set of m(3,k,l) points. If P has a k-hole, an l-hole and the remaining points S, then some point p of S can see some edge of these holes. Therefore,

(i) m(3,3,3)=5 holds by m(3,3)=4, (ii) m(3,3,4)=6 holds by m(3,4)=5,

(iii) m(3,3,5)=10 holds by m(3,5)=10, and (iv) m(3,4,5)=10 holds by m(4,5)=10.

We show m(3,4,4)=7. By m(4,4)=7, if the remaining point exists, the point sees some edge of 4-holes. Otherwise, two 4-holes have only the common vertex p, namely (v1v2v3p)4 and (u1u2u3p)4. Then we have a 3-hole of (v1pu3)3 or (u1pv3)3. Hence we can show the existence of desired holes.

The next result shows a set of nine points has three 4-holes with disjoint interiors, and this value is tight.

Theorem 4

m(4,4,4)=9.

Proof

The lower bound of m(4,4,4) realizes an 8-point set as shown in , so m(4,4,4)9.

To prove the upper bound, let P be a set of nine points. We have the cases according to the number of vertices of P. If |V(P)|5, there exists a cutting line dividing P into a 4-hole and the remaining 7-point set S by Lemma 2. Then we have two 4-holes of S using m(4,4)=7.

Case 1: |V(P)|=4. If an edge of V(P) has no friend, we have a cutting line dividing into a 4-hole and the remaining seven points by Lemma 1, and we are done by m(4,4)=7.

Otherwise, every edge of P has its friend. Let V(P)={v1,v2,v3,v4} and pi,i+1 is a friend to an edge vivi+1¯ for any i,1i4. We consider the position of the remaining point p of P.

Subcase 1A: p lies in some Ti=pi1,ivipi,i+1 for i=1,2,3,4, say p1,2v2p2,3. If p lies in H(v2p4,1;v1), then we have a 4-hole of (pp4,1v1p1,2)4 and H(pp4,1;v3) has seven points. Otherwise, p lies in H(v2p4,1;v3). Then we have (pp2,3v3p3,4)4, (pp3,4v4p4,1)4 and (pp4,1v1p1,2)4.

Subcase 1B: p lies inside the quadrilateral p1,2p2,3p3,4p4,1. We obtain (p4,1v1p1,2p)4, (p1,2v2p2,3p)4 and (p2,3v3p3,4p)4 with disjoint interiors.

Case 2: |V(P)|=3. We only consider the case in which every edge of V(P) has its friend by Lemma 1. Let V(P)={v1,v2,v3} and denote Ti=pi1,ivipi,i+1 for i=1,2,3. There are two subcases.

Subcase 2A: Some Ti, say T2, is empty.

(i) v1p1,2p2,3 is not empty: Since we have q1=α(p2,3;p1,2,v1), there exists a cutting line l(q1,p2,3) dividing into a 4-hole (p1,2v2p2,3q1)4 and the remaining seven points.

(ii) v1p1,2p2,3 is empty: Since there is q2=α(v1;p2,3,p3,1), we consider the convex cone γ=γ(q2;v1,p2,3), see . If γ is not empty, for q3=α(q2;v1,p2,3) we have a cutting line l(q3,q2) dividing into two 4-holes of (v1p1,2q2q3)4 and (p1,2v2p2,3q2)4, and the remaining five points. There is a 4-hole of the 5-point set by n(4)=5 and we are done.

If γ is empty, q2 is a friend to the edge v1p2,3¯ of V(P) for P=P{v2,p1,2}. We remark that v1p2,3q2 cannot be contained in the convex hull of any 4-hole of P. Thus we obtain (p1,2v2p2,3q2)4 and two 4-holes of the 7-point set P.

Subcase 2B: Ti contains only the point wi of P for every i=1,2,3.

We consider the position of w2. If w2 lies in H(v2w1;v1), we have a cutting line l(w1,w2) dividing into (v1p1,2w2w1)4 and the 7-point set. Also, by the symmetry, if w2 lies in H(v2w3;v3), l(w2,w3) is the cutting line. Otherwise, we have three 4-holes of (p1,2v2w2w1)4, (w2v2p2,3w3)4 and (w1w2w3p3,1)4 with disjoint interiors.

Fig. 5 We consider the convex cone γ(q2;v1,p2,3).

Fig. 5 We consider the convex cone γ(q2;v1,p2,3′).

We next consider the case of two 4-holes and one 5-hole with disjoint interiors, that is not exact value.

Theorem 5

11m(4,4,5)12.

Proof

The lower bound realizes a 10-point set P such that |V(P)|=5 and each edge of V(P) has its friend. To show m(4,4,5)12, let P be a 12-point set. If |V(P)|7, there exists a cutting line dividing P into a 4-hole and the remaining 10-point set S by Lemma 2. We have both a 4-hole and a 5-hole of S by m(4,5)=10.

For 3|V(P)|6, we discuss under the condition in which every edge of V(P) has its friend by Lemma 1. Recall that V(P)={vi}i1 in clockwise order and pi,i+1 is the friend to an edge vivi+1¯. We consider a triangle Ti=pi1,ivipi,i+1 for any i.

Case 1: Some Ti, say T2, is empty.

Subcase 1A: If v1p1,2p2,3 is not empty, we have a cutting line l(q1,p2,3) for q1=α(p2,3;p1,2,v1) dividing into a 4-hole (p1,2v2p2,3q1)4 and the remaining ten points.

Subcase 1B: v1p1,2p2,3 is empty.

(i) v1p2,3v3 is not empty: Since there is q2=α(v1;p2,3,v3), we consider γ=γ(q2;v1,p2,3). If γ is not empty, for q3=α(q2;v1,p2,3) we have a cutting line l(q3,q2) dividing into a 5-hole of (v1p1,2p2,3q2q3)5 and the remaining eight points. There are two 4-hole of the 8-point set by m(4,4)=7. If γ is empty, since q2 is a friend to v1p2,3¯ of V(P) for P=P{v2,p1,2}, v1p2,3q2 cannot be contained in the convex hull of any 4-hole of P. We obtain (p1,2v2p2,3q2)4 and both a 4-hole and a 5-hole of the 10-point set P.

(ii) v1p2,3v3 is empty: Note that |V(P)|4. For q3=α˜(v1;v3,v4), we have a cutting line l(v1,q3) dividing into (v1p1,2p2,3v3q3)5 and the remaining eight points.

Case 2: No Ti is empty for any i. Since |P|3|V(P)|, we consider the following two subcases.

Subcase 2A: |V(P)|=4. Let wi be only the point of P inside Ti for each i. If w2 lies in H(v2w1;v1), we have a 4-hole (w2w1v1p1,2)4 and l(w1,w2) is the cutting line. Otherwise, w2 is in H(v2w1;v3). Then we have (w1p1,2v2w2)4, (w2p2,3w3p3,4)4 and (w1w2p3,4w4p4,1)5.

Subcase 2B: |V(P)|=3. There are two cases according to the number of points of P inside Ti.

(i) Some Ti, say T2 contains only the point w: If p1,2p2,3v1 is empty, we have a cutting line l(p2,3,q1) for q1=α(p2,3;v1,p3,1) dividing into (p1,2wp2,3q1v1)5 and the remaining eight points. If it is not empty, we consider γ=γ(q2;p2,3,p1,2) for q2=α(p2,3;p1,2,v1). Then if γ is not empty, we have a cutting line l(q3,q2) for q3=α(q2;p2,3,p1,2) dividing into six points containing (p1,2wp2,3q3q2)5 and the remaining eight points. If γ is empty, then q2 is a friend to p1,2p2,3¯ of V(P) for P=P{v2,w}. Hence we obtain (p1,2wp2,3q2)4, and a 4-hole and a 5-hole of the 10-point set P.

(ii) Every triangle Ti contains exactly two points of P: If some Ti, say T2 contains {w1,w2} such that Q={p1,2,w1,w2,p2,3} is in convex position, for q4=α˜(p1,2;p2,3,v3) we have a cutting line l(p1,2,q4) dividing into six points containing a 5-hole formed by Q{q4} and the remaining eight points. Otherwise, we have a configuration as shown in and we can obtain the desired holes.

Fig. 6 The final configuration in case 2.

Fig. 6 The final configuration in case 2.

5 Conclusions

1. We showed several results for m(k,l). In fact, the condition of integers k and l is for 3kl6, so the number of types for m(k,l) are ten cases. However, 30n(6)463 means that the function m(k,6) is not valid. Therefore, we checked out all the cases of m(k,l) for l5.

For a set of three holes, we can easily show the following results by a simple method. Let v1,v2 be consecutive vertices on the convex hull of a given point set. We consider a point p such that the closed convex cone γ(v1;v2,p) contains exactly ten points. Then we have a 5-hole on this convex cone by n(5)=10. Therefore, m(3,5,5)18, m(4,5,5)m(4,5)+8=18 and m(5,5,5)m(5,5)+826. The lower bounds of m(3,5,5) and m(4,5,5) are shown by 14m(5,5)m(3,5,5)m(4,5,5). And the lower bound of m(5,5,5) realizes the configuration as shown in , which implies n(5,5)17.

Fig. 7 m(5,5,5)17.

Fig. 7 m(5,5,5)≥17.

Proposition 5

14m(3,5,5)m(4,5,5)18, 17m(5,5,5)26.

Hence, for a set of three holes, we estimated all the cases except for m(k,l,6).

2. The following theorem is announced in [Citation22] without proof.

Theorem A

Any point set P with n=2k+3 elements in general position contains the vertices of k empty convex quadrilaterals with disjoint interiors.

Using this result, both m(4,4)=7 in Theorem 1 and m(4,4,4)=9 in Theorem 4 are derived. However, because the proof has been not published for ten years, we prove only our results in this article to introduce a new problem. In addition, we can show that the lower bound of m(4,4,,4) for k 4-holes is realized in the configuration of a 2k+2 point set P such that |V(P)|=k+1 and each edge has its friend. Therefore,

Proposition 6

m(4,4,,4k)2k+3

References

  • ErdősP., SzekeresG., A combinatorial problem in geometry, Compos. Math., 2 1935 463–470
  • ErdősP., SzekeresG., On some extremum problems in elementary geometry, Ann. Univ. Sci. Budapest, Eötvös, Sect, Math., 3/4 1960 53–62
  • KalbfleishJ.D., KalbfleischJ.D., StantonR.G., A combinatorial problem on convex regions, Proc. La. Conf. Comb., Graph Theory Comput., Lousiana State Univ., 1 1970 180–188
  • SzekeresG., PetersL., Computer solution to the 17-point Erdős-Szekeres problem, ANZIAM J., 48 2006 151–164
  • TóthG., ValtrP., The Erdős-Szekeres theorem: Upper bounds and related results GoodmanJ.E., PachJ., WelzlE.Cominatorial and Computational Geometry, vol. 522005557–568
  • BárányI., KárolyiGy., Problems and results around the Erdős–Szekeres convex polygon theorem AkiyamaJ., KanoM., UrabeM.Discrete and Computational Geometry, Japanese Conference, JCDCG 2000 Lecture Notes in Computer Science, vol. 2098 2001Springer91–105
  • MorrisW., SoltanV., The Erdős–Szekeres problem on points in convex position–A survey, Bull. Amer. Math. Soc., 3742000 437–458
  • Erdos̋P., Some combinatorial problems in geometry Proceedings Conference University Haifa Lecture Notes in Mathematics, vol. 792 1979 46–53
  • HarborthH., Konvexe fünfecke in ebenen Punktmengen, Elem. Math., 33 1978 116–118
  • HortonJ., Sets with no empty 7-gons, Canad. Math. Bull., 26 1983 482–484
  • OvermarsM.H., Finding sets of points without empty convex 6-gons, Discrete Comput. Geom., 29 2003 153–158
  • GerkenT., Empty convex hexagons in planar point sets, Discrete Comput. Geom., 39 2008 239–272
  • NicolásC.M., The empty hexagon theorem, Discrete Comput. Geom., 38 2007 389–397
  • ValtrP., On empty hexagons GoodmanJ.E., PachJ., PollackR.Surveys on Discrete and Computational Geometry, Twenty Years Later2008AMS433–441
  • KoshelevV.A., On Erdős-Szekeres problem for empty hexagons in the plane, Model. Anal. Iform. Sist., 1622009 22–74
  • UrabeM., On a partition into convex polygonss, Discrete Appl. Math., 64 1996 179–191
  • HosonoK., UrabeM., On the number of disjoint convex quadrilaterals for a plannar point set, Comput. Geom., Theory Appl., 20 2001 97–104
  • HosonoK., UrabeM., On the minimum size of a point set containing two non-intersecting empty convex polygons AkiyamaJ., KanoM., TanX.Discrete and Computational Geometry, Japanese Conference, JCDCG 2004 Lecture Notes in Computer Science, vol. 3742 2005 117–122
  • HosonoK., UrabeM., A minimal planar point set with specified disjoint empty convex subsets ItoH., KanoM., KatohN., UnoY.Computational Geometry and Graph Theory, International Conference, KyotoCGGT 2007 Lecture Notes in Computer Science, vol. 4535 2008 90–100
  • BhattacharyaB.B., DasS., On the minimum size of a point set containing a 5-hole and a disjoint 4-hole, Stud. Sci. Math. Hung., 48 2011 445–457
  • BhattacharyaB.B., DasS., Disjoint empty convex pentagons in planar point sets, Period. Math. Hungar., 66 2013 73–86
  • M. Lomeli-Haro, T. Sakai, J. Urrutia, Convex quadrilaterals of point sets with disjoint interiors, in: The Collection of Extended Abstracts, in: Computational Geometry and Graph Theory, International Conference, KyotoCGGT 2007.