Solutions to Exercises Chapter 10: Ramsey s Theorem

Similar documents
1. Suppose that a, b, c and d are four different integers. Explain why. (a b)(a c)(a d)(b c)(b d)(c d) a 2 + ab b = 2018.

Pigeonhole Principle and Ramsey Theory

The Pigeonhole Principle

Bichain graphs: geometric model and universal graphs

EGMO 2016, Day 1 Solutions. Problem 1. Let n be an odd positive integer, and let x1, x2,..., xn be non-negative real numbers.

Sydney University Mathematical Society Problems Competition Solutions.

4 a b 1 1 c 1 d 3 e 2 f g 6 h i j k 7 l m n o 3 p q 5 r 2 s 4 t 3 3 u v 2

(q 1)p q. = 1 (p 1)(q 1).

Thus, X is connected by Problem 4. Case 3: X = (a, b]. This case is analogous to Case 2. Case 4: X = (a, b). Choose ε < b a

Ramsey Theory. May 24, 2015

Copyright 2010 Pearson Education, Inc. Publishing as Prentice Hall.

Generalized Pigeonhole Properties of Graphs and Oriented Graphs

Olympiad Combinatorics. Pranav A. Sriram

37th United States of America Mathematical Olympiad

13 th Annual Harvard-MIT Mathematics Tournament Saturday 20 February 2010

SOLUTIONS TO ADDITIONAL EXERCISES FOR II.1 AND II.2

0 Sets and Induction. Sets

Triangle-free graphs with no six-vertex induced path

Covering the Convex Quadrilaterals of Point Sets

K 4 -free graphs with no odd holes

International Mathematics TOURNAMENT OF TOWNS. Senior A-Level Paper Fall 2012.

13 = m m = (C) The symmetry of the figure implies that ABH, BCE, CDF, and DAG are congruent right triangles. So

The Advantage Testing Foundation Solutions

MAT 570 REAL ANALYSIS LECTURE NOTES. Contents. 1. Sets Functions Countability Axiom of choice Equivalence relations 9

Baltic Way 2003 Riga, November 2, 2003

F 2k 1 = F 2n. for all positive integers n.

Schur s Theorem and Related Topics in Ramsey Theory

MA103 Introduction to Abstract Mathematics Second part, Analysis and Algebra

Cographs; chordal graphs and tree decompositions

Basic Combinatorics. Math 40210, Section 01 Fall Homework 8 Solutions

40th International Mathematical Olympiad

Math 421, Homework #6 Solutions. (1) Let E R n Show that = (E c ) o, i.e. the complement of the closure is the interior of the complement.

Name (please print) Mathematics Final Examination December 14, 2005 I. (4)

41st International Mathematical Olympiad

HW2 Solutions Problem 1: 2.22 Find the sign and inverse of the permutation shown in the book (and below).

Notes for Math 290 using Introduction to Mathematical Proofs by Charles E. Roberts, Jr.

The edge-density for K 2,t minors

Putnam Greedy Algorithms Cody Johnson. Greedy Algorithms. May 30, 2016 Cody Johnson.

Generalized Quadrangles Weakly Embedded in Finite Projective Space

Week 2: Sequences and Series

Packing cycles with modularity constraints

Partial characterizations of clique-perfect graphs II: diamond-free and Helly circular-arc graphs

Out-colourings of Digraphs

2007 Shortlist JBMO - Problems

Topological properties

Disjoint paths in unions of tournaments

Exercises for Unit V (Introduction to non Euclidean geometry)

M-saturated M={ } M-unsaturated. Perfect Matching. Matchings

Topology Exercise Sheet 2 Prof. Dr. Alessandro Sisto Due to March 7

Solutions Manual for Homework Sets Math 401. Dr Vignon S. Oussa

INTERNATIONAL MATHEMATICAL OLYMPIADS. Hojoo Lee, Version 1.0. Contents 1. Problems 1 2. Answers and Hints References

GENERALIZED PIGEONHOLE PROPERTIES OF GRAPHS AND ORIENTED GRAPHS

Theorem (Special Case of Ramsey s Theorem) R(k, l) is finite. Furthermore, it satisfies,

ALGEBRA. 1. Some elementary number theory 1.1. Primes and divisibility. We denote the collection of integers

2 BAMO 2015 Problems and Solutions February 26, 2015

PUTNAM TRAINING EASY PUTNAM PROBLEMS

Sequences. Chapter 3. n + 1 3n + 2 sin n n. 3. lim (ln(n + 1) ln n) 1. lim. 2. lim. 4. lim (1 + n)1/n. Answers: 1. 1/3; 2. 0; 3. 0; 4. 1.

34 th United States of America Mathematical Olympiad

1. Groups Definitions

On non-hamiltonian circulant digraphs of outdegree three

Discrete Geometry. Problem 1. Austin Mohr. April 26, 2012

SUMS PROBLEM COMPETITION, 2000

Putnam Pigeonhole Principle October 25, 2005

International Mathematics TOURNAMENT OF THE TOWNS

1. Prove that for every positive integer n there exists an n-digit number divisible by 5 n all of whose digits are odd.

PRIMES Math Problem Set

Baltic Way 2008 Gdańsk, November 8, 2008

2018 Best Student Exam Solutions Texas A&M High School Students Contest October 20, 2018

Solutions of APMO 2017

SOLUTIONS FOR IMO 2005 PROBLEMS

Compact hyperbolic Coxeter n-polytopes with n + 3 facets

The CENTRE for EDUCATION in MATHEMATICS and COMPUTING cemc.uwaterloo.ca Euclid Contest. Wednesday, April 15, 2015

Latin squares: Equivalents and equivalence

USA Mathematical Talent Search Round 1 Solutions Year 27 Academic Year

HMMT February 2018 February 10, 2018

Near-domination in graphs

First selection test

Multi-coloring and Mycielski s construction

MATH 409 Advanced Calculus I Lecture 7: Monotone sequences. The Bolzano-Weierstrass theorem.

Lines, parabolas, distances and inequalities an enrichment class

32 nd United States of America Mathematical Olympiad Recommended Marking Scheme May 1, 2003

2 Sequences, Continuity, and Limits

Square 2-designs/1. 1 Definition

Solutions I.N. Herstein- Second Edition

The 3rd Olympiad of Metropolises

The Advantage Testing Foundation Olympiad Solutions

Algorithms: Background

On Some Polytopes Contained in the 0,1 Hypercube that Have a Small Chvátal Rank

Groups. 3.1 Definition of a Group. Introduction. Definition 3.1 Group

arxiv: v1 [math.co] 22 Apr 2014

SMT 2013 Power Round Solutions February 2, 2013

SOLUTIONS FOR 2011 APMO PROBLEMS

The random graph. Peter J. Cameron University of St Andrews Encontro Nacional da SPM Caparica, 14 da julho 2014

Note that Todd can only move north or west. After this point, Allison executes the following strategy based on Todd s turn.

Packing Congruent Bricks into a Cube

MAT 137Y: Calculus! Problem Set 4 Sample Solutions

Introduction to Real Analysis Alternative Chapter 1

Examples: The (left or right) cosets of the subgroup H = 11 in U(30) = {1, 7, 11, 13, 17, 19, 23, 29} are

Finite Generalized Quadrangles

Ahlswede Khachatrian Theorems: Weighted, Infinite, and Hamming

Transcription:

Solutions to Exercises Chapter 10: Ramsey s Theorem 1 A platoon of soldiers (all of different heights) is in rectangular formation on a parade ground. The sergeant rearranges the soldiers in each row of the rectangle in decreasing order of height. He then rearranges the soldiers in each column in decreasing order of height. Using the Pigeonhole Principle, or otherwise, prove that it is not necessary to rearrange the rows again; that is, the rows are still in decreasing order of height. Let a i j denote the height of the soldier in row i and column j after the first rearrangement (within the rows). Thus, the numbers a i1,a i2,...,a in form a decreasing sequence. Let b i j denote the height of the soldier in row i and column j after the second rearrangement. Then the numbers b 1 j,b 2 j,...,b n j form a decreasing sequence, and in fact are the numbers a 1 j,a 2 j,...,a m j arranged in decreasing order. We have to prove that the rows are also decreasing; that is, that b i j > b i j+1 for all i and j. Suppose not, and suppose that b i j < b i j+1 for some i and j. Choose a number x such that b i j < x < b i j+1. Now b i j is the ith largest of the numbers in the set A j = {a 1 j,...,a m j }, while b i j+1 is the ith largest number in the set A j+1 = {a 1 j+1,...,a m j+1 }. So at least i members of A j+1 are greater than x, whereas fewer than i members of A j are greater than x. But we can pair off the elements of A j and A j+1 so that the member of A j is always the larger of the pair: for we have a k j > a k j+1 for all k. So there must be at least as many members of A j greater than x than there are in A j+1, a contradiction. 2 Show that any finite graph contains two vertices lying on the same number of edges. Let G be a graph with n vertices v 1,...,v n. Place v i in a pigeonhole labelled m if it has exactly m neighbours. The possible labels on the pigeonholes are 0, 1, 2,..., n 1. So n vertices are placed in n pigeonholes. However, the pigeonholes labelled 0 and n 1 cannot both be occupied. For, if v i were in pigeonhole zero and v j in pigeonhole n 1, then v i is joined to no other vertices (in particular, not to v j ), but v j is joined to all other vertices (in particular, to v i ), a contradiction. Now the n vertices lie in at most n 1 pigeonholes, and so some pigeonhole contains at least two vertices. 1

3 (a) Show that, given five points in the plane with no three collinear, the number of convex quadrilaterals formed by these points is odd. (b) Prove Fact 2 in the proof of (10.5.2). (a) There are several cases. First, suppose that the five vertices form a convex pentagon. Then any four of them form a convex quadrilateral; so there are five convex quadrilaterals. Next, suppose that four points form a convex quadrilateral (say p 1, p 2, p 3,, p 4 in anti-clockwise order) which contains the fifth in its interior. Let q be the intersection of the diagonals p 1 p 3 and p 2 p 4. Then p 5 lies on neither diagonal, so is in one of the four triangles into which they dissect the quadrilateral, say qp 1 p 2. Then the quadrilaterals p 1 p 3 p 4 p 5 and p 2 p 3 p 4 p 5 are convex but the other two are not; so there are three convex quadrilaterals in all. Finally, suppose that three points (say p 1, p 2, p 3 ) form a triangle with p 4 and p 5 in its interior. The line p 4 p 5 meets two of the three sides, say p 1 p 2 and p 1 p 3. Then p 2 p 3 p 4 p 5 is the only convex quadrilateral. (b) Suppose that any four points form a convex quadrilateral, but the entire set is not a convex n-gon. Consider a line passing through p n and rotating around this point. Suppose its starting position is the line p n p 1. As it rotates, it will successively pass through all the other points, and after rotating through π, it passes again through p 1. We record whether the points lie on the same end of the line as p 1 originally did. After rotating through π, the point p 1 is on the other end of the line. So it must occur that two successive points, say p i and p j, lie on opposite ends. It cannot occur that all the other points lie in the sector bounded by the lines p n p i and p n p j. For this sector is convex, so the convex hull of the points would be contained in the sector, and would not contain p n. By construction, no point lies in the sector between p i p n produced beyond p n and the line p n p j ; or in the sector between p j p n produced and p n p i. So there is a point p k in the sector between p i p n produced and p j p n produced. Then p n lies in the convex hull of p i p j p k, and the quadrilateral p i p j p k p n is not convex. 4 Show that, if N > mnp, then any sequence of N real numbers must contain either a strictly increasing subsequence with length greater than m, a strictly decreasing subsequence with length greater than n, or a constant subsequence of length greater than p. Show also that this result is best possible. Let N real numbers be given, with N > mnp. By the Pigeonhole Principle, if only mn or fewer distinct values occur, then some value must be taken by more 2

than p numbers, so there is a constant subsequence of length greater than p. Otherwise, more than mn values occur, and the second proof of (10.5.1) shows that there is either a strictly increasing sequence of length greater than m, or a strictly decreasing sequence of length greater than n. To show that the result is best possible, take the sequence on page 155, and repeat each term p times. 5 (a) Show that any infinite sequence of real numbers contains an infinite subsequence which is either constant or strictly monotonic. (b) Using the Principle of the Supremum, prove that every increasing sequence of real numbers which is bounded above is convergent. (c) Hence prove the Bolzano Weierstrass Theorem: Every bounded sequence of real numbers has a convergent subsequence. (a) Given an infinite sequence x 1,x 2,..., colour the pair {i, j} of positive integers (where i < j) red if x i < x j, green if x i > x j, and blue if x i = x j. By Ramsey s Theorem, there is an infinite monochromatic subset, which indexes an infinite monotone or constant subsequence of the original sequence. (b) Suppose that y 1,y 2,... is an increasing sequence of real numbers which is bounded above. Let l be its supremum, or least upper bound. We claim that l is the limit of the sequence. Let ε > 0 be given. Then l ε is not an upper bound, so some term y n of the series exceeds it. Since the series is increasing, we have y m > l ε for all m n. But l + ε is a strict upper bound, so y m < l + ε for all m. Combining these two statements, we see that y m l < ε for all m n, where n was calculated from ε. Thus, by definition, l is the limit of the sequence, which is therefore convergent. (c) Let x 1,x 2,... be a bounded sequence of real numbers By (a), there is a constant or monotone subsequence. A constant subsequence is trivially convergent; an increasing subsequence was shown to be convergent in part (b); and dually a decreasing subsequence is convergent. So the Bolzano Weierstrass Theorem is proved. 3

6 (a) Let X be the set of residues modulo 17. Colour the 2-element subsets of X by assigning to {x,y} the colour red if x y ±1,±2,±4 or ± 8 (mod 17), blue otherwise. Show that there is no monochromatic 4-set. [HINT: By symmetry, we may assume that the 4-set contains 0 and 1; this greatly reduces the number of cases to be considered!] (b) Find a colouring of the 2-subsets of an 8-set red and blue so that there is no red 3-set and no blue 4-set. (c) Let X consist of all subsets of {0,1,2,3,4} of even cardinality. Colour the 2-subsets {x,y} of X as follows: red if x y = {i, j} and i j ±1 (mod 5); blue if x y = {i, j} and i j ±2 (mod 5); green if x y = 4. Show that there is no monochromatic 3-set. (a) Note, as in the solution on page 342, that the operation of adding a fixed element of GF(17) doesn t change the colour of an edge (since (y+a) (x+a) = y x). Also, the values ±1,±2,±4,±8 are precisely the quadratic residues in GF(17). Hence multiplication by a fixed residue doesn t change the colour of an edge. (We have ry rx = r(y x); and, by Exercise 13 of Chapter 9, if r is a residue, then r(y x) is a residue if and only if y x is a residue.) Similarly, if n is a non-residue, then n(y x) is a residue if and only if y x is a non-residue; so multiplication by n changes the colour of every edge. Suppose that there is a red 4-clique. If a is one of its points, then adding a gives a red 4-clique containing 0. Then, if b is a non-zero element, then b is a residue, and multiplication by the residue b 1 gives a 4-clique containing 0 and 1. The other two points are joined to both 0 and 1 by red edges, and hence are two of the three such points 2,9,16. But any two of these are joined by blue edges. So no red 4-clique can exist. If a blue 4-clique exists, then multiplication by a non-residue transforms it into a red 4-clique, an impossibility. So no blue 4-clique can exist either. (b) Take the eight points joined to 0 by red edges in the example in part (a). This set can contain no blue 4-set, and no red 3-set (since such a set, together with 0, would form a red 4-set in the configuration of (a)). 4

(c) First we check that the definition is exhaustive. If A and B are even, then A B = A + B 2 A B is also even. Also, (A X) (B X) = A B, so the operation of taking the symmetric difference with a fixed set preserves the colours of edges. Thus, if there is a monochromatic 3-set, then there is one containing /0. Thus, it suffices to prove that, among the points joined to /0 by red edges, all the edges are green and blue, and similarly for the other colours. Furthermore, the permutation induced by shifting 0,1,2,3,4 cyclically also preserves the colours. The red neighbours of /0 are 01,12,23,34,40 (where 01 is short for {0,1}, and so on). Now the edges joining 01 to 12 and 04 are both blue (since, for example, 01 12 = 02), while those joining it to 23 and 34 are both green. Similarly for the other colours. 7 (a) Prove the following theorem of Schur: Schur s TheoremThere is a function f on the natural numbers with the property that, if the numbers {1,2,..., f (n)} are partitioned into n classes, then there are two numbers x and y such that x,y and x+y all belong to the same class. (In other words, the numbers {1,2,..., f (n)} cannot be partitioned into n sumfree sets.) (b) State and prove an infinite version of Schur s Theorem. (a) Take N to be one less than the Ramsey number R(n,2,3). Now suppose that the set {1,2,...,N} is partitioned into n classes C 1,...,C n. Colour the 2-subsets of {1,2,...,N +1} with n colours c 1,...,c n by the following rule: the 2-set {x,y} (with x < y) has colour c i if (and only if) y x C i. (Note that 1 y x N.) Since N + 1 = R(n, 2, 3), there is a monochromatic triangle {x, y, z} (with x < y < z) of colour c i. This means that y x,z y,z x C i. But (y x)+(z y) = z x; so C i contains two numbers and their sum. Thus the theorem is proved, with f (n) = R(n, 2, 3) 1. (b) If the set of natural numbers is partitioned into a finite number n of subsets, then there are two numbers x and y such that x, y and x + y belong to the same subset. Proof. As in the previous proof, we colour the 2-sets of natural numbers with colours c 1,...,c n by the rule that the 2-set {x,y} (with x < y) has colour c i if and only if y x belongs to the ith subset C i. By the infinite Ramsey theorem, there exists a monochromatic triangle {x,y,z} (with x < y < z). Then y x,z y,z x belong to the same subset and (y x) + (z y) = z x. 5

8 A delta-system is a family of sets whose pairwise intersections are all equal. (So, for example, a family of pairwise disjoint sets is a delta-system.) Prove the existence of a function f of two variables such that any family F of at least f (n,k) sets of cardinality n contains k sets forming a delta-system. State and prove an infinite version of this theorem. Do you regard this theorem as part of Ramsey theory? (a) Since there just 2 n subsets of a n-set, as long as the family of k-sets contains more than 1 + (2 n 1)(r 1) members, for any set A in the family, there must be r sets all having the same intersection with A. If A i 1 and F i 1 have been chosen, let A i be any member of F i 1, and let F i be a subfamily all of whose members have the same intersection with A i. By the preceding paragraph, we can ensure that F i > F i 1 /2 k. So the process can be continued for m steps provided the original family has more than 2 mk members. Now suppose that the series continues for (k 1)(n + 1) + 1 terms. Now A i A j depends only on i, not j (if i < j); call this intersection B i. If i < j < k, then both A j and A k contain B i, and so B j B i. But B i n, so B i+1 B i can hold at most n + 1 times. By the Pigeonhole Principle, there is a run of at least k consecutive values of i for which B i remains constant. The corresponding sets A i for a delta-system with k terms. (b) An infinite family of n-sets must contain an infinite subfamily which is a delta-system. Proof. In the above argument, there are infinitely many sets all having the same intersection with A, and so we can choose an infinite sequence A1,A 2,... such that all the families F i are infinite. Again, the intersection B i = A i A j (for i < j) can change at most n times, so it remains constant after some value i 0, and the sets A i0 +1,A i0 +2,... form an infinite delta-system. (c) These results assert that a large structure (here a family of n-sets) must contain a smaller, more homogeneous substructure (a delta-system). This is the general form of a theorem of Ramsey theory. 9 Why are constructive existence proofs more satisfactory than non-constructive ones? 9. This is an essay question. Some points: An application (especially one outside mathematics) may require an explicit example of the object whose existence is asserted. For example, an 6

existence theorem for a graph describing the connections on a chip would be of no use to the chip manufacturer unless it has a constructive proof. Some schools of philosophy of mathematics reject the law of excluded middle (and hence proof by contradiction), and accept only constructive proofs. On the other hand, a non-constructive proof may give a better bound, and may suggest how random search could be directed to increase the chance of finding the object in question. 7