Szemeredi-Trotter Polynomials and Incidences

Szemerédi-Trotter: Polynomials and Incidences
Maxwell Fishelson
1 Preliminaries
Definition 1 We let O signify an asymptotic upper bound. Say f (n) and g(n) are two functions with
codomain R. To assert that f (n) = O(g(n)) is to assert that there exists a positive constant c such that,
for all n, f (n) < c · g(n).
Definition 2 Similarly, we let Ω signify an asymptotic lower bound. To assert that f (n) = Ω(g(n)) is
to assert that there exists a positive constant c such that, for all n, f (n) > c · g(n).
Definition 3 For a point p and a curve c, we say p is “incident” to c if p lies on c. For a set of points
P and a set of curves C, we define the incidence graph of P × C to be the following bipartite graph: a
vertex for each point in P on one side, a vertex for each curve in C on the other, and an edge between
a point and a curve if that point is incident to that curve.
2 The Szemerédi-Trotter Theorem and Its Extensions

The Szemerédi-Trotter Theorem states:
(Szemerédi, Trotter [1]) Let P be a set of m points and let L be a set of n lines, both in R2 . Then, the
number of incidences between points in P and lines in L = I(P, L) = O(m2/3 n2/3 + m + n)
The original proof provided by Szemerédi and Trotter used a technique known as cell decomposi-
tion. Later, Székely came up with a simple combinatorial proof of Szemerédi-Trotter [2]. His argument
converted the points, lines, and incidences into a graph theory diagram, and used a crossing number
inequality [3].
With the help of Székely’s simpler proof, Pach and Sharir were able to generalize his arguments using
a multigraph. They derived a theorem that bounded incidences between points and algebraic curves in
general, not just incidences between points and lines:
(Pach, Sharir [4]) Let P be a set of m points and let C be a set of n constant-degree algebraic curves,
both in R2 , such that the incidence graph of P × C does not contain a copy of Ks,t . Then, I(P, C) =
s 2s−2
O(m 2s−1 n 2s−1 + m + n).
3 Using Szemerédi-Trotter with Polynomials

Let A and B be sets of real numbers. We define the set A + B = {a + b : a ∈ A, b ∈ B}. We define
the set A · B = {ab : a ∈ A, b ∈ B}. Let f (x) be a polynomial. We define the set f (A) = {f (a) : a ∈ A}.
Say we want to lower bound |A + A| in terms of |A|. Unfortunately, the tightest bound we have is
|A + A| = Ω(|A|), which comes from the trivial fact that |A + A| > |A| since 2a ∈ A + A for every a ∈ A.
To show that this is the strongest bound we have, we look at the case when the terms of |A| form an
arithmetic sequence (say A = {x, 2x, · · · , nx}). For all A of this form, we have |A + A| = 2|A| − 1.
However, we can get non-trivial bounds when we try to minimize the number of elements in 2 different
sets at the same time, such as max(|A + A|, |A ∗ A|). A famous unproven conjecture of Erdős and
Szemerédi states:
(Erdős, Szemerédi [5]) For every set of reals A, max(|A + A|, |A ∗ A|) = Ω(|A|2− ) for every positive
arbitrarily close to 0.
Elekes [6] was able to prove the bound max(|A+A|, |A∗A|) = Ω(|A|5/4 ) using an argument involving
Szemerédi-Trotter. The following result also serves to bound two sets at the same time, except we will
be considering sets constructed with polynomials. We use a Szemerédi-Trotter argument similar to that
of Elekes.
Theorem 1. 1 Let A be a set of real numbers and let f (x), g(x) be polynomials with d = deg(f ) ≤ deg(g).
2d+1
Then, max(|A + A|, |f (A) + g(A)|) = Ω(|A| 2d ).
Proof. Let P be the set of points in R2 : {(x, y) : x ∈ A + A, y ∈ f (A) + g(A)}. Therefore, |P | =

|A + A| · |f (A) + g(A)|. Let C be the set of polynomial curves in R2 : {y = f (x − u) + g(v) : u, v ∈ A}.
|C| = O(|A|2 ) since there are |A|2 choices for u, v. We may have duplicate polynomials in C if g(v) takes
on the same value for different v in A. However, for any real r, there are at most deg(g) reals v such
that g(v) = r. Therefore, each polynomial in {y = f (x − u) + g(v) : u, v ∈ A} is duplicated at most
|A|2
deg(g) times over the |A|2 choices for u, v. Thus, |C| ≥ deg(g) ⇒ |C| = Ω(|A|2 ).
We want to bound the number of incidences I(P, C) between points in P and curves in C. For each
curve y = f (x − u) + g(v) in C, the point (k + u, f (k) + g(v)) is incident to it for every real k. For
every k ∈ A, the point (k + u, f (k) + g(v)) ∈ P . Therefore, each curve in C must have at least |A|
points in P incident to it: (k + u, f (k) + g(v)) ∈ y = f (x − u) + g(v) for each of the |A| ways to pick k.
Consequently, I(P, C) > |A| · |C| = Ω(|A|3 ).
We finish by using the generalized version of Szemerédi-Trotter for curves:
(Pach, Sharir [4]) Let P be a set of m points and let C be a set of n constant-degree algebraic curves,
both in R2 , such that the incidence graph of P × C does not contain a copy of Ks,t . Then, I(P, C) =
s 2s−2
O(m 2s−1 n 2s−1 + m + n).
All of the polynomials in C have degree d and the same leading coefficient. Therefore, any two
distinct curves in C intersect in at most d − 1 points. Thus, the incidence graph of P × C does
d 2s−2 d
not contain a copy of Kd,2 . Consequently, I(P, C) = O(|P | 2d−1 · |C| 2s−1 + |P | + |C|) = O(|P | 2d−1 ·
4s−4
|A| 2s−1 + |P | + |A|2 ). Since I(P, C) is asymptotically lower bounded by |A|3 and asymptotically up-
d 4s−4 d 4s−4
per bounded by |P | 2d−1 · |A| 2s−1 + |P | + |A|2 , we must have |P | 2d−1 · |A| 2s−1 + |P | + |A|2 = Ω(|A|3 ).
2d+1 d 4s−4
Therefore, |P | = Ω(|A| d ) since one of the terms in the sum |P | 2d−1 · |A| 2s−1 + |P | + |A|2 must be
2d+1
asymptotically greater than |A|3 . q And since |P | = |A + A| · |f (A) + g(A)| = Ω(|A| d ), we must have
2d+1 2d+1
max(|A + A|, |f (A) + g(A)|) = Ω( |A| d ) = Ω(|A| 2d ).
2d+1
Note: The bound Ω(|A| 2d ) seems relatively weak, as it is close to Ω(|A|). However, we realize that
d is solely determined by the minimum degree between our two polynomials, meaning we can get decent
bounds from this generalization as long as one of the two polynomials has relatively small degree.
1
From my 2016 Intel STS submission
2
4 References
[1] E. Szemerdi and W. Trotter, Extremal problems in discrete geometry, Combinatorica 3 (1983), 381–
392.
[2] L. Székely, Crossing numbers and hard Erdős problems in discrete geometry, Combinat. Probab.
Comput. 6 (1997), 353–358.
[3] M. Ajtai, V. Chvátal, M. Newborn, and E. Szemerédi, Crossing-free subgraphs, Annals of Discrete
Mathematics 12 (1982), 9–12.
[4] J. Pach and M. Sharir, On the number of incidences between points and curves, Combinat. Probab.
Comput. 7 (1998), 121–127.
[5] J. Zahl, An improved bound on the number of point-surface incidences in three dimensions, Contrib.
Discrete Math. 8 (2013), 100–121.
[6] G. Elekes, On the number of sums and products, Acta Arithmetica LXXXI.4 (1997), 365–367.

Szemeredi-Trotter Polynomials and Incidences

Uploaded by

Copyright:

Available Formats

Szemeredi-Trotter Polynomials and Incidences

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Szemeredi-Trotter Polynomials and Incidences

Uploaded by

Copyright:

Available Formats

Szemerédi-Trotter: Polynomials and Incidences

2 The Szemerédi-Trotter Theorem and Its Extensions

3 Using Szemerédi-Trotter with Polynomials

Proof. Let P be the set of points in R2 : {(x, y) : x ∈ A + A, y ∈ f (A) + g(A)}. Therefore, |P | =

We finish by using the generalized version of Szemerédi-Trotter for curves:

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.