We just finished the Erdős-Stone Theorem, and ex(n, F ) (1 1/(χ(F ) 1)) ( n

Similar documents
The Zarankiewicz problem in 3-partite graphs

Lecture 2-1 Kinematics in One Dimension Displacement, Velocity and Acceleration Everything in the world is moving. Nothing stays still.

Some Ramsey results for the n-cube

Rainbow saturation and graph capacities

23.2. Representing Periodic Functions by Fourier Series. Introduction. Prerequisites. Learning Outcomes

Notes for Lecture 17-18

Finish reading Chapter 2 of Spivak, rereading earlier sections as necessary. handout and fill in some missing details!

BEng (Hons) Telecommunications. Examinations for / Semester 2

Physics for Scientists & Engineers 2

EXERCISES FOR SECTION 1.5

Let us start with a two dimensional case. We consider a vector ( x,

2.7. Some common engineering functions. Introduction. Prerequisites. Learning Outcomes

Inventory Analysis and Management. Multi-Period Stochastic Models: Optimality of (s, S) Policy for K-Convex Objective Functions

Monochromatic Infinite Sumsets

Solutions from Chapter 9.1 and 9.2

MATH 4330/5330, Fourier Analysis Section 6, Proof of Fourier s Theorem for Pointwise Convergence

15. Vector Valued Functions

Approximation Algorithms for Unique Games via Orthogonal Separators

Matrix Versions of Some Refinements of the Arithmetic-Geometric Mean Inequality

Optimality Conditions for Unconstrained Problems

Chapter 2 Summary. Carnegie Learning

Chapter 7: Solving Trig Equations

LINEAR INVARIANCE AND INTEGRAL OPERATORS OF UNIVALENT FUNCTIONS

23.5. Half-Range Series. Introduction. Prerequisites. Learning Outcomes

THE MATRIX-TREE THEOREM

Echocardiography Project and Finite Fourier Series

WEEK-3 Recitation PHYS 131. of the projectile s velocity remains constant throughout the motion, since the acceleration a x

11!Hí MATHEMATICS : ERDŐS AND ULAM PROC. N. A. S. of decomposiion, properly speaking) conradics he possibiliy of defining a counably addiive real-valu

Hedgehogs are not colour blind

Weyl sequences: Asymptotic distributions of the partition lengths

Extremal graph theory II: K t and K t,t

Undetermined coefficients for local fractional differential equations

4 Sequences of measurable functions

2. Nonlinear Conservation Law Equations

A problem related to Bárány Grünbaum conjecture

Exercises: Similarity Transformation

INDEPENDENT SETS IN GRAPHS WITH GIVEN MINIMUM DEGREE

Math 334 Test 1 KEY Spring 2010 Section: 001. Instructor: Scott Glasgow Dates: May 10 and 11.

Extremal colorings and independent sets

The minimum number of nonnegative edges in hypergraphs

Math 116 Practice for Exam 2

Two Coupled Oscillators / Normal Modes

Physics 235 Chapter 2. Chapter 2 Newtonian Mechanics Single Particle

KINEMATICS IN ONE DIMENSION

On Measuring Pro-Poor Growth. 1. On Various Ways of Measuring Pro-Poor Growth: A Short Review of the Literature

on the interval (x + 1) 0! x < ", where x represents feet from the first fence post. How many square feet of fence had to be painted?

DIFFERENTIAL GEOMETRY HW 5

Matlab and Python programming: how to get started

Congruent Numbers and Elliptic Curves

T L. t=1. Proof of Lemma 1. Using the marginal cost accounting in Equation(4) and standard arguments. t )+Π RB. t )+K 1(Q RB

Randomized Perfect Bipartite Matching

Lab #2: Kinematics in 1-Dimension

Chapter 2. First Order Scalar Equations

Math 315: Linear Algebra Solutions to Assignment 6

5.1 - Logarithms and Their Properties

t 2 B F x,t n dsdt t u x,t dxdt

1 Review of Zero-Sum Games

Linear Response Theory: The connection between QFT and experiments

Kinematics Vocabulary. Kinematics and One Dimensional Motion. Position. Coordinate System in One Dimension. Kinema means movement 8.

Challenge Problems. DIS 203 and 210. March 6, (e 2) k. k(k + 2). k=1. f(x) = k(k + 2) = 1 x k

Families with no matchings of size s

d 1 = c 1 b 2 - b 1 c 2 d 2 = c 1 b 3 - b 1 c 3

Chapter 15: Phenomena. Chapter 15 Chemical Kinetics. Reaction Rates. Reaction Rates R P. Reaction Rates. Rate Laws

SUBSPACES OF MATRICES WITH SPECIAL RANK PROPERTIES

4.1 - Logarithms and Their Properties

SOME MORE APPLICATIONS OF THE HAHN-BANACH THEOREM

IB Physics Kinematics Worksheet

Bernoulli numbers. Francesco Chiatti, Matteo Pintonello. December 5, 2016

IMPLICIT AND INVERSE FUNCTION THEOREMS PAUL SCHRIMPF 1 OCTOBER 25, 2013

arxiv: v1 [math.fa] 9 Dec 2018

Radical Expressions. Terminology: A radical will have the following; a radical sign, a radicand, and an index.

U( θ, θ), U(θ 1/2, θ + 1/2) and Cauchy (θ) are not exponential families. (The proofs are not easy and require measure theory. See the references.

Topics in Combinatorial Optimization May 11, Lecture 22

The Maximal Subgroups of The Symplectic Group Psp(8, 2)

Average Number of Lattice Points in a Disk

Lecture 10: The Poincaré Inequality in Euclidean space

EECE251. Circuit Analysis I. Set 4: Capacitors, Inductors, and First-Order Linear Circuits

Chapters 6 & 7: Trigonometric Functions of Angles and Real Numbers. Divide both Sides by 180

Hamilton- J acobi Equation: Weak S olution We continue the study of the Hamilton-Jacobi equation:

10. State Space Methods

Final Spring 2007

Chapter Three Systems of Linear Differential Equations

HOMEWORK # 2: MATH 211, SPRING Note: This is the last solution set where I will describe the MATLAB I used to make my pictures.

MA Study Guide #1

Languages That Are and Are Not Context-Free

A Special Hour with Relativity

( ) ( ) if t = t. It must satisfy the identity. So, bulkiness of the unit impulse (hyper)function is equal to 1. The defining characteristic is

Mathcad Lecture #8 In-class Worksheet Curve Fitting and Interpolation

ACE 562 Fall Lecture 5: The Simple Linear Regression Model: Sampling Properties of the Least Squares Estimators. by Professor Scott H.

Introduction to Probability and Statistics Slides 4 Chapter 4

Elements of Computer Graphics

Math 2142 Exam 1 Review Problems. x 2 + f (0) 3! for the 3rd Taylor polynomial at x = 0. To calculate the various quantities:

Lecture 4 Notes (Little s Theorem)

On a problem of Graham By E. ERDŐS and E. SZEMERÉDI (Budapest) GRAHAM stated the following conjecture : Let p be a prime and a 1,..., ap p non-zero re

Lecture 20: Riccati Equations and Least Squares Feedback Control

SMT 2014 Calculus Test Solutions February 15, 2014 = 3 5 = 15.

Transform Techniques. Moment Generating Function

Solutions to Assignment 1

Math 105 Second Midterm March 16, 2017

Overview. COMP14112: Artificial Intelligence Fundamentals. Lecture 0 Very Brief Overview. Structure of this course

Transcription:

Lecure 3 - Kövari-Sós-Turán Theorem Jacques Versraëe jacques@ucsd.edu We jus finished he Erdős-Sone Theorem, and ex(n, F ) ( /(χ(f ) )) ( n 2). So we have asympoics when χ(f ) 3 bu no when χ(f ) = 2 i.e. F is biparie. We focus on his case now, and his lecure looks a F = K s,, he complee biparie graph wih pars of sizes s and. Here is a warmup exercise. Exercise A. Deermine ex(n, K, ). The exremal problem for K s, is relaed o he Zarankiewicz problem (95). This problem asks for he maximum number z(m, n, s, ) of s in an m n 0- marix ha does no conain an s minor filled wih s. Observe ha if A is an m n 0- marix, hen A is he incidence marix of an m n biparie graph: he rows of he marix are labelled by he verices in he par of size m, and he columns are labelled by he verices in he par of size n. If A has no s minor filled wih s, hen he graph has no K s, wih s verices in he par of size m and verices in he par of size n. Theorem. (Kövari-Sós-Turán, 954) For m, n, s,, z(m, n, s, ) (s ) / nm / + ( )m. Proof. We wrie he proof in he language of graphs. Le A be an m n marix wih z = z(m, n, s, ) ones and no s minor filled wih s, and le G be he corresponding m n biparie graph, say wih pars R of size m and C of size n. They key observaion is d(v) C (s ). By convexiy, his implies v R E(G) / R C R (s ). Using (x + ) /! ( x ) x /! for x, we ge This gives he required bound. m(z/m + ) (s )n. The proof of he bound above also gives an upper bound on he Turán numbers ex(n, K s, ):

Theorem 2. For s,, ex(n, K s, ) 2 (s )/ n 2 / + ( )n. 2 Proof. If G = (V, E) is a K s, -free n-verex graph, hen d(v) n (s ). v V Since v V d(v) = 2e(G), he convexiy of binomial coefficiens gives n(2e(g)/n + ) (s )n which gives he heorem. In general, he order of magniude of ex(n, K s, ) and z(m, n, s, ) are no known when s, 4, and specifically, K 4.4. The bes lower bounds in general come from random graphs and random marices: Theorem 3. For all s,, ex(n, K s, ) = Ω(n 2 (s+ 2)/(s ) ). The same holds for z(n, n, s, ). Laer on we shall look a consrucions of K s, -free graphs. Righ now we concenrae on he case s = = 2. Quadrilaerals. The case s = = 2 of he Kővari-Sós-Turán heorem is exensively researched. From he las secion, we have: ex(n, C 4 ) 2 n n 3 4 + n and z(m, n, 2, 2) mn /2 + m. 4 Exremal consrucions come from projecive geomery. Recall a projecive plane of order q consiss of q 2 + q + poins and q 2 + q + lines, such ha each line is a se of q + poins, each poin is conained in q + lines, every pair of lines inersecs in exacly one poin, and every wo poins are conained in exacly one line. The exisence of projecive planes has been esablished only when q is a prime power, and remains one of he main open problems in projecive geomery. Direcly from projecive planes, by considering he incidence marix of poins and lines, one obains he exac value of z(n, n, 2, 2): Theorem 4. Le q be a posiive ineger and n = q 2 + q +. Then z(n, n, 2, 2) = (q + )n if and only if here exiss a projecive plane of order q. 2

Equivalenly, he biparie incidence graph of a projecive plane has (q + )(q 2 + q + ) edges, and so a projecive plane of order q exiss if and only if here exiss a (q + )-regular n by n biparie C 4 -free graph, where n = q 2 + q +. Exercise B. Prove his heorem. This heorem gives he asympoic formula z(n, n, 2, 2) n 3/2 as n, due o he exisence of projecive planes of all prime orders and he following lemma from number heory: Lemma. For all large enough n, here exiss a prime p beween n and n + n 2/40. Erdős, Rényi and Sós (966) were he firs o describe asympoically exremal C 4 -free graphs. The resuls of he las secion show ex(n, C 4 ) 2 n3/2 + o(n 3/2 ), and heir consrucion maches his upper bound asympoically. I is described as follows. Le V be a 3-dimensional vecor space over F q, and le ER q be he graph whose verex se is he se of -dimensional subspaces of V, and whose edge se consiss of pairs of orhogonal -dimensional subspaces of V. Then ER q is known as a polariy graph of a projecive plane, and one checks ha ER q is C 4 -free and has 2 q(q + )2 edges. Exercise C. Verify ha e(er q ) = 2 q(q + )2 and ha ER q is C 4 -free. In general, given a projecive plane wih poin se P and line se L, a polariy of he projecive plane is an involuion π : P L P L ha preserves incidences. We refer he reader o books on projecive geomery for more informaion on planes and polariies. This allows us in essense o fold he biparie incidence graph of a projecive plane ono one of is pars o obain a C 4 -free graph: if G q is he biparie incidence graph of a projecive plane, wih pars P and L corresponding o poins and lines, hen form a graph H q wih V (H q ) = P and where {p, p 2 } E(H q ) if p is adjacen o π(p 2 ). I is a challenging ask o prove ex(q 2 +q +, C 4 ) 2 q(q +)2 ; noe ha when n = q 2 +q +, he couning argumens of he las secion only give ex(n, C 4 ) (n/2) n 3/4 + n/4 = 2 (q + )(q2 + q + ). Füredi (988) showed ha in fac polariy graphs are exremal C 4 -free graphs, and showed ha any C 4 -free graph wih q 2 + q + verices where q > 3 is a prime power mus be he polariy graph of some projecive plane. In paricular, Theorem 5. Le q > 3 be a prime power and n = q 2 + q +. Then ex(n, C 4 ) = 2 q(q + )2. Once more, he disribuion of primes according o he lemma shows ha for all n, ex(n, C 4 ) 2 n3/2 + o(n 3/2 ). Deermining ex(n, C 4 ) when n is no of he form q 2 + q + seems o be a very difficul problem. There are many near exremal consrucions of C 4 -free graphs, we give an example here. Exercise D. Le G be he graph whose verex se is F q F q \{(0, 0)} where q is prime, and whose edge se consiss of pairs {(x, y), (a, b)} where b + y = xa. Verify ha G q is C 4 -free and has v(g) = q 2 and e(g) 2 (q )(q2 ). 3

Füredi (996) also deermined he asympoic value of ex(n, K 2, ) and z(n, n, 2, ), showing ex(n, K 2, ) 2 n 3/2 for all 2 and z(n, n, 2, ) n 3/2 for all 2. In addiion, Füredi (996) showed ex(n, K 3,3 ) 2 n5/3 + o(n 5/3 ). We end wih he following dense K 3,3 - free consrucion due o Brown (966) which gives he lower bound maching Füredi s upper bound on ex(n, K 3,3 ): Exercise E. Le G q be he graph whose verex se is F q F q F q. Join a riple (a, b, c) o (x, y, z) if (x a) 2 + (y b) 2 + (z c) 2 =. Prove ha G q conains no K 3,3 and deermine he number of edges in G q. Conclude ex(n, K 3,3 ) 2 n5/3 + o(n 5/3 ). Quick Applicaions. The firs quick applicaion is geomeric. The famous Erdős uni disance problem (946) asks for he maximum number of uni disances ha can be defined by n poins in he plane in oher words, if P is he se of poins, how many pairs {x, y} P have d(x, y) = in he Euclidean meric. Erdős conjecured ha he answer is n +o() as n, and his remains wide open despie a remendous amoun of work in rying o find good upper bounds. The saring poin is in he following heorem: Theorem 6. Any se of n poins in he plane defines a mos n 3/2 / 2+o(n 3/2 ) uni disances. Proof. We reduce o an exremal problem. Noe ha if x, y are poins in he plane, hen here exis a mos wo poins z such ha d(x, z) = d(y, z) =. Form he graph G whose verex se is he se of n poins in he plane, and where {x, y} E(G) if d(x, y) =. Then G is K 2,3 -free. By he Kövari-Sós-Turán Theorem, e(g) n 3/2 / 2 + o(n 3/2 ). This elemenary argumen is no very difficul o improve o O(n 4/3 ), however he conjecure sill appears ou of reach. Exercise F. Prove ha he number of uni disances beween n poins in R 3 is O(n 5/3 ). Wha lower bound can you give? The Szemerédi-Troer Theorem (984) on planar poin-line incidences saes ha m poins and n lines in he plane define a mos abou (mn) 2/3 + m + n poin-line incidences. This provides he saed upper bound of O(n 4/3 ) for uni disances in he plane. This resul is special o he plane, since in he projecive plane, we may have n poins and n lines and he oal number of poin-line incidences is abou n 3/2. We can prove a general upper bound as follows: Theorem 7. A se of m poins and n lines in he plane define a mos mn /2 + m + n incidences. 4

We leave he proof as an exercise. A final applicaion is o Sidon ses. A se S in an abelian group Γ is called a Sidon se if x + y = z + w wih x, y, z, w S implies {x, y} = {z, w}. The exremal size of Sidon ses has been sudied for a leas half a cenury, wih roos in projecive geomery. I urns ou ha a projecive plane may be defined by aking special Sidon ses S in Z q 2 +q+ hese are called Singer cycles or perfec difference ses and hey are Sidon ses of size q + in Z q 2 +q+. Their special propery is ha every non-zero ineger mod q 2 + q + can be represened uniquely as a difference of elemens in S. To ge a projecive plane, le he lines of he plane be he se of ranslaes of S, namely {S + λ : λ Z q 2 +q+}. Theorem 8. If Γ is an abelian group of order n, and S Γ is a Sidon se, hen S n /2 +. Proof. Form a biparie graph G wih pars X = Y = Γ, and join x X o y Y if x + a = y for some a S. One checks ha because S is a Sidon se, G is C 4 -free. Therefore if we wrie n = q 2 + q + for a posiive real q, hen e(g) z(n, n, 2, 2) (q + )n n 3/2 + n. On he oher hand, e(g) = S Γ = S n, and so S n /2 +. Exercise G. Le Γ = F q F q. Prove ha if q is odd, hen S = {(x, x 2 ) : x F q } is a Sidon se, and if q is even, hen S = {(x, x 3 ) : x F q } is a Sidon se. Exercise H. Le S be a Sidon se in an abelian group Γ. Le G be he graph wih V (G) = Γ and E(G) = {{x, y} : x + y S}. Deermine he number of edges in G, and wheher G conains a cycles of lengh hree and four. 5