On distinct distances in homogeneous sets in the Euclidean space
József SolymosiDepartment of Mathematics, University of British Columbia, Vancouver, BC, Canada V6T 1Z2, Email: solymosi@math.ubc.ca The research was supported by OTKA and NSERC grants.
Csaba D. TóthDepartment of Mathematics, MIT, Cambridge, MA 02139, USA, Email: toth@math.mit.edu
Abstract
A homogeneous set of
points in the
-dimensional Euclidean space determines at least
distinct distances for a constant
. In three-space, we slightly improve our general bound and show that a homogeneous set of
points determines at least
distinct distances.
1 Introduction
The history of the distinct distance problem goes back to Erdős [10] who asked the question: What is the minimal number
of distinct distances determined by
points in the
-dimensional Euclidean space
? The example of
points in the
-dimensional integer grid
shows that
for any
and, in particular,
.
Erdős conjectured that these bounds are essentially optimal.
Research efforts on the distinct distance problem have lead to powerful methods (such as the crossing theory [24] , the
-cutting theory [6] ) that found innumerable applications in discrete and computational geometry. An excellent survey by Pach and Sharir [18] elaborate on the history of the distinct distance problem and its connections to other fields of discrete mathematics. The currently known best lower bound for the distinct distance problem in the plane,
, is due to Katz and Tardos [16] . Their proof combines results of Solymosi and Tóth [20] with additive number theory.
In higher dimensions, fewer results are known. Aronov et al. [2] showed that the number of distinct distances determined by a set of
points in three-dimensional space is
for any
. Solymosi and Vu [22] proved a general lower bound of
for any
.
In this paper, we consider the number
of distinct distances in homogeneous sets of
points in
. A finite point set
is homogeneous if the following two conditions hold:
lies in the interior of an axis-aligned
-dimensional cube
of volume
, and any unit cube in
contains at most
points of
. Homogeneous sets represent an important special case for the distinct distance problem because the best known upper bound constructions (the
-dimensional integer grids) are homogeneous, and because of numerous connections to analysis [13, 11] . Iosevich [12] studied the distinct distance problem for homogeneous sets (with additional restrictions). He showed that
for
. Solymosi and Vu [21] proved a general bound
for
. In the case
, they have also obtained a slightly better bound
. In this paper, we improve all previous lower bounds on the number of distinct distances in homogeneous sets of
points in
.
Theorem 1
For any
, every homogeneous set
of
points in
determines at least
distinct distances. Moreover, there is a point
from which there are this many distinct distances to other points of
.
For
, and
, our general lower bound is
,
, and
. In three-dimensions, we slightly improve on our general bound for
and prove the following.
Theorem 2
Every homogeneous set
of
points in
determines at least
distinct distances. Moreover, there is a point
from which there are this many distinct distances to other points of
.
We prove Theorem 1 in Section 3 . The proof of Theorem 2 can be found in Section 4 . In the next section, we present a key lemma on the number of
-flats incident to many points in a homogeneous point set in
, for
.
2 Rich hyperplanes in homogeneous sets
Consider a set
of
points in
. We say that a
-flat (a
-dimensional affine subspace) is
-rich if it is incident to at least
points of
. The celebrated Szemerédi-Trotter Theorem [23] states that for
points in the plane, the number of
-rich lines (
-flats) cannot exceed
, and this bound is tight in the worst case.
The number of
-rich
-flats in
has been intensely studied. The Szemerédi-Trotter type results have widespread applications in discrete and combinatorial geometry. The Szemerédi-Trotter Theorem's multi-dimensional generalizations [7, 1, 9] always impose some kind of restriction on the point set or on the set of
-flats, otherwise
points on a line give rise to an infinity of
-rich
-flats for any
.
We say that a
-flat
is
-degenerate for a constant
if any
-flat contains no more than
points of
. A set of
points is affine independent if it is contained in a unique
-flat, which is said to be spanned by the point set. We recall a result of Beck [4] on
-degenerate hyperplanes.
Theorem 3 (Beck)
For any
, there are constants
with the following property: Given a point set
, if an
-rich
-flat
is
-degenerate, then
contains at least
distinct
-flats spanned by
.
Elekes and Tóth [9] found that there is a constant
for every dimension
such that the number of
-rich
-degenerate hyperplanes in
is at most
. Here, the second term,
, is dominant if
. We show below a much stronger upper bound for homogeneous sets: The number of
-rich hyperplanes is at most
.
Let
denote the maximal number of
-rich
-flats in a homogeneous set
of
points in
, and let
For the number of
-rich lines in homogeneous sets in
(that is, for
), Solymosi and Vu [21] established the following lemma.
Lemma 4 (Solymosi & Vu)
For every
, the number of
-rich lines in a homogeneous set of
points in
is bounded by
We extend their result for arbitrary
,
.
Lemma 5
For every
,
, the number of
-rich
-flats in a homogeneous set of
points in
is bounded by
The example of the
-dimensional integer grid
shows that this bound is best possible for every
.
Proof. Let
be a homogeneous set of
points in
for some
. We proceed by induction on
. The base case,
, is equivalent to Lemma 4 .
Let us assume that
and that
for every
,
. We count separately the
-rich
-flats that are
-degenerate and those that are not.
If an
-rich
-flat is not
-degenerate, then it contains an
-rich
-flat. By induction, the number of
-rich
-flats is
.
Every
-rich
-flat
can be extended to a (degenerate)
-rich
-flat in
different ways:
together with a point of
spans a
-flat. This gives an upper bound of
on the not
-degenerate
-rich
-flats.
Next, we consider the
-degenerate
-rich
-flats. Let
be a family of all affine independent
-element subsets of
. In a homogeneous set of
points, there are
such subsets.
For every
, independently, we subdivide
into
regions, for some constant
to be specified later: Let
denote the
-flat spanned by
, and let
denote the
-flat orthogonal to
. The set of points in
at unit distance from
is the Minkowski sum
, where
is a unit sphere
in
. A projection
can be defined as
, where
is the
-flat spanned by
and
. Let us partition
into
regions
such that the regions have the same volume and every
-flat through the center of
(that is, every great circle on the sphere
) intersects at most
regions. We then subdivide
into
regions
,
, such that
Since the
-flat
intersects the cube
, the intersection
has volume
for every
. By homogeneity, we have
points of
in every region
. Observe that every
-flat that contains
intersects at most
regions.
For an
, we estimate the cardinality of the set
of point pairs
such that
-
∙
is an affine independent set and spans an
-rich
-flat of
;
-
∙
and
lie in a region
, for some
.
By counting all pairs
satisfying the second condition, we obtain an upper bound for
:
Let
be the set of triples
such that
and
. By combining the upper bounds for
and for
, we have
Next, we compute a lower bound for the quantity
. Consider an
-degenerate
-rich
-flat
. For a set
of
affine independent points in
, let
denote a maximal size subset of
whose elements have distinct projections under
. By Beck's Theorem 3 , there is a family
of
subsets
such that the number
of distinct projections under
is at least
.
For an
, we denote by
the set of point pairs
such that
spans
. By construction,
intersects at most
subsets
of the partition associated with
. If the constant
is sufficiently large, then the average number of points of
over the regions
intersecting
is at least
. Any two points in
determine a pair of
, and so
We can now give a lower bound for
. There are
distinct
-rich
-flats. Each such
-flat
contains
affine independent subsets of size
for which
. For each such
, we have
. Therefore,
By contrasting the lower and upper bounds for
, we obtain
as required.
Corollary 6
For every
,
, the number of incidences of points and
-rich
-flats in a homogeneous set of
points in
is at most
Proof. In any homogeneous point set of size
in
, the number of incidences is bounded by
3 Proof of Theorem 1
We are given a homogeneous set
of
points in
. We may assume without loss of generality that every coordinate of every point in
is irrational, while the enclosing cube
has rational coordinates. Let
denote the maximum number of distinct distances measured from a point of
(including distance
). For every
, the points of
lie on
concentric spheres centered at
.
We subdivide
into
congruent subcubes
, where
Every hyperplane and every sphere intersects the interior of at most
subcubes.
Let
be a set of triples
such that
-
(i)
,
-
(ii)
and
lie in a subcube
for some
,
-
(iii)
and
are equidistant from
.
All points are located on
spheres centered at the
points of
. There are
sphere-point incidences. The cubes
,
, subdivide each sphere into patches. Since every sphere intersects at most
subcubes
, there are at most
patches, where each patch lies entirely in a subcube
. The average number of points on a patch is at least
. If
points lie on a sphere patch centered at
, then this patch contributes
triples
to
. We conclude that the number of triples is
.
For every
, let
denote the set of triples
such that the bisector hyperplane of the segment
is incident to at least
but less then
points of
. Since every bisector plane is incident to less than
points, we can partition
into
subsets
There is a value
for some
, such that
.
For a pair
,
, all points of the set
lie on the bisector hyperplane of the line segment
. One bisector hyperplane intersects at most
subcubes, and in each subcube
it can bisect at most
point pairs. So the number of pairs
bisected by the same hyperplane is at most
Let
denote the set of all bisector hyperplanes that bisect the pair
for some
. By definition, any hyperplane in
is incident to at least
but less than
points of
. By Lemma 5 , we have
We can now give an upper bound for
. In a triple
, point
lies on a bisector hyperplane of
. Each bisector hyperplane is incident to less than
points of
and bisects at most
pairs
. Therefore
|
(1)
|
We obtain another upper bound for
by the following argument: In a triple
, both
and
lie in the same subcube
. There are
subcubes, and each subcube contains less than
point pairs. Hence, there are less than
such pairs
. For each pair
, where
, there are at most
points
on the bisector hyperplane of
. We conclude that
Using the upper bound for
from Inequality ( 1 ), we have
as required. This completes the proof of Theorem 1
4 Proof of Theorem 2
Consider a homogeneous set
of
points in
. Similarly to the previous section, we assume that all coordinates of every point in
are irrational, and the vertices of the bounding cube
have integer coordinates. We subdivide
into
congruent cubes
, for
where
is a constant to be specified later.
By Theorem 3 ,
contains
affine independent point pairs.
This implies that there is a subset
such that
and every
is incident to
distinct lines spanned by
. For every
, let
be a set of
points such that the lines
,
, are distinct. For every point
, let
be a unit sphere centered at
.
We project the points of
into
distinct point in
, we denote by
the projection of
. The set of projections to is let
We consider a simplicial partition for
defined as follows.
Definition 7
Given a set
points on a sphere
and an integer
, the simplicial partition of
is a partition of
into
sets
with the following properties
-
∙
for every
,
lies in set
whose boundary consists of a finite number of circular arcs;
-
∙
for every
,
;
-
∙
every circle in
crosses at most
sets
.
A circle crosses a set
, if it intersects it but does not contain it.
By a result of Matoušek [17, 5] , there exists a simplicial partition for any
and
. (A general result of Matoušek applies: The range space of disks in
have the properties that its primal shatter function is quadratic and every disk can be approximated with a finite set of ranges with respect to a finite point set.) We consider a simplicial partition for
with integer
.
Let
be a set of quadruples
such that,
-
(i)
the points
,
, and
are distinct;
-
(ii)
, and
lie in a subcube
for some
,
-
(iii)
, and
lie in a subset
, for some
, when projected from center
;
-
(iv)
, and
are equidistant from
.
We give a lower bound on the number of quadruples in
. Consider a sphere
centered at
. We partition the point set
into groups in the following way. Two points of
are in the same group if and only if they are in the same cube
,
, and their projections with respect to
are in the same subset
. We can partition
into the subcubes
,
, by
planes. These planes partition the sphere
along
circles. Every circle intersects at most
regions
,
. If
circles intersect a region
, they can partition
into
pieces. Hence, the number of groups
is partitioned into is at most
On the
spheres centered at points of
, we have a total of at most
groups. We choose constant
(in the definition of
) such that a group contains
points on average. If
points lie on a group, then this group contributes
quadruples
to
. We conclude that the total number of quadruples is
.
The multiplicity of a pair
is defined as
We choose a parameter
to be specified later, and distinguish two types of quadruples in
: A quadruple
is low if at least one edge of the triangle
have multiplicity at most
. A quadruple
is high if the multiplicity of all three edges of
are above
. Let
and
denote the sets of low and high quadruples, respectively. We distinguish two cases:
First we consider the case that
, then we proceed with the case
.
Case
. There are at least
low quadruples in
. We define a set of triples
We have extracted
triples from
. Similarly, to the previous section, we compute an upper bound on
. Every pair
from a triple of
lies in one of the
subcubes of
, and for every pair
there are at most
centers
. Therefore, we have an upper bound
|
(2)
|
Case
. At least half of the quadruples in
are high, and so
. For every
, project the points
to the sphere
. We denote by
the projection
of a point
. If
, then the intersection of the bisector plane of
and
is the bisector (great circle) of the segment
in the sphere
. A (possibly degenerate) triangle
defines three distinct bisectors. The bisectors of a triangle
meet in two antipodal points on the sphere. The triangles that determine the same triple of bisectors are similar (the center of similarity is the intersection of the bisectors). Specifically, if the triangles
determine the same triple of bisectors, then the points
are collinear (the points
and
are also collinear). Every triple of bisectors determines a family of triangles. A family of quadruples is a collection of quadruples
with a common center
if the triangles
form a family.
For every
, let
denote the set of high quadruples
such that the triangle
on the sphere
is part of a family of size at least
but less than
. Since the size of any family is less than
, we can partition
into
subsets
There is a value
,
, such that
.
Next, we derive an upper bound for the parameter
. Each quadruple
uniquely determine a family
of quadruples of size at least
and at most
. There are
disjoint quadruple families in
. A family
corresponds to at least
spherical triangles on spheres centered at
such that their
vertices lie on three planes incident to
. For
, we call one of these planes a plane spanned by the family.
We denote by
the set of pairs
such that
and
is a plane spanned by a family of triangles on spheres centered at
. Note, though, that several families of triangles with a common center
can span the same plane
. Let
denote the set of pairs
such that
is spanned by at least
but less than
families in
centered at
. There is a value
,
, such that the pairs in
correspond to at least
quadruple families of
. Since each pair
belongs to at least
families, the number of point-plane pairs in
is at least
.
The pair
is an incidence between the point
and a
-rich plane
. All incidences are distinct. By Corollary 6 , we have an upper bound on incidences of
-rich planes in a homogeneous set
:
|
(3)
|
contains
quadruples
where
is a point from the set
of size
. There is a set
of size
such that each
is a center of at least
quadruples of
. For every
, we define a set of
triangles in the sphere
by
For every
, let
denote the set of
-rich planes incident to
. Let
denote the set of intersections (i.e., great circles) of
and
-rich planes in
. Note that for every edge
of a triangle in
, the bisector of
is in
.
Consider the simplicial partition of the set
for a point
. Every set
,
, contains at most
triangles of
. Since there are
subsets
and
triangles, at least
subset must contain
triangles of
. For these sets
, the
triangles determine at least
distinct bisectors in
. A bisector crosses at most
regions, and so we obtain the same bisector of
in at most
regions. We conclude that the number of bisectors determined by the
triangles of
is
| |
| |
Using the estimate
from Inequality ( 3 ), we obtain that
Each of the
points of
is incident to
distinct
-rich planes. This gives
incidences on
-rich planes of
. By Corollary 6 , we have
|
(4)
|
In both cases, we have derived lower bounds for
in terms of
and
.
We choose
such that we obtain the same result in both cases. By comparing Inequalities ( 2 ) and ( 4 ), we have
|
(5)
|
The choice
establishes Inequality ( 5 ) in both cases.
This completes the proof of Theorem 2
References
-
P. K. Agarwal and B. Aronov, Counting facets and incidences, Discrete Comput. Geom. 7 (1992) 359–369.
-
B. Aronov, J. Pach, M. Sharir, and G. Tardos, Distinct distances in three and higher dimensions, Combin. Probab. Comput. 13 (2004), 283–293.
-
B. Aronov and M. Sharir, Cell complexities in hyperplane arrangements, Discrete Comput. Geom. 32 (2004), 107–115.
-
J. Beck, On the lattice property of the plane and some problems of Dirac, Motzkin and Erdős, Combinatorica 3 (3-4) (1983), 281–297.
-
B. Chazelle. The Discrepancy Method. Cambridge University Press, 2000.
-
K. L. Clarkson, H. Edelsbrunner, L. J. Guibas, M. Sharir, and E. Welzl, Combinatorial complexity bounds for arrangements of curves and spheres, Discrete Comput. Geom. 5 (1990), 99–160.
-
H. Edelsbrunner and M. Sharir, A hyperplane incidence problem with applications to counting distances, in Proc. SIGAL International Symposium on Algorithms (T. Asano et al., eds.), vol. 450 of LNCS, Springer-Verlag, Berlin, 1990, pp. 419–428.
-
Gy. Elekes, A note on the number of distinct distances, Period. Math. Hungar. 38 (1999), 173–177.
-
Gy. Elekes and Cs. D. Tóth, Incidences of not-too-degenerate hyperplanes, in Proc. 21st ACM Sympos. Comput. Geom. (Pisa, 2005), ACM Press, to appear.
-
P. Erdős, On sets of distances of
points, Amer. Math. Monthly 53 (1946), 248–250.
-
S. Hofmann and A. Iosevich, Circular averages and Falconer-Erdős distance conjecture in the plane for random metrics, Proc. Amer. Math. Soc. 133 (1) (2005), 133–143
-
A. Iosevich, Curvature, combinatorics, and the Fourier transform, Notices Amer. Math. Soc. 48 (2001), 577–583.
-
A. Iosevich, N. Katz, and S. Pedersen, Fourier basis and the Erdős distance problem, Math. Research Letters 6 (2) (1999), 251–255.
-
A. Iosevich and I. Łaba, Distance sets of well-distributed planar point sets, Discrete Comput. Geom. 31 (2004), 243–250.
-
S. Konyagin and I. Łaba, Distance sets of well-distributed planar sets for polygonal norms, Israel J. Math. to appear.
-
N. H. Katz and G. Tardos, A new entropy inequality for the Erdős distance problem, in: Towards a theory of geometric graphs, vol. 342 of Contemp. Math., Amer. Math. Soc., Providence, RI, 2004, pp. 119–126.
-
J. Matoušek, Efficient partition trees, Discrete Comput. Geom. 8 (1992), 315–334.
-
J. Pach and M. Sharir, Geometric incidences, in Towards a Theory of Geometric Graphs, vol. 342 of Contemporary Mathematics, Amer. Math. Soc., Providence, RI, 2004, pp. 185–223.
-
J. Solymosi, G. Tardos, and Cs. D. Tóth, The
most frequent distances in the plane, Discrete Comput. Geom. 28 (2002), 639–648.
-
J. Solymosi and Cs. D. Tóth, Distinct distances in the plane, Discrete Comput. Geom. 25 (2001), 629–634.
-
J. Solymosi and V. Vu, Distinct distances in high dimensional homogeneous sets, in: Towards a theory of geometric graphs, vol. 342 of Contemp. Math., Amer. Math. Soc., Providence, RI, 2004, pp. 259–268.
-
J. Solymosi and V. Vu, Near optimal bounds for the Erdős distinct distance problem in high dimensions, Combinatorica to appear.
-
E. Szemerédi and W. T. Trotter Jr., Extremal problems in Discrete Geometry, Combinatorica 3 (3–4) (1983), 381–392.
-
L. A. Székely, Crossing numbers and hard Erdős problems in discrete geometry, Combinatorics, Probability & Computing 6 (3) (1997), 353–358.