Problem Books in Mathematics

EdwardJ.Barbeau

University
of Toronto
Mathematics
Competition
(20012015)
Problem Books in Mathematics

Series editor:

Peter Winkler
Department of Mathematics
Dartmouth College
Hanover, NH 03755
USA

More information about this series at http://www.springer.com/series/714

Edward J. Barbeau

University of Toronto
Mathematics Competition
(20012015)

123
Edward J. Barbeau
University of Toronto
Toronto, Ontario, Canada

ISSN 0941-3502 ISSN 2197-8506 (electronic)

Problem Books in Mathematics
ISBN 978-3-319-28104-9 ISBN 978-3-319-28106-3 (eBook)
DOI 10.1007/978-3-319-28106-3

Library of Congress Control Number: 2016930169

Springer International Publishing Switzerland 2016

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to the material contained herein or for any
errors or omissions that may have been made.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG Switzerland
 Preface

Students from the University of Toronto have participated in the annual Put-
nam Competition since its inception in 1938. Many of them have competed
with distinction, but others have not fared so well. Because of the way in
which the Putnam scripts are marked, it is necessary to make signicant
progress on a problem to earn one or two points.
In 2001, I started a mathematics competition for undergraduates at the
University of Toronto to oer the experience of solving problems under a
time constraint. My goal was to provide a paper that would be more widely
accessible while maintaining some challenge and interest. A secondary goal
was to produce a set of problems that might nd a place in regular college and
university courses. So, I have collected them here in the hope that teachers
and students may nd them useful.
The problems have many sources. Some are original; some are side results
in research papers; some have been contributed; some have appeared in the
American Mathematical Monthly, and some, of unknown provenance, have
been lying in folders in my oce. Each problem is graded out of 10, with
opportunity for part marks.
So that students have some choice, there are eight to ten problems on
each paper. I expected that students normally should attempt no more than
ve, and to enforce this, they were told that only their best ve problems
would be marked unless the score on these ve was at least 30. Inevitably,
there were students who correctly solved more than ve.
The rst two years of college mathematics should provide sucient back-
ground. Students should have knowledge of secondary level algebra and ge-
ometry, elementary linear algebra, basic number theory, elementary combi-
natorics, techniques of single variable calculus, as well as fundamental results
on limits, continuous functions, and convergence of sequences and series.
In Chap. 1, I present the questions as they were given. However, the
solutions are in the following chapters, arranged according to their subject
matter in chronological order. Where a problem can be classied in more

v
vi PREFACE

than one way, I have provided the statement of the problem in each relevant
chapter, but the solution appears in only one chapter, indicated by a number
in parentheses at the beginning of its statement in Chap. 1. At the beginning
of some solutions, there is a name in square brackets of either a competitor
or a colleague, to acknowledge the source of an approach to the problem.
However, I take full responsibility for the actual presentation.
The book concludes with two appendices. The rst establishes notation
and conventions and lists denitions and results that might not be readily
available to every student. The second lists the names of students who per-
formed well in the competition.
I am grateful to my colleague, Victor Ivrii, for advice on diagrams; to
graduate student Tyler Holden for valuable advice in solving technical anom-
alies; to Emile LeBlanc and Marco De la Cruz-Heredia, who look after the
computer systems in the mathematics department, for technical help; to grad-
uate students Ian Greig and Samer Seraj for checking over some of the so-
lutions; and to my wife, Eileen, for her sharp eyes and continuous support.
The problems from the American Mathematical Monthly appear with the
permission, freely given, of the Mathematical Association of America.

Toronto, Canada Edward J. Barbeau

 Contents

Preface v
Chapter 1. Problems of the Contests 1

Chapter 2. Algebra 23

Chapter 3. Inequalities 49

Chapter 4. Sequences and Series 59

Chapter 5. Calculus and its Applications 75

Chapter 6. Other Topics in Analysis 85

Chapter 7. Linear Algebra 107

Chapter 8. Geometry 133

Chapter 9. Group Theory 149

Chapter 10. Combinatorics and Finite Mathematics 153

Chapter 11. Number Theory 171

Appendix A. Denitions, Conventions, Notation and Basics 187

1. General 187
2. Algebra 187
3. Inequalities 189
4. Sequences and Series 190
5. Analysis 190
6. Linear Algebra 191
7. Geometry 194
8. Group Theory 196

vii
viii CONTENTS

9. Combinatorics 197
10. Number Theory 198

Appendix B. Top Ranking Students 201

Index 205
 CHAPTER 1

Problems of the Contests

Except for the rst contest, for which 3 h was allotted, the time allowed for
each contest was 3 12 h.
The chapter in which the solution appears is given in parentheses at the
beginning of the problem.

Sunday, March 18, 2001

1. (3) Let a, b, c > 0, a < bc and 1 + a3 = b3 + c3 . Prove that

1 + a < b + c.
2. (5) Let O = (0, 0) and Q = (1, 0). Find the point P on the line
with equation y = x + 1 for which the angle OP Q is a maximum.
3. (10) (a) Consider the innite integer lattice in the plane (i.e., the
set of points with integer coordinates) as a graph, with the edges
being the lines of unit length connecting nearby points. What is
the minimum number of colours that can be used to colour all the
vertices and edges of this graph, so that
(i) each pair of adjacent vertices gets two distinct colours;
(ii) each pair of edges that meet at a vertex gets two distinct
colours; and
(iii) an edge is coloured dierently than either of the two vertices
at the ends?
(b) Extend this result to lattices in real n-dimensional space.
4. (7) Let V be the vector space of all continuous real-valued func-
tions dened on the open interval (/2, /2), with the sum of two
functions and the product of a function and a real scalar dened in
the usual way.
(a) Prove that the set {sin x, cos x, tan x, sec x} is linearly
independent.
(b) Let W be the linear space generated by the four trigonometric
functions given in (a), and let T be the linear transformation

Springer International Publishing Switzerland 2016 1

E.J. Barbeau, University of Toronto Mathematics Competition
(20012015), Problem Books in Mathematics,
DOI 10.1007/978-3-319-28106-3 1
2 1. PROBLEMS OF THE CONTESTS

determined on W into V by T (sin x) = sin2 x, T (cos x) =

cos2 x, T (tan x) = tan2 x and T (sec x) = sec2 x. Determine a
basis for the kernel of T .
5. (2) Let n be a positive integer and x a real number not equal to a
nonnegative integer. Prove that
n n(n 1) n(n 1)(n 2) n(n 1)(n 2) 1
+ + + +
x x(x 1) x(x 1)(x 2) x(x 1)(x 2) (x n + 1)
n
= .
xn+1
6. (4) Prove that, for each positive integer n, the series

 kn
2k
k=1
converges to twice an odd integer not less than (n + 1)!.
7. (6) Suppose that x 1 and that x = x + {x}, where x is the
greatest integer not exceeding x and the fractional part {x} satises
0 x < 1. Dene
 
x + {x}
f (x) = .
x
(a) Determine the supremum, i.e., the least upper bound, of the
values of f (x) for 1 x.
(b) Let x0 1 be given, and for n 1, dene xn = f (xn1 ).
Prove that limn xn exists.
8. (2) A regular heptagon (polygon with seven equal sides and seven
equal angles) has diagonals of two dierent lengths. Let a be the
length of a side, b be the length of a shorter diagonal and c be the
length of a longer diagonal of a regular heptagon (so that a < b < c).
Prove ONE of the following relationships:
a2 b2 c2
2
+ 2 + 2 =6
b c a
or
b2 c2 a2
2
+ 2 + 2 = 5.
a b c

Saturday, March 9, 2002

1. (8) Let A, B, C be three pairwise orthogonal faces of a tetrahedron

meeting at one of its vertices and having respective areas a, b, c.
Let the face D opposite this vertex have area d. Prove that
d 2 = a 2 + b2 + c 2 .
2. (2) Angus likes to go to the movies. On Monday, standing in line,
he noted that the fraction x of the line was in front of him, while
1/n of the line was behind him. On Tuesday, the same fraction x of
 1. PROBLEMS OF THE CONTESTS 3

the line was in front of him, while 1/(n + 1) of the line was behind
him. On Wednesday, the same fraction x of the line was in front of
him, while 1/(n + 2) of the line was behind him. Determine a value
of n for which this is possible.
3. (11) In how many ways can the rational 2002/2001 be written as
the product of two rationals of the form (n + 1)/n, where n is a
positive integer?
4. (5) Consider the parabola of equation y = x2 . The normal is con-
structed at a variable point P and meets the parabola again in
Q. Determine the location of P for which the arc length along the
parabola between P and Q is minimized.
5. (6) Let n be a positive integer. Suppose that f is a function dened
and continuous on [0, 1] that is dierentiable on (0, 1) and satises
f (0) = 0 and f (1) = 1. Prove that, there exist n [distinct] numbers
xi (1 i n) in (0, 1) for which
n
 1
= n.
i=1
f  (xi )

6. (3) Let x, y > 0 be such that x3 +y 3 xy. Prove that x2 +y 2 1.

7. (7) Prove that no vector space over R is a nite union of proper
subspaces.
8. (7) (a) Suppose that P is an n n nonsingular matrix and that u
and v are column vectors with n components. The matrix v t P 1 u
is 11, and so can be identied with a scalar. Suppose that its value
is not equal to 1. Prove that the matrix P + uv t is nonsingular
and that
1
(P + uv t )1 = P 1 P 1 uv t P 1

where v t denotes the transpose of v and = 1 + v t P 1 u.
(b) Explain the situation when = 0.
9. (10) A sequence whose entries are 0 and 1 has the property that,
if each 0 is replaced by 01 and each 1 by 001, then the sequence
remains unchanged. Thus, it starts out as 010010101001 . . . . What
is the 2002th term of the sequence?

Sunday, March 16, 2003

1. (4) Evaluate

  
2
arctan .
n=1
n2
4 1. PROBLEMS OF THE CONTESTS

2. (3) Let a, b, c be positive real numbers for which a + b + c = abc.

Prove that
1 1 1 3
+ + .
1+a 2 1+b 2 1+c 2 2
3. (5) Solve the dierential equation
y  = yy  .

4. (4) Show that the positive integer n divides the integer nearest to
(n + 1)!
.
e
5. (6) For x > 0, y > 0, let g(x, y) denote the minimum of the three
quantities, x, y + 1/x and 1/y. Determine the maximum value of
g(x, y) and where this maximum is assumed.
6. (10) A set of n lightbulbs, each with an on-o switch, numbered
1, 2, . . . , n are arranged in a line. All are initially o. Switch 1 can
be operated at any time to turn its bulb on or o. Switch 2 can
turn bulb 2 on or o if and only if bulb 1 is o; otherwise, it does
not function. For k 3, switch k can turn bulb k on or o if and
only if bulb k 1 is o and bulbs 1, 2, . . . , k 2 are all on; otherwise
it does not function.
(a) Prove that there is an algorithm that will turn all of the bulbs
on.
(b) If xn is the length of the shortest algorithm that will turn on
all n bulbs when they are initially o, determine the largest
prime divisor of 3xn + 1 when n is odd.
7. (6) Suppose that the polynomial f (x) of degree n 1 has all real
roots and that > 0. Prove that the set {x R : |f (x)| |f  (x)|}
is a nite union of closed intervals whose total length is equal to
2n.
8. (7) Three m n matrices A, B and A + B have rank 1. Prove that
either all the rows of A and B are multiples of one and the same
vector, or that all of the columns of A and B are multiples of one
and the same vector.
9. (5) Prove that the integral

sin2 x
dx
0 2 x2
exists and evaluate it.
10. (9) Let G be a nite group of order n. Show that n is odd if and
only if each element of G is a square.
 1. PROBLEMS OF THE CONTESTS 5

Sunday, March 14, 2004

1. (2) Prove that, for any nonzero complex numbers z and w,

 
 z w 

(|z| + |w|)  + 2|z + w|.
|z| |w| 

2. (4) Prove that

 1
1 1 1 1
xx dx = 1 2 + 3 4 + 5 + .
0 2 3 4 5
3. (11) Suppose that u and v are positive integer divisors of the positive
integer n and that uv < n. Is it necessarily so that the greatest
common divisor of n/u and n/v exceeds 1?
4. (3) Let n be a positive integer exceeding 1. How many permutations
{a1 , a2 , . . . , an } of {1, 2, . . . , n} are there which maximize the value
of the sum
|a2 a1 | + |a3 a2 | + + |ai+1 ai | + + |an an1 |
over all permutations? What is the value of this maximum sum?
5. (7) Let A be a n n matrix with determinant equal to 1. Let B
be the matrix obtained by adding 1 to every entry of A. Prove that
the determinant of B is equal to 1 + s, where s is the sum of the n2
entries of A1 .
6. (5) Determine
 1   1 
dt dt
.
0 1 t4 0 1 + t4
7. (2) Let a be a parameter. Dene the sequence {fn (x) : n =
0, 1, 2, . . . } of polynomials by
f0 (x) 1
and
fn+1 (x) = xfn (x) + fn (ax)
for n 0.
(a) Prove that, for all n, x,
fn (x) = xn fn (1/x).
(b) Determine a formula for the coecient of xk (0 k n) in
fn (x).
8. (7) Let V be a complex n-dimensional inner product space. Prove
that
1
|u|2 |v|2 |u v|2 |u + v|2 |(u, v)|2 |u|2 |v|2 .
4
6 1. PROBLEMS OF THE CONTESTS

9. (8) Let ABCD be a convex quadrilateral for which all sides and
diagonals have rational length and AC and BD intersect at P .
Prove that AP , BP , CP , DP all have rational length.

Saturday, March 12, 2005

1. (5) Show that, if /2 < < /2, then

log(1 + tan tan x) dx = log sec .
0

2. (6) Suppose that f is continuously dierentiable on [0, 1] and that

1
0
f (x)dx = 0. Prove that
 1  1  1
2 f (x)2 dx |f  (x)| dx |f (x)| dx.
0 0 0

3. (7) How many n n invertible matrices A are there for which all
the entries of both A and A1 are either 0 or 1?
4. (7) Let a be a nonzero real and u and v be real 3-vectors. Solve
the equation
2ax + (v x) + u = O
for the vector x.
5. (6) Let f (x) be a polynomial with real coecients, evenly many
of which are nonzero, which is palindromic. This means that the
n read the same in either direction, i.e. ak = ank if
coecients
f (x) = k=0 ak xk or, alternatively, f (x) = xn f (1/x), where n is
the degree of the polynomial. Prove that f (x) has at least one root
of absolute value 1.
6. (9) Let G be a subgroup of index 2 contained in Sn , the group of
all permutations of n elements. Prove that G = An , the alternating
group of all even permutations.
7. (11) Let f (x) be a nonconstant polynomial that takes only integer
values when x is an integer, and let P be the set of all primes that
divide f (m) for at least one integer m. Prove that P is an innite
set.
8. (7) Let AX = B represent a system of m linear equations in n
unknowns, where A = (aij ) is an m n matrix, X = (x1 , . . . , xn )t
is an n 1 vector and B = (b1 , . . . , bm )t is an m 1 vector. Suppose
that there exists at least one solution for AX = B. Given 1 j n,
prove that the value of the jth component is the same for every
solution X of AX = B if and only if the rank of A is decreased if
the jth column of A is removed.
9. (6) Let S be the set of all real-valued functions that are dened,
positive and twice continuously dierentiable on a neighbourhood
 1. PROBLEMS OF THE CONTESTS 7

of 0. Suppose that a and b are real parameters with ab = 0, b < 0.

Dene operators from S to R as follows:
A(f ) = f (0) + af  (0) + bf  (0);
G(f ) = exp A(log f ).
(a) Prove that A(f ) G(f ) for f S;
(b) Prove that G(f + g) G(f ) + G(g) for f, g S;
(c) Suppose that H is the set of functions in S for which G(f )
f (0). Give examples of nonconstant functions, one in H and
one not in H. Prove that, if > 0 and f, g H, then f , f + g
and f g all belong to H.
10. (10) Let n be a positive integer exceeding 1. Prove that, if a graph
with 2n+1 vertices has at least 3n+1 edges, then the graph contains
a circuit (i.e., a closed non-self-intersecting chain of edges whose
terminal point is its initial point) with an even number of edges.
Prove that this statement does not hold if the number of edges is
only 3n.

Sunday, March 12, 2006

1. (10) (a) Suppose that a 6 6 square grid of unit squares (chess-
board) is tiled by 1 2 rectangles (dominoes). Prove that it can
be decomposed into two rectangles, tiled by disjoint subsets of the
dominoes.
(b) Is the same thing true for an 8 8 array?
2. (7) Let u be a unit vector in R3 and dene the operator P by
P (x) = u x for x R3 (where denotes the cross product).
(a) Describe the operator I + P 2 .
(b) Describe the action of the operator I + (sin )P + (1 cos )P 2 .
3. (2) Let p(x) be a polynomial of positive degree n with n distinct
real roots a1 < a2 < < an . Let b be a real number for which
2b < a1 + a2 . Prove that
2n1 |p(b)| |p (a1 )(b a1 )|.

4. (8) Two parabolas have parallel axes and intersect in two points.
Prove that their common chord bisects the segment whose endpoints
are the points of contact of their common tangent.
5. (10) Suppose that you have a 3 3 grid of squares. A line is a
set of three squares in the same row, the same column or the same
diagonal; thus, there are eight lines.
Two players A and B play a game. They take alternate turns,
A putting a 0 in any unoccupied square of the grid and B putting
a 1. The rst player is A, and the game cannot go on for more
than nine moves. (The play is similar to noughts and crosses, or
8 1. PROBLEMS OF THE CONTESTS

tic-tac-toe.) A move is legitimate if it does not result in two lines of

squares being lled in with dierent sums. The winner is the last
player to make a legitimate move.
(For example, if there are three 0s down the diagonal, then B
can place a 1 in any vacant square provided it completes no other
line, for then the sum would dier from the diagonal sum. If there
are two zeros at the top of the main diagonal and two ones at the
left of the bottom line, then the lower right square cannot be lled
by either player, as it would result in two lines with dierent sums.)
(a) What is the maximum number of legitimate moves possible in
a game?
(b) What is the minimum number of legitimate moves possible in a
game that would not leave a legitimate move available for the
next player?
(c) Which player has a winning strategy? Explain.
6. (6) Suppose that k is a positive integer and that
f (t) = a1 e1 t + a2 e2 t + + ak ek t
where a1 , . . . , ak , 1 , . . . , k are real numbers with 1 < 2 < <
k . Prove that f (t) = 0 has nitely many real solutions. What is
the maximum number of solutions possible, as a function of k?
7. (7) Let A be a real 3 3 invertible matrix for which the sums of
the rows, columns and two diagonals are all equal. Prove that the
rows, columns and diagonal sums of A1 are all equal.
8. (6) Let f (x) be a real function dened and twice dierentiable on
an open interval containing [1, 1]. Suppose that 0 < and
that |f (x)| and |f  (x)| for 1 x 1. Prove that

|f  (x)| 2
for 1 x 1. (Part marks are possible for the weaker inequality
|f  (x)| + .)
9. (2) A high school student asked to solve the surd equation

3x 2 2x 3 = 1
gave the following answer: Squaring both sides leads to
3x 2 2x 3 = 1
so x = 6. The answer is, in fact, correct.
Show that there are innitely many real quadruples (a, b, c, d)
for which this method leads to a correct solution of the surd equation

ax b cx d = 1.
 1. PROBLEMS OF THE CONTESTS 9

10. (8) Let P be a planar polygon that is not convex. The vertices can
be classied as either convex or concave according as to whether the
angle at the vertex is less than or greater than 180 respectively.
There must be at least two convex vertices. Select two consecutive
convex vertices (i.e., two interior angles less than 180 for which all
interior angles in between exceed 180 ) and join them by a segment.
Reect the edges between these two convex angles in the segment
to form along with the other edges of P a polygon P1 . If P1 is not
convex, repeat the process, reecting some of the edges of P1 in a
segment joining two consecutive convex vertices, to form a polygon
P2 . Repeat the process. Prove that, after a nite number of steps,
we arrive at a polygon Pn that is convex.

Sunday, March 11, 2007

1. (10) An m n rectangular array of distinct real numbers has the

property that the numbers in each row increase from left to right.
The entries in each column, individually, are rearranged so that the
numbers in each column increase from top to bottom. Prove that
in the nal array, the numbers in each row will increase from left to
right.
2. (11) Determine distinct positive integers a, b, c, d, e such that the
ve numbers a, b2 , c3 , d4 , e5 constitute an arithmetic progression.
(The dierence between adjacent pairs is the same.)
3. (10) Prove that the set {1, 2, . . . , n} can be partitioned into k sub-
sets with the same sum if and only if k divides 12 n(n + 1) and
n 2k 1.
4. (6) Suppose that f (x) is a continuous real-valued function dened
on the interval [0, 1] that is twice dierentiable on (0, 1) and satises
(i) f (0) = 0 and (ii) f  (x) > 0 for 0 < x < 1.
(a) Prove that there exists a number a for which 0 < a < 1 and
f  (a) < f (1);
(b) Prove that there exists a unique number b for which a < b < 1
and f  (a) = f (b)/b.
5. (6) For x 1 and x = 0, let

8[1 (1 x)1/2 ]3
f (x) = .
x2
(a) Prove that limx0 f (x) exists. Take this as the value of f (0).
(b) Determine the smallest closed interval that contains all values
assumed by f (x) on its domain.
(c) Prove that f (f (f (x))) = f (x) for all x 1.
6. (4) Let h(n) denote the number of nite sequences {a1 , a2 , . . . , ak }
of positive integers exceeding 1 for which k 1, a1 a2 ak
10 1. PROBLEMS OF THE CONTESTS

and n = a1 a2 ak . (For example, if n = 20, there are four such

sequences {20}, {10, 2}, {5, 4} and {5, 2, 2}, and h(20) = 4.)
Prove that

h(n)
= 1.
n=1
n2

7. (7) Find the Jordan canonical form of the matrix uvt where u and
v are column vectors in Cn .
8. (8) Suppose that n points are given in the plane, not all collinear.
Prove that there are at least n distinct straight lines that can be
drawn through pairs of the points.
9. (11) Which integers can be written in the form
(x + y + z)2
xyz
where x, y, z are positive integers?
10. (5) Solve the following dierential equation
2y  = 3|y|1/3
subject to the initial conditions
y(2) = 1 and y(3) = 1.
Your solution should be everywhere dierentiable.

Sunday, March 9, 2008

1. (8) Three angles of a heptagon (7-sided polygon) inscribed in a
circle are equal to 120 . Prove that at least two of its sides are
equal.
2. (6) (a) Determine a real-valued function g dened on the real num-
bers that is decreasing and for which g(g(x)) = 2x + 2.
(b) Prove that there is no real-valued function f dened on the real
numbers that is decreasing and for which f (f (x)) = x + 1.
3. (4) Suppose that a is a real number and the sequence {an } is dened
recursively by a0 = a and
an+1 = an (an 1)
for n 0. Find the values of a for which the sequence {an } con-
verges.
4. (2) Suppose that u, v, w, z are complex numbers for which u + v +
w + z = u2 + v 2 + w2 + z 2 = 0. Prove that
(u4 + v 4 + w4 + z 4 )2 = 4(u8 + v 8 + w8 + z 8 ).
 1. PROBLEMS OF THE CONTESTS 11

5. (7) Suppose that a, b, c C with ab = 1. Evaluate the determinant

of
c a a2 an1
b c a an2
2
b b c an3
.
.. .. .. ..
. . . .
bn1 bn2 c
6. (10) 2008 circular coins, possibly of dierent diameters, are placed
on the surface of a at table in such a way that no coin is on top of
another coin. What is the largest number of points at which two of
the coins could be touching?
7. (9) Let G be a group of nite order and identity e. Suppose that
is an automorphism of G onto itself with the following properties:
(1) (x) = x if and only if x = e; (2) ((x)) = x for each element
x of G.
(a) Give an example of a group and automorphism for which these
conditions are satised.
(b) Prove that G is commutative (i.e., xy = yx for each pair x, y of
elements in G).
8. (11) Let b 2 be an integer base of numeration and let 1 r b1.
Determine the sum of all rdigit numbers of the form
ar1 ar2 . . . a2 a1 a0 ar1 br1 + ar2 br2 + + a1 r + a0
whose digits increase strictly from left to right: 1 ar1 < ar2 <
< a1 < a0 b 1.
9. (2) For each positive integer n, let
n
2k
Sn = .
k2
k=1
Prove that Sn+1 /Sn is not a rational function of n.
10. (8) A point is chosen at random (with the uniform distribution) on
each side of a unit square. What is the probability that the four
points are the vertices of a quadrilateral with area exceeding 12 ?

Sunday, March 8, 2009

1. (5) Determine the supremum and the inmum of
(x 1)x1 xx
(x (1/2))2x1
for x > 1.
2. (7) Let n and k be integers with n 0 and k 1. Let x0 , x1 , . . . ,
xn be n + 1 distinct points in Rk and let y0 , y1 , . . . , yn be n + 1
real numbers (not necessarily distinct). Prove that there exists a
12 1. PROBLEMS OF THE CONTESTS

polynomial p of degree at most n in the coordinates of x with respect

to the standard basis for which p(xi ) = yi for 0 i n.
3. (11) For each positive integer n, let p(n) be the product of all posi-
tive integral divisors of n. Is it possible to nd two distinct positive
integers m and n for which p(m) = p(n)?
4. (4) Let {an } be a real sequence for which

an
n=1
n

converges. Prove that

a1 + a2 + + an
lim = 0.
n n

5. (7) Find a 3 3 matrix A with elements in Z2 for which A7 = I

and A = I. (Here, I is the identity matrix and Z2 is the eld of
two elements 0 and 1 where addition and multiplication are dened
modulo 2.)
6. (11) Determine all solutions in nonnegative integers (x, y, z, w) to
the equation
2x 3y 5z 7w = 1.

7. (3) Let n 2. Minimize a1 + a2 + + an subject to the constraints

0 a1 a2 an and a1 a2 + a2 a3 + + an1 an + an a1 = 1.
(When n = 2, the latter condition is a1 a2 = 1; when n 3, the
sum on the left has exactly n terms.)
8. (7) Let a, b, c be members of a real inner-product space (V,
, )
whose norm is given by x2 =
x, x. (You may assume that V is
Rn if you wish.) Prove that
a + b + b + c + c + a a + b + c + a + b + c
for a, b, c, V .
9. (11) Let p be a prime congruent to 1 modulo 4. For each real number
x, let {x} = x x denote the fractional part of x. Determine
  k 2  1

: 1 k (p 1) .
p 2

10. (10) Suppose that a path on an m n grid consisting of the lattice

points {(x, y) : 1 x m, 1 y n} (x and y both integers)
consisting of mn 1 unit segments begins at the point (1, 1), passes
through each point of the grid exactly once, does not intersect itself
and nishes at the point (m, n). Show that the path partitions the
rectangle bounded by the lines x = 1, x = m, y = 1, y = n into two
 1. PROBLEMS OF THE CONTESTS 13

subsets of equal area, the rst consisting of regions opening to the

left or up, and the second consisting of regions opening to the right
or down.

Sunday, March 7, 2010

1. (8) Let F1 and F2 be the foci of an ellipse and P be a point in
the plane of the ellipse. Suppose that G1 and G2 are points on the
ellipse for which P G1 and P G2 are tangents to the ellipse. Prove
that F1 P G1 = F2 P G2 .
2. (10) Let u0 = 1, u1 = 2 and un+1 = 2un + un1 for n 1. Prove
that, for every nonnegative integer n,
  (i + j + k)! 
un = : i, j, k 0, i + j + 2k = n .
i!j!k!

3. (7) Let a and b, the latter nonzero, be vectors in R3 . Determine

the value of for which the vector equation
a (x b) = b
is solvable, and then solve it.
4. (10) The plane is partitioned into n regions by three families of
parallel lines. What is the least number of lines to ensure that
n 2010?
5. (2) Let m be a natural number, and let c, a1 , a2 , . . . , am be complex
numbers for which |ai | = 1 for i = 1, 2, . . . , m. Suppose also that
m

lim ani = c.
n
i=1

Prove that c = m and that ai = 1 for i = 1, 2, . . . , m.

6. (2) Let f (x) be a quadratic polynomial. Prove that there exist
quadratic polynomials g(x) and h(x) for which
f (x)f (x + 1) = g(h(x)).

7. (6) Suppose that f is a continuous real-valued function dened on

the closed interval [0, 1] and that
 1 2  1   1 
xf (x) dx = f (x) dx x2 f (x) dx .
0 0 0

Prove that there is a point c (0, 1) for which f (c) = 0.

8. (7) Let A be an invertible symmetric nn matrix with entries {aij }
in Z2 . Prove that there is an n n matrix with entries in Z2 such
that A = M t M only if aii = 0 for some i.
14 1. PROBLEMS OF THE CONTESTS

9. (6) Let f be a real-valued function dened on R with a continuous

third derivative, let S0 = {x : f (x) = 0}, and, for k = 1, 2, 3,
Sk = {x : f (k) (x) = 0}, where f (k) denotes the kth derivative of f .
Suppose also that R = S0 S1 S2 S3 . Must f be a polynomial
of degree not exceeding 2?
10. (6) Prove that the set Q of rationals can be written as the union of
countably many subsets of Q each of which is dense in the set R of
real numbers.

Sunday, March 6, 2011

1. (6) Let S be a nonvoid set of real numbers with the property that,
for each real number x, there is a unique real number f (x) belonging
to S that is farthest from x, i.e., for each y in S distinct from f (x),
|x f (x)| > |x y|. Prove that S must be a singleton.
2. (3) Let u and v be positive reals. Minimize the larger of the two
values
1 1
2u + 2 and 2v + 2 .
v u
3. (10) Suppose that S is a set of n nonzero real numbers such that
exactly p of them are positive and exactly q are negative. Determine
all the pairs (n, p) such that exactly half of the threefold products
abc of distinct elements a, b, c of S are positive.
4. (4) Let {bn : n 1} be a sequence of positive real numbers such
that
3bn+2 bn+1 + 2bn
for every positive integer n. Prove that either the sequence con-
verges or that it diverges to innity.
5. (2) Solve the system
x + xy + xyz = 12
y + yz + yzx = 21
z + zx + zxy = 30.

6. (2) [The problem posed on the competition is not available for pub-
lication, as it appeared as Enigma problem 1610 in the August 25,
2010 issue of the New Scientist. It concerned the determination of
the nal score of a three game badminton match where the score of
each competitor was in arithmetic progression. The full statement
of the problem is available on the internet.]
7. (7) Suppose that there are 2011 students in a school and that each
student has a certain number of friends among his schoolmates. It
is assumed that if A is a friend of B, then B is a friend of A, and
also that there may exist certain pairs that are not friends. Prove
 1. PROBLEMS OF THE CONTESTS 15

that there is a nonvoid subset S of these students for which every

student in the school has an even number of friends in S.
8. (9) The set of transpositions of the symmetric group S5 on
{1, 2, 3, 4, 5} is
{(12), (13), (14), (15), (23), (24), (25), (34), (35), (45)}
where (ab) denotes the permutation that interchanges a and b and
leaves every other element xed. Determine a product of all trans-
positions, each occuring exactly once, that is equal to the identity
permutation , which leaves every element xed.
9. (7) Suppose that A and B are two square matrices of the same
order for which the indicated inverses exist. Prove that
(A + AB 1 A)1 + (A + B)1 = A1 .

10. (2) Suppose that p is an odd prime. Determine the number of

subsets S contained in {1, 2, . . . , 2p 1, 2p} for which (a) S has
exactly p elements, and (b) the sum of the elements of S is a multiple
of p.

Saturday, March 10, 2012

1. (8) An equilateral triangle of side length 1 can be covered by ve
equilateral triangles of side length u. Prove that it can be covered
by four equilateral triangles of side length u. (A triangle is a closed
convex set that contains its three sides along with its interior.)
2. (11) Suppose that f is a function dened on the set Z of integers
that takes integer values and satises the condition that f (b) f (a)
is a multiple of ba for every pair a, b, of integers. Suppose also that
p is a polynomial with integer coecients such that p(n) = f (n) for
innitely many integers n. Prove that p(x) = f (x) for every positive
integer x.
3. (2) Given the real numbers a, b, c not all zero, determine the real
solutions x, y, z, u, v, w for the system of equations:
x 2 + v 2 + w 2 = a2
u 2 + y 2 + w 2 = b2
u2 + v 2 + z 2 = c2
u(y + z) + vw = bc
v(x + z) + wu = ca
w(x + y) + uv = ab.
16 1. PROBLEMS OF THE CONTESTS

4. (11) (a) Let n and k be positive integers. Prove that the least
common multiple of {n, n + 1, n + 2, . . . , n + k} is equal to
 
n+k
rn
k
for some positive integer r.
(b) For each positive integer k, prove that there exist innitely many
positive integers n, for which the number r dened in part (a) is
equal to 1.
5. (8) Let C be a circle and Q a point in the plane. Determine the
locus of the centres of those circles that are tangent to C and whose
circumference passes through Q.
6. (6) Find all continuous real-valued functions dened on R that sat-
isfy f (0) = 0 and
f (x) f (y) = (x y)g(x + y)
for some real-valued function g(x).
7. (6) Consider the following problem:
Suppose that f (x) is a continuous real-valued function dened
on the interval [0, 2] for which
 2  2
f (x) dx = (f (x))2 dx.
0 0
Prove that there exists a number c [0, 2] for which either f (c) = 0
or f (c) = 1.
(a) Criticize the following solution:
2
Solution. Clearly 0 f (x)dx 0. By the Extreme Value Theo-
rem, there exist numbers u and v in [0, 2] for which f (u) f (x)
f (v) for 0 x 2. Hence
 2  2  2
f (u) f (x) dx f (x)2 dx f (v) f (x) dx.
0 0 0
2 2
Since 0 f (x) dx = 1 0 f (x) dx, by the Intermediate Value The-
2

orem, there exists a number c [0, 2] for which f (c) = 1. 

(b) Show that there is a nontrivial function f that satises the con-
ditions of the problem but that never assumes the value 1.
(c) Provide a complete solution of the problem.
8. (8) Determine the area of the set of points (x, y) in the plane that
satisfy the two inequalities:
x2 + y 2 2
x4 + x3 y 3 xy + y 4 .
 1. PROBLEMS OF THE CONTESTS 17

9. (10) In a round-robin tournament of n 2 teams, each pair of

teams plays exactly one game that results in a win for one team
and a loss for the other (there are no ties).
(a) Prove that the teams can be labelled t1 , t2 , . . . , tn , so that, for
each i with 1 i n 1, team ti beats ti+1 .
(b) Suppose that a team t has the property that, for each other team
u, one can nd a chain u1 , u2 , . . . , um of (possibly zero) distinct
teams for which t beats u1 , ui beats ui+1 for 1 i m 1 and
um beats u. Prove that all of the n teams can be ordered as in
(a) so that t = t1 and each ti beats ti+1 for 1 i n 1.
(c) Let T denote the set of teams that can be labelled as t1 in an
ordering of teams as in (a). Prove that, in any ordering of teams
as in (a), all the teams in T occur before all the teams that are
not in T .
10. (7) Let A be a square matrix whose entries are complex numbers.
Prove that A = A if and only if AA = A2 .

Saturday, March 9, 2013

1. (11) (a) Let a be an odd positive integer exceeding 3, and let n be
a positive integer. Prove that
n
a2 1
has at least n + 1 distinct prime divisors.
(b) When a = 3, determine all the positive integers n for which the
assertion in (a) is false.
2. (8) ABCD is a square; points U and V are situated on the re-
spective sides BC and CD. Prove that the perimeter of triangle
CU V is equal to twice the sidelength of the square if and only if
U AV = 45 .
3. (6) Let f (x) be a convex increasing real-valued function dened on
the closed interval [0, 1] for which f (0) = 0 and f (1) = 1. Suppose
that 0 < a < 1 and that b = f (a).
(a) Prove that f is continuous on (0, 1).
(b) Prove that
 1
0ab2 (x f (x)) dx 1 4b(1 a).
0

4. (11) Let S be the set of integers of the form x2 + xy + y 2 , where x

and y are integers.
(a) Prove that any prime p in S is either equal to 3 or is congruent
to 1 modulo 6.
(b) Prove that S includes all squares.
(c) Prove that S is closed under multiplication.
18 1. PROBLEMS OF THE CONTESTS

5. (8) A point on an ellipse is joined to the ends of its major axis.

Prove that the portion of a directrix intercepted by the two joining
lines subtends a right angle at the corresponding focus.
6. (2) Let p(x) = x4 + ax3 + bx2 + cx + d be a polynomial with rational
coecients. Suppose that p(x) has exactly one real root r. Prove
that r is rational.
7. (7) Let (V,
) be a two-dimensional inner product space over the
complex eld C and let z1 and z2 be unit vectors in V . Prove that
1
sup{|
z, z1 
z, z2 | : z = 1}
2
with equality if and only if
z1 , z2  = 0.
8. (7) For any real square matrix A, the adjugate matrix, adj A, has
as its elements the cofactors of the transpose of A, so that
A adj A = adj A A = (det A)I.
(a) Suppose that A is an invertible square matrix. Show that
(adj (At ))1 = (adj (A1 ))t .
(b) Suppose that adj (At ) is orthogonal (i.e., its inverse is its trans-
pose). Prove that A is invertible.
(c) Let A be an invertible nn square matrix and let det (tI A) =
tn + c1 tn1 + + cn1 t + cn be the characteristic polynomial
of the matrix A. Determine the characteristic polynomial of adj
A.
9. (9) Let S be a set upon whose elements there is a binary operation
(x, y) xy which is associative (i.e. x(yz) = (xy)z). Suppose that
there exists an element e S for which e2 = e and that for each
a S, there is at least one element b for which ba = e and at most
one element c for which ac = e. Prove that S is a group with this
binary operation.
10. (6) (a) Let f be a real-valued function dened on the real number
eld R for which |f (x)f (y)| < |xy| for any pair (x, y) of distinct
elements of R. Let f (n) denote the nth composite of f dened by
f (1) (x) = f (x) and f (n+1) (x) = f (f (n) (x)) for n 2. Prove that
exactly one of the following situations must occur:
(i) limn+ f (n) (x) = + for each real x;
(ii) limn+ f (n) (x) = for each real x;
(iii) there is a real number z such that
lim f (n) (x) = z
n+

for each real x.

(b) Give examples to show that each of the three cases in (a) can
occur.
 1. PROBLEMS OF THE CONTESTS 19

Sunday, March 9, 2014

1. (7) The permanent, per A, of an n n matrix A = (ai,j ), is equal to
the sum of all possible products of the form a1,(1) a2,(2) an,(n) ,
where runs over all the permutations on the set {1, 2, . . . , n}.
(This is similar to the denition of determinant, but there is no
sign factor.) Show that, for any n n matrix A = (ai,j ) with
positive real terms,
n1

per A n! ai,j .
1i,jn

2. (11) For a positive integer N written in base 10 numeration, N 

denotes the integer with the digits of N written in reverse order.
There are pairs of integers (A, B) for which A, A , B, B  are all dis-
tinct and A B = B  A . For example,
3516 8274 = 4728 6153.
(a) Determine a pair (A, B) as described above for which both A
and B have two digits, and all four digits involved are distinct.
(b) Are there any pairs (A, B) as described above for which A has
two and B has three digits?
3. (11) Let n be a positive integer. A nite sequence {a1 , a2 , . . . , an }
  ai is said to be tight if and only if 1 a1 < a2 <
of positive integers
< an , all n2 dierences aj ai with i < j are distinct, and an
is as small as possible.
(a) Determine a tight sequence for n = 5.
(b) Prove that there is a polynomial p(n) of degree not exceeding 3
such that an p(n) for every tight sequence {ai } with n entries.
4. (6) Let f (x) be a continuous real-valued function on [0, 1] for which
 1  1
f (x) dx = 0 and x f (x)dx = 1.
0 0
(a) Give an example of such a function.
(b) Prove that there is a nontrivial open interval I contained in
(0, 1) for which |f (x)| > 4 for x I.
5. (4) Let n be a positive integer. Prove that
n  n n
1 1 1  2k1
n = = n1
k k k2nk 2 k
k=1 k=1 k=1
  
1  n
= n1 : k odd , 1 k n .
2 k
20 1. PROBLEMS OF THE CONTESTS

6. (2) Let f (x) = x6 x4 + 2x3 x2 + 1.

(a) Prove that f (x) has no positive real roots.
(b) Determine a nonzero polynomial g(x) of minimum degree for
which all the coecients of f (x)g(x) are nonnegative rational
numbers.
(c) Determine a polynomial h(x) of minimum degree for which all
the coecients of f (x)h(x) are positive rational numbers.
7. (3) Suppose that x0 , x1 , . . . , xn are real numbers. For 0 i n,
dene
yi = max (x0 , x1 , . . . , xi ).
Prove that
n1

yn2 4x2n 4 yi (xi+1 xi ).
i=0

When does equality occur?

8. (8) The hyperbola with equation x2 y 2 = 1 has two branches, as
does the hyperbola with equation y 2 x2 = 1. Choose one point
from each of the four branches of the locus of (x2 y 2 )2 = 1 such
that area of the quadrilateral with these four vertices is minimized.
9. (4) Let {an } and {bn } be positive real sequences such that
an
lim =u>0
n n

and
 n
bn
lim = v > 0.
n an

Prove that
 
bn
lim =1
n an

and
lim (bn an ) = u log v.
n

10. (6) Does there exist a continuous real-valued function dened on R

for which f (f (x)) = x for all x R?

Sunday, March 8, 2015

1. (6) Suppose that u and v are two real-valued functions dened on

the set of reals. Let f (x) = u(v(x)) and g(x) = u(v(x)) for each
real x. If f (x) is continuous, must g(x) also be continuous?
2. (8) Given 2n distinct points in space, the sum S of the lengths of
all the segments joining pairs of them is calculated. Then n of the
points are removed along with all the segments having at least one
 1. PROBLEMS OF THE CONTESTS 21

endpoint from among them. Prove that the sum of the lengths of
all the remaining segments is less that 12 S.
3. (6) Let f : [0, 1] R be continuously dierentiable. Prove that
  1 
 f (0) + f (1)  1
 f (x) dx sup{|f  (x)| : 0 x 1}.
 2 4
0

4. (3) Determine all the values of the positive integer n 2 for which
the following statement is true, and for each, indicate when equality
holds.
For any nonnegative real numbers x1 , x2 , . . . , xn ,
(x1 + x2 + + xn )2 n(x1 x2 + x2 x3 + + xn1 xn + xn x1 ),
where the right side has n summands.
5. (5) Let f (x) be a real polynomial of degree 4 whose graph has two
real inection points. There are three regions bounded by the graph
and the line passing through these inection points. Prove that two
of these regions have equal area and that the area of the third region
is equal to the sum of the other two areas.
6. (11) Using the digits 1, 2, 3, 4, 5, 6, 7, 8, each exactly once, create
two numbers and form their product. For example, 472 83,156 =
39,249,632. What are the smallest and the largest values such a
product can have?
7. (5) Determine
 2
ex dx
.
0 e1x + ex1
8. (4) Let {an } and {bn } be two decreasing positive real sequences for
which

an =
n=1
and


bn = .
n=1
Let I be a subset of the natural numbers, and dene the sequence
{cn } by

an , if n I;
cn =
bn , if n I.

Is it possible for n=1 cn to converge?
9. (7) What is the dimension of the vector subspace of Rn generated
by the set of vectors
((1), (2), (3), . . . , (n))
22 1. PROBLEMS OF THE CONTESTS

where runs through all n! of the permutations of the rst n natural

numbers?
10. (2) (a) Let
g(x, y) = x2 y + xy 2 + xy + x + y + 1.
We form a sequence {x0 } as follows: x0 = 0. The next term x1
is the unique solution 1 of the linear equation g(t, 0) = 0. For
each n 2, xn is the solution other than xn2 of the equation
g(t, xn1 ) = 0.
Let {fn } be the Fibonacci sequence determined by f0 = 0,
f1 = 1 and fn = fn1 + fn2 for n 2. Prove that, for any
nonnegative integer k,
fk fk+2
x2k = and x2k+1 = .
fk+1 fk+1
(b) Let
h(x, y) = x2 y + xy 2 + xy + (x + y) +
be a polynomial with real coecients , , . We form a bilateral
sequence {xn : n Z} as follows. Let x0 = 0 be given arbitrar-
ily. We select x1 and x1 to be the two solutions of the quadratic
equation h(t, x0 ) = 0 in either order. From here, we can dene in-
ductively the terms of the sequence for positive and negative values
of the index so that xn1 and xn+1 are the two solutions of the
equation h(t, xn ) = 0. We suppose that at each stage, neither of
these solutions is zero.
Prove that the sequence {xn } has period 5 (i.e. xn+5 = xn for
each index n) if and only if 3 + 2 = 0.
 CHAPTER 2

Algebra

2001:5. Let n be a positive integer and x a real number not equal to a

nonnegative integer. Prove that
n n(n 1) n(n 1)(n 2) n(n 1)(n 2) 1
+ + + +
x x(x 1) x(x 1)(x 2) x(x 1)(x 2) (x n + 1)
n
= .
xn+1
Solution 1. The result holds for n = 1 and x = 0. Suppose, as an
induction hypothesis, the result holds for n = k and x equal to any real
number that is not a nonnegative integer; note that in this case, the left side
has k terms. When n = k + 1,
k + 1 (k + 1)k (k + 1)k(k 1) (k + 1)k(k 1) 1
+ + + +
x x(x 1) x(x 1)(x 2) x(x 1)(x 2) (x k)
 
k+1 k+1 k k(k 1) k(k 1) 1
= + + + +
x x x 1 (x 1)(x 2) (x 1) (x k)
   
k+1 k+1 k k+1 k
= + = 1+
x x x1k+1 x xk
 
k+1 x k+1
= =
x xk x (k + 1) + 1
as desired. The result follows by induction.
a
Solution 2. [P. Gyrya] For real a and positive integer m, let m denote
a(a 1) (a m + 1)/m! and let a0 be 1. It is straightforward to establish
that, for each positive integer k:
     
x x1 x1
= +
k k k1

Springer International Publishing Switzerland 2016 23

E.J. Barbeau, University of Toronto Mathematics Competition
(20012015), Problem Books in Mathematics,
DOI 10.1007/978-3-319-28106-3 2
24 2. ALGEBRA

and
   k  
x xi1
= .
k i=0
ki
The left side of the required equality is equal to
     
n! x1 x2 xn
+ + +
x (x n + 1) n 1 n2 nn
 
n! x n
= =
x (x n + 1) n 1 xn+1
as desired.
2001:8 A regular heptagon (polygon with seven equal sides and seven
equal angles) has diagonals of two dierent lengths. Let a be the length of
a side, b be the length of a shorter diagonal and c be the length of a longer
diagonal of a regular heptagon (so that a < b < c). Prove ONE of the
following relationships:

a2 b2 c2
2
+ 2 + 2 =6
b c a
or

b2 c2 a2
2
+ 2 + 2 = 5.
a b c
Solution 1. Let A, B, C, D, E be consecutive vertices of the regular
heptagon. Let AB, AC and AD have respective lengths a, b, c, and let
BAC = . Then = /7, the length of BC, of CD and of DE is a,
the length of AE is c, CAD = DAE = , since the angles are sub-
tended by equal chords of the circumcircle of the heptagon, ADC = 2,
ADE = AED = 3 and ACD = 4. Triangles ABC and ACD can be
glued together along BC and DC (with C on C) to form a triangle similar
to ABC, whence
a+c b
(2.1) = .
b a
Triangles ACD and ADE can be glued together along CD and ED (with D
on D) to form a triangle similar to triangle ABC, whence
b+c b
(2.2) = .
c a
Equation (2.2) can be rewritten as 1b = a1 1c . whence
ac
b= .
ca
Substituting this into (2.1) yields
(c + a)(c a) c
=
ac ca
 2. ALGEBRA 25

which simplies to
(2.3) a3 a2 c 2ac2 + c3 = 0.
Note also from (2.1) that b2 = a2 + ac.
a2 b2 c2 a4 c2 + b4 a2 + c4 b2 6a2 b2 c2
2
+ 2 + 2 6=
b c a a 2 b2 c 2
a c + (a + 2a c + a c )a + c4 (a2 + ac) 6a2 c2 (a2 + ac)
4 2 4 3 2 2 2
=
a 2 b2 c 2
a6 + 2a5 c 4a4 c2 6a3 c3 + a2 c4 + ac5
=
a 2 b2 c 2
a(a + 3ac + c )(a3 a2 c 2ac2 + c3 )
2 2
= = 0.
a 2 b2 c 2

b2 c2 a2
+ + 5
a2 b2 c2
(a4 + 2a3 c + a2 c2 )c2 + a2 c4 + a4 (a2 + ac) 5a2 c2 (a2 + ac)
=
a 2 b2 c 2
a + a c 4a c 3a c + 2a2 c4
6 5 4 2 3 3
=
a 2 b2 c 2
a2 (a + 2c)(a3 a2 c 2ac2 + c3 )
= = 0.
a 2 b2 c 2
Solution 2 (of the rst result). [J. Chui] Let the heptagon be ABCDEF G
and = /7. Using the Law of Cosines in the indicated triangles ACD and
ABC, we obtain the following:
 
a 2 + c 2 b2 1 a c b2
cos 2 = = +
2ac 2 c a ac
 2
2a2 b2 1 b
cos 5 = =1
2a2 2 a
from which, since cos 2 = cos 5,
 2  
1 b 1 a c b2
1 + = +
2 a 2 c a ac
or
b2 a c b2
(2.4) 2
=2+ + .
a c a ac
Examining triangles ABC and ADE, we nd that cos = b/2a and
cos = (2c2 a2 )/(2c2 ) = 1 (a2 /2c2 ), so that
a2 b
(2.5) 2
=2 .
c a
26 2. ALGEBRA

Examining triangles ADE and ACF , we nd that cos 3 = a/2c and cos 3 =
(2b2 c2 )/(2b2 ), so that

c2 a
(2.6) 2
=2 .
b c

Adding Eqs. (2.4), (2.5), (2.6) yields

b2 c2 a2 c2 bc b2
+ + = 6 + .
a2 b2 c2 ac
By Ptolemys Theorem, the sum of the products of pairs of opposite sides of
a concylic quadrilaterial is equal to the product of the diagonals. Applying
this to the quadrilaterals ABDE and ABCD, respectively, yields c2 = a2 +bc
and b2 = ac + a2 , whence c2 bc b2 = a2 + bc bc ac a2 = ac and we
nd that
b2 c2 a2
+ + = 6 1 = 5.
a2 b2 c2

Solution 3. There is no loss of generality in assuming that the vertices of

the heptagon are placed at the seventh roots of unity on the unit circle in the
complex plane. Let = cos(2/7) + i sin(2/7) be the fundamental seventh
root of unity. Then 7 = 1, 1 + + 2 + + 6 = 0 and (, 6 ), ( 2 , 5 ),
( 3 , 4 ) are pairs of complex conjugates. We have that

a = | 1| = | 6 1|

b = | 2 1| = | 9 1|

c = | 3 1| = | 4 1|.

It follows from this that

b c a
= | + 1| = | 2 + 1| = | 3 + 1|,
a b c
whence
b2 c2 a2
2
+ 2 + 2 = ( + 1)( 6 + 1) + ( 2 + 1)( 5 + 1) + ( 3 + 1)( 4 + 1)
a b c
= 2 + + 6 + 2 + 2 + 5 + 2 + 3 + 4
= 6 + ( + 2 + 3 + 4 + 5 + 6 ) = 6 1 = 5 .

Also
a b c
= | 4 + 2 + 1| = | 6 + 3 + 1| = | 2 + + 1|,
b c a
 2. ALGEBRA 27

whence
a2 b2 c2
2
+ 2 + 2 = ( 4 + 2 + 1)( 3 + 5 + 1) + ( 6 + 3 + 1)( + 4 + 1)
b c a
+ ( 2 + + 1)( 5 + 6 + 1)
= (3 + 2 2 + 3 + 4 + 2 5 ) + (3 + + 2 3 + 2 4 + 6 )
+ (3 + 2 + 2 + 5 + 2 6 )
= 9 + 3( + 2 + 3 + 4 + 5 + 6 ) = 9 3 = 6.

Solution 4. Suppose that the circumradius of the heptagon is 1. By

considering isosceles triangles with base equal to the sides or diagonals of the
heptagon and apex at the centre of the circumcircle, we see that

a = 2 sin = 2 sin 6 = 2 sin 8

b = 2 sin 2 = 2 sin 9
c = 2 sin 3 = 2 sin 4

where = /7 is half the angle subtended at the circumcentre by each side

of the heptagon. Observe that
1 1 2 1 3
cos 2 = ( + 6 ) cos 4 = ( + 5 ) cos 6 = ( + 4 )
2 2 2
where is the fundamental primitive root of unity. We have that

b c a
= 2 cos = 2 cos 6 = 2 cos 2 = 2 cos 4
a b c
whence
b2 c2 a2
2
+ 2 + 2 = 4 cos2 6 + 4 cos2 2 + 4 cos2 4
a b c
= ( 3 + 4 )2 + ( + 6 )2 + ( 2 + 5 )2
= 6 + 2 + + 2 + 2 + 5 + 4 + 2 + 3 = 6 1 = 5.

Also
a sin 6
= = 4 cos2 2 1 = ( + 6 )2 1 = 1 + 2 + 5
b sin 2
b sin 9
= = 4 cos2 3 1 = 4 cos2 4 1
c sin 3
= ( 2 + 5 )2 1 = 1 + 4 + 3
c sin 3
= = 4 cos2 6 1 = ( 3 + 4 )2 1 = 1 + 6 + ,
a sin
28 2. ALGEBRA

whence
a2 b2 c2
+ + = (3 + 2 2 + 3 + 4 + 2 5 ) + (3 + + 2 3 + 2 4 + 6 )
b2 c2 a2
+ (3 + 2 + 2 + 5 + 2 6 )
= 9 + 3( + 2 + 3 + 4 + 5 + 6 ) = 9 3 = 6.
Comment. The regular heptagon is a rich source of interesting relation-
ships. See the paper Golden elds; a case for the heptagon by Peter Steinbach
in Mathematics Magazine 70:1 (February, 1997), 2231.
2002:2. Angus likes to go to the movies. On Monday, standing in line,
he noted that the fraction x of the line was in front of him, while 1/n of
the line was behind him. On Tuesday, the same fraction x of the line was in
front of him, while 1/(n + 1) of the line was behind him. On Wednesday, the
same fraction x of the line was in front of him, while 1/(n + 2) of the line
was behind him. Determine a value of n for which this is possible.
Answer. When x = 5/6, he could have 1/7 of a line of 42 behind him,
1/8 of a line of 24 behind him and 1/9 of a line of 18 behind him. When
x = 11/12, he could have 1/14 of a line of 84 behind him, 1/15 of a line of 60
behind him and 1/16 of a line of 48 behind him. When x = 13/15, he could
have 1/8 of a line of 120 behind him, 1/9 of a line of 45 behind him and 1/10
of a line of 30 behind him.
Solution. The strategy in this solution is to try to narrow down the search
by considering a special case. Suppose that x = (u 1)/u for some positive
integer exceeding 1. Let 1/(u + p) be the fraction of the line behind Angus.
Then Angus himself represents this fraction of the line:
 
u1 1 p
1 + = ,
u u+p u(u + p)
so that there would be u(u + p)/p people in line. To make this an integer, we
can arrange that u is a multiple of p. To get an integer for p = 1, 2, 3, take
u to be any multiple of 6. Thus, we can arrange that x is any of 5/6, 11/12,
17/18, 23/24, and so on.
Comment 1. The solution indicates how we can select x for which the
amount of the line behind Angus is represented by any number of consecutive
integer reciprocals. For example, in the case of x = 11/12, he could also have
1/13 of a line of 156 behind him. Another strategy might be to look at
x = (u 2)/u, i.e. successively at x = 3/5, 5/7, 7/9, . In this case, we
assume that 1/(u p) of the line is behind him, and need to ensure that
u 2p is a positive divisor of u(u p) for three consecutive values of u p.
If u is odd, we can achieve this with u any odd multiple of 15 and with
p = 12 (u 1), 12 (u 3), 12 (u 5).
 2. ALGEBRA 29

Comment 2. With the same fraction in front on 2 days, suppose that

1/n of a line of u people is behind Angus on the rst day, and 1/(n + 1) of a
line of v people is behind him on the second day. Then
1 1 1 1
+ = +
u n v n+1
so that uv = n(n + 1)(u v). This yields both (n2 + n v)u = (n2 + n)v
and (n2 + n + u)v = (n2 + n)u, leading to
u2 v2
uv = = 2 .
n2 +n+u n +nv
Two immediate possibilities are (n, u, v) = (n, n + 1, n) and (n, u, v) =
(n, n(n + 1), 12 n(n + 1)). To get some more, taking u v = k, we get the
quadratic equation
u2 ku k(n2 + n) = 0
with discriminant
= k 2 + 4(n2 + n)k = [k + 2(n2 + n)]2 4(n2 + n)2 ,
a pythagorean relationship when is square and the equation has integer
solutions. Select , , so that = n2 +n and let k = (2 + 2 2) =
( )2 ; this will make the discriminant equal to a square.
Taking n = 3, for example, yields the possibilities (u, v) = (132, 11),
(60, 10), (36, 9), (24, 8), (12, 6), (6, 4),(4, 3). In general, we nd that
(n, u, v) = (n, ( ), ( )) when n2 + n = with > . It turns
out that k = u v = ( )2 .
2003:7. Suppose that the polynomial f (x) of degree n 1 has all real
roots and that > 0. Prove that the set {x R : |f (x)| |f  (x)|} is a
nite union of closed intervals whose total length is equal to 2n.
The solution to this problem appears in Chap. 6.
2004:1. Prove that, for any nonzero complex numbers z and w,
 
z w 

(|z| + |w|) + 2|z + w| .
|z| |w| 

Solution 1.
 
z w 

(|z| + |w|)  +
|z| |w| 
 
 |z|w |w|z 

= z + w + +
|w| |z| 
1
|z + w| + |
z zw + wzw|

|z||w|
|zw|
= |z + w| + |
z + w|
= 2|z + w|.
|z||w|
30 2. ALGEBRA

Solution 2. Let z = aei and w = bei , with a and b real and positive.
Then the left side is equal to
|(a + b)(ei + ei )| = |aei + aei + bei + bei |
|aei + bei | + |aei + bei |.

Observe that
|z + w|2 = |(aei + bei )(aei + bei )|
= a2 + b2 + ab[ei() + ei() ]
= |(aei + bei )(aei + bei )|
from which we nd that the left side does not exceed
|aei + bei | + |aei + bei | = 2|aei + bei | = 2|z + w|.

Solution 3. Let z = aei and w = bei , where a and b are positive reals.
Then the inequality is equivalent to
 
 1 i 
 (e + ei ) |ei + (1 )ei |
2 

where = a/(a + b). But this simply says that the midpoint of the segment
joining ei and ei on the unit circle in the Argand diagram representing
complex numbers as points in the plane is at least as close to the origin as
another point on the segment.
Solution 4. [G. Goldstein] The inequality is equivalent to
 
t 
(|t| + 1)  + 1 2|t + 1|

|t|
where t = z/w.
Let t = r(cos + i sin ). Then the inequality becomes
 
(r + 1) (cos + 1) + sin 2 (r cos + 1)2 + r2 sin2
2 2

= 2 r2 + 2r cos + 1.
Now,
4(r2 + 2r cos + 1) (r + 1)2 (2 + 2 cos )
= 2r2 (1 cos ) + 4r(cos 1) + 2(1 cos )
= 2(r 1)2 (1 cos ) 0,
from which the inequality follows.
Solution 5. [R. Mong] Consider complex numbers as vectors in the plane.
q = (|z|/|w|)w is a vector of magnitude z in the direction w and p = (|w|/|z|)z
is a vector of magnitude w in the direction z. A reection about the angle
 2. ALGEBRA 31

bisector of vectors z and w interchanges p and w, q and z. Hence |p + q| =

|w + z|. Therefore
 
z w 

(|z| + |w|)  +
|z| |w| 
= |z + q + p + w| |z + w| + |p + q|
= 2|z + w|.
2004:7. Let a be a parameter. Dene the sequence {fn (x) : n = 0, 1, 2, }
of polynomials by
f0 (x) 1
and
fn+1 (x) = xfn (x) + fn (ax)
for n 0.
(a) Prove that, for all n, x,
fn (x) = xn fn (1/x).
(b) Determine a formula for the coecient of xk (0 k n) in fn (x).

Solution. The polynomial fn (x) has degree n for each n, and we will
write
n
fn (x) = b(n, k)xk .
k=0
Then
n
 n

xn fn (1/x) = b(n, k)xnk = b(n, n k)xk .
k=0 k=0
Thus, (a) is equivalent to b(n, k) = b(n, n k) for 0 k n.

n When
n a = 1, it can be established by induction that fn (x) = (x + 1)n =
n n n1
k=0 k x . Also, when a = 0, fn (x) = x + x + + x + 1 = (xn+1
1

1)(x  1) . Thus, (a) holds in these cases and b(n, k) is respectively equal
to nk and 1.
Suppose, henceforth, that a = 1. For n 0,
n
 n

fn+1 (x) = b(n, k)xk+1 + ak b(n, k)xk
k=0 k=0
n n

= b(n, k 1)xk + b(n, n)xn+1 + b(n, 0) + ak b(n, k)xk
k=1 k=1
n

= b(n, 0) + [b(n, k 1) + ak b(n, k)]xk + b(n, n)xn+1 ,
k=1
32 2. ALGEBRA

whence b(n + 1, 0) = b(n, 0) = b(1, 0) and b(n + 1, n + 1) = b(n, n) = b(1, 1)

for all n 1. Since f1 (x) = x + 1, b(n, 0) = b(n, n) = 1 for each n. Also
(2.7) b(n + 1, k) = b(n, k 1) + ak b(n, k)
for 1 k n.
We conjecture what the coecients b(n, k) are from an examination of
the rst few terms of the sequence:
f0 (x) = 1; f1 (x) = 1 + x; f2 (x) = 1 + (a + 1)x + x2 ;
f3 (x) = 1 + (a2 + a + 1)x + (a2 + a + 1)x2 + x3 ;
f4 (x) = 1+(a3 +a2 +a+1)x+(a4 +a3 +2a2 +a+1)x2 +(a3 +a2 +a+1)x3 +x4 ;
f5 (x) = (1+x5 )+(a4 +a3 +a2 +a+1)(x+x4 )+(a6 +a5 +2a4 +2a3 +2a2 +a+1)
(x2 + x3 ).
We make the empirical observation that
(2.8) b(n + 1, k) = an+1k b(n, k 1) + b(n, k)
which, with (2.7), yields
(an+1k 1)b(n, k 1) = (ak 1)b(n, k)
so that
   n+1 
ak 1 k a 1
b(n + 1, k) = + a b(n, k) = b(n, k)
an+1k 1 an+1k 1
for n k. This leads to the conjecture that, when n > k,
 n 
(a 1)(an1 1) (ak+1 1)
(2.9) b(n, k) = b(k, k)
(ank 1)(ank1 1) (a 1)
where b(k, k) = 1.
We establish this conjecture. Let c(n, k) be the right side of (2.9) for
1 k n 1 and c(n, n) = 1. Then c(n, 0) = b(n, 0) = c(n, n) = b(n, n) = 1
for each n. In particular, c(n, k) = b(n, k) when n = 1.
We show that
c(n + 1, k) = c(n, k 1) + ak c(n, k)
for 1 k n, which will, through an induction argument, imply that
b(n, k) = c(n, k) for 0 k n. The right side is equal to
 n   k+1  
a 1 a 1 ak 1 k
+ a
ank 1 a1 ank+1 1
(an+1 1)(an 1) (ak+1 1)
= = c(n + 1, k)
(an+1k 1)(ank 1) (a 1)
as desired. Thus, we now have a formula for b(n, k) as required in (b).
Finally, (a) can be established in a straightforward way, either from the
formula (2.9) or using the pair of recursions (2.7) and (2.8).
 2. ALGEBRA 33

2005:5. Let f (x) be a polynomial with real coecients, evenly many

of which are nonzero, which is palindromic. This means that
the coecients
n
read the same in either direction, i.e. ak = ank if f (x) = k=0 ak xk , or,
n
alternatively, f (x) = x f (1/x), where n is the degree of the polynomial.
Prove that f (x) has at least one root of absolute value 1.
The solution of this problem appears in Chap. 6.
2005:7. Let f (x) be a nonconstant polynomial that takes only integer
values when x is an integer, and let P be the set of all primes that divide
f (m) for at least one integer m. Prove that P is an innite set.
The solution to this problem appears in Chap. 11.
2006:3. Let p(x) be a polynomial of positive degree n with n distinct
real roots a1 < a2 < < an . Let b be a real number for which 2b < a1 + a2 .
Prove that
2n1 |p(b)| |p (a1 )(b a1 )|.

Solution. Wolog, let p(x) have leading coecient 1. Observe that, for
i > 1,
ai a1 = ai b + b a1 < ai b + a2 b < 2(ai b)
 
and that p(x) = (x ai ), from which p (a1 ) = i2 (a1 ai ). Then

|p(b)| = |b a1 | |b ai |
i2
1
|b a1 | (ai a1 ) = |b a1 ||p (ai )|2(n1) .
2
i2

2006:9. A high school student asked to solve the surd equation

3x 2 2x 3 = 1
gave the following answer: Squaring both sides leads to
3x 2 2x 3 = 1
so x = 6. The answer is, in fact, correct.
Show that there are innitely many real quadruples (a, b, c, d) for which
this method leads to a correct solution of the surd equation

ax b cx d = 1.

Solution 1. Solving the general equation properly leads to

ax b cx d = 1 = ax b = 1 + cx d + 2 cx d

= (a c)x = (b + 1 d) + 2 cx d.
To make the manipulation simpler, specialize to a = c + 1 and d = b + 1.
Then the equation becomes
x2 = 4(cx d) = 0 = x2 4cx + 4d.
34 2. ALGEBRA

Using the students method to solve the same equation gives ax b

cx d = 1 which yields x = (1 + b + d)/(a c) = 2d. So, for the method
to work, we need
0 = 4d2 8cd + 4d = 4d(d 2c + 1)
which can be achieved by making 2c = d + 1. So we can take
(a, b, c, d) = (t + 1, 2(t 1), t, 2t 1)
for some real t. The original problem corresponds to t = 2.
The equation
 
(t + 1)x 2(t 1) tx (2t 1) = 1
is satised by x = 2 and x = 4t 2. The rst solution works for all values
of t, while the second is valid if and only if t 12 . The equation (t + 1)x
2(t 1) tx (2t 1) = 1 is equivalent to x = 4t 2.
Solution 2. [G. Goldstein] Analysis. We want to solve simultaneously
the equations

(2.10) ax b cx d = 1
and
(2.11) ax b cx d = 1.
From (2.10), we nd that

(2.12) ax b = 1 + (cx d) + 2 cx d.

From (2.11) and (2.12), we obtain that d = cx d, so that x = (d2 + d)/c.
From (2.11), we have that x = (1 + b + d)/(a c).
Select a, c, d so that d > 0 and ac(a c) = 0, and choose b to satisfy
d2 + d 1+b+d
= .
c ac
Let
d2 + d 1+b+d (d2 + d) + (1 + b + d) (d + 1)2 + b
x= = = = .
c ac c + (a c) a
Then

ax b cx d = (d + 1)2 d2 = (d + 1) d = 1
and
ax b cx d = (d + 1)2 + b b d2 d d = 1.
Comments. In Solution 2, if we take c = d = 1, we get the family of
parameters (a, b, c, d) = (a, 2a 4, 1, 1). R. Barrington Leigh found the set of
parameters given in Solution 1. A. Feizmohammadi provided the parameters
(a, b, c, d) = (2c, 0, c, 1).
 2. ALGEBRA 35

2008:4. Suppose that u, v, w, z are complex numbers for which u + v +

w + z = u2 + v 2 + w2 + z 2 = 0. Prove that
(u4 + v 4 + w4 + z 4 )2 = 4(u8 + v 8 + w8 + z 8 ).
Solution. Let u, v, w, z be the roots of a monic quartic polynomial f (t).
Since u + v + w + z = 0 and
1
uv + uw + uz + vw + vz + wz = [(u + v + w + z)2 (u2 + v 2 + w2 + z 2 )] = 0,
2
we must have that f (t) = t4 at b for some complex numbers a and b.
Suppose that sn = un +v n +wn +z n for positive integers n. Then s4 = 4b
and sk+4 = ask+1 +bsk for k 1. Therefore, s5 = 0 and s8 = as5 +bs4 = 4b2 ,
so that 4s8 = (4b)2 = s24 , as desired.
2008:9. For each positive integer n, let
n
2k
Sn = .
k2
k=1
Prove that Sn+1 /Sn is not a rational function of n.
Solution 1. Assume that Sn+1 /Sn is a rational function of n. Then
Sn Sn 1
= =
n
2 n 2 Sn Sn1 1 Sn1 Sn

is a rational function. From this, it follows that Sn = r(n)2n n2 for some

rational function r. Since
2n+1 2n+1 2n
2
= Sn+1 Sn = r(n + 1) 2
r(n) 2 ,
(n + 1) (n + 1) n
we have that
2n2 r(n + 1) (n + 1)2 r(n) = 2n2 .
Since this equation holds for innitely many values of n, we have the corre-
sponding rational function identity. Letting r(x) = f (x)/g(x), where f and
g are coprime polynomials, we obtain the equation
2x2 f (x + 1)g(x) (x + 1)2 f (x)g(x + 1) = 2x2 g(x)g(x + 1).
The polynomial g(x) cannot be a constant (look at the leading coe-
cients).
There exists an integer n, namely 0, for which the polynomials g(x) and
g(x + n) have a common divisor of positive degree. However, if n exceeds the
maximum absolute value of a root of g(x), then g(x) and g(x+n) are coprime.
Let N be the largest integer for which g(x) and g(x + N ) have a common
irreducible divisor u(x) of positive degree. Then u(x N ) divides g(x) and
so divides (x + 1)2 f (x)g(x + 1). Since f and g are coprime, u(x N ) divides
(x + 1)2 g(x + 1). Suppose that u(x N ) divides g(x + 1); then u(x) must
divide g(x + N + 1), which contradicts the determination of N . Therefore,
u(x N ) divides (x + 1)2 , and so u(x + 1) divides (x + N + 2)2 .
36 2. ALGEBRA

Since u(x + 1) divides g(x + 1), u(x + 1) must divide 2x2 f (x + 1)g(x).
Since u(x + 1) does not divide f (x + 1), u(x + 1) divides 2x2 g(x). Since
u(x + 1) divides g(x + N + 1), u(x + 1) cannot divide g(x). Therefore,
u(x + 1) must divide x2 , as well as (x + N + 2)2 . But this is an impossibility.
This contradiction yields the result of the problem.
Solution 2. [C. Ochanine] Since
Sn+1 Sn+1 Sn 2n+1
1= = ,
Sn Sn (n + 1)2 Sn
it is enough to show that Sn /2n is not a rational function. Let f (n) = Sn /2n .
Then
1 1
f (n + 1) = f (n) +
2 (n + 1)2
for every positive integer n.
If f were rational, then we would have
1 1
f (x + 1) = f (x) +
2 (x + 1)2
for all real x. Since, by the denition, f (1) is nite, so also is f (0). Substi-
tuting x = 1 in the foregoing equation, we see that 1 is a pole of f (x).
Since
1 1 1 1 1
f (x + 2) = f (x + 1) + = f (x) + + ,
2 (x + 2)2 4 2(x + 1)2 (x + 2)2
we see that 2 is also a pole of f (x). Continuing on, we nd that every
negative integer is a pole of f (x), contradicting its rationality.
The desired result follows.
[Problem 2008:9 was contributed by Franklin Vera Pacheco.]
2009:2. Let n and k be integers with n 0 and k 1. Let x0 , x1 , . . . ,
xn be n + 1 distinct points in Rk and let y0 , y1 , . . . , yn be n + 1 real numbers
(not necessarily distinct). Prove that there exists a polynomial p of degree
at most n in the coordinates of x for which p(xi ) = yi for 0 i n.
The solution to this problem appears in Chap. 7.
2010:5. Let m be a natural number, and let c, a1 , a2 , , am be complex
numbers for which |ai | = 1 for i = 1, 2, , m. Suppose also that
m

lim ani = c.
n
i=1

Prove that c = m and that ai = 1 for i = 1, 2, , m.

Solution. If ai = eii , then either the sequence {ani } is periodic and

assumes the value 1 innitely often (when i is a rational multiple of ) or
has a subsequence whose limit is 1 (when i is not a rational multiple of ).
 2. ALGEBRA 37

[In the latter case, we can nd an increasing subsequence {nk } of natural

n n
numbers for which ani k converges, so that ai k+1 k converges to 1.]
We prove that there is a subsequence {nk } of natural numbers for which
limk ani k = 1 for each 1 i m. Proceed by induction on m. When
m = 1, the limit of any subsequence of {an1 } is equal to the limit of the
whole sequence, so that c = 1 in this case. In fact, we can go further:
a1 = lim an+1
1 = lim an1 = 1.
Suppose the induction hypothesis holds for m 1. Then there is a sub-
sequence S1 of natural numbers such that {ani } has limit 1 along this subse-
quence for 1 i m 1. We can nd a subsequence S2 along which the
sequence {anm } converges to some limit b on the unit circle. Let S3 be the
sequence of dierences of consecutive terms in the sequence S2 (so that the
sequence along S3 consists of quotients of consecutive terms of the sequences
along S2 ). Then, for 1 i m, the sequence {ani } converges to 1 along S3 .
It follows from this that c = m. Also, along S3 , we have that
m m

ai = lim an+1
i = m.
i=1 i=1
m
Therefore i=1 Re ai = m. But the real part of each ai does not exceed 1,
with equality if and only if ai = 1, it follows that ai = 1 for each i.
[Problem 2010:5 was contributed by Bamdad R. Yahaghi.]
2010:6. Let f (x) be a quadratic polynomial. Prove that there exist
quadratic polynomials g(x) and h(x) for which
f (x)f (x + 1) = g(h(x)).

Solution 1. [A. Remorov] Let f (x) = a(x r)(x s). Then

f (x)f (x + 1) = a2 (x r)(x s + 1)(x r + 1)(x s)
= a2 (x2 + x rx sx + rs r)(x2 + x rx sx + rs s)
= a2 [(x2 (r + s 1)x + rs) r][(x2 (r + s 1)x + rs) s]
= g(h(x)),
where g(x) = a2 (x r)(x s) = af (x) and h(x) = x2 (r + s 1)x + rs.
Solution 2. Let f (x) = ax2 + bx + c, g(x) = px2 + qx + r and h(x) =
2
ux + vx + w. Then
f (x)f (x + 1) = a2 x4 + 2a(a + b)x3 + (a2 + b2 + 3ab + 2ac)x2 + (b + 2c)(a + b)x
+c(a + b + c)
g(h(x)) = p(ux2 + vx + w)2 + q(ux2 + vx + w) + r
= pu2 x4 + 2puvx3 + (2puw + pv 2 + qu)x2
+ (2pvw + qv)x + (pw2 + qw + r).
38 2. ALGEBRA

Equating coecients, we nd that pu2 = a2 , puv = a(a+b), 2puw+pv 2 +qu =

a2 +b2 +3ab+2ac, (b+2c)(a+b) = (2pw +q)v and c(a+b+c) = pw2 +qw +r.
We need to nd just one solution of this system. Let p = 1 and u = a. Then
v = a + b and b + 2c = 2pw + q from the second and fourth equations. This
yields the third equation automatically. Let q = b and w = c. Then from the
fth equation, we nd that r = ac.
Thus, when f (x) = ax2 + bx + c, we can take g(x) = x2 + bx + ac and
h(x) = ax2 + (a + b)x + c.
Solution 3. [S. Wang] Suppose that
f (x) = a(x + h)2 + k = a(t (1/2))2 + k,
where t = x + h + 12 . Then f (x + 1) = a(x + 1 + h)2 + k = a(t + (1/2))2 + k,
so that
f (x)f (x + 1) = a2 (t2 (1/4))2 + 2ak(t2 + (1/4)) + k 2
 2   2 
a a ak
= a t + + 2ak t +
2 4 2
+ 2
+k .
2 16 2
Thus, we can achieve the desired representation with h(x)=t2 = x2 +(2h +
2
a2
1)x + 14 and g(x) = a2 x2 + ( a ak 2
2 + 2ak)x + ( 16 + 2 + k ).
Solution 4. [V. Krakovna] Let f (x) = ax2 + bx + c = au(x) where
u(x) = x2 + dx + e, where b = ad and c = ae. If we can nd functions v(x)
and w(x) for which u(x)u(x + 1) = v(w(x)), then f (x)f (x + 1) = a2 v(w(x)),
and we can take h(x) = w(x) and g(x) = a2 v(x).
Dene p(t) = u(x + t), so that p(t) is a monic quadratic in t. Then,
noting that p (t) = u (x + t) = 2, we have that
u (x) 2
p(t) = u(x + t) = u(x) + u (x)t + t = t2 + u (x)t + u(x),
2
from which we nd that
u(x)u(x + 1) = p(0)p(1) = u(x)[u(x) + u (x) + 1]
= u(x)2 + u (x)u(x) + u(x) = p(u(x)) = u(x + u(x)).
Thus, u(x)u(x + 1) = v(w(x)) where w(x) = x + u(x) and v(x) = u(x).
Therefore, we get the desired representation with
 
2 b c
h(x) = x + u(x) = x + 1 + x+
a a
and
g(x) = a2 v(x) = a2 u(x) = af (x) = a2 x2 + abx + ac.
Comments. The second solution can also be obtained by looking at
special cases, such as when a = 1 or b = 0, getting the answer and then
making a conjecture.
There is an interesting story behind this problem. Originally it was noted
that the product of two consecutive oblong numbers (products of consecutive
 2. ALGEBRA 39

integers) was also oblong, since for each integer n, (n 1)n n(n + 1) =
(n2 1)n2 . The same is true for integers one more than a perfect square,
since [(n 1)2 + 1][n2 + 1] = (n2 n + 1)2 + 1, and numbers of the form
n2 + n + 1 since (n2 n + 1)(n2 + n + 1) = n4 + n2 + 1. It was not dicult
to conjecture and prove the result that f (x)f (x + 1) = f (x + f (x)) for monic
quadratics f (x).
This problem was also given to high school students participating in
a problems correspondence program. One solver, James Rickards, then in
Grade 9 at an Ottawa, Canada high school, noted that, if the roots of f (x)
were r and s, then those of f (x + 1) were r 1 and s 1. Thus, f (x)f (x + 1)
was a quartic polynomial two of whose roots had the same sum as the other
two. He proved that this root property characterized quartic polynomials
representable as the composite of two quadratics.
In fact, he generalized to polynomials of higher degree and showed that
a polynomial of degree mn could be written as the composite g(h(x)) of
polynomials g of degree m and h of degree n if and only if its set of roots could
be partitioned into m sets of n elements (not necessarily distinct) for which
all the fundamental symmetric functions of degrees up to m 1 had the same
values. I was unable to nd any indication of this result in the literature,
and so felt that it should be published in a journal with wide circulation.
It was accepted and published under the title: James Rickards, When is
a polynomial a composition of other polynomials? American Mathematical
Monthly 118:4 (April, 2011), 358363.
2011:5. Solve the system

x + xy + xyz = 12
y + yz + yzx = 21
z + zx + zxy = 30.

Solution 1. Let u = xyz. Then x + xy = 12 u so that z + z(12 u) = 30

and z = 30/(13u). Similarly, y = 21/(31u) and x = 12/(22u). Plugging
these expressions (with xyz = u) into any one of the three equations yields
that

0 = u3 65u2 + 1306u 7560 = (u 10)(u 27)(u 28) .

We get the three solutions

 
12 21 15
(x, y, z) = (1, 1, 10), , , , (2, 7, 2),
5 4 7

all of which satisfy the system.

40 2. ALGEBRA

Solution 2. Dene u and obtain the expressions for x, y, z as in Solution 1.

Substitute these into the equation xyz = u. This leads to the quartic equation
0 = u(u 13)(u 22)(u 31) + (12)(21)(30)
= u4 66u3 + 1371u2 8866u + 7560
= (u 1)(u3 65u2 + 1306u 7560).
Apart from the three values of u already identied, we have u = 1. This
leads to  
4 7 5
(x, y, z) = , , .
7 10 2
While, indeed, xyz = 1, we nd that x + xy + xyz = 69/35, y + yz + xyz =
69/20, z + zx + xyz = 69/14, so the solution u = 1 is extraneous.
[Problem 2011:5 was posed by Stanley Rabinowitz in the Spring, 1982
issue of AMATYC Review.]
2011:6. [The problem posed on the competition is not available for
publication, as it appeared as Enigma problem 1610 in the August 25, 2010
issue of the New Scientist. It concerned the determination of the nal score
of a three game badminton match where the score of each competitor was in
arithmetic progression. The full statement of the problem is available on the
internet.]
Solution 1. [K. Ng] Recall the winner of a badminton game must obtain
at least 21 points and lead the opponent by at least 2 points. Let the scores
for the three games be a, a+u, a+2u for player A and b, b+v, b+2v for player
B. Suppose, wolog, that A wins the rst game and that B wins the second.
Then a b 2 and (b + v) (a + u) 2, whence v u 2 + (a b) 4.
Hence
(b + 2v) (a + 2u) = [(b + v) (a + u)] + (v u) 6 .
Thus the third game is won by B by at least 6 points. Therefore, B must
have exactly 21 points in the third game. It follows that B has more than 21
points in each of the previous games, as A does in the rst game.
Hence a b = 2 and
2 = (b + v) (a + u) = (v u) (a b) = (v u) 2,
so that v u = 4. Therefore, the dierence in the scores of the third game is
exactly 2(v u) (a b) = 6 and the nal scores are (15, 21).
Solution 2. There are two dierent ways the score can present at the
conclusion of a game: (a) The winner can have 21 points and the other no
more than 19 points; (b) The winner can have at least 22 points and the loser
two points fewer than the winner.
It is not possible for the scores in the rst and third games to be of type
(b). For the signed dierences of the scores is also an arithmetic progression
of three numbers, the rst and third of which are both 2, both 2, or 2 and
 2. ALGEBRA 41

2. In the rst two cases, this would mean that the same person won all
three games, and in the third case that the scores in the second game were
the same and there would be no winner of the second game.
Therefore either the opening or closing game has one person with a score
of 21 and the other with a score s less than 20. Let the scores of the rst
player be (21, 21+x, 21+2x) and of the second be (s, s+y, s+2y), possibly in
reverse order. Note that x cannot be negative, since otherwise s s + 2y =
21 s + y. Since not all three games are won by the same person, we
must have y > x, s + y + 2 = 21 + x and 21 + 2x + 2 = s + 2y. Therefore
s = 19 (y x) = 23 2(y x), whence s = 15. Since the winner of the game
with scores (21, 15) is the same as that of the second game and dierent from
that of the third, this game must be the nal game of the match. Therefore,
the scores of the nal game are (21, 15).
Note that y x = 4, and that, whenever x 1, sets of scores (21 +
2x, 23 + 2x), (21 + x, 19 + x), (21, 15) all satisfy the conditions.
2011:10. Suppose that p is an odd prime. Determine the number of
subsets S contained in {1, 2, . . . , 2p 1, 2p} for which (a) S has exactly p
elements, and (b) the sum of the elements of S is a multiple of p.
Solution. Let be a primitive pth root of unity, so that is a power of
cos(2/p) + i sin(2/p) with the exponent not divisible by p. Let

2p
f (x) = (x i ) = (xp 1)2 = x2p 2xp + 1.
i=1

We can also write


f (x) = x2p + {(1)p i1 ++ip : {i1 , i2 , , ip }

{1, 2, . . . , 2p} xp + + 1
= x2p + (a0 + a1 + a2 2 + + ap1 p1 )xp + + 1,
for some coecients ai . We are required to determine the value of a0 .
Let g(x) = (a0 2) + a1 x + a2 x2 + + ap1 xp1 . Observe that g(x)
vanishes when x is any primitive
  pth root of unity and that the sum 2pdening
the coecient of xp has 2pp terms, so that a0 + a1 + + ap1 = p . Since
g(x) is a polynomial of degree p 1 whose zeros are the primitive pth roots
of unity, g(x) must be a constant multiple of 1 + x + + xp1 , so that
a0 2 = a1 = a2 = = ap1 .
 
Thus a0 + (p 1)(a0 2) = 2pp , so that
  
1 2p
a0 = 2 + 2.
p p

[Problem 2011:10 was contributed by Ali Feiz Mohammadi.]

42 2. ALGEBRA

2012:2. Suppose that f is a function dened on the set Z of integers that

takes integer values and satises the condition that f (b) f (a) is a multiple
of b a for every pair a, b, of integers. Suppose also that p is a polynomial
with integer coecients such that p(n) = f (n) for innitely many integers n.
Prove that p(x) = f (x) for every positive integer x.
The solution to this problem appears in Chap. 11.
2012:3. Given the real numbers a, b, c not all zero, determine the real
solutions x, y, z, u, v, w for the system of equations:
x 2 + v 2 + w 2 = a2
u 2 + y 2 + w 2 = b2
u2 + v 2 + z 2 = c2
u(y + z) + vw = bc
v(x + z) + wu = ca
w(x + y) + uv = ab.

Solution 1. Evaluating b2 c2 in two ways, we nd that

0 = (u2 + y 2 + w2 )(u2 + v 2 + z 2 ) (uy + uz + vw)2
= (u2 yz)2 + (uv wz)2 + (wu vy)2 .

Hence u2 = yz. Similarly, v 2 = zx and w2 = xy. Inserting these three values

into the rst three equations yields that
x(x + y + z) = a2
y(x + y + z) = b2
z(x + y + z) = c2 .
It follows that not all of x, y, z can vanish and that (x+y +z)2 = a2 +b2 +c2 .
Also
x : a2 = y : b2 = z : c2 = 1 : (x + y + z) = (x + y + z) : (a2 + b2 + c2 ).
Therefore
(x, y, z, u, v, w) = (a2 d, b2 d, c2 d, bcd, cad, abd)
where d2 (a2 + b2 + c2 ) = 1.
Solution 2. [Y. Wu; P.J. Zhao] From the given equations, we see that
 
(x, v, w) (w, u, y) = w(x + y) + uv = ab = x2 + v 2 + w2 w2 + u2 + y 2 ;
 
(w, u, y) (v, z, u) = u(y + z) + vw = bc = w2 + u2 + y 2 v 2 + z 2 + u2 ;
 
(v, z, u) (x, v, w) = v(x + z) + uw = ca = v 2 + z 2 + u2 x2 + v 2 + w2 .
 2. ALGEBRA 43

It follows from the conditions for equality in the Cauchy-Schwarz Inequality

that (x, v, w), (w, u, y) and (v, z, u) are all constant multiples of the same
unit vector (p, q, r). Indeed
(x, v, w) = a(p, q, r); (w, u, y) = b(p, q, r); (v, z, u) = c(p, q, r).
Thus, x = ap, y = br, z = cq, u = bq = cr, v = aq = cp and w = ar = bp,
whence
u2 = bcqr = yz; v 2 = acpq = xz; w2 = abpr = xy.
The solution now can be completed as in Solution 1.
2013:6. Let p(x) = x4 + ax3 + bx2 + cx + d be a polynomial with rational
coecients. Suppose that p(x) has exactly one real root r. Prove that r is
rational.
Solution. Since nonreal roots occur in pairs, p(x) must have an even
number of real roots counting multiplicity. Therefore r must be a double
or quadruple root. If r is a quadruple root, then p(x) = (x r)4 and r =
a/4 is rational. Suppose, otherwise, that p(x) = (x r)2 q(x) where q(x)
is an irreducible quadratic. The derivative p (x) is equal to (x r)f (x)
where f (x) = 2q(x) + (x r)q  (x). Since q(r) does not vanish, f (r) = q(r)
so that f (x) and q(x) must be distinct and coprime. The monic greatest
common divisor of p(x) and p (x) must therefore be x r. Since (by the
Euclidean algorithm) this is a polynomial with rational coecients, therefore
r is rational.
2014:6. Let f (x) = x6 x4 + 2x3 x2 + 1.
(a) Prove that f (x) has no positive real roots.
(b) Determine a nonzero polynomial g(x) of minimum degree for which
all the coecients of f (x)g(x) are nonnegative rational numbers.
(c) Determine a polynomial h(x) of minimum degree for which all the
coecients of f (x)h(x) are positive rational numbers.

Solution. (a) Note that

f (x) = (x2 1)(x4 1) + 2x3 = (x2 1)2 (x2 + 1) + 2x3
from which we see that f (x) > 0 for all x > 0. Alternatively,
f (x) = x6 + x3 + (x3 x4 ) + (1 x2 ) = x4 (x2 1) + x2 (x 1) + x3 + 1
from which we see that there are no roots in either of the intervals [0, 1] and
[1, ).
(b) It is straightforward to see that multiplying f (x) by a linear polyno-
mial will not achieve the goal. We have that
f (x)(x2 + bx + c) = x8 + bx7 + (c 1)x6 + (2 b)x5 + (2b c 1)x4
+ (2c b)x3 + (1 c)x2 + bx + c
from which we see that taking c = 1 and 1 b 2 will yield the desired
polynomial g(x). Examples of suitable g(x) are x2 + x + 1 and (x + 1)2 .
44 2. ALGEBRA

(c) From (b), we note that no quadratic polynomial will serve. However,
taking h(x) = x3 + 2x2 + 2x + 1, we nd that
f (x)h(x) = x9 + 2x8 + x7 + x6 + x5 + x4 + x3 + x2 + 2x + 1.

Comment. It is known that for a given real polynomial f (x), there exists
a polynomial g(x) for which all the coecients of f (x)g(x) are nonnegative
(resp. positive) if and only if none of the roots of f (x) are positive (resp.
nonnegative).

Another example is f (x) = x3 x + 1. Since f (x) = x3 + (1 x) =

x(x 1) + 1, we see that there are no roots in [0, 1] and [1, ). No linear
2

polynomial x + c will do for g(x). Since

f (x)(x2 + bx + c) = x5 + bx4 + (c 1)x3 + (1 b)x2 + (b c)x + c,
we require that 1 c b and 0 b 1. The only possibility is that
g(x) = x2 + x + 1.
For strictly positive coecients in the product, set g(x) = ax3 + bx2 +
cx + d. Then
f (x)g(x) = ax6 + bx5 + (c a)x4 + (a + d b)x3 + (b c)x2 + (c d)x + d.
For g to be suitable, we require that c > a > 0, a + d > b and b > c > d > 0.
We can take g(x) = 3x3 + 5x2 + 4x + 3.
[The examples in Problem 2014:6 are due to Horst Brunotte in D usseldorf,
Germany.]
2015:10. (a) Let
g(x, y) = x2 y + xy 2 + xy + x + y + 1.
We form a sequence {x0 } as follows: x0 = 0. The next term x1 is the unique
solution 1 of the linear equation g(t, 0) = 0. For each n 2, xn is the
solution other than xn2 of the equation g(t, xn1 ) = 0.
Let {fn } be the Fibonacci sequence determined by f0 = 0, f1 = 1 and
fn = fn1 + fn2 for n 2. Prove that, for any nonnegative integer k,
fk fk+2
x2k = and x2k+1 = .
fk+1 fk+1
(b) Let
h(x, y) = x2 y + xy 2 + xy + (x + y) +
be a polynomial with real coecients , , . We form a bilateral sequence
{xn : n Z} as follows. Let x0 = 0 be given arbitrarily. We select x1 and
x1 to be the two solutions of the quadratic equation h(t, x0 ) = 0 in either
order. From here, we can dene inductively the terms of the sequence for
positive and negative values of the index so that xn1 and xn+1 are the two
 2. ALGEBRA 45

solutions of the equation h(t, xn ) = 0. We suppose that at each stage, neither

of these solutions is zero.
Prove that the sequence {xn } has period 5 (i.e. xn+5 = xn for each
index n) if and only if 3 + 2 = 0.
(a) Solution. Observe that
g(x, y) = (xy + 1)(x + y + 1).
For each value of y = 0, the equation g(x, y) = 0 has two solutions: y = 1/x
and y = (x + 1). Observe that g(x, y) is symmetrical in x and y, so that,
for each consecutive pair xn , xn+1 of terms in the sequence g(xn , xn+1 ) =
g(xn+1 , xn ) = 0. Consider the equation 0 = g(t, x1 ) = g(t, 1) = (t + 1)t.
One of its solutions is x0 = 0 and the other is x2 = 1.
For the equation 0 = g(t, x2 ), we have that g(x1 , x2 ) = g(x2 , x1 ) = 0 and
x1 x2 + 1 = 0. Therefore x2 + x3 + 1 = 0, so that x3 = (x2 + 1). Continuing
on in this way, we nd that, for each positive integer k, x2k1 x2k = 1 and
x2k + x2k+1 = 1, whereupon
1 1 x2k1
x2k+1 = 1 + = .
x2k1 x2k1
When k = 1, we nd that x2k1 = x1 = 1 = f2 /f1 . Suppose, for k 1,
we have that x2k1 = fk+1 /fk . Then
fk fk+1 + fk fk+2
x2k+1 = 1 = = .
fk+1 fk+1 fk+1
By induction, we obtain the desired expression for x2k+1 . Also x2k =
1/x2k1 = fk /fk+1 .
(b) Solution 1. Observe that from the sum of the roots of h(t, xn ) = 0,
we have that  2 
x + xn +
xn1 + xn+1 = n ,
xn
or

xn1 + xn + xn+1 =
xn
for each n. The sequence will have period 5 if and only if the sum of any ve
consecutive terms is constant.
Since, for each integer n,

xn+2 + xn+1 + xn =
xn+1
and

xn + xn1 + xn2 = ,
xn1
we have that
 
xn+1 + xn1
xn+2 + xn+1 + xn + xn1 + xn2 = 2 xn
xn+1 xn1
46 2. ALGEBRA

 
x2n + xn +
= 2 + xn
xn +
( + )xn + ( 2 2)
=
xn +
   3 
+ 2
= + + .
2 xn +
This sum is independent of n if and only if the term involving xn vanishes
identically, i.e., if and only if the required condition holds.
Solution 2. From the formula for the product of the roots, we obtain that
xn +
xn1 xn+1 =
xn
so that
xn1 xn xn+1 = xn +
for each index n. Therefore
xn+2 xn+1 x2n xn1 xn2 = (xn+1 + )(xn1 + )
   
2 xn +
= xn + + + 2
xn xn
= ( 3 + 2 ) xn
whence  
3 + 2
xn+2 xn+1 xn xn1 xn2 = + .
xn
The result again follows.
Comments. If x0 = 0, then h(t, x0 ) is linear and there is a single root
/. We can extend the sequence in only one direction, and it begins with
the terms 0, /, ( 3 + 2 )/(), . . . .
The study of the type of sequence given in this problem began with the
recursions given by
xn + c
xn+1 =
xn1
or, equivalently, if you want to continue the sequence backwards,
xn + c
xn1 = ,
xn+1
where c is a parameter and n ranges over the integers. We suppose that
the terms of the sequence are selected to avoid division by 0. When c = 0
and c = 1, any bilateral sequence satisfying the recursion is periodic with
respective periods 6 and 5. For other values of c, certain but not all sequences
are periodic. The examination of the periodic cases lead to the identication
of an invariant for the sequence.
 2. ALGEBRA 47

When
x2 y + xy 2 + x2 + (c + 1)x + y 2 + (c + 1)y + c
fc (x, y) = ,
xy
it turns out that  
y+c
f (x, y) = f y,
x
with the result that f (xn , xn+1 ) is constant with respect to the index n.
The equation fc (x, y) = k can be rewritten as hc,k (x, y) = 0 where
hc,k (x, y) = x2 y + xy 2 + x2 + y 2 kxy + (c + 1)(x + y) + c
= xy(x + y) + (x + y)2 (k + 2)xy + (c + 1)(x + y) + c
= (y + 1)x2 + (y 2 ky + (c + 1))x + (y + 1)(y + c) .

The function hc,k (x, y) is a symmetric polynomial quadratic in each variable.

For each n, the terms xn1 and xn+1 are the roots of the quadratic equation
hc,k (x, xn ) = 0, provided xn = 1. Indeed, from the relation between the
coecients and the product of the roots of a quadratic, we corroborate the
relation
(2.13) xn1 xn+1 = xn + c.
Moreover, we have
x2n kxn + c + 1
(2.14) xn1 + xn+1 = .
xn + 1
These sequences, for various values of the parameter c, are studied in the
paper Ed Barbeau, Boaz Gelbord and Steve Tanny, Periodicities of solutions
of the generalized Lyness recursion, J. Dierence Equations and Applications
1 (1995), 291306.
We look at a slight generalization and point out how two types of recur-
sions are related to each other. Let
h(x, y) = x2 y + xy 2 + (x2 + y 2 ) + xy + (x + y) +
= (y + )x2 + (y 2 + y + )x + (y 2 + y + ).
As before, we can seed the recursion by specifying x0 and arranging that, for
each integer n, xn1 and xn+1 are the solutions of the equation h(x, xn ) = 0.
To avoid complications, we suppose that xn never assumes the value .
Sequences determined in this way satisfy both the recursions:
 2 
xn + xn +
(2.15) xn+1 + xn1 =
xn +
and
x2n + xn +
(2.16) xn+1 xn1 = .
xn +
48 2. ALGEBRA

Remarkably, if we are given any sequence that satises the rst of these
recursion relations, we can select a constant so that it satises the second.
Similarly, given any sequences given by the second recursion relationship, we
can select a value of so that it satises the rst. Thus, either recursion
individually allows us to introduce a function h(x, y) that will allow us to
generate the sequence. This follows from the following proposition.
We have
 2 
y + y + y 2 + y +
h ,y = h(x, y)
x(y + ) x2 (y + )
and   2  
y + y +
h x, y = h(x, y)
y+
with the result that, when x and y are consecutive terms, h(x,y)
xy is invariant
along any recursion satisfying (2.16) and h(x, y) is invariant along any recur-
sion satisfying (2.15). Thus, any recursion satisfying either (2.15) or (2.16)
will satisfy the other with a suitable choice of parameters.
Proof.
 2 
y + y + (y 2 + y + )2 2 y 2 + y +
h , y =(y + ) + (y + y + )
x(y + ) x2 (y + )2 x(y + )
+ (y 2 + y + )
y 2 + y +  
= 2
(y 2 + y + ) + x(y 2 + y + ) + x2 (y + )
x (y + )
y 2 + y +  2 
= 2
x y + y 2 x + (x2 + y 2 ) + xy + (x + y) +
x (y + )
y 2 + y +
= h(x, y).
x2 (y + )
Also
  2  
y + y +
h x, y
y+
 
2 2x(y 2 + y + ) (y 2 + y + )2
= (y + ) x + +
y+ (y + )2
 2 
y + y +
(y + y + )
2
+ x + (y 2 + y + )
y+
= x2 y + x2 + 2xy 2 + 2xy + 2x xy 2 xy x + y 2 + y +
= h(x, y).
 CHAPTER 3

Inequalities

2001:1. Let a, b, c > 0, a < bc and 1 + a3 = b3 + c3 . Prove that 1 + a < b + c.

Solution 1. Since (1 + a)(1 a + a2 ) = (b + c)(b2 bc + c2 ), and since
1 a + a2 and b2 bc + c2 = 12 (b c)2 + 12 (b2 + c2 ) are positive, we have that
1 + a < b + c 1 a + a2 > b2 bc + c2 .
Suppose, if possible, that 1 + a b + c. Then
b2 bc + c2 1 a + a2
(b + c)2 3bc (1 + a)2 3a > (1 + a)2 3bc
(b + c)2 > (1 + a)2 b + c > 1 + a
which is a contradiction.
Solution 2. [J. Chui] Let u = (1 + a) (b + c). Then
(1 + a)3 (b + c)3 = u[(1 + a)2 + (1 + a)(b + c) + (b + c)2 ]
= u[(1 + a)2 + (1 + a)(b + c) + b2 + 2bc + c2 ].
But also
(1 + a)3 (b + c)3 = (1 + a3 ) (b3 + c3 ) + 3a(1 + a) 3bc(b + c)
= 0 + 3[a(1 + a) bc(b + c)] < 3bcu.
It follows from these that
0 > u[(1 + a)2 + (1 + a)(b + c) + b2 bc + c2 ].
Since the quantity in square brackets is positive, we must have that u < 0,
as desired.

Springer International Publishing Switzerland 2016 49

E.J. Barbeau, University of Toronto Mathematics Competition
(20012015), Problem Books in Mathematics,
DOI 10.1007/978-3-319-28106-3 3
50 3. INEQUALITIES

Solution 3. [A. Momin; N. Martin] Suppose, if possible, that (1 + a)

(b + c). Then
0 (1 + a)2 (b + c)2 = (1 + a2 ) (b2 + c2 ) 2(bc a) < (1 + a2 ) (b2 + c2 ).
Hence 1 + a2 > b2 + c2 . It follows that
(1 a + a2 ) (b2 bc + b2 ) = (1 + a2 ) (b2 + c2 ) + (bc a) > 0
so that
(1 a + a2 ) > (b2 bc + c2 ).
However
(1 + a)(1 a + a2 ) = 1 + a3 = b3 + c3 = (b + c)(b2 bc + c2 ),
from which it follows that 1 + a < b + c, yielding a contradiction. Hence, the
desired result follows.
Solution 4. [P. Gyrya] Let p(x) = x3 3ax.Checking the rst derivative

yields that p(x)
is strictly increasing for x > a. Now 1 + a 2 a > a

and b + c 2 bc > 2 a > a, so both 1 + a and b + c lie in the part of the
domain of p(x) where it strictly increases. Now
p(1 + a) = (1 + a)3 3a(1 + a) = 1 + a3 = b3 + c3
= (b + c)3 3bc(b + c) < (b + c)3 3a(b + c) = p(b + c)
from which it follows that 1 + a < b + c.
Solution 5. Consider the function g(x) = x(1 + a3 x) = x(b3 + c3 x).
Then g(1) = g(a3 ) = a3 and g(b3 ) = g(c3 ) = (bc)3 . Since a3 < (bc)3 and
the graph of g(x) is a parabola opening down, it follows that b3 and c3 lie
between 1 and a3 .
Now consider the function h(x) = x1/3 +(b3 +c3 x)1/3 = x1/3 +(1+a3
x) for 0 x 1+a3 . Then h(1) = h(a3 ) = 1+a and h(b3 ) = h(c3 ) = b+c.
1/3

Since
2
h (x) = (x5/3 + (1 + a3 x)5/3 ),
9
the function h(x) is concave. Since b3 and c3 lie between 1 and a3 , it follows
that 1 + a < b + c, as desired.
2002:6. Let x, y > 0 be such that x3 +y 3 xy. Prove that x2 +y 2 1.
Solution 1. Let y = tx. Since x > y > 0, we have that 0 < t < 1. Then
x3 (1 + t3 ) x(1 t) x2 (1 + t3 ) (1 t). Therefore,
 
1t
x2 + y 2 = x2 (1 + t2 ) (1 + t2 )
1 + t3
1 t + t2 t3 t(1 t + 2t2 )
= 3
=1 .
1+t 1 + t3
Since 1 t + 2t2 , having negative discriminant, is always positive, the desired
result follows.
 3. INEQUALITIES 51

Solution 2. [J. Chui] Suppose, if possible, that x2 + y 2 = r2 > 1. We can

write x = r sin and y = r cos for 0 < < /2. Then
x3 + y 3 (x y) = r3 sin3 + r3 cos3 r sin + r cos
> r sin (sin2 1) + r cos3 + r cos
= r sin cos2 + r cos3 + r cos
 
1
2
= r cos cos + sin
cos
> r cos2 (2 sin ) > 0,
contrary to hypothesis. The result follows by contradiction.
2003:2. Let a, b, c be positive real numbers for which a + b + c = abc.
Prove that
1 1 1 3
+ + .
1+a 2 1+b 2 1+c 2 2
Solution 1. Let a = tan , b = tan , c = tan , where , , (0, /2).
Then
tan + tan + tan tan tan tan
tan( + + ) =
1 tan tan tan tan tan tan
a + b + c abc
= = 0,
1 ab bc ca
whence + + = . Then, the left side of the inequality is equal to
cos + cos + cos = cos + cos cos( + )
     
+ +
= 2 cos cos 2 cos2 +1
2 2 2
   
+ +
2 cos 2 cos2 +1
2 2
   2
3 1 + 3
= 2 cos 1 ,
2 2 2 2
with equality if and only if = = = /3.
Solution 2. Dene , and and note that + + = as in Solution 1.
Since cos x is a concave function on [0, /2], we have from Jensens inequality
that  
cos + cos + cos ++ 1
cos = cos = ,
3 3 3 2
from which the result follows.
2004:1. Prove that, for any nonzero complex numbers z and w,
 
z w 

(|z| + |w|) + 2|z + w| .
|z| |w| 

The solution to this problem appears in Chap. 2.

52 3. INEQUALITIES

2004:4. Let n be a positive integer exceeding 1. How many permutations

{a1 , a2 , . . . , an } of {1, 2, . . . , n} are there which maximize the value of the sum
|a2 a1 | + |a3 a2 | + + |ai+1 ai | + + |an an1 |
over all permutations? What is the value of this maximum sum?
Solution. First, suppose that n is odd. Then
   
 n + 1   n + 1 

|ai+1 ai | ai+1 + ai
2   2 
with equality if and only if 12 (n + 1) lies between ai+1 and ai .
Hence
n1
 n      
   n + 1     
|ai+1 ai | 2 a i  a 1 n + 1  + a n n + 1 
 2   2   2 
i=1 i=1
 n     
  n + 1   n + 1 
= |2ai (n + 1)| a1  + an
i=1
2   2 
= [(n 1) + (n 3) + + 2 + 0 + 2 + + (n 1)]
   
 n + 1   n + 1 

a 1 + an
2   2 
     
n1  n + 1   n + 1 
= 4 1 + 2 + + 
a 1 + an
2 2   2 
     
((n 1)/2) ((n + 1)/2)  n + 1   n + 1 
=4 a1 + a
2   2 
n
2
n2 1 n2 3
1=
2 2
since |a1 ((n + 1)/2)| + |an ((n + 1)/2)| 1. Equality occurs when one
of a1 and an is equal to 12 (n + 1) and the other is equal to 12 (n + 1) 1.
We get a permutation giving the maximum value of 12 (n2 3) if and
only if the foregoing conditions on a1 and an are satised (in four possible
ways) and 12 (n + 1) lies between ai and ai+1 for each i. For example, if
a1 = 12 (n + 1) + 1, then we require that the 12 (n 3) numbers a3 , a5 , . . . , an2
exceed 12 (n + 1) + 1 and the 12 (n 1) numbers a2 , a4 , . . . , an1 are less than
1
2 (n + 1). Thus, there are
   
n3 n1
4 ! !
2 2
ways of achieving the maximum.
Now suppose that n is even. As before,
   
 n + 1   n + 1 

|ai+1 ai | ai+1 + ai
2   2 
with equality if and only if 12 (n + 1) lies between ai+1 and ai .
 3. INEQUALITIES 53

We have that
n1    
  n + 1   n + 1 
|ai+1 ai | 2(1 + 3 + + (n 1)) a1 + a
2   2 
n
i=1
n2 n2 2
1= ,
2 2
with the latter inequality becoming equality if and only if {a1 , an } = {n/2, (n+
2)/2}. Suppose, say, that a1 = n/2 and an = (n + 2)/2. Then, to achieve
the maximum, we require that {a3 , a5 , . . . , an1 } = {1, 2. . . . , (n 2)/2} and
{a2 , a4 , . . . , an2 } = {(n/2) + 2, (n/2) + 3, . . . , n}. The maximum value of
(n2 2)/2 can be achieved with 2[(n 2)/2]!2 permutations.
2004:8. Let V be a complex n-dimensional inner product space. Prove
that
1
|u|2 |v|2 |u v|2 |u + v|2 |(u, v)|2 |u|2 |v|2 .
4
The solution to this problem appears in Chap. 7.
2005:2. Suppose that f is continuously dierentiable on [0, 1] and that
1
0
f (x)dx = 0. Prove that
 1  1  1
2 f (x)2 dx |f  (x)|dx |f (x)|dx.
0 0 0

The solution to this problem appears in Chap. 6.

2005:9 Let S be the set of all real-valued functions that are dened, pos-
itive and twice continuously dierentiable on a neighbourhood of 0. Suppose
that a and b are real parameters with ab = 0, b < 0. Dene operators from
S to R as follows:
A(f ) = f (0) + af  (0) + bf  (0);
G(f ) = exp A(log f ).
(a) Prove that A(f ) G(f ) for f S;
(b) Prove that G(f + g) G(f ) + G(g) for f, g S;
(c) Suppose that H is the set of functions in S for which G(f ) f (0).
Give examples of nonconstant functions, one in H and one not in
H. Prove that, if > 0 and f, g H, then f , f + g and f g all
belong to H.
The solution to this problem appears in Chap. 6.
2006:3. Let p(x) be a polynomial of positive degree n with n distinct
real roots a1 < a2 < < an . Let b be a real number for which 2b < a1 + a2 .
Prove that
2n1 |p(b)| |p (a1 )(b a1 )|.

The solution to this problem appears in Chap. 2.

54 3. INEQUALITIES

2009:7. Let n 2. Minimize a1 +a2 + +an subject to the constraints

0 a1 a2 an and a1 a2 + a2 a3 + + an1 an + an a1 = 1. (When
n = 2, the latter condition is a1 a2 = 1; when n 3, the sum on the left has
exactly n terms.)

Solution. If n 3 and ai = 1/ n for each i, then a1 +a2 + +an = n.
If a1 = a2 = = an2 = 0 and an1 = an = 1, a1 + a2 + + an = 2.
Therefore, when n 3, the minimum does not exceed the lesser of n and 2.

Let n = 2. By the arithmetic-geometric means inequality, we have that

a1 + a2 2 a1 a2
so that the minimum is 2.
Let n = 3. Then
(a1 + a2 + a3 )2 = a21 + a22 + a23 + 2(a1 a2 + a2 a3 + a3 a1 )
1 1 1
= (a21 + a22 ) + (a22 + a23 ) + (a23 + a21 ) + 2
2 2 2
a1 a2 + a2 a3 + a3 a1 + 2 = 3,

so that the minimum is 3.
Let n = 4. Then
(a1 + a2 + a3 + a4 )2 = a21 + a22 + a23 + a24 + 2 + 2a1 a3 + 2a2 a4
= (a1 + a3 )2 + (a2 + a4 )2 + 2
2[(a1 + a3 )(a2 + a4 )] + 2 = 4.
Therefore the minimum value of a1 + a2 + a3 + a4 is 2.
When n 5, we have that
a1 a3 + a2 a4 + + an1 a1 + an a2
a1 a2 + a2 a3 + + an1 [(a1 an ) + an ] + an a1
= a1 a2 + a2 a3 + + an1 an + an a1 + an1 (a1 an ).

Hence
(a1 + a2 + + an )2
= a21 + a22 + + a2n1 + a2n + 2(a1 a2 + a2 a3 + + an1 an + an a1 )
+ 2(a1 a3 + a2 a4 + + an1 a1 + an a2 )

+ {ai aj : i j 1, 2 (mod n)}
a21 + a22 + + a2n1 + a2n + 4(a1 a2 + a2 a3 + + an1 an + an a1 )
+ 2an1 (a1 an )
 3. INEQUALITIES 55

= a21 + a22 + + a2n1 + a2n + 4 + [an1 + (a1 an )]2

a2n1 (a1 an )2
= a22 + + a2n2 + 4 + (an1 + a1 an )2 + 2a1 an 4.
Therefore the minimum value of a1 + a2 + + an is equal to 2.
2011:2. Let u and v be positive reals. Minimize the larger of the two
values
1 1
2u + 2 and 2v + 2 .
v u

Solution 1. Observe that

       
1 1 1 1
2u + 2 + 2v + 2 = u + u + 2 + v + v + 2
v u u v
 1/3  1/3
1 1
3 uu 2 +3 vv 2 =6
u v
by the arithmetic-geometric means inequality. It follows that one of 2u + v 2
and 2v + u2 must be not less than 3. Since both assume the value 3 when
u = v = 1, the required minimum value is 3.
Solution 2. Wolog, we may assume that u v, whereupon
1 1 1
2u + 2v + 2 2v + 2 ,
v2 v u
with equality if and only if u = v. By checking the rst derivative, it can be
veried that 2v + v 2 achieves its minimum value of 3 when v = 1, so that
2u + v 2 3 for all positive values of u and v, with equality if and only if
u = v = 1.
2012:8. Determine the area of the set of points (x, y) in the plane that
satisfy the two inequalities:
x2 + y 2 2
x4 + x3 y 3 xy + y 4 .

The solution to this problem appears in Chap. 8.

2014:7. Suppose that x0 , x1 , , xn are real numbers. For 0 i n,
dene
yi = max(x0 , x1 , . . . , xi ).
Prove that
n1

yn2 4x2n 4 yi (xi+1 xi ).
i=0

When does equality occur?

56 3. INEQUALITIES

Solution. We can simplify the problem. Suppose that for some i < j,
xi = yi = yi+1 = = yj < yj+1 . Then yj+1 = xj+1 > xi and
yi (xi+1 xi ) + yi+1 (xi+2 xi+1 ) + + yj (xj+1 xj ) = yi (xj+1 xi ),
so that all of xi+1 , , xj do not gure in the terms of the right side. There-
fore, with no loss of generality, we can assume that
x0 < x1 < < xn1
so that yi = xi for 0 i n 1, and yn is equal to the maximum of xn1
and xn .
The right side of the inequality is equal to
4x2n 4[x0 (x1 x0 ) + x1 (x2 x1 ) + + xn1 (xn xn1 )]
= 2x20 + 2(x1 x0 )2 + 2(x2 x1 )2 + + 2(xn xn1 )2 + 2x2n
= 2x20 + 2(x1 x0 )2 + 2(x2 x1 )2 + + 2(xn1 xn2 )2 + x2n1
+ (2xn xn1 )2 .
Therefore, the right side is greater than or equal to both of 2x2n and x2n1 .
The desired result follows.
Equality occurs if and only if xi = 0 for each i.

Comment. It is natural to try a proof by induction. When n = 1, the

right side is equal to
4x21 4x0 x1 + 4x20 = 2x21 + 2x20 + 2(x1 x0 )2
which is clearly greater than or equal to either of x21 and x20 .

Suppose that the inequality holds for n = m 1. Then

m

4x2m+1 4 yi (xi+1 xi )
i=0
m1
!

= 4x2m+1 4x2m + 4x2m 4 yi (xi+1 xi ) 4ym (xm+1 xm )
i=0
4x2m+1 4x2m + ym
2
4ym (xm+1 xm ) K.
Since ym+1 is the greater of xm+1 and ym , there are two cases to consider.
When ym+1 = xm+1 , then, since xm ym ym+1 and 2ym+1 ym ym ,
K = (2ym+1 ym )2 + 4xm (ym xm ) (2ym+1 ym )2 ym+1
2
.
However, when ym+1 = ym , then
2
K = ym + 4(xm+1 xm )(xm+1 + xm ym )
and it is not clear whether the second term is always nonnegative. Can an
induction argument be made to work?
 3. INEQUALITIES 57

[Problem 2014:7 is due to Mathias Beiglb

ock of the University of Vienna,
and was communicated to the contest by Florian Herzig of the University of
Toronto.]
2015:4. Determine all the values of the positive integer n 2 for which
the following statement is true, and for each, indicate when equality holds.
For any nonnegative real numbers x1 , x2 , . . . , xn ,
(x1 + x2 + + xn )2 n(x1 x2 + x2 x3 + + xn1 xn + xn x1 ),
where the right side has n summands.
Solution. Let x1 = x2 = 1 and x3 = x4 = = xn = 0. Then the
left side of the inequality is equal to 4 and the right side to n. Therefore a
necessary condition for the inequality to hold for all sets of xi is n 4.
For n = 2, we nd that
(x1 + x2 )2 2(x1 x2 + x2 x1 ) = (x1 x2 )2 0,
so the inequality holds with equality if and only if x1 = x2 .
For n = 3, we nd that
2[(x1 +x2 +x3 )2 3(x1 x2 +x2 x3 +x3 x1 )] = (x1 x2 )2 +(x2 x3 )2 +(x3 x1 )2 0,
so the inequality holds with equality if and only if x1 = x2 = x3 .
For n = 4, we nd that
(x1 + x2 + x3 + x4 )2 4(x1 x2 + x2 x3 + x3 x4 + x4 x1 )
= x21 + x22 + x23 + x24 + 2x1 x3 + 2x2 x4 2x1 x2
2x2 x3 2x3 x4 2x4 x1
= (x1 x2 )2 + (x3 x4 )2 + 2(x1 x2 )(x3 x4 )
= (x1 x2 + x3 x4 )2 0,

so the inequality holds with equality if and only if x1 + x3 = x2 + x4 .

Comment. Another case that might be tried is xi = i 1 for 1 i n.
Then x1 + + xn = 12 n(n 1) and x1 x2 + + xn1 xn + xn x1 = 13 n(n
1)(n 2). The left side minus the right is equal to 12 n (n 1)(5 n), which
1 2

establishes that n 5. In fact, in the n = 5 case, equality occurs whenever

the xi form an arithmetic progression, so a counterexample has to be found
elsewhere.
 CHAPTER 4

Sequences and Series

2001:6. Prove that, for each positive integer n, the series


kn
2k
k=1
converges to twice an odd integer not less than (n + 1)!.
Solution 1. Convergence of the series results from either the ratio or the
root test. For nonnegative integers n, let

 kn
Sn = .
2k
k=1

Then S0 = 1 and

 kn  kn
1
Sn Sn =
2 2k 2k+1
k=1 k=1
 n 
k (k 1)n
=
2k 2k
k=1 k=1

k (k 1)n
n
=
2k
k=1
   n1   n2   n3 
n k n k n k n1 1
= + + (1)
1 2k 2 2k 3 2k 2k
k=1

whence
      
n n n n1
Sn = 2 Sn1 Sn2 + Sn3 + (1) .
1 2 3
An induction argument establishes that Sn is twice an odd integer.

Springer International Publishing Switzerland 2016 59

E.J. Barbeau, University of Toronto Mathematics Competition
(20012015), Problem Books in Mathematics,
DOI 10.1007/978-3-319-28106-3 4
60 4. SEQUENCES AND SERIES

Observe that S0 = 1, S1 = 2, S2 = 6 and S3 = 26. We prove by induction

that, for each n 0,

Sn+1 (n + 2)Sn

from which the desired result will follow. Suppose that we have established
this for n = m 1. Now
     
m+1 m+1 m+1
Sm+1 = 2 Sm Sm1 + Sm2
1 2 3

  
m+1
Sm3 + .
4

For each positive integer r not exceeding 12 (m + 1),

   
m+1 m+1
Sm2r+2 Sm2r+1
2r 1 2r
   
m+1 m+1
(m 2r + 3) Sm2r+1
2r 1 2r
   
m+1 m 2r + 2
= (m 2r + 3) Sm2r+1 0 .
2r 1 2r

When r = 1, we get inside the square brackets the quantity

m m+2
(m + 1) =
2 2

while when r > 1, we get

 
m 2r + 2
(m 2r + 3) > (m 2r + 3) (m 2r + 2) = 1.
2r

Hence
    
m+1 m+1
Sm+1 2 Sm Sm1
1 2
 
m(m + 1) 1
2 (m + 1)Sm Sm
2 m+1
 m
=2 m+1 Sm = (m + 2)Sm .
2
 4. SEQUENCES AND SERIES 61

Solution 2. Dene Sn as in the foregoing solution. Then, for n 1,

1  kn
Sn = +
2 2k
k=2

1 1  (k + 1)n
= +
2 2 2k
k=1

n n1  n 
1 1k + 1 k
n
+ + n1 k+1
= +
2 2 2k
k=1
     
1 1 n n
= + Sn + Sn1 + + S1 + 1
2 2 1 n1
whence S1 = 2 and
     
n n n
Sn = Sn1 + Sn2 + + S1 + 2
1 2 n1
for n 2. It is easily checked that Sk 2 (mod 4) for k = 1. As an induction
hypothesis, suppose this holds for 1 k n 1. Then, modulo 4, the right
side is congruent to
n  
!
 n
2 2 + 2 = 2(2n 2) + 2 = 2n+1 2,
k
k=0

and the desired result follows.

For n 1,
n+1 n+1  
Sn+1 1 Sn + Sn1 + + n+1
2 n S1 + 2
=
Sn Sn
n+1  
2 Sn1 + n+1 Sn2 + + (n + 1)S1 + 2
= (n + 1) + n
 3n
1 Sn1 + 2 Sn2 + + nS1 + 2
(n + 1) + 1 = n + 2,

since each term in the numerator of the latter fraction exceeds each corre-
sponding term in the denominator.
Solution 3 (of the rst part). [P. Gyrya] Let f (x) be a dierentiable
function and let D be the dierentiation operator. Dene the operator L by

L(f )(x) = x D(f )(x).

k
Suppose that f (x) = (1 x)1 = k=0 x . Then, it is standard that
n
L (f )(x) has a power series expansion obtained by term-by-term dierentia-
tion that converges absolutely for |x| < 1. By induction, it can be shown that
the series given in the problem is, for each nonnegative integer n, Ln (f )(1/2).
62 4. SEQUENCES AND SERIES

It is straightforward to verify that

L((1 x)1 ) = x(1 x)2
L2 ((1 x)1 ) = x(1 + x)(1 x)3
L3 ((1 x)1 ) = x(1 + 4x + x2 )(1 x)4
L4 ((1 x)1 ) = x(1 + 11x + 11x2 + x3 )(1 x)5 .
In general, a straightforward induction argument yields that for each positive
integer n,
Ln (f )(x) = x(1 + an,1 x + + an,n2 xn2 + xn1 )(1 x)(n+1)
for some integers an,1 , , an,n2 . Hence
Ln (f )(1/2) = 2(2n1 + an,1 2n2 + + an,n2 2 + 1),
yielding the desired result.
2003:1. Evaluate  

 2
arctan .
n=1
n2

Solution 1. Let an = arctan n for n 0. Then

(n + 1) (n 1) 2
tan(an+1 an1 ) = = 2
1 + (n 1)
2 n
for n 1. Then
m
2
arctan 2 = arctan(m + 1) + arctan m arctan 1 arctan 0.
n=1
n
Letting m yields the answer /2 + /2 /4 0 = 3/4.
Solution 2. Let bn = arctan(1/n) for n 0. Then
2
tan(bn1 bn+1 ) =
n2
for n 2, whence
m
 m
2
arctan = arctan 2 + (bn1 bn+1 )
n=1
n2 n=2
1 1 1
= arctan 2 + arctan 1 + arctan arctan arctan
2 m m+1
1 1
= (arctan 2 + arccot 2) + arctan 1 arctan arctan
m m+1
1 1
= + arctan arctan
2 4 m m+1
for m 3, from which the result follows by letting m tend to innity.
n 2
Solution 3. [S. Huang] Let sn = k=1 arctan(2/k ) and tn = tan sn .
Then {tn } = {2, , 9/2, 14/5, 20/9, . . . } where the numerators of the
 4. SEQUENCES AND SERIES 63

fractions are {2, 5, 9, 14, 20, . . . } and the denominators are {1, 0,
2, 5, 8, . . . }. We conjecture that
n(n + 3)
tn =
(n 2)(n + 1)
for n 1. This is true for 1 n 5. Suppose that it holds to n = k 1 5,
so that tk1 = (k 1)(k + 2)/(k 3)k. Then
tk1 + 2k 2
tk =
1 2tk1 k 2
k 2 (k 1)(k + 2) + 2(k 3)k
=
k 3 (k 3) + 2(k 1)(k + 2)
k(k + 3)(k 2 2k + 2) k(k + 3)
= = .
(k 2)(k + 1)(k 2k + 2)
2 (k 2)(k + 1)
The desired expression for tn holds by induction and so limn tn = 1.
For n 3, tn < 0 and tan1 (2/n2 ) < /2, so we must have /2 < sn <
and sn = tan1 tn . Therefore
lim sn = arctan( + lim tn ) = (/4) = (3)/4.
n n

2003:4. Show that the positive integer n divides the integer nearest to
(n + 1)!
.
e
Solution. By Taylors Theorem, we have that
1 1 1 1 ec
e1 = 1 + + + (1)n+1 + (1)n+2
1! 2! 3! (n + 1)! (n + 2)!
where 1 < c < 0. Hence
(n + 1)! (n + 1)!
(n + 1)!e1 = (n + 1)! (n + 1)! +
2! 3!
ec
+ + (1)n [(n + 1) 1] + (1)n2 .
n+2
The last term does not exceed 1/(n + 2), which is less than 1/2. The sec-
ond last term is equal to n, and each previous term has a factor n in the
numerator that is not cancelled out by the denominator. Since the sum of
all but the last term is an integer divisible by n and is the nearest integer to
(n + 1)!/e, the result holds.
2004:2. Prove that
 1
1 1 1 1
xx dx = 1 2 + 3 4 + 5 + .
0 2 3 4 5
Solution. First, let
 1
I(m, n) = xm (log x)n dx
0
64 4. SEQUENCES AND SERIES

for nonnegative integers m and n. Then I(0, 0) = 1 and I(m, 0) = 1/(m + 1)

for every nonnegative integer m. Taking the parts u = (log x)n , dv = xm dx
and noting that limx0 xm+1 (log x)n = 0, we nd that I(m, n) = (n/(m+1))
I(m, n 1) for n 1, whence
(1)n n!
I(m, n) =
(m + 1)n+1
for each nonnegative integer n. In particular,
(1)k k!
I(k, k) =
(k + 1)(k+1)
for each nonnegative integer k.

Using the fact that the series is uniformly convergent and term-by-term
integration is possible, we nd that
 1  1  1
 (x log x)k
xx dx = ex log x dx = dx
0 0 0 k!
k=0

 k
(1) 1 1
=1+ (k+1)
= 1 2 + 3 .
(k + 1) 2 3
k=1

2007:6. Let h(n) denote the number of nite sequences {a1 , a2 , . . . , ak }

of positive integers exceeding 1 for which k 1, a1 a2 ak and
n = a1 a2 ak . (For example, if n = 20, there are four such sequences {20},
{10, 2}, {5, 4} and {5, 2, 2} and h(20) = 4.)
Prove that

h(n)
= 1.
n=1
n2

Solution 1. We have that

 
 h(n)   1
1+ =
n=2
n2 r=2
r2k
k=0

 
1 r2
= =
r=2
1 1/r2 r=2
r2 1
n 
  
r r 2n
= lim = lim = 2,
n
r=2
r1 r+1 n n + 1

from which the result follows.

Solution 2. [J. Kramar] Observe that

h(n) 
2
= {(a1 a2 ak )2 : a1 a2 ak 2}.
n=1
n
 4. SEQUENCES AND SERIES 65

For m 2, let

bm = {(a1 a2 ak )2 : a1 = m a2 ak 2}.
Since
1 1
bm = + 2 (bm + bm1 + + b2 ),
m2 m
then, for m 3,
(m2 1)bm = 1 + b2 + + bm1 .
Note that
1 1 1 1
+ 2 + 2 + = .
b2 =
22 4 8 3
Assume as an induction hypothesis that, for 2 k m 1, bk =
2/(k(k + 1)). Then
m1

1
(m + 1)(m 1)bm = 1 + b2 + + bm1 = 1 + 2
k(k + 1)
k=2
m1   
 1 1 1 1
=1+2 =1+2
k k+1 2 m
k=2
 
2 m1
=2 =2 ,
m m
so that bm = 2/(m(m + 1)).
Hence
    
h(n) 1 1
= b m = 2 = 1.
n=1
n2 m=2 m=2
m m+1

2008:3. Suppose that a is a real number and the sequence {an } is dened
recursively by a0 = a and
an+1 = an (an 1)
for n 0. Find the values of a for which the sequence {an } converges.

Solution. When the sequence converges, the limit b must satisfy the equa-
tion b = b(b1) so that b = 0 or b = 2. It is clear that the sequence converges
when a = 1, 0, 1, 2. If a > 2, then an induction argument shows that {an }
is increasing and unbounded. If a < 1, then a1 > 2 and the sequence again
diverges.

If 1 < a < 2, we show that the sequence will have an entry in the
interval [0, 1]. Suppose that 1 < a < 2, then a(a 1) < a, so that an will
initially decrease until it arrives in the interval (0, 1]. If 1 < a < 0, then
a1 = (a)(1 a) will lie in (0, 2).
66 4. SEQUENCES AND SERIES

y = x(x1)

Fig. 4.1. A cobweb diagram

It remains to analyze the situation that 0 < a < 1. Then 1/4 a1 < 0
and 0 < a2 5/16 < 1. Thus, the sequence alternates between the intervals
[1/4, 0) and (0, 1). Observe that
an+2 = an+1 (an+1 1) = an (an 1)(a2n an 1)
= an (1 a2n (2 an )).
When n is even, then an and an+2 are both positive and an+2 < an . When
n is odd, then an and an+2 are both negative and 1 a2n (2 an ) (0, 1),
so that the sequence {a2m } decreases to a limit u and {a2m+1 } increases
to a limit v. We must have that u = v(v 1) and v = u(u 1) so that
u = u(u 1)(u2 u 1) or u3 (2 u) = 0. Since u = 2, we must have
u = v = 0.
Hence the sequence converges if and only if 1 a 2; the limit is 2
when a = 1, 2 and 0 when 1 < a < 2.

Comment. The situation can also be analyzed using a cobweb diagram

on the graph of the function y = x(x 1). On the same axes, plot y = x and
y = f (x), and then follow the progress of the points (a1 , a2 ) = (a1 , f (a1 )),
(a2 , a2 ), (a2 , a3 ) = (a2 , f (a2 )), (a3 , a3 ), and so on. The positions of the
abscissae will indicate the behaviour of an as n increases. See Fig. 4.1.
 4. SEQUENCES AND SERIES 67

2008:9. For each positive integer n, let

n
2k
Sn = .
k2
k=1
Prove that Sn+1 /Sn is not a rational function of n.
The solution to this problem appears in Chap. 2.
2009:4. Let {an } be a real sequence for which

an
n=1
n
converges. Prove that
a1 + a2 + + an
lim = 0.
n n
Solution. Recall Abels Partial Summation Formula:
n
 n

xk yk = (x1 + x2 + + xn )yn+1 + (x1 + x2 + + xk )(yk yk+1 ).
k=1 k=1

Applying this to xn = an /n and yn = n. we obtain that

n

a1 + a2 + + an = (x1 + x2 + + xn )(n + 1) (x1 + x2 + + xk ),
k=1

whence
 n
  n
a1 + a 2 + + a n  1 1
= xk 1+ (x1 + x2 + + xk ).
n n n
k=1 k=1

Let
n "
ak #
n
 
L = lim xk = lim .
n n k
k=1 k=1

Then
  
n
 n
1 1
L = lim 1+ xk = lim (x1 + x2 + + xk ).
n n n n
k=1 k=1

To see the latter equality, suppose that un = x1 + x2 + + xn and vn =

(1/n)(u1 + u2 + + un ), so that
1
v n un = [(u1 un ) + (u2 un ) + (un un )].
n
68 4. SEQUENCES AND SERIES

Let  > 0 be given. First select n1 so that |uk un | < 12  for n k n1 .

Since lim un = L, there exists a positive integer M for which |un | < M for
each n. Select n2 > n1 such that, for all n > n2 , 2M n1 /n < /2. Then, for
all n > n2 ,
n1 n
!
1  
|vn un | |uk un | + |uk un |
n
k=1 k=n1 +1
 
1 2M n1 n n1 "  # "  #
+ < + = .
n n n 2 2 2
It follows that lim vn = L, and the desired result obtains.
2010:2. Let u0 = 1, u1 = 2 and un+1 = 2un + un1 for n 1. Prove
that, for every nonnegative integer n,
  (i + j + k)! 
un = : i, j, k 0, i + j + 2k = n .
i!j!k!

The solution to this problem appears in Chap. 10.

2011:4. Let {bn : n 1} be a sequence of positive real numbers such
that
3bn+2 bn+1 + 2bn
for every positive integer n. Prove that either the sequence converges or that
it diverges to innity.
Solution 1. For n 1, let cn = min(bn , bn+1 ). Then bn+1 cn and
bn+2 13 bn+1 + 23 bn cn , so that cn+1 cn for n 1.
Since bn cn for all n, {bn } diverges to innity if {cn } does.
Otherwise, let {cn } be bounded. Since the sequence increases, it increases
to a nite limit m. Let  > 0 be given. Suppose that N is selected so that
m 12  < cn m for n N . If bn is equal to either cn or cn1 , then
m  < bn < m for n > N . If bn is equal to neither cn1 nor cn , then
cn1 = bn1 and cn = bn+1 , whence
bn 3bn+1 2bn1 = 3cn 2cn1 3m 2(m /2) = m + .
Thus, m  < bn < m +  for n > N . It follows that limn bn = m.
Solution 2. [A. Remorov] Let u = lim inf n bn and v = lim supn bn .
If {bn } is unbounded, then, given M > 0, there exists r such that br > 3M .
Then br+1 > 13 br > M and br+2 > 23 br > M , whereupon it can be established
by induction that bn > M for n > r. Hence u M for each M > 0 and
limn bn = +.
Otherwise, {bn } is bounded and v is nite. Suppose, if possible, that
v > u. Let s be selected so that bn > u 13 (v u) for n s. If it happens
that bn+1 > u + 34 (v u) for some n s, then
u 1 2u 1 1
bn+2 > + (v u) + (v u) = u + (v u),
3 4 3 6 12
 4. SEQUENCES AND SERIES 69

from which it follows that bn+k > u + 121

(v u) for all k > 0. But this would
entail that u u + 12 (v u), which is false. Therefore bn+1 < u + 34 (v u)
1

for all n s. Hence v u + 34 (v u), or v u < 34 (v u), which is false.

Therefore v = u and the result follows.
2013:10.
(a) Let f be a real-valued function dened on the real number eld
R for which |f (x) f (y)| < |x y| for any pair (x, y) of distinct
elements of R. Let f (n) denote the nth composite of f dened by
f (1) (x) = f (x) and f (n+1) (x) = f (f (n) (x)) for n 2. Prove that
exactly one of the following situations must occur:
(i) limn+ f (n) (x) = + for each real x;
(ii) limn+ f (n) (x) = for each real x;
(iii) there is a real number z such that
lim f (n) (x) = z
n+

for each real x.

(b) Give examples to show that each of the three cases in (a) can occur.
The solution to this problem appears in Chap. 6.
2014:3. Let n be a positive integer. A nite sequence {a1 , a2 , . . . , an } of
positive
n integers ai is said to be tight if and only if 1 a1 < a2 < < an , all
2 dierences aj ai with i < j are distinct, and an is as small as possible.
(a) Determine a tight sequence for n = 5.
(b) Prove that there is a polynomial p(n) of degree not exceeding 3 such
that an p(n) for every tight sequence {ai } with n entries.
The solution to this problem appears in Chap. 11.
2014:5. Let n be a positive integer. Prove that
    n 
1 
n n n
1 1 1 2k1 k
 n = = = n1 : k odd , 1 k n .
k k
k2nk 2n1 k 2 k
k=1 k=1 k=1

Solution. The values of each of the sums for 1 n 5 are 1, 1, 5/6,

2/3, 8/15.

The equality of the second and third sums

  is straightforward
  to establish.
Let Sn denote the leftmost sum. Since k nk = n n1k1 , we have that
n    1
n1
Sn = n .
k1
k=1

Using the fact that

  1   1   1
n n n1
(n + 1) + (n + 1) = n ,
k1 k k1
70 4. SEQUENCES AND SERIES

for 1 k n, it can be shown that S1 = 1 and that

Sn 1
Sn+1 = +
2 n+1
for n 1.
Therefore
1 1 1 1 1
Sn = + + + + + +
2n122 n2 32 n3 k2nk n
n
 n
1 1  2k1
= = n1 .
k 2nk 2 k
k=1 k=1

For each positive integer n, let

n
1 
k
Tn = .
2n1 k
k odd

Then T1 = 1 and
n+1   n 
1  nk 1  k1
k
2Tn+1 Tn = = n1
2n1 k 2 k
k odd k odd
1  n + 1 2n
2
= = = .
(n + 1)2n1 k (n + 1)2n1 n+1
k odd

Therefore the recursions for Sn and Tn agree when n = 1 and satisfy the
same equations. Thus the result holds.
2014:9. Let {an } and {bn } be positive real sequences such that
an
lim =u>0
n n
and
 n
bn
lim = v > 0.
n an
Prove that
 
bn
lim =1
n an
and
lim (bn an ) = u log v.
n

Solution. Observe that

 n
bn 1 bn
lim = lim exp log = exp(0 v) = exp 0 = 1.
n an n n an
 4. SEQUENCES AND SERIES 71

Suppose that v = 1. Then for suciently large values of n, bn = an and

we have that
     n
an bn a n an bn
(bn an ) log 1 + = log .
bn a n an n an
Note that
 
bn a n bn
lim = lim 1 =0
n an n an

and that limt0 t1 log(1 + t) = 1, so that

 
lim (bn an ) 1 = u log v
n

as desired.
If v = 1, we need to exercise more care, as bn could equal an innitely
often. In this case, the indices can be partitioned into two sets A and B,
where respectively bn = an for n A and bn = an for n B; either of these
sets could be innite or nite. If innite, taking the limit over n A and
over n B yields the result 0 = u log v, and the problem is solved.
[Problem 2014:9 was contributed by the AN-anduud Problem Solving
Group in Ulaanbataar, Mongolia.]
2015:8. Let {an } and {bn } be two decreasing positive real sequences for
which

an =
n=1
and


bn = .
n=1
Let I be a subset of the natural numbers, and dene the sequence {cn } by

an , if n I
cn =
bn , if n I.

Is it possible for n=1 cn to converge?
Solution. The answer is yes. Let s1 = 0, and for n 0, let
n
 k
sn = 22 .
k=0

For n 0, let In be the set of positive integers k for which sn1 < k sn .
Dene
2n 2n
ak = 22 n and bk = 22 when k I2n ;
2n+1 2n+1
ak = 22 and bk = 22 n
when k I2n+1 .
It can be veried that both {an } and {bn } are decreasing.
72 4. SEQUENCES AND SERIES

Let

$
I= I2n .
n=0

Then, for each nonnegative n,

 
ak = bk = 1.
kI2n+1 kI2n

Therefore

an = 2;
nI


an = ;
nI


bn = ;
nI

and

bn = 2.
nI

[Problem 2015:8 was contributed by Franklin Vera Pacheco.]

2015:10. (a) Let

g(x, y) = x2 y + xy 2 + xy + x + y + 1.

We form a sequence {x0 } as follows: x0 = 0. The next term x1 is the unique

solution 1 of the linear equation g(t, 0) = 0. For each n 2, xn is the
solution other than xn2 of the equation g(t, xn1 ) = 0.
Let {fn } be the Fibonacci sequence determined by f0 = 0, f1 = 1 and
fn = fn1 + fn2 for n 2. Prove that, for any nonnegative integer k,
fk fk+2
x2k = and x2k+1 = .
fk+1 fk+1

(b) Let
h(x, y) = x2 y + xy 2 + xy + (x + y) +
be a polynomial with real coecients , , . We form a bilateral sequence
{xn : n Z} as follows. Let x0 = 0 be given arbitrarily. We select x1 and
x1 to be the two solutions of the quadratic equation h(t, x0 ) = 0 in either
order. From here, we can dene inductively the terms of the sequence for
positive and negative values of the index so that xn1 and xn+1 are the two
 4. SEQUENCES AND SERIES 73

solutions of the equation h(t, xn ) = 0. We suppose that at each stage, neither

of these solutions is zero.
Prove that the sequence {xn } has period 5 (i.e. xn+5 = xn for each
index n) if and only if 3 + 2 = 0.
The solution to this problem appears in Chap. 2.
 CHAPTER 5

Calculus and its Applications

2001:2. Let O = (0, 0) and Q = (1, 0). Find the point P on the line with
equation y = x + 1 for which the angle OP Q is a maximum.
Solution 1. For the point P to maximize OP Q, it must be a point of
tangency of a circle with chord OQ and the line of equation y = x + 1. The
general circle through O and Q has equation
 2
1 2 1
x + (y k) = + k 2
2 4
or
x2 x + y 2 2ky = 0.
Solving this with y = x + 1 yields

2x2 + (1 2k)x + (1 2k) = 0.

The roots of this quadratic equation are coincident if and only if 2k = 1 or

1 2k = 8. When k = 12 , we get the point of tangency (0, 1). When k = 72 ,
we get the point of tangency (2, 1).
The rst solution corresponds to a circle with chord OQ touching the
line of equation y = x + 1 at the point P (0, 1). This circle has diameter
P Q and the angle OP Q is equal to 45 . This angle certainly exceeds angle
OXQ, where X is any point other than P on the line in the upper half plane.
The second solution corresponds to a circle touching the line of equation
y = x + 1 below the x-axis at the point R(2, 1). The angle ORQ exceeds
angle OY Q, where Y is any point other than R on the line in the lower half-
plane. Now, the line y = x + 1 makes an angle of 45 with the line y = 1,
and this angle exceeds ORQ. Hence P is the desired point.

Springer International Publishing Switzerland 2016 75

E.J. Barbeau, University of Toronto Mathematics Competition
(20012015), Problem Books in Mathematics,
DOI 10.1007/978-3-319-28106-3 5
76 5. CALCULUS AND ITS APPLICATIONS

Solution 2. Let (t, t + 1) be a typical point P on the line y = x + 1. The

slope of OP is (t + 1)/t and of QP is (t + 1)/(t 1). Hence
t+1
tan OP Q = .
2t2 + t + 1
Let f (t) = (t + 1)/(2t2 + t + 1), so that (2t2 + t + 1)f (t) = t + 1. Then
(4t + 1)f (t) + (2t2 + t + 1)f  (t) = 1, whence
2t(t + 2)
f  (t) = .
(2t2 + t + 1)2
Now f  (t) < 0 when t > 0 or t < 2 while f  (t) > 0 when 2 < t < 0. Thus,
f (t) assumes its minimum value of 1/7 when t = 2 and its maximum
value of 1 when t = 0. Hence |f (t)| 1 with equality if and only if t = 0.
The desired point P is (0, 1).

Comment. The workings can be made a little easier by considering

cot OP Q = 2t 1 + 2(t + 1)1 . The derivative of this is 2[1 (t + 1)2 ], and
this is positive if and only if t > 0 and t < 2.

Solution 3. Let P be (x, x + 1) and let = OP Q. From the Law of

Cosines, we have that

1 = x2 + (x + 1)2 + (x 1)2 + (x + 1)2 2 (2x2 + 2x + 1)(2x2 + 2) cos ,
whence
2x2 + x + 1
cos =  .
(2x2 + 2x + 1)(2x2 + 2)
Let
(2x2 + 2x + 1)(2x2 + 2)
f (x) = sec2 =
(2x2 + x + 1)2
4x4 + 4x3 + 6x2 + 4x + 2
= 4
4x + 4x3 + 5x2 + 2x + 1
(x + 1)2
=1+ = 1 + g(x)2 ,
(2x2 + x + 1)2
where g(x) = (x+1)(2x2 +x+1)1 . Since g  (x) = [2x(x+2)](2x2 +x+1)2 ,
we see that g(x) has its maximum value when x = 0 and its minimum when
x = 2, and vanishes when x = 1. Hence g(x)2 and thus sec assume
relative maxima when x = 0 and x = 2. A quick check reveals that x = 0
gives the overall maximum, and so that the angle is maximized when P is
located at (0, 1).
2002:4. Consider the parabola of equation y = x2 . The normal is con-
structed at a variable point P and meets the parabola again in Q. Determine
the location of P for which the arc length along the parabola between P and
Q is minimized.
 5. CALCULUS AND ITS APPLICATIONS 77

Solution. Wolog, we may assume that u > 0, as the arc length for u and
u is the same. The tangent to the parabola at (u, u2 ) has slope 2u, and so
the normal has slope 1/2u. The equation of the normal is
1
y u2 = (x u)
2u
and this intersects the parabola at the point
 
1 2 1
u ,u + 1 + 2 .
2u 4u
The arc length is given by
 u 
f (u) = 1 + 4x2 dx = F (u) F (u (1/2u)),
u(1/2u)

where F is a function for which F  (x) = (1 + 4x2 )1/2 . Then

 
1
f  (u) = F  (u) F  (u (1/2u)) 1 + 2
2u
 
1
= (1 + 4u2 )1/2 (1 + 4u2 + 4 + u2 )1/2 1 + 2
2u
 
1 1
= (1 + 4u2 )1/2 (4u4 + 5u2 + 1)1/2 + 3
u 2u
 
(u + 1) (2u 1)
2 1/2 2
= (1 + 4u2 )1/2 1 + .
2u3
f  (u) is negative when u is close to 0, and positive when u is very large. It
vanishes if and only if 2u3 = (u2 + 1)1/2 (2u2 1). Thus 4u6 = (u2 +1)

(4u 4u + 1) 0 = 3u + 1,and we have that f (1/ 3) = 0. Hence
4 2 2
f (u)
decreases on the interval (0, 1/ 3) and increases on the interval (1/ 3, ).
Hence, the arc length is minimized when P is the one of the points
   
1 1 1 1
, , , .
3 3 3 3
2003:3. Solve the dierential equation

y  = yy  .

Solution. The equation can be rewritten

dy  
y = yy 
dy
from which either y  = 0 and y is a constant, or dy  = ydy, whence
dy 1
= y  = (y 2 + k)
dx 2
for some constant k. (One can also get to the same place by noting that the
equation can be rewritten as 2y  = (y 2 ) and integrating.)
78 5. CALCULUS AND ITS APPLICATIONS

In case k = 0, we have that 2dy/y 2 = dx, whence 2/y = x + c and

y = 2(x + c)1 for some constant c.
In the case that k = a2 > 0, we have that
2dy
= dx
y 2 + a2
whence (2/a) tan1 (y/a) = x + c and
a(x + c)
y = a tan
2
for some constant c.
In the case that k = b2 < 0, we have that
 
2dy 1 dy dy
= = dx
y 2 b2 b yb y+b
whence
 
1  y b 
ln =x+c
b y + b
for some constant c. Solving this for y yields that
   
1 + eb(x+c) 1 eb(x+c)
y=b and y = b .
1 eb(x+c) 1 + eb(x+c)

2003:9. Prove that the integral


sin2 x
dx
0 2 x2
exists and evaluate it.

Solution. Since limx (sin2 x)/( 2 x2 ) exists and equals 0 (by

lHopitals Rule), there is a removable singularity in the integrand at x = .
The integral on the innite interval converges by comparison with the integral
of 1/x2 .

First, note that, for k 0,

 (k+1)  (k+1)  (k+1) !
sin2 x 1 sin2 x sin2 x
dx = dx dx
k 2 x2 2 k x k x+
 k  (k+2)
1 sin2 x 1 sin2 x
= dx + dx.
2 (k1) x 2 (k+1) x

Hence
 n   (n+1)
sin2 x 1 sin2 x 1 sin2 x
dx = dx + dx.
0 2 x2 2 x 2 (n1) x
 5. CALCULUS AND ITS APPLICATIONS 79

The rst integral on the right vanishes, because the integrand is odd, and so
   (n+1)  (n+1)
 n
sin2 x  sin2 x
 dx = 1 dx
1 1
dx
1
,
 2 x2  2 (n1) x 2 (n1) x (n 1)
0

 sin2 x
with the result that 0 2 x2 dx = 0.
2004:2. Prove that
 1
1 1 1 1
xx dx = 1 + 3 4 + 5 + .
0 22 3 4 5

The solution to this problem appears in Chap. 4.

2004:6. Determine
 1   1 
dt dt
.
0 1 t4 0 1 + t4

Solution. The substitution t2 = sin leads to

 1  /2  /2
dt cos d 1 d
= = .
0 1 t4 0 2 sin cos 2 0 sin

The substitution t2 = tan followed by the substitution = 2 leads to

 1  /4
dt sec2 d
=
0 1 + t4 0 2 tan sec
 /4
1 d
=
2 0 sin cos
 /4

1 2 d 1 /2 2 d
= =
2 0 sin 2 4 0 sin
 /2
1 d
= .
2 2 0 sin

Thus the answer is 2.

2005:1. Show that, if /2 < < /2, then


log(1 + tan tan x) dx = log sec .
0
80 5. CALCULUS AND ITS APPLICATIONS

Solution 1. [D. Han]

 
log(1 + tan tan x) dx = log(1 + tan tan( u)) du
0 0
   
tan tan u
= log 1 + tan du
0 1 + tan tan u

= [log(1 + tan2 ) log(1 + tan tan u)] du
0

= log sec2 log(1 + tan tan u) du
0

= 2 log sec log(1 + tan tan x) dx.
0

The desired result follows.

Solution 2.

log(1 + tan tan x) dx
0

= [log(cos cos x + sin sin x) log cos log cos x] dx
0

= [log cos( x) + log sec log cos x] dx
0
 
= log cos( x) dx + log sec log cos x dx
0 0
 
= log cos xdx + log sec log cos xdx = log sec .
0 0


Solution 3. Let F () = 0 log(1 + tan tan x) dx. Then, using the sub-
stitution u = tan x, we nd that

sec2 tan x
F  () = log(1 + tan2 ) + dx
0 1 + tan tan x
 tan  
tan u tan
= log sec2 + + 2 + 2 du
0 1 + (tan )u u + 1 u + 1
 tan
1
= +2 log sec + log(1 + (tan )u) + log(1 + u2 ) + tan [arctan u]
2 0
1
= 2 log sec log(1 + tan2 ) + log sec2 + tan
2
= log sec + tan .
 5. CALCULUS AND ITS APPLICATIONS 81

Also, the derivative of G() log sec is equal to log sec + tan . Since
F  () = G () and F (0) = G(0) = 0, it follows that F () = G() for 2 <
< 2 , as desired.
Comment. It is interesting to observe that

log 1 + log sec2
log(1 + tan tan x) dx = ,
0 2
the length of the interval times the average of the integrand values at the
endpoint.
2006:4. Two parabolas have parallel axes and intersect in two points.
Prove that their common chord bisects the segments whose endpoints are the
points of contact of their common tangent.
The solution to this problem appears in Chap. 8.
2007:10. Solve the following dierential equation
2y  = 3|y|1/3
subject to the initial conditions
y(2) = 1 and y(3) = 1.
Your solution should be everywhere dierentiable.
Solution. Depending on the sign of y in any region, separation of vari-
ables leads to the solution
y 2/3 = x + c or y = (x + c)3/2
when y 0 and to
y 2/3 = (x + c) or y = [(x + c)]3/2
when y < 0. The desired solution is


[(x + 1)] ,
3/2
if x < 1;
y(x) = 0, if 1 x 2;


(x 2)3/2 , if x > 2.
[Problem 2007:10 was contributed by Victor Ivrii.]
2009:1. Determine the supremum and the inmum of
(x 1)x1 xx
(x (1/2))2x1
for x > 1.
Solution. Let g(x) be the function in question and let
f (x) = log g(x)
= (x 1) log(x 1) + x log x (2x 1) log((2x 1)/2)
= (x 1) log(x 1) + x log x (2x 1) log(2x 1) + (2x 1) log 2.
82 5. CALCULUS AND ITS APPLICATIONS

Then
f  (x) = log(x 1) + 1 + log x + 1 2 log(2x 1) 2 + 2 log 2
   
4x(x 1) 1
= log = log 1
(2x 1)2 (2x 1)2
< log 1 = 0.
Therefore, f (x), and hence g(x) is a decreasing function on (1, ).

It is straightforward to check that limx1 f (x) = log 2, so that

limx1 g(x) = 2.
To check behaviour for large values of x, let

h(u) = g(u + (1/2))

(u (1/2))u(1/2) (u + (1/2))u+(1/2)
=
u2u
 u  u  1/2  1/2
1 1 1 1
= 1 1+ 1 1+
2u 2u 2u 2u
 u  1/2
1 1 + (1/2u)
= 1 2 .
4u 1 (1/2u)

When v = 1/(2u), the logarithm of the rst factor is

log(1 v 2 )
,
2v
and an application of lHopitals Rule yields that its limit is 0 as v 0. It
follows that
lim h(u) = 1.
u

Therefore, the desired supremum is 2 and inmum is 1.

2012:8. Determine the area of the set of points (x, y) in the plane that
satisfy the two inequalities:

x2 + y 2 2
x4 + x3 y 3 xy + y 4 .

The solution to this problem appears in Chap. 8.

2015:5. Let f (x) be a real polynomial of degree 4 whose graph has two
real inection points. There are three regions bounded by the graph and the
line passing through these inection points. Prove that two of these regions
have equal area and that the area of the third region is equal to the sum of
the other two areas.
 5. CALCULUS AND ITS APPLICATIONS 83

Solution. By scaling and translating, we may assume that the two inec-
tion points are located at x = 1 and x = 1 and that f (x) is monic. Since
f  (x) is a multiple of (x + 1)(x 1) = x2 1, we have that
f (x) = (x4 6x2 + 5) + (bx + c)
where 2b = f (1) f (1) and 2c = f (1) + f (1). The line passing through
the inection points (1, f (1)) and (1, f (1)) is y = g(x) with g(x) = bx + c.
Since f (x) g(x) = (x2 5)(x2 1), the curves with equations y = f (x)
and y = g(x) intersect when x = 5, 1. The three areas in question are
given by
 1   5 

   
 
 (x 6x + 5) dx =  (x4 6x2 + 5) dx
4 2
5 1 
 
 5  16
=  x5 /5 2x3 + 5x 1  =
5
and  1
 1
(x4 6x2 + 5) dx = x5 /5 2x3 + 5x 1 = 32/5.
1
The result follows.
[Problem 2015:5 is #E817 from the American Mathematics Monthly
55:5 (May, 1948), 317; 56:2 (February, 1949), 106108.]
2015:7. Determine
 2
ex dx
.
0 e1x + ex1
Solution 1.
 2   1 2u
ex dx 1
eu du e du
=e =e
0 e1x + ex1 1 e u + eu
1 e 2u + 1
  2 
e 2u
1 e e +1
= log(e + 1) 1 = log
2 2 e2 + 1
e
= log e2 = e.
2
Solution 2. Setting u = ex , we nd that the integral is equal to
 e2  e2
du u du e  e2
=e 2 2
= log(e2 + u2 ) 1
1 (e/u) + (u/e) 1 e +u 2
 2 
e e + e4 e
= log 2
= log e2 = e.
2 e +1 2
Solution 3. We rst establish a general result: Suppose that f is contin-
uous on the interval [a, b]. Then
 b
f (x a) dx 1
= (b a).
a f (x a) + f (b x) 2
84 5. CALCULUS AND ITS APPLICATIONS

Making the substitution x = a + b u, we see that the given integral is

equal to
 b
f (b u) du
.
a f (u a) + f (b u)
b
Adding the two integrals together yields a dx = b a, from which the result
is found.
Apply this result to f (x) = ex , a = 0 and b = 2, to obtain
 2  2
ex dx ex dx
1x + ex1
= e 2x + ex
= e.
0 e 0 e
 CHAPTER 6

Other Topics in Analysis

2001:7. Suppose that x 1 and that x = x+{x}, where x is the greatest
integer not exceeding x and the fractional part {x} satises 0 {x} < 1.
Dene  
x + {x}
f (x) = .
x
(a) Determine the supremum, i.e., the least upper bound, of the values
of f (x) for 1 x.
(b) Let x0 1 be given, and for n 1, dene xn = f (xn1 ). Prove
that limn xn exists.
Solution. (a) Let x = y + z, where y = x and z = {x}. Then

2 yz
f (x)2 = 1 + ,
y+z

which is less than 2 because y = z and yz < 12 (y + z) by the arithmetic-

geometric means inequality. Hence 0 f (x) < 2 for each value of x. Taking
y = 1, we nd that
 
2 2 z
lim f (x) = lim 1 + = 2,
x2 z1 1+z

whence sup{f (x) : x 1} = 2.
(b) In determining the fate of {xn }, note that after the rst entry, the
sequence lies in the interval [1, 2). So, wolog, we may assume that 1 x0 < 2.
If x0 = 1, then each
xn = 1 and the limit is 1. For the rest, note that f (x)
simplies to (1 + x 1)/ x on (1, 2). The key point now is to observe that
there is exactly one value v between 1 and 2 for which f (v) = v.
Let x =1 + u with u > 0. Then it can be checked that f (x) = x if and
only if 1 + 2 u + u = 1 + 3u + 3u2 + u3 or u5 + 6u4 + 13u3 + 12u2 + 4u 4 = 0.
Since the left side is strictly increasing in u, takes the value 4 when u = 0

Springer International Publishing Switzerland 2016 85

E.J. Barbeau, University of Toronto Mathematics Competition
(20012015), Problem Books in Mathematics,
DOI 10.1007/978-3-319-28106-3 6
86 6. OTHER TOPICS IN ANALYSIS

and the value 32 when u = 1, the equation is satised for exactly one value
of u in (0, 1); now let v = 1 + u for this value of u. The value of v turns out
to be approximately 1.375.
Since
1 3/2
f  (x) =x (x 1)1/2 [(2x 1) (x 1)1/2 ] > 0,
2
thefunction f (x) is strictly increasing on (1, 2). Observing that f (5/4) =
3/ 5 > 5/4 and f (25/16) = 7/5 < 25/16, we conclude that, when 1 < x < v,
x < f (x) < f (v) = v, and when v < x < 2, then v = f (v) < f (x) < x. Thus,
when 1 < x0 < v, the iterates {xn } constitute a bounded increasing sequence
whose limit is a xed point of f , and when v < x0 < 2, the iterates {xn }
constitute a bounded decreasing sequence whose limit is a xed point of f .
In either case, limn xn = v.
2002:5. Let n be a positive integer. Suppose that f is a function dened
and continuous on [0, 1] that is dierentiable on (0, 1) and satises f (0) = 0
and f (1) = 1. Prove that, there exist n [distinct] numbers xi (1 i n) in
(0, 1) for which
 n
1
 (x )
= n.
i=1
f i

Solution. Since f (x) is continuous on [0, 1], it assumes every value be-
tween 0 and 1 inclusive. Select points 0 = u0 < u1 < u2 < < un1 <
un = 1 in [0, 1] for which f (ui ) = i/n for 0 i n. Then, by the Mean
Value Theorem, for each i = 1, 2, . . . , n, there exists xi (ui1 , ui ) for which
1 f (ui ) f (ui1 )
= = f  (xi ).
n(ui ui1 ) ui ui1
Therefore,
n
 n
1
= n (ui ui1 ) = n.
i=1
f  (xi ) i=1

2003:5. For x > 0, y > 0, let g(x, y) denote the minimum of the three
quantities, x, y + 1/x and 1/y. Determine the maximum value of g(x, y) and
where this maximum is assumed.
Solution 1. Consider Fig. 6.1 showing the curves of equations x = y+1/x,
x= 1/y and y + 1/x = 1/y, all three of which contain the point P
( 2, 1/ 2), and the regions in which g(x, y) is each one of the three given
functions. In region A, g(x, y) = x. In region B, g(x, y) = 1/y. In region C,
g(x, y) = y + 1/x.

When g(x, y) = x, the maximum value of g(x, y) is equal to 2, assumed
at
the point P . When g(x, y) = 1/y, the maximum value of g(x, y) is equal
to 2, also assumed at P . Suppose (x, y) is such that g(x, y) = y +1/x. Note
 6. OTHER TOPICS IN ANALYSIS 87

that the curve y = 2 (1/x) passes through the point P , where it intersects
each of the curves y +(1/x) = (1/y) and y +(1/x) = x. It intersects neither of
these curves at any other point, and so
lies vertically above the region
where
g(x, y) = y + (1/x). In this region, y 2 (1/x) and so g(x, y) 2, with
equality only at the point
P. Hence the required maximum is 2, and it is
assumed at (x, y) = ( 2, 1/ 2).
Solution 2. [R. Appel] If x 1 and y 1, then g(x, y) x 1. If y 1,
then g(x, y) 1/y 1. It remains to examine the case x > 1 and y < 1, so
that y + (1/x) < 2. Suppose that min (x, 1/y) = a and max (x, 1/y) = b.
Then max (1/x, y) = 1/a and min (1/x, y) = 1/b, so that
1 1 1 a+b
y+ = + = .
x a b ab

y + 1/x = x

y + 1/x = 1/y

A C
x = 1/y
x

Fig. 6.1. In region A, g(x, y) = x. In region B, g(x, y) =

1/y. In region C, g(x, y) = y + 1/x
88 6. OTHER TOPICS IN ANALYSIS

Hence g(x, y) = min (a, (a + b)/(ab)). Either a2 2 or a2 2. But in the

latter case,
a+b 2b
= 2.
ab 2b

In
either
case, g(x, y) 2. This maximum value is attained when (x, y) =
( 2, 1/ 2).
Solution 3. [D. Varodayan] By the continuity of the functions, each of
the regions {(x, y) : 0 < x < y + (1/x), xy < 1}, {(x, y) : 0 < x, y + (1/x) <
x, y + (1/x) < (1/y)}, and {(x, y) : 0 < (1/y) < x, (1/y) < y + (1/x)} is
an open subset of the plane; using partial derivatives, we see that none of
the three functions being minimized have any critical values there. It follows
that any extreme values of g(x, y) must occur on one of the curves dened
by the equations
(6.1) x = y + (1/x)

(6.2) x = 1/y

(6.3) y + (1/x) = (1/y)

On the curve (6.1), x 1 and
 
x
g(x, y) = min x, 2
x 1

x, if x 2;
= x

x2 1 , if x 2.
On the curve (6.2),
g(x, y) = min(x, 2/x)

x, if x 2;
=
2/x, if x 2.
On the curve (6.3), 0 < y < 1 and
 
y 1
g(x, y) = min ,
1 y2 y
 y
2, if 0 < y < 12 ;
= 1y
1/y, if 12 y 1.

On each of these
curves, g(x, y) reaches its maximum value of 2 when
(x, y) = ( 2, 1/ 2).
2003:7. Suppose that the polynomial f (x) of degree n 1 has all real
roots and that > 0. Prove that the set {x R : |f (x)| |f  (x)|} is a
nite union of closed intervals whose total length is equal to 2n.
 6. OTHER TOPICS IN ANALYSIS 89

Solution. Wolog, we may assume that the leading coecient is 1. Let

k k
f (x) = i=1 (x ri )mi , where n = i=1 mi . Then
k
f  (x)  mi
= .
f (x) i=1
x ri
k
Note that the derivative of this function, i=1 mi (x ri )2 is less than 0,
so that it decreases on each interval upon which it is dened. By considering
the graph of f  (x)/f (x), we see that f  (x)/f (x) 1/ on nitely many
intervals of the form (ri , si ], where ri < si and the ri and sj interlace, and
f  (x)/f (x) 1/ on nitely many intervals of the form [ti , ri ), where ti < ri
and the ti and rj interlace. For each i, we have ti < ri < si < ti+1 .

The equation f  (x)/f (x) = 1/ can be rewritten as

k

0 = (x r1 )(x r2 ) (x rk ) mi (x r1 ) (x
ri ) (x rk )
i=1
 k k

 
k
=x ri mi xk1 + .
i=1 i=1

The sum of the solutions of this equation is

s1 + s2 + + sk = r1 + + rk + n,
m
so that i=1 (si ri ) = n. This is the sum of the lengths of the intervals
(ri , si ] on which f  (x)/f (x) 1/. Similarly, we can show that f  (x)/f (x)
1/ on a nite collection of intervals of total length n. The set on which
the inequality of the problem holds is equal to the union of all of these half-
open intervals and the set {r1 , r2 , . . . , rk }. The result follows.
2005:2. Suppose that f is continuously dierentiable on [0, 1] and that
1
0
f (x)dx = 0. Prove that
 1  1  1
2 f (x)2 dx |f  (x)| dx |f (x)| dx.
0 0 0

Solution 1.
 1  1
2 2
f (x) dx = f (x)[2f (x) f (0) f (1)] dx
0 0
 1
= f (x)[(f (x) f (0)) (f (1) f (x))] dx
0
 1  x  1 
= f (x) f  (t)dt f  (t) dt dx
0 0 x
 1  x  1 
 
|f (x)|  f  (t) dt f  (t) dt dx
0 0 x
90 6. OTHER TOPICS IN ANALYSIS

 1  x  1 
|f (x)| |f  (t)| dt + |f  (t)|dt dx
0 0 x
 1  1 

= |f (x)| |f (t)| dt dx
0 0
 1  1
= |f  (x)| dx |f (x)| dx.
0 0
x
Solution 2. Let F (x) = 0 f (t) dt. Then, integrating by parts, we nd
that
 1  1  1
f  (x)F (x) dx = f  (x)F (x) dx.
1
f 2 (x) dx = [f (x)F (x)]0
0 0 0
Hence
 1  1  1 
f 2 (x) dx |f  (x)||F (x)| dx |f  (x)| dx sup{|F (x)| : 0x1}.
0 0 0
Now
 x   1 
   
2|F (x)| = |F (x) F (0)| + |F (1) F (x)| =  f (t) dt +  f (t) dt
0 x
 1  
 
f (t) dt,
0
from which the result follows.
2005:5. Let f (x) be a polynomial with real coecients, evenly many
of which are nonzero, which is palindromic. This means that the
n coecients
read the same in either direction, i.e. ak = ank if f (x) = k=0 ak xk , or
alternatively, f (x) = xn f (1/x), where n is the degree of the polynomial.
Prove that f (x) has at least one root of absolute value 1.
Solution. If n = deg f (x) is odd, then k and n k have opposite parity
and 
f (1) = {ak ((1)k + (1)nk ) : 0 k < n/2} = 0.

Suppose that n = 2m. Then f (x) has 2m + 1 coecients: m pairs

(ak , ank ) of equal coecients (0 k m 1) and am . Since evenly many
coecients are nonzero, we must have that am = 0. Hence
m1

f (x) = ak (xk + x2mk )
k=0
m1

= xm ak (x(mk) + xmk )
k=0
m

= xm amj (xj + xj ).
j=1
 6. OTHER TOPICS IN ANALYSIS 91

m m
Then f (ei ) = 2emi j=1 amj cos j. Let P () = j=1 amj cos j. Then
ei is a root of f if and only if P () = 0.
 2
Now 0 P ()d = 0 and P () is not constant (otherwise f would be
identically 0), so P () is a continuous real function that assumes both positive
and negative values. Hence, by the Intermediate Value Theorem, P must
vanish somewhere on the interval [0, 2], and the result follows.
2005:9. Let S be the set of all real-valued functions that are dened,
positive and twice continuously dierentiable on a neighbourhood of 0. Sup-
pose that a and b are real parameters with ab = 0, b < 0. Dene operators
from S to R as follows:
A(f ) = f (0) + af  (0) + bf  (0);
G(f ) = exp A(log f ).
(a) Prove that A(f ) G(f ) for f S;
(b) Prove that G(f + g) G(f ) + G(g) for f, g S;
(c) Suppose that H is the set of functions in S for which G(f ) f (0).
Give examples of nonconstant functions, one in H and one not in
H. Prove that, if > 0 and f, g H, then f , f + g and f g all
belong to H.
Solution. (a) The result if clear if A(f ) 0. Suppose that A(f ) > 0. We
have that (log f ) = f  /f and (log f ) = (f  /f ) (f  /f )2 , so that
  2
af  (0) bf  (0) f (0)
A(log f ) = log f (0) + + b .
f (0) f (0) f (0)
We have that
 
af  (0) bf  (0)
log A(f ) = log f (0) + log 1 + +
f (0) f (0)
 
af (0) bf (0)
log f (0) + +
f (0) f (0)
  2
f (0)
= A(log f ) + b A(log f ).
f (0)

(b) Note that, for any functions u, v S,

G(uv) = exp A(log u + log v) = exp[A(log u) + A(log v)]
= exp A(log u) exp A(log v) = G(u)G(v).
Similarly G(u/v) = G(u)/G(v). From (a), we have the inequalities
       
f f g g
A G and A G .
f +g f +g f +g f +g
Adding these two inequalities yields
G(f ) + G(g)
1
G(f + g)
92 6. OTHER TOPICS IN ANALYSIS

from which the desired result follows.

(c) Suppose that > 0 and f, g H. Then
G(f ) = G()G(f ) = G(f ) f (0);
G(f g) = G(f )G(g) f (0)g(0) = (f g)(0);
G(f + g) G(f ) + G(g) f (0) + g(0) = (f + g)(0).

Suppose that a = 0, and let f (x) = ekx . Then G(f ) = eka . If ka > 0,
2
then f H; if ka < 0, then f H. Suppose that a = 0, and let f (x) = ekx .
Then G(f ) = e2kb . If k < 0, then f H; if k > 0, then f H.
2006:3. Let p(x) be a polynomial of positive degree n with n distinct
real roots a1 < a2 < < an . Let b be a real number for which 2b < a1 + a2 .
Prove that
2n1 |p(b)| |p (a1 )(b a1 )|.
The solution to this problem appears in Chap. 2.
2006:6. Suppose that k is a positive integer and that
f (t) = a1 e1 t + a2 e2 t + + ak ek t
where a1 , . . . , ak , 1 , . . . , k are real numbers with 1 < 2 < < k . Prove
that f (t) = 0 has nitely many real solutions. What is the maximum number
of solutions possible, as a function of k?

Solution. We prove by induction that the equation can have at most

k 1 solutions. It holds for k = 1. Suppose that it holds up to k, and that
a1 e1 t + + ak+1 ek+1 t vanishes for m distinct values of t. Then
g(t) a1 e(1 k+1 )t + a2 e(2 k+1 )t + + ak e(k k+1 )t + ak+1
has the same roots, so that, by Rolles theorem,
g  (t) a1 (1 k+1 )e(1 k+1 )t + a2 (2 k+1 )e(2 k+1 )t
+ + ak (k k+1 )e(k k+1 )t
has m 1 distinct roots. Hence m 1 k 1 and m k.

For k 2, let pk (x) = (x 1)(x 2)(x 3) (x k 1) a1 + a2 x +

+ ak xk1 . Then fk (t) = et pk (et ) = a1 et + + ak ekt has k 1 roots,
namely ti = log i (1 i k 1). Hence k 1 roots are possible.
2006:8. Let f (x) be a real function dened and twice dierentiable on an
open interval containing [1, 1]. Suppose that 0 < and that |f (x)|
and |f  (x)| for 1 x 1. Prove that

|f  (x)| 2
for 1 x 1. (Part marks are possible for the weaker inequality |f  (x)|
+ .)
 6. OTHER TOPICS IN ANALYSIS 93

Solution. Let |f  (x)| attain its maximum value on [1, 1] at x = u. Then,

for 1 a u b 1, we have, by Taylors theorem, for some v, w in [a, b],

1
f (a) f (u) = (a u)f  (u) + (a u)2 f  (v),
2

1
f (b) f (u) = (b u)f  (u) + (b u)2 f  (w).
2
Therefore
1
(b a)f  (u) = f (b) f (a) + [(a u)2 f  (v) (b u)2 f  (w)]
2
so that
1
(b a)|f  (u)| |f (a)| + |f (b)| + [(b u)2 |f  (w)| + (a u)2 |f  (v)|]
2

2 + [(b u)2 + (a u)2 ] 2 + (b a)2 ,
2 2
since
(b a)2 (b u)2 (a u)2 = 2(b u)(u a) 0.

Select a and b so that b a = 2 /. Then
)  )

|f (u)| () + = 2 .

Comment. If a = 1 and b = 1, then we get in the above that 2|f  (u)|

2 + 2.

2007:4. Suppose that f (x) is a continuous real-valued function dened

on the interval [0, 1] that is twice dierentiable on (0, 1) and satises (i)
f (0) = 0 and (ii) f  (x) > 0 for 0 < x < 1.
(a) Prove that there exists a number a for which 0 < a < 1 and f  (a) <
f (1);
(b) Prove that there exists a unique number b for which a < b < 1 and
f  (a) = f (b)/b.

Solution. (a) By the Mean Value Theorem, there exists c (0, 1) for
which f  (c) = f (1). Since f  (x) is increasing, when 0 < a < c, f  (a) < 1.
(b) For a (0, c), let g(x) = f (x) xf  (a). Then g(0) = 0 and g  (x) =
f (x) f  (a). For 0 < u < a < v < 1, g  (u) < 0 < g  (v) (since f  (x)


increases). Therefore g(a) < 0 and g(1) > 0. Since g  (x) > 0 for a < x < 1,
there is a unique number b for which g(b) = 0, and the result follows.
94 6. OTHER TOPICS IN ANALYSIS

2007:5. For x 1 and x = 0, let

8[1 (1 x)1/2 ]3
f (x) = .
x2
(a) Prove that limx0 f (x) exists. Take this as the value of f (0).
(b) Determine the smallest closed interval that contains the set of all
values assumed by f (x) on its domain.
(c) Prove that f (f (f (x))) = f (x) for all x 1.

Solution. Suppose that x = 1 t2 for t 0 and let g(t) = f (1 t2 ).

Then
8(1 t)3 8(t 1) 8 16
g(t) = = = .
(1 t)2 (1 + t)2 (t + 1)2 (t + 1) (t + 1)2
(a) Since g(1) = 0, its follows that limx0 f (x) = 0.
(b) Since
 2  
4 8 16
0 1 =1 ,
t+1 t + 1 (t + 1)2
g(t) assumes its maximum value of 1 at t = 3. Indeed,
g  (t) = 8(t + 1)2 + 32(t + 1)3 = 8(t + 1)3 (3 t),
so that g(t) increases from 8 at t = 0 to its maximum at t = 3 and then
decreases to the limit 0 as t tends to innity. Therefore the smallest closed
interval containing the image of f is [8, 1]. Observe that this interval gets
mapped onto itself one-to-one, with f (8) = 1 and f (1) = 8.
(c) Let x = 8(t 1)(t + 1)2 = f (1 t2 ). Then
1 x = (t 3)2 (t + 1)2 ,
so that 
3t
1+t , if 0 t 3;
(1 x) 1/2
= t3
1+t , if t > 3.
and 
2(t1)
t+1 , if 0 t 3;
1 (1 x) 1/2
= 4
t+1 , if t > 3.
Thus   
8(t 1) 1 t2 , if 0 t 3;
f (f (1 t )) = f
2
2
= 8(t+1)
(t + 1) (t1)2 , if t > 3.
Thus, f (f (x)) = x when 8 x 1, i.e. when x = 1 t2 for 0 t 3.

Since 8 f (x) 1 for all x 1, it follows that f (f (f (x))) = f (x) for

all x 1.

Comment. It is of slight interest to note that f (3) = f (24) = 8/9.

 6. OTHER TOPICS IN ANALYSIS 95

2008:2.
(a) Determine a real-valued function g dened on the real numbers that
is decreasing and for which g(g(x)) = 2x + 2.
(b) Prove that there is no real-valued function f dened on the real
numbers that is decreasing and for which f (f (x)) = x + 1.
(a) Solution. Let r > 0, r = 1. We determine a decreasing composite
square root of r2 x + r2 . Let g(x) = rx + b. Then g(g(x)) = r2 x + b(1 r).
When b = r2 (1r)1 , g is a decreasing function for which g(g(x)) = r2 x+r2 .
The particular case in the problem can be thus dealt with; an answer is

g(x) = 2x 2( 2 + 1) = 2(x + 2) 2.

(b) Solution 1. [C. Ochanine] Suppose that f (f (x)) = x+1 (f (x) = x+ 12

denes such a function). Then

f (x + 1) = f (f (f (x))) = f (x) + 1

for all x, whence f (x + 1) > f (x). Therefore, f cannot be decreasing.

Solution 2. If there is such a function f , then it must be one-one and
onto R. It must be continuous, since the only discontinuities a monotone
function can have are jump discontinuities. Observe that u <> f (u) if and
only if v = f (u) >< f (f (u)) = f (v). It follows that x f (x) can take both
positive and negative values. By the Intermediate Value Theorem, there is a
number z for which z = f (z). But then z = f (f (z)) = z + 1, a contradiction.
Therefore there is no such function f
Solution 3. Suppose, if possible, there is a decreasing continuous function
f for which f (f (x)) = x + 1. Since x + 1 has no xpoint, neither does the
function f (x). Therefore f (x) x never vanishes. As f is continuous on R,
either f (x) > x for all real x or f (x) < x for all real x.
Suppose the former. Let M exceed the maximum of 0 and f (0). Then
f (0) > f (M ) > M > f (0), a contradiction. Therefore, f (x) > x for all x
cannot occur. Similarly, f (x) < x for all x cannot occur. Therefore, no such
function f exists and x + 1 has no continuous decreasing composite square
root.
2010:7. Suppose that f is a continuous real-valued function dened on
the closed interval [0, 1] and that
 1 2  1   1 
xf (x) dx = f (x) dx x2 f (x) dx .
0 0 0

Prove that there is a point c (0, 1) for which f (c) = 0.

Solution 1. Suppose, if possible, that f never vanishes on the interval,
then it must be everywhere positive or negative. By replacing f by f is
96 6. OTHER TOPICS IN ANALYSIS

necessary, wolog we can assume that f (x) > 0 on [0, 1]. Let f (x) = [g(x)]2
for some positive function g. Then the equation becomes
 1 2  1   1 
xg 2 (x) dx = g 2 (x) dx (xg(x))2 dx .
0 0 0

This is the equality situation in the Cauchy-Schwarz Inequality, whence xg(x)

must be a constant multiple of g(x). But this is not the case. Therefore, by
a contradiction argument, the result follows.
Solution 2. The condition and conclusion is satised by the  1zero function.
Suppose, henceforth, that f is not identically zero and let 0 f (x) dx = u,
1
0
xf (x) dx = uv. If u = 0, then f must assume both positive and negative
values in (0, 1), and so, by the Intermediate Value Theorem, must vanish.
1
Assume u = 0. Then the condition implies that 0 x2 f (x) dx = uv 2 . Then
 1  1  1  1
(x v) f (x) dx =
2
x f (x) dx 2v
2
xf (x) dx + v 2
f (x) dx
0 0 0 0
= uv 2 2uv 2 + uv 2 = 0,
whence (x v)2 f (x), and also f (x) must assume both positive and negative
values on (0, 1). The result follows.
2010:9. Let f be a real-valued function dened on R with a continuous
third derivative, let S0 = {x : f (x) = 0}, and, for k = 1, 2, 3, Sk = {x :
f (k) (x) = 0}, where f (k) denotes the kth derivative of f . Suppose also that
R = S0 S1 S2 S3 . Must f be a polynomial of degree not exceeding 2?
Solution. Observe that, because f and its derivatives are continuous,
each set Sk is closed. Note also that, if U is an open subset upon which f or
either of its rst two derivatives vanish, then all derivatives of higher order
(in particular, f (3) must also vanish on U ).
First, we show that S3 is dense in R. Otherwise, there is an open interval
J upon which f (3) never vanishes, so that J S0 S1 S2 . The set J \ S2
must be a nonvoid open set, and so contain an open interval J1 upon which
f  never vanishes. Similarly, there is a nonvoid open interval J2 J1 upon
which f  never vanishes and a nonvoid open interval J3 J2 upon which f
never vanishes. But then, none of f , f  . f  , f (3) would vanish on J3 contrary
to hypothesis.
Let I be an open real interval and let T0 = I \ S0 . If T0 = , then f , and
so f (3) , must vanish on I. Otherwise, T0 is a nonvoid open set and there is an
open interval I1 T0 I. Let T1 = I1 \ S1 . Then, as before, either f  , and so
f (3) , vanishes on I1 or else there is a nonvoid open interval I2 T1 I1 I.
Let T2 = I2 \ S2 . Either f  , and so f (3) , vanishes on I2 or there is an nonvoid
open subset I3 T2 I2 I. Then either f (3) vanishes on I3 or there is a
nonvoid open interval I4 T3 I3 , where T3 is the open set I3 \ S3 . But this
would contradict the density of S3 in R.
 6. OTHER TOPICS IN ANALYSIS 97

Therefore, f (3) is identically 0 on R, and so therefore, by the Mean Value

Theorem, f  must be constant, f  at most linear and f at most quadratic.
2010:10. Prove that the set Q of rationals can be written as the union
of countably many subsets of Q each of which is dense in the set R of real
numbers.

Solution. Let {r1 , r2 , . . . , ri , . . . } = {2, 3, 5, 6, 7, 10, 11, 12, 13, 14, 15, 17,
. . . } be the increasing sequence of all positive integers that are not the mth
power of any integer for any integer exponent exceeding 1. Let Xk be the set
of all rationals of the form a/rkc where a is an integer coprime with rk and c
is a positive integer. We include also in X1 all integers. Then it can be seen
that every rational lies in one of the Xk and that the Xk are disjoint.
(c+2)
Let k 1 and  > 0; select c such that 1/rc < . Then {(ark +1)/rk :
a Z} is a set of elements of Xk equally spaced at intervals of 1/rkc+1 in R,
so that an element of the set lies inside each real open interval of length .
Hence Xk is dense in R.
2011:1. Let S be a nonvoid set of real numbers with the property that,
for each real number x, there is a unique real number f (x) belonging to S that
is farthest from x, i.e., for each y in S distinct from f (x), |x f (x)| > |x y|.
Prove that S must be a singleton.
Solution 1. Let a be an arbitrary real, and let b be the unique point of
S that is farthest from a. Then every number in S lies on the same side of b
as a. Let c be the unique number in S that is farthest from b. Then S must
be in the closed interval bounded by b and c. It follows that both b and c
are the farthest points in S from 12 (b + c). By the unicity, we must have that
b = c and the result follows.
Solution 2. Since the condition applies to 0 in particular, |x| |f (0)| for
all x S. Therefore S is bounded. Let a = inf S and b = sup S. Then a b
and S [a, b]. Since a x f (a) b, for all x S, it follows that f (a) = b.
Similarly f (b) = a, and both a and b belong to S.
For each x S, |x 12 (a + b)| |a 12 (a + b)| = |b 12 (a + b)|, so that
a = f ( 12 (a + b)) = b. The desired result follows.
[Problem 2011:1 was contributed by Bamdad Yahaghi.]
2012:6. Find all continuous real-valued functions dened on R that
satisfy f (0) = 0 and
f (x) f (y) = (x y)g(x + y)
for some real-valued function g(x).
Solution 1. Setting y = 0, we nd that f (x) = xg(x) for all real x, so
that g(x) is continuous for all nonzero values of x. Therefore
(6.4) (x + y)[f (x) f (y)] = (x y)f (x + y)
98 6. OTHER TOPICS IN ANALYSIS

for all real x, y. Hence, when (x + y)(x y) = 0,

f (x) f (y) f (x + y)
= .
xy x+y
Suppose that x = 0. Then
f (x) f (y) f (2x)
f  (x) = lim = .
yx xy 2x
Taking y = 1 x, we nd that
f (x) f (1 x) = (2x 1)f (1).
Therefore, when x = 0,
f  (x) + f  (1 x) = 2f (1),
so that
f (2x) f (2 2x)
+ = 2f (1) =
2x 2(1 x)
(1 x)f (2x) + xf (2 2x) = 4x(1 x)f (1).
Substituting (2x, 2 2x) for (x, y) in the Eq. (6.4), we nd that
f (2)
f (2x) f (2 2x) = (4x 2) = (2x 1)f (2).
2
Hence
f (2x) = 4x(1 x)f (1) + x(2x 1)f (2)
so that
   
1 f (2) f (2)
f (x) = x(2 x)f (1) + x(x 1)f (2) = f (1) x + 2f (1)
2
x
2 2 2
for x = 0. Since f (x) is continuous, this should hold for all real x, so that
f (x) is a linear combination of the functions x and x2 .
It is straightforward to check that f (x) = x and f (x) = x2 satisfy the
equation (with g(x) respectively equal to 1 and x), along with any linear
combination of these. Hence the general solution is
f (x) = ax2 + bx
where a and b are arbitrary constants.
Solution 2. [J. Zung] Since
1  "x # "x #
g(x) = f +1 f 1 ,
2 2 2
we see that g(x) is a continuous function. As in Solution 1, we show that
f  (x) = g(2x).
We can verify that the vector space generated by the polynomials x
and x2 consists of solutions of the equation. Given a solution, we can add
a polynomial of degree not exceeding 2 to it to get a solution for which
f (1) = f (0) = f (1) = 0, so there is no loss of generality in assuming this
to be the case.
 6. OTHER TOPICS IN ANALYSIS 99

Since f (1) = f (0) = f (1), we have that g(0) = 0, f (x) = xg(x)

and g(x) = g(x) for each x. (For the last, observe that for x = 0,
f (x) f (x) = 0 so that xg(x) = f (x) = f (x) = xg(x).) We have that
f (x) f (1 x) = (2x 1)g(1)
and
f (x) f (1 x) = (2x + 1)g(1) = (2x + 1)g(1),
whence
f (1 x) f (1 x) = 4xg(1).
However, from the given equation, we have also that
f (1 x) f (1 x) = 2g(2x) = 2g(2x),
so that g(2x) = 2xg(1) for all x. Therefore f (x) = g(1)x2 . Since f (1) = 0,
we must have g(1) = 0, so that f (x) 0.
The signicance of this is that the polynomial we add to a solution f of
the equation to get a solution that vanishes at 1, 0, 1 is the negative of the
solution itself, so that the solution must be itself a polynomial of degree 1 or 2.
2012:7. Consider the following problem:
Suppose that f (x) is a continuous real-valued function dened on the
interval [0, 2] for which
 2  2
f (x) dx = (f (x))2 dx.
0 0
Prove that there exists a number c [0, 2] for which either f (c) = 0 or
f (c) = 1.
(a) Criticize the following solution:
2
Clearly 0 f (x) dx 0. By the Extreme Value Theorem, there exist
numbers u and v in [0, 2] for which f (u) f (x) f (v) for 0 x 2. Hence
 2  2  2
f (u) f (x) dx f (x)2 dx f (v) f (x) dx.
0 0 0
2 2
Since 0 f (x)2 dx = 1 0 f (x) dx, by the Intermediate Value Theorem, there
exists a number c [0, 2] for which f (c) = 1. 
(b) Show that there is a nontrivial function f that satises the condi-
tions of the problem but that never assumes the value 1.
(c) Provide a complete solution of the problem.
Solution. (a) The solution does not apply when f (x) is identically zero.
However, there is a more fundamental diculty with it. The solution is
correct for a nonzero function which is nonnegative at every point of the
interval. However, it fails whenever f (x) assumes a negative value. In this
case, there are values of x for which f (u) f (x) < 0, so that f (u)f (x)
f (x)2 and so  
f (u) f (x) dx (f (x))2 dx
N N
100 6. OTHER TOPICS IN ANALYSIS

where N is the subset of [0, 2] upon which f (x) is negative. This puts into
question the displayed inequality in the purported solution.
2 2
(b) We observe that when f (x) = 12 x, then 0 f (x) dx > 0 (f (x))2 dx,
2 2
while when f (x) = 12 (x 2), then 0 f (x) dx < 0 < 0 (f (x))2 dx. This
suggests that we can satisfy the condition of the problem with a function of
the form f (x) = 12 (x ) for some value of in the interval (0, 2). Any such
function will never assume the value 1 on the interval.

We have that
 2
f (x) dx = 1
0

and
 2
1
[3(1 )2 + 1].
(f (x))2 dx =
0 6

These two expressions are equal when = 2/3. Thus, we can take

x 1
f (x) = .
2 6

Comment. J. Love showed that the function

1 3 1
f (x) = x = (2 3x)
2 4 4
satises the condition of the problem but does not assume the value 1.
(c) On the one hand, we can complete the proof in (a) by considering the
possibility that f (x) takes negative values. Since the integral of f (x) over
[0, 2] is nonnegative, f (x) must assume positive values as well. Therefore,
by the Intermediate Value Theorem, it must vanish somewhere within the
interval.
Alternatively, we can get a more direct argument by considering the
function g(x) = f (x)2 f (x). By hypothesis, g(x) is a continuous function
2
for which 0 g(x) dx = 0. Therefore, either g(x) vanishes identically or it
takes both positive and negative values. By the Intermediate Value Theorem,
there exists a number c for which g(c) = 0, whence f (c) is equal to 0 or 1.
One can also argue by contradiction. Suppose that f (x) takes neither of
the values 0 and 1. Then one of the following must occur for each x in the
interval [0, 2]:
Case 1: f (x) < 0, in which case f (x) < f (x)2 ;
Case 2: 0 < f (x) < 1, in which case f (x)2 < f (x);
Case 3: 1 < f (x), in which case f (x) < f (x)2 .
 6. OTHER TOPICS IN ANALYSIS 101

In Cases 1 and 3, the left side of the given equation is less than the
right, while in Case 2, the left side is greater than the right. In all cases, the
equality cannot hold, and we obtain the result.
2013:3. Let f (x) be a convex increasing real-valued function dened
on the closed interval [0, 1] for which f (0) = 0 and f (1) = 1. Suppose that
0 < a < 1 and that b = f (a).
(a) Prove that f is continuous on (0, 1).
(b) Prove that
 1
0ab2 (x f (x)) dx 1 4b(1 a).
0

Solution 1. (a) Let 0 x < y 1. Then y = (1 t)x + t where

t = (y x)/(1 x) and f (y) (1 t)f (x) + tf (1). Therefore
 
yx
0 f (y) f (x) t[f (1) f (x)] = [1 f (x)].
1x
If follows that limyx f (y) = f (x) so that f is right continuous at x. A similar
argument shows that f is left continuous when 0 < x 1. Therefore f is
continuous on [0, 1].
(b) Let 0 < x < 1. Then f (x) (1 x)f (0) + xf (1) = x, so that the
graph of y = f (x) over [0, 1] lies below the graph of y = x. Thus a b.
Let O = (0, 0), I = (1, 1), P = (a, b) and Q = (a, a). The area between
the graphs of y = f (x) and y = x is not less than the combined areas of the
triangles OP Q and IP Q, namely 12 (a b)a + 12 (a b)(1 a) = 12 (a b).
Therefore
 1
1
(x f (x)) dx (a b).
0 2
1
The rightmost inequality is equivalent to 0 f (x) dx 2b(1 a). We rst
show that there is a line of equation y = m(x a) + b passing through (a, b)
that passes under the graph of y = f (x).
Suppose that 0 u < x < v 1. Then (v u)x = (v x)u + (x u)v
and
(v x)f (x) + (x u)f (x) = (v u)f (x) (v x)f (u) + (x u)f (v).
This is equivalent to
f (x) f (u) f (v) f (x)
.
xu vx
Let m be any number between the supremum of the left side over u and the
inmum of the right side over v.
The line y = m(x a) + b contains the points (a b/m, 0) and (1, m(1
a) + b). By the convexity of f , it cannot contain any point of the graph of
102 6. OTHER TOPICS IN ANALYSIS

y = f (x) and so lies underneath the graph. The area under the line is equal
to
    2
1 b 2 m(1 a) + b
1a+ (m(1 a) + b) = 2b(1 a),
2 m m 2
by the arithmetic-geometric means inequality. This gives the desired result.
Solution 2 (to part of b). First, we have that b = f (a) (1 a)f (0) +
af (1) = a, so that a b 0. Making use of the respective substitutions
x = ta and x = (1 t)a + t = a + t(1 a) we nd that
 a  1  1
(x f (x)) dx = a (ta f (ta)) dt a (ta tf (a)) dt
0 0 0
 1
a(a b)
=a t(a b) dt = ,
0 2
and that
 1  1
(x f (x)) dx = (1 a) [(1 t)a + t f ((1 t)a + t)] dt
a 0
 1
(1 a) [(1 t)a + t ((1 t)f (a) + tf (1))] dt
0
 1
(1 a)(a b)
= (1 a) (1 t)(a b) dt = .
0 2
Adding these two inequalities together yields
 1
ab
(x f (x)) dx .
0 2

2013:10. (a) Let f be a real-valued function dened on the real number

eld R for which |f (x) f (y)| < |x y| for any pair (x, y) of distinct elements
of R. Let f (n) denote the nth composite of f dened by f (1) (x) = f (x) and
f (n+1) (x) = f (f (n) (x)) for n 2. Prove that exactly one of the following
situations must occur:
(i) limn+ f (n) (x) = + for each real x;
(ii) limn+ f (n) (x) = for each real x;
(iii) there is a real number z such that
lim f (n) (x) = z
n+

for each real x.

(b) Give examples to show that each of the three cases in (a) can occur.
Solution. (a) Note that the condition on f implies that f is continuous.
Suppose that there exists a real number z for which f (z) = z. Let x = z.
Then |f (x) f (z)| = |f (x) z| < |x z|, so that f (x) = x. Dene xn =
f (n) (x) and let u be the limit of the decreasing sequence {|xn z|}. Note
that either xn z u or xn > z + u for each n, so that as n increases, xn is
 6. OTHER TOPICS IN ANALYSIS 103

either arbitrarily close to z u on the left or arbitrarily close to z + u on the

right.

If limn xn = z u, then

z u = lim f (n+1) (xn ) = lim f (f n (xn )) = f (z u)

n n

from which u = 0 and limn xn = z. Similarly, if limn xn = z + u, then

u = 0 and limn xn = z.

Otherwise, given  > 0, we can nd a positive integer n for which z

u 12  < xn z u and z + u xn+1 < z + u + 12 . Therefore
1
|f (z u) xn+1 | = |f (z u) f (xn )| < |z u xn | < ,
2
from which
1 1
z + u  < xn+1  < f (z u) < xn+1 +  < z + u + .
2 2
Since this holds for all positive , f (z u) = f (z + u). Similarly, f (z + u) =
f (z u). Hence

2u = |(z + u) (z u)| = |f (z u) f (z + u)| = |f (z + u) f (z u),

so that u = 0 and limn xn = z.

Suppose that there is no z for which f (z) = z. Then the function g(x) =
f (x) x is continuous and vanishes nowhere on R. If g(x) > 0 for all x, then
for each x, f (n) (x) is an increasing sequence. It cannot have a nite limit,
since this limit would be a xpoint of f . Therefore the sequence must diverge
to innity. The case that g(x) < 0 for all x can be similarly handled.
(b) (i) Let f (x) = (1 + x2 )1/2 . Then f (x) > x and so f (n) (x) is a strictly
increasing function of n. Since f  (x) = x(1 + x2 )1/2 , we nd that, given
x, y, by the Mean Value Theorem, there exists a number z for which

|f (x) f (y)| = |f  (z)||x y| < |x y|.

(ii) Let f (x) = (1 + x2 )1/2 .

(iii) Let f (x) = 12 x.
2014:4. Let f (x) be a continuous real-valued function on [0, 1] for which
 1  1
f (x)dx = 0 and xf (x) dx = 1.
0 0

(a) Give an example of such a function.

(b) Prove that there is a nontrivial open interval I contained in (0, 1)
for which |f (x)| > 4 for x I.
104 6. OTHER TOPICS IN ANALYSIS

Solution. (a) Trying a function of the form c(x 12 ) yields the examples
f (x) = 12x 6. Other examples are f (x) = 21 2 cos(x) and f (x) =
2 sin(2x).
(b) If there existed no such interval I, then |f (x)| 4 for all x [0, 1].
Then
 1  12  1
1= f (x)(x(1/2)) dx= (f (x))((1/2)x) dx+ f (x)(x(1/2)) dx
1
0 0 2
 1
2
 1
<4 ((1/2) x)dx + 4 (x (1/2)) dx = 4 (1/8) + 4 (1/8) = 1.
1
0 2

Note that the inequality is strict. Equality would require that f (x) = 4 on
(0, 12 ) and f (x) = 4 on ( 12 , 1), an impossibility for a continuous function.
Comment. B. Yaghani draws attention to Problem 25 of Chap. 11
appearing on page 184 of Michael Spivak, Calculus (W.A. Benjamin, 1967),
where it is required to show that, if g is twice dierentiable on [0, 1] with
g(0) = g  (0) = g  (1) = 0 and g(1) = 1, then |g  (c)| 4 for some c
t t
[0, 1]. Setting g(t) = 0 (x t)f (x)dx, whereupon g  (t) = 0 f (x)dx and
g  (t) = f (t), yields the desired result.
[Problem 2014:4 is based on #4520 in the American Mathematical
Monthly 60:1 (January, 1953), 48; 61:4 (April, 1954), 266268.]
2014:10. Does there exist a continuous real-valued function dened on
R for which f (f (x)) = x for all x R?
Solution 1. The answer is no. Since f (f (x)) = x denes a one-one
function, it follows that f (x) itself is one-one. Therefore f (x) is either strictly
increasing or strictly decreasing on R. But this is impossible, since in either
case, f (f (x)) would be increasing.
Solution 2. The answer is no. Suppose that f is such a function and
that f (0) = u. Then f (u) = f (f (0)) = 0 and so u = f (0) = f (f (u)) = u.
Hence u = 0 and f (0) = 0.
Suppose that f (v) = 0. Then f (0) = f (f (v)) = v. Therefore f (v) = 0
if and only if v = 0.
Let a = 0 and b = f (a). Then f (b) = a and f (a) = b. Consider
the four numbers a, b, a, b repeated cyclically. The alternate ones have
opposite signs, so there must be two numbers of the same sign followed by
one of the opposite sign. With no loss of generality, we can assume that a
and b are positive while a and b are negative. Then f (a) and f (b) have
opposite signs, so by the Intermediate Value Theorem, there must exist c
between a and b for which f (c) = 0. But this is a contradiction. Therefore,
there is no continuous function with the desired property.
2015:1. Suppose that u and v are two real-valued functions dened on
the set of reals. Let f (x) = u(v(x)) and g(x) = u(v(x)) for each real x. If
f (x) is continuous, must g(x) also be continuous?
 6. OTHER TOPICS IN ANALYSIS 105

Solution 1. The answer is no. Let


0, when x 0;
v(x) =
1, when x > 0.

Suppose that u is any function for which u(1) = 1 and u(0) = u(1) = 0.
Then f (x) = 0 for all real x while

0, when x 0;
g(x) =
1, when x > 0.

Solution 2. (Xuan Tu) An example for which g(x) fails to be continuous

at 1 is (u(x), v(x)) = ((1 + x)1 , x2 ).
2015:3. Let f : [0, 1] R be continuously dierentiable. Prove that
  1 
 f (0) + f (1)  1
 f (x) dx  sup{|f  (x)| : 0 x 1}.
 2  4
0

Solution 1. Integrating by parts, we nd that

 1    1  1
1 1
x f  (x) dx = x f (x) f (x)dx
0 2 2 0 0

 1
f (1) + f (0)
= f (x) dx.
2 0

Since
     1 
 1
1   1  1
 x f  (x) dx sup |f  (x)| x dx = sup |f  (x)|,
 2  2  4
0 0

the desired result follows.

Solution 2. Note that
 1  1
f (1) f (0) f (0) + f (1)
(f (x) f (0)) dx = f (x) dx.
2 0 2 0

Because of this relation and because f (t) f (0) and f (t) have the same
derivative, there is no loss in generality in assuming that f (0) = 0. Then
 1  1  1  x
f (1) 1
f (x) dx = f  (x) dx f  (y) dy dx
2 0 2 0 0 0
 1  1  1
1  
= f (y) dy f (y) dx dy
2 0 0 y
106 6. OTHER TOPICS IN ANALYSIS

  1
1 1 
= f (y) dy f  (y)(1 y) dy
2 0 0
 1 
1 
= y f (y) dy
0 2
and we can conclude as in Solution 1.
Comment. A weaker result can be obtained by noting that the  1 Integral
Mean Value Theorem provides a value c [0, 1] for which f (c) = 0 f (x)dx.
The left side is thus equal to
1 1
|(f (1) f (c)) + (f (0) f (c))| (|f (1) f (c)| + |f (c) f (0)|)
2 2
   c #
1 "  1   
+ 


=  f (x) dx   f (x) dx 
2 c 0
 1
1
|f  (x)| dx sup |f  (x)|.
0 2
[Problem 2015:3 was contributed by Omran Kouba, Higher Institute for
Applied Sciences and Technology, Damascus, Syria.]
 CHAPTER 7

Linear Algebra

2001:4. Let V be the vector space of all continuous real-valued functions

dened on the open interval (/2, /2), with the sum of two functions and
the product of a function and a real scalar dened in the usual way.
(a) Prove that the set {sin x, cos x, tan x, sec x} is linearly independent.
(b) Let W be the linear space generated by the four trigonometric
functions given in (a), and let T be the linear transformation det-
ermined on W into V by T (sin x) = sin2 x, T (cos x) = cos2 x,
T (tan x) = tan2 x and T (sec x) = sec2 x. Determine a basis for the
kernel of T .
Solution 1.
(a) Suppose that a sin x + b cos x + c tan x + d sec x = 0 identically for
all x ( 2 , 2 ). Then evaluating this equation at 0, /6, /4, /3 leads to
a system of four linear equations in a, b, c, d whose sole solution is given by
a = b = c = d = 0.
(b) Since sin2 x + cos2 x = 1 and 1 + tan2 x = sec2 x, it is clear from
the linearity of T that T (sin x + cos x + tan x sec x) = 0. On the other
hand, by evaluating at three of the values of x used in (a), we can see that
{sin2 x, cos2 x, tan2 x} is linearly independent. Thus the nullity of T is at
least 1 and the rank of T is at least 3. Hence the kernel of T has dimension
1 and must be the span of {sin x + cos x + tan x sec x}.
Solution 2.
(a)
a sin x + b cos x + c tan x + d sec x = 0 (x)
a sin x cos x + b cos2 x + c sin x + d = 0 (x)
sin x(a cos x + c) = (b cos2 x + d) (x)
(1 cos2 x)(a2 cos2 x + 2ac cos x + c2 ) = b2 cos4 x + 2bd cos2 x + d2 (x)
(b2 + a2 ) cos4 x + 2ac cos3 x + (c2 a2 + 2bd) cos2 x 2ac cos x + (d2 c2 ) = 0 (x)
a = b = c = d = 0,

Springer International Publishing Switzerland 2016 107

E.J. Barbeau, University of Toronto Mathematics Competition
(20012015), Problem Books in Mathematics,
DOI 10.1007/978-3-319-28106-3 7
108 7. LINEAR ALGEBRA

from which the linear independence follows. (Note that the left expression is
a polynomial in cos x which vanishes at innitely many values.)
(b) a sin x + b cos x + c tan x + d sec x belongs to the kernel of T
a sin2 x + b cos2 x + c tan2 x + d sec2 x = 0 (x)
a + (b a) cos2 x + (c + d) sec2 x c = 0 (x)
(a c) + (b a) cos2 x + (c + d) sec2 x = 0 (x)
(b a) cos4 x + (a c) cos2 x + (c + d) = 0 (x)
b = a = c = d,
which occurs if and only if a sin x + b cos x + c tan x + d sec x is a multiple of
sin x + cos x + tan x sec x. Hence the nullity of T is 1 and the rank of T is 3.
A basis for the kernel of T consists of the function sinx + cos x + tan x sec x.
2002:7. Prove that no vector space over R is a nite union of proper
subspaces.
Solution. Suppose that the statement is false. Let r be the minimum
number of proper subspaces whose union is the whole of V . Suppose that
S1 , S2 , . . . , Sr are subspaces for which V = ri=1 Sr . Note that r 2. Since
r is minimal, no subspace is contained in a union of the others. Select v
S1 \ ri=2 Si and w S2 \S1 . For each real , v + w S1 . By the pigeonhole
principle, there is an index j and two distinct reals and for which v + w
and v + w both belong to Sj . Hence, ( )v Sj , so that v Sj . But
this contradicts the choice of v.
[Problem 2002:7 is #10707 in the American Mathematical Monthly 106:1
(January, 1999), 67; 107:10 (December, 2000), 951952.]
2002:8. [(a) Suppose that P is an n n nonsingular matrix and that u
and v are column vectors with n components. The matrix v t P 1 u is 1 1,
and so can be identied with a scalar. Suppose that its value is not equal
to 1. Prove that the matrix P + uv t is nonsingular and that
1
(P + uv t )1 = P 1 P 1 uv t P 1

where v t denotes the transpose of v and = 1 + v t P 1 u. (b) Explain the
situation when = 0.
Solution 1. (a) Observe that uv t is an n n matrix and that for each
column vector w, v t w is a 1 1 matrix or scalar. Thus, (uv t )w = u(v t w) =
(v t w)u, and so uv t is a rank 1 matrix, whose range is spanned by the vector
u. It follows that the matrix uv t P 1 is also rank 1 whose range is the span
of u. Note that, for every real scalar ,
(uv t P 1 )u = u(v t P 1 u) = (v t P 1 u)u
since v t P 1 u, being a 1 1 matrix, is essentially a scalar. Thus, on the span
of u, (uv t P 1 ) behaves like (v t P 1 u)I, a multiple of the identity. It follows
that
(v t P 1 uI uv t P 1 )(uv t P 1 ) = O.
 7. LINEAR ALGEBRA 109

Now we are ready to establish the result.

1 1 t 1
(P + uv t )(P 1 P uv P )

1 1
=I + uv t P 1 uv t P 1 uv t P 1 uv t P 1

1
=I + [( 1)uv P uv P uv t P 1 ]
t 1 t 1

1
=I + [(v t P 1 u)uv t P 1 uv t P 1 uv t P 1 ]

1
=I + [(v t P 1 uI uv t P 1 )(uv t P 1 )] = I,

as desired.
(b) Suppose that v t P 1 u = 1. Then v t P 1 uv t = v t = v t P 1 P , so
that v t P 1 (uv t + P ) = O. Since v t P 1 = O, we cannot have an inverse for
uv t + P .
Solution 2. This is similar to Solution 1, with the inverse checked on the
right instead of the left. Let z = P 1 u. Since uv t is a rank 1 matrix whose
range is spanned by u (see Solution 1), P 1 uv t is a rank 1 matrix whose
range is spanned by z. We have that
(P 1 uv t )z = (P 1 uv t P 1 )u = (P 1 u)(v t P 1 u)
= (v t P 1 u)(P 1 u) = (v t P 1 u)z

so that, on the span of z, P 1 uv t behaves like (v t P 1 u)I. Thus

(v t P 1 uI P 1 uv t )(P 1 uv t ) = 0.

Note that
1 1 t 1
(P 1 P uv P )(P + uv t )

1 1 t 1 1 t 1 t
= I + P 1 uv t P uv P uv P uv

1
=I+ [( 1)P 1 uv t P 1 uv t P 1 uv t ]

1
= I + [(v t P 1 u)P 1 uv t P 1 uv t P 1 uv t ]

1
= I + [(v t P 1 uI P 1 uv t )P 1 uv t ]

from which the result follows.

Comment. This is known as the Sherman-Morrison Formula (see

G.H.Golub, C.F.Van Loon, Matrix computation, third edition, 1996, page 51).
110 7. LINEAR ALGEBRA

2003:8. Three m n matrices A, B and A + B have rank 1. Prove that

either all the rows of A and B are multiples of one and the same vector, or
that all of the columns of A and B are multiples of one and the same vector.

Solution 1. Let A = (aij ) and B = (bij ). Since the image each of the
matrix A and the matrix B as an operator on Fn (where F is the underlying
eld) is the span of its column vectors, there exist nonzero vectors (ai ), (bi ),
(uj ), (vj ) for which
aij = ai uj and bij = bi vj .
Hence aij + bij = ai uj + bi vj . Similarly, since A + B = (aij + bij ) has rank
1, there are vectors (ci ) and (wj ) for which
aij + bij = ci wj .
Thus, for all i and j,
a i u j + bi v j = c i w j .
If the rows of A and B are multiples of the same row, then there exists
a constant k for which vj = kuj for each j, and so ci wj = (ai + kbi )uj , and
so the ith row of A + B is also a multiple of (uj ).
Suppose that the vectors (u1 , . . . , uj , . . . , un ) and (v1 , . . . , vj , . . . , vn ) are
linearly independent. Wolog, we may assume that (u1 , u2 ) and (v1 , v2 ) are
linearly independent. Then u1 v2 u2 v1 = 0, and for each i, the system
u 1 a i + v 1 bi = w 1 c i
u 2 a i + v 2 bi = w 2 c i
has the solution ai = pci , bi = qci , where
w1 v 2 w2 v 1
p=
u1 v 2 u2 v 1
and
u 1 w2 u 2 w1
q=
u1 v 2 u2 v 1
are independent of i. The result follows.
Solution 2. Suppose that A = uat and B = vbt where a = (aj ), b = (bj ),
u = (ui ) and v = (vi ) are column vectors. If the set {u, v} is linearly
dependent, then the columns of A + B, A and B are multiplies of each other
and the result follows. Otherwise, let {u, v} be linearly independent. The
ith columns of A and B are respectively ai u and bi v. Since A + B is of rank
1, there exists a column vector w and scalars ci for which
a i u + bi v = c i w
 7. LINEAR ALGEBRA 111

for each i. If ci = 0, then ai = bi = 0. Since not all ci vanish, we may

suppose wolog that c1 = 0 and that a1 = 0.
For each i, j,
ci (aj u + bj v) cj (ai u + bi v) = ci (cj w) cj (ci w) = 0
= (ci aj cj ai )u + (ci bj cj bi )v = 0
= ci aj cj ai = ci bj cj bi = 0
= ci aj = cj ai and ci bj = cj bi
= ci cj (aj bi ai bj ) = 0.
If ci = 0, then taking j = 1, we nd that bi = (ai /a1 )b1 as well as ai =
(ai /a1 )a1 . Since this is true for i = 1 and when ci = 0, we conclude that the
vectors b and a are parallel, and the rows of A + B are sums of multiples of
these parallel vectors. The result follows.
2004:5. Let A be a n n matrix with determinant equal to 1. Let
B be the matrix obtained by adding 1 to every entry of A. Prove that the
determinant of B is equal to 1+s, where s is the sum of the n2 entries of A1 .

Solution 1. First, we make a general observation. Let U = (uij ) and

V = (vij ) be two n n matrices. Then det (U + V ) is the sum of the
determinants of 2n n n matrices WS , where S is a subset of {1, 2, . . . n} and
the (i, j)th element of WS is equal to uij when i S and vij when i S. In
the special case that U = A and V = E, the matrix whose every entry is equal
to 1, WS is equal to 0 except when S = {1, 2, . . . , n} or S = {1, 2, . . . , n}\{k}
for some integer k. In the former case, det WS = det A = 1, and in the latter
case, all rows of S except the kth agree with the corresponding rows of A and
the kth row of WS consists solely of 1s, so that the determinant of S is equal
to Ak,1 + Ak,2 + + Ak,n , where Aij is the cofactor of aij in the expansion
of det A. Thus,
n
 n 
  
det(A + E) = det A + Akl = 1 + Aij ,
k=1 l=1 i,j

as desired.
n
Solution 2. [R. Barrington Leigh] Let A1 = (cij ) and dj = i=1 cij , the
jth column sum of A1 for 1 j n. Let E be the n n matrix all of whose
entries are 1, so that B = A + E. Observe that E = (1, 1, . . . , 1)t (1, 1, . . . , 1),
where t denotes the transpose. Then
det B = (det B)(det A)1 = (det B)(det A1 ) = det(BA1 )
= det[(A + E)A1 ] = det[I + (1, 1, . . . , 1)t (1, 1, . . . , 1)(cij )]
= det[I + (1, 1, . . . , 1)t (d1 , d2 , . . . , dn )].
112 7. LINEAR ALGEBRA

Thus, we need to calculate the determinant of the matrix

1 + d1 d2 dn
d1 1 + d2 dn
.

d1 d2 1 + dn
This is equal to the determinant of the matrix

1 + d1 d 2 dn
1 1 0
.

1 0 1
which in turn is equal to the determinant of a matrix whose top row is
(1 + d1 + d2 + + dn , 0, 0, . . . , 0) and for which the cofactor of the top left
element is the (n 1) (n 1) identity matrix. The result follows.
Solution 3. [M.-D. Choi] Lemma: If C is a rank 1 n n matrix, then the
determinant of I + C is equal to 1 + trace C.
Proof of Lemma. The sum of the eigenvalues of C is equal to the trace
of C (sum of the diagonal elements). Since 0 is an eigenvalue of algebraic
multiplicity n 1, the remaining eigenvalue is equal to the trace of C. The
eigenvalues of I + C are 1 with algebraic multiplicity n 1 and 1 + trace C,
so that the determinant of I + C, which is the product of its eigenvalues, is
equal to 1 + trace C. 
In the problem, with E dened as in the previous solutions,
det(A + E) = det(A) det(I + A1 E) = det(I + A1 E).
Since E is of rank 1, so is A1 E. The diagonal elements of A1 E are the row
sums of A1 and so the trace of A1 E is equal to the sum of all the elements
of A1 . The desired result follows.
2004:8. Let V be a complex n-dimensional inner product space. Prove
that
1
|u|2 |v|2 |u v|2 |u + v|2 |(u, v)|2 |u|2 |v|2 .
4
Solution. The right inequality is the Cauchy-Schwarz Inequality. We
have that
4|(u, v)|2 4|u|2 |v|2 + |u v|2 |u + v|2
= 4(u, v)(v, u) 4(u, u)(v, v) + (u v, u v)(u + v, u + v)
= 4(u, v)(v, u) 4(u, u)(v, v) + [(u, u) + (v, v) (u, v) + (v, u)]
[(u, u) + (v, v) + (u, v) + (v, u)]
= 4(u, v)(v, u) 4(u, u)(v, v) + [(u, u)2 + (v, v)2 + 2(u, u)(v, v)]
[(u, v)2 + 2(u, v)(v, u) + (v, u)2 ]
 7. LINEAR ALGEBRA 113

= [(u, u) (v, v)]2 [(u, v) (v, u)]2

= [|u|2 |v|2 ]2 [2i Im (u, v)]2
= [|u|2 |v|2 ]2 + 4[Im (u, v)]2 0.

[Problem 2004:8 was posed by Peter Rosenthal in a problems newsletter

in the University of Toronto Mathematics Department in 1967.]
2005:3. How many n n invertible matrices A are there for which all
the entries of both A and A1 are either 0 or 1?
Solution 1. Let A = (aij ) and A1 = (bij ). Since A and A1 are both
invertible, there is at least one 1 in each row and in each
n column of either of
these matrices. Suppose that 1 k n. Since 1 = j=1 akj bjk , there is a
n
unique index m for which akm = bmk = 1. If l = k, then 0 = j=1 alj bjk =
n
j=1 akj bjl , so that alm = bml = 0. Thus, the mapping k m is one-one
and each row and each column of each matrix contains exactly one 1. Thus
A is a permutation matrix (as is A1 ), and there are n! possibilities.
Solution 2. Let e = (1, 1, 1, . . . , 1)t . Then each component of Ae is a
integer. The kth component of e is the kth component of A1 (Ae),
positive
n
namely j=1 bkj (Ae)j , where bij is the (i, j)th component of A1 . Thus
n

1= bkj (Ae)j
j=1

so that there is at most one value of i for which bki = 0 and bki = (Ae)i = 1.
Since A1 is invertible, there is at least one 1 in each row, so there must be
exactly one in each row. Similarly, there is at least one 1 in each column, so
that A1 must be a permutation matrix, as must be A. Therefore, there are
exactly n! possibilities.
1
Solution 3. [E. Redelmeier] n Let A = (aij ) and A = (bij ). Consider
any value of i. Since 1 = j=1 aij bji , there is a unique value of j, say
j = (i)
n for which a ij = b ji = 1. Suppose, for i = l, that (i) = (l). Then
0 = j=1 aij bjl ai(i) b(l)l = 1, a contradiction. Hence is one-one and
so a permutation.
Suppose, for k = (i), aik = ai(i) . Then k = (l), for some index l
n
distinct from i and so j=1 aij bjl aik bkl ai(i) b(l)l = 1, a contradiction.
Hence each row contains exactly one 1 and, moreover, A is a permutation
matrix. So also is A1 . Conversely, each permutation matrix has the desired
property, and so the number of matrices is n!.
2005:4. Let a be a nonzero real and u and v be real 3-vectors. Solve
the equation
2ax + (v x) + u = O
for the vector x.
114 7. LINEAR ALGEBRA

Solution 1. If v = 0, then x = (2a)1 u. Let v = 0. Take the dot and

cross products by v to get
2a(v x) + v u = 0
and
2a(v x) + v (v x) + v u = O.
The second equation can be manipulated to
2a(v x) + (v x)v (v v)x + (v u) = O
whence
1
vx= [(v x)v (v v)x + (v u)]
2a
1 1 1
= 2 (v u)v + (v v)x (v u).
4a 2a 2a

Therefore
1 1 1
2ax + u = (v u)v (v v)x + (v u)
4a2 2a 2a
which can be rearranged to give
[8a3 + 2a(v v)]x = (v u)v + 2a(v u) 4a2 u.
Observe that, since v v > 0, the coecient of x is nonzero and the same
sign as a.
Solution 2. Suppose that v = u for some real . Then, since v x is
orthogonal to both v and x, we must have that x = (2a)1 u.
Suppose that v is not a multiple of u. Then we can write u = v + w
where w = O and v w. Then {v, w, v w} is an orthogonal basis for
3-space. Dene the operator
T (x) = 2ax + (v x).
Then T (v) = 2av, T (w) = 2aw + (v w), T (v w) = 2a(v w) + v
(v w) = 2a(v w) |v|2 w. Then, if x = xv + yw + z(v w), the given
equation is rendered

2a 0 0 x
0 2a |v|2 y = 1 .
0 1 2a z 0
Hence, the components of x are given by
2a 1
x= y= 2 z= .
2a 4a + |v|2 4a2 + |v|2
Comment. Taking u = v + w, we can rewrite the result in Solution 1 as
2a(4a2 + |v|2 )x = |v|2 v + 2a(v w) 4a2 v 4a2 w
= (4a2 + |v|2 )v (2a)2 w + 2a(v w),
which agrees with the result in Solution 2.
 7. LINEAR ALGEBRA 115

Solution 3. Denote the three components of the vectors by subscripts in

the usual way. Since v x = (v2 x3 v3 x2 , v3 x1 v1 x3 , v1 x2 v2 x1 ), the
given equation is equivalent to the system of scalar equations:
2ax1 v3 x2 + v2 x3 = u1
v3 x1 + 2ax2 v1 x3 = u2
v2 x1 + v1 x2 + 2ax3 = u3 .
The determinant D of the coecients is equal to 2a(4a2 + |v|2 ), and so is
nonzero. By Cramers Rule, we obtain the solution:
Dx1 = (u1 v1 + u2 v2 + u3 v3 )v1 + 2a(v2 u3 v3 u2 ) 4a2 u1 ,
Dx2 = (u1 v1 + u2 v2 + u3 v3 )v2 + 2a(v3 u1 v1 u3 ) 4a2 u2 ,
Dx3 = (u1 v1 + u2 v2 + u3 v3 )v3 + 2a(v1 u2 v2 u1 ) 4a2 u3 .

2005:7. Let f (x) be a nonconstant polynomial that takes only integer

values when x is an integer, and let P be the set of all primes that divide
f (m) for at least one integer m. Prove that P is an innite set.
The solution to this problem appears in Chap. 11.
2005:8. Let AX = B represent a system of m linear equations in n
unknowns, where A = (aij ) is an m n matrix, X = (x1 , . . . , xn )t is an n 1
vector and B = (b1 , . . . , bm )t is an m 1 vector. Suppose that there exists
at least one solution for AX = B. Given 1 j n, prove that the value of
the jth component is the same for every solution X of AX = B if and only
if the rank of A is decreased if the jth column of A is removed.
Solution 1. Let C1 , C2 , . . . , Cn be the n columns of A. Suppose that the
dimension of the span of these columns is reduced when Cj is removed. Then
Cj is not a linear combination of the other Ci . Thus, if AX = AY = B and
Z = X Y , then AZ = O. This means that z1 C1 + z2 C2 + + zj Cj + +
zn Cn = O. Because Cj is not a linear combination of the other Ci , we must
have that zj = 0, i.e. that xj = yj .
On the other hand, if the dimension of the span is the same whether or
not Cj is present, then Cj must be a linear combination of the remaining
Ci . In this case, there is a vector Z with a nonzero value of zj for which
AZ = z1 C1 + + zj Cj + + zn Cn = O. Thus if X is a solution of
AX = B, then X + Z is a second solution with a dierent value of the jth
component.
Solution 2. As in the rst solution, we observe that every solution of
AX = B has the same jth entry if and only if every solution of AX = O has
zero in the jth position.
Suppose that every solution of AX = O has 0 in the jth position. Let Aj
be the matrix A with the jth column removed and Xj the vector X with the
jth entry removed. Then there is a linear isomorphism between the solutions
116 7. LINEAR ALGEBRA

of AX = 0 and Aj Xj = 0, where X Xj is dened by suppressing the jth

entry. Since the target space of Aj has dimension n 1. we nd that
n rank A = nullity A = nullity Aj = (n 1) rank Aj ,
whence rank Aj = rank A 1.
On the other hand, suppose that there are solutions of AX = O for
which the jth component fails to vanish. Each solution of Aj Y = 0 lifts to
a solution of AX = 0 by inserting the entry 0 after the (j 1)th position of
Y . However, this lifting does not get all the solutions, so that
n rankA = nullityA > nullityAj = (n 1) rankAj
whence rankAj > rankA 1. Since, always, the rank of Aj does not exceed
that of A, we must have that the ranks of Aj and A are equal. The desired
result follows.
2006:2. Let u be a unit vector in R3 and dene the operator P by
P (x) = u x for x R3 (where denotes the cross product).
(a) Describe the operator I + P 2 .
(b) Describe the action of the operator I + (sin )P + (1 cos )P 2 .
Solution 1. (a) Recall that a(bc) = (ac)b(ab)c for any 3-vectors
a, b, c.
We have that
P 2 (x) = u (u x) = (u x)u (u u)x = (u x)u x
from which we see that (I + P 2 )(x) = (u x)u. Thus, I + P 2 is an orthogonal
projection onto the one-dimensional spaces spanned by {u}.
(b) Let R = I + (sin )P + (1 cos )P 2 . Observe that P (u) = O. Then
R(u) = u, so that the one-dimensional space spanned by {u} is xed under
R. Let v be a unit vector orthogonal to u and let w = u v. Then {u, v, w}
is an orthonormal basis for R3 . We have that
R(v) = v + (sin )w + (1 cos )(v) = (cos )v + (sin )w
and
R(w) = w + (sin )(u w) + (1 cos )(w) = ( sin )v + (cos )w.
It follows that the plane through O orthogonal to u is rotated about the axis
u through an angle ; by linearity, the same holds for all of R3 .
Solution 2. Wolog, we may assume that u = (1, 0, 0). Let x = (x, y, z).
Then u x = (0, z, y) and u (u x) = (0, y, z). Hence (I + P 2 )x =
(x, 0, 0) and

(I + (sin )P + (1 cos )P 2 )x = (x, 0, 0) + (0, (cos )y (sin )z, (sin )y + (cos )z).

Thus, I +P 2 is an orthogonal projection onto u and I +(sin )P +(1cos )P 2

is a rotation about the axis u through the angle .
 7. LINEAR ALGEBRA 117

Solution 3. Let v be the image of u after a 90 rotation and let w = uv.

Then u, v, w is an orthonormal basis of R3 and u w = v. Hence, if
x = xu + yv + zw, then we nd that (I + P 2 )x = xu = (x u)u and (I +
(sin )P +(1cos )P 2 )(x) = xu+[(cos )y(sin )z]v+[(cos )z+(sin )y]w,
and we reach the same conclusion as in the previous solutions.
2006:7. Let A be a real 3 3 invertible matrix for which the sums of the
rows, columns and two diagonals are all equal. Prove that the rows, columns
and diagonal sums of A1 are all equal.
Solution 1. Let r be the common sum and let e = (1, 1, 1)t . Then
Ae = re. Since A is invertible, r = 0 and A1 e = r1 e. The other eigenvalues
(counting repetitions) of A are u and u, since the trace of A is equal to r.
Since A is invertible, u = 0.
The eigenvalues of A1 are r1 , u1 and u1 , and the trace of A1 is
equal to r1 + u1 u1 = r1 . Since also A1 e = r1 e, the row and main
diagonal sums are r1 .
Since (A1 )t = (At )1 , we can apply the same reasoning to At to nd
that the row sums of (A1 )t and, hence, the column sums of A1 are equal
to r1 .
Finally, let B = P A, where

0 0 1
P = 0 1 0 .
1 0 0
Then B 1 = A1 P and we can apply the result to B, which has the same
row, column and diagonal sums as A. The main diagonals of B and B 1 are
respectively equal to the cross diagonals of A and A1 , so we nd that the
cross diagonal of A1 is equal to r1 . The desired result follows.
Solution 2. Since the common sum of the rows, columns and diagonals
of a magic square is three times the central entry, the general form of the
matrix A is
a 3c a b b
c + b a c c + a b
2c b a+bc 2c a
The adjugate, adj A of this matrix is

3c2 a2 + b2 ac 2bc b2 a2 6c2 + 5ac + bc 3c2 + b2 a2 + 2ac 5bc
5ac 5bc + b2 a2 b2 a2 + 2ac 2bc b2 a2 ac + bc .
b2 a2 3c2 + 2ac + bc 6c + b2 a2 ac 5bc
2 2 2 2
b a 3c + 5ac 2bc
The sum of each row, column and diagonal of adj A is 3(b a)(b + a 2c).
Since (det A)A1 = adj A, the desired result follows.

Comment. A direct solution for the rows and columns can be obtained
from the relationships between the elements of A and A1 . Let A = (aij )
and A1 = (bij ). Then, for 1 i, k 3,
118 7. LINEAR ALGEBRA


3
aij bjk = ik ,
j=1

where ij is the Kronecker delta. Let the common sums for A be s. Then


3 
3  3 
3  3 
3
s bjk = ( aij )bjk = ( aij bjk ) = ik = 1,
j=1 j=1 i=1 i=1 j=1 i=1

3
whence s = 0 and j=1 bjk = 1/s. Hence, each column sum of A1 is 1/s.
A similar argument holds for the row sums.

2007:7. Find the Jordan canonical form of the matrix uvt where u and
v are column vectors in Cn .

Solution. Suppose rst that vt u = 0. Then

(uvt )u = u(vt u) = (vt u)u

with u = O, so that uvt has the nonzero eigenvalue vt u. Since uvt has rank
1, so also does its Jordan form, which is then
t
v u 0 0 0
0 0 0 0

.. .. .. . . ..
. . . . .
0 0 0 0

If vt u = 0, u = O, v = O, then (uvt )2 = O and uvt has rank 1, so that its

Jordan form is

0 1 0 0
0 0 0 0

.. .. .. .. .. .
. . . . .
0 0 0 0

2008:5. Suppose that a, b, c C with ab = 1. Evaluate the determi-

nant of

c a a2 an1
b c a an2
2
b b c an3
.
.. .. .. ..
. . . .
bn1 bn2 c
 7. LINEAR ALGEBRA 119

Solution 1. With u = (1, b, b2 , , bn1 ) and v = (1, a, a2 , , an1 ), we

have to nd det(ut v + (c 1)In ) with uvt = n. Since ut v is similar to

n 0 0 0
0 0 0 0

.. .. .. ..
. . . .
0 0 0 0

we are led to the answer (n + c 1)(c 1)n1 .

Solution 2. Setting c = 1 yields n columns that are proportional. Hence
(c 1)n1 is a factor of the determinant. The leading term is cn , so that the
determinant must be (c + k)(c 1)n1 . Since the coecient of cn1 vanishes,
k = n 1.
Solution 3. [C. Ochanine] Denote the determinant of the n n matrix by
n (a, b, c). By subtracting a suitable multiple of the rst row, we nd that
n (a, b, c) is equal to the determinant of

c a a2 an1
n2
0 c 1
a ac an2 a c
c
0 b b
c 1c an3 a c
n3
c
.. .. ..
0 . . .

Noting that c 1c = (1 1c )(c + 1) and pulling out the factor 1 1

c from the
last n 1 rows yields
 n1
1
n (a, b, c) = c 1 n1 (a, b, c + 1).
c

An induction argument shows that n (a, b, c) = (n + c 1)(c 1)n1 .

Solution 4. [S. Wong] The eigenvalues of the given matrix are those
numbers for which

c a a2 an1
b c a an2
2
c an3
0 = det b b

. . . .
.. .
.
.
.
.
.
bn1 b n2 c

1 ( c + 1) a a2 an1
b 1 ( c + 1) a an2

b2 b 1 ( c + 1) an3
= det .
. . . .
.. .. .. ..
b n1 b n2 1 ( c + 1)
120 7. LINEAR ALGEBRA

Hence is an eigenvalue of the given matrix if and only if c + 1 is an

eigenvalue with the same multiplicity of the matrix

1 a a2 an1
b 1 a an2
2
b b 1 an3
M .
.. .. .. ..
. . . .
bn1 bn2 1

Since all the columns of M are in proportion, the rank of M is 1 and so 0 is an

eigenvalue of multiplicity n 1. By examining the trace (which is the sum of
the eigenvalues), we see that the remaining eigenvalue is n (an eigenvector is
(1, b, b2 , , bn1 )t ). Thus, the eigenvalues of the given matrix are c 1 with
multiplicity n 1 and n + c 1 with multiplicity 1. Since the determinant of
the given matrix is the product of its eigenvalues, we nd that the required
answer is (n + c 1)(c 1)n1 .
Solution 5. [E. Flat] Denote the determinant to be found by Dn . Observe
that D1 = c and that D2 = c2 1 = (c 1)(c + 1). By taking b times each
row from the next, we nd that

c a a2 an2 an1
b(1 c) c1 0 0 0

0 b(1 c) c 1 0 0

Dn = det .. .. .. .. .. .
. . . . .

0 0 0 c1 0
0 0 0 b(1 c) c 1

Expanding along the last column yields, for n 2,

Dn = (c 1)Dn1 + (1)n1 an1 [b(1 c)]n1 = (c 1)Dn1 + (c 1)n1 .

An induction argument then establishes that

Dn = (c 1)n1 (c + n 1).

Solution 6. [J. Kramar; O. Ivrii] The (i, j)th term in the given matrix is
cij aji , where ij is the Kronecker delta that takes the value 1 when i = j
and 0 otherwise. The expansion of the determinant of this matrix is the sum
of n! terms of the form
n
 n
 n

c () ai(i) = c () ai a(i) = c () .
i=1 i=1 i=1

where is a permutation of {1, 2, . . . , n} and () is the number of xed

elements of . Thus, the value of the determinant is independent of a and b,
so that, wolog, we may restrict ourselves to the case that a = b = 1.
 7. LINEAR ALGEBRA 121

But then the required determinant is the product of the eigenvalues

(counting multiplicity) of

c 1 1 1
1 c 1 1

1 1 c 1
.
.. .. .. ..
. . . .
1 1 1 c
The (n 1)-dimensional subspace
{(x1 , x2 , . . . , xn )t : x1 + x2 + + xn = 0}
consists of eigenvectors with eigenvalue c 1. A complement of this subspace
is the span of the vector (1, 1, . . . , 1)t , which is an eigenvector with eigenvalue
c + n 1. Therefore the product of the eigenvalues and determinant of the
matrix is (c + n 1)(c 1)n1 .
2009:2. Let n and k be integers with n 0 and k 1. Let x0 , x1 , . . . ,
xn be n + 1 distinct points in Rk and let y0 , y1 , . . . , yn be n + 1 real numbers
(not necessarily distinct). Prove that there exists a polynomial p of degree at
most n in the coordinates of x with respect to the standard basis for which
p(xi ) = yi for 0 i n.

Solution. For 0 i < j n, let Hij be the hyperplane {z : (xi xj )

z = 0}. Since there are nitely many such hyperplanes, their union is not all
of Rk . (See Problem 2002:7.) Therefore, there exists a vector u for which
(xi xj ) u = 0 for all distinct i, j. Therefore, the real numbers ti = xi u
are all distinct 0 i n. There is a real polynomial q of degree at most n
for which q(ti ) = yi (0 i n). The polynomial p(x) = q(x u) has the
desired property.

2009:5. Find a 3 3 matrix A with elements in Z2 for which A7 = I

and A = I. (Here, I is the identity matrix and Z2 is the eld of two elements
0 and 1 where addition and multiplication are dened modulo 2.)

Solution. The minimum polynomial of A has degree at most 3 and divides

the polynomial
t7 1 = (t 1)(t6 + t5 + t4 + t3 + t2 + t + 1) .
Since A is not the identity matrix, its minimum polynomial divides the latter
factor. This minimum polynomial cannot be any of the irreducibles t, t + 1
(by the Factor Theorem) and t2 + t + 1, so it must be one of the irreducibles
t3 + t2 + 1, t3 + t + 1. Indeed
t6 + t5 + t4 + t3 + t2 + t + 1 = (t3 + t + 1)(t3 + t2 + 1),
over Z2 .
122 7. LINEAR ALGEBRA

We let A be the companion matrix of the rst factor on the right, namely

0 1 0
A = 0 0 1 .
1 1 0
Indeed, it can be checked that A3 = I + A, whence A6 = I + A2 and A7 =
A + A3 = A + I + A = I, as desired.
[Problem 2009:5 was contributed by Alfonso Grazia-Saz.]
2009:8. Let a, b, c be members of a real inner-product space (V,
, )
whose norm is given by x2 =
x, x. (You may assume that V is Rn if you
wish.) Prove that
a + b + b + c + c + a a + b + c + a + b + c
for a, b, c V .

Solution. From the inner product representations of the squares of the

norms, we have that
a + b2 + b + c2 + c + a2 = a2 + b2 + c2 + a + b + c2 .
Squaring both sides of the desired inequality, we see that it suces to prove
the following
(7.1) p + rq + r pq + rp + q + r
for p, q, r V and apply this to permutations of a, b, c.

Making the substitution x = p + r, y = q + r, z = p + q + r, so that

p = z y, q = z x, r = x + y z, we see that (7.1) is equivalent to
(7.2) xy z xz y + zx + y z
for x, y, z, V .
Let w be the orthogonal projection of z onto the span of x and y. Then
z = w + v where v is orthogonal to x, y and w, so that
z x2 = (w x) + v2 = w x2 + v2 w x2
and z x w x. Similarly, z y w y , z w and
x + y z x + y w. Thus, it is enough to prove that
xy w xw y + wx + y w
where x, y, w belong to a two-dimensional real inner product space.
But such a space is isometric to C with the usual absolute value, so we
may suppose that x, y, w C and can be multiplied. Since
xy = (w x)(w y) + w(x + y w),
an application of the triangle inequality yields the result.
 7. LINEAR ALGEBRA 123

Comment. Geometrically, equation (7.2) can be formulated as: sup-

pose that OABC is a parallelogram and that P is a point in space; then
|OA||OC| |P A||P C| + |P O||P B|.
2010:3. Let a and b, the latter nonzero, be vectors in R3 . Determine
the value of for which the vector equation
a (x b) = b
is solvable, and then solve it.
Solution 1. If there is a solution, we must have a b = |b|2 , so that
= (a b)/|b|2 . On the other hand, suppose that has this value. Then
0 = b a b (x b)
= b a [(b b)x (b x)b]
so that
b a = |b|2 x (b x)b.
A particular solution of this equation is
ba
x=u .
|b|2
Let x = z be any other solution. Then
|b|2 (z u) = |b|2 z |b|2 u
= (b a + (b z)b) (b a + (b u)b)
= (b z)b
so that z u = b for some scalar .
We check when this works. Let x = u + b for some scalar . Then
(b a) b
a (x b) = a (u b) = a
|b|2
b (b a)
=a+
|b|2
(b a)b (b b)a
=a+
|b|2
 
ba
=a+ b a = b,
|b|2
as desired. Hence, the solution is
ba
x= + b,
|b|2
where is an arbitrary scalar.
Solution 2. [B. Yahagni] Suppose, to begin with, that {a, b} is linearly
dependent. Then a = [(a b)/|b|2 ]b. Since (x b) b = 0 for all x, the
equation has no solutions except when = (a b)/|b|2 . In this case, it
becomes x b = 0 and is satised by x = b, where is any scalar.
124 7. LINEAR ALGEBRA

Otherwise, {a, b, a b} is linearly independent and constitutes a basis

for R3 . Let a solution be
x = a + b + (a b).
Then
x b = (a b) + [(a b) b] = (a b) + (a b)b (b b)a
and the equation becomes
(1 + |b|2 )a (a b)b (a b) = b.
Therefore = 0, is arbitrary, = 1/|b|2 and = (a b) = (a b)/|b|2 .
Therefore, the existence of a solution requires that = (a b)/|b|2 and
the solution then is
1
x = b (a b).
|b|2
Solution 3. Writing the equation in vector components yields the system
b3 x2 b2 x3 = a1 b1 ;
b3 x1 + b1 x3 = a2 b2 ;
b2 x1 b1 x2 = a3 b3 .
The matrix of coecients of the left side is of rank 2, so that the corre-
sponding homogeneous system of equations has a single innity of solutions.
Multiplying the three equations by b1 , b2 and b3 respectively and adding
yields
0 = a1 b1 + a2 b2 + a3 b3 (b21 + b22 + b23 ).
Thus, for a solution to exist, we require that
a 1 b1 + a 2 b2 + a 3 b3
= .
b21 + b22 + b23
In addition, we learn that the corresponding homogeneous system is satis-
ed by
(x1 , x2 , x3 ) = (b1 , b2 , b3 )
where is an arbitrary scalar.
It remains to nd a particular solution for the nonhomogeneous system.
Multiplying the third equation by b2 and subtracting the second multiplied
by b3 , we obtain that
(b22 + b23 )x1 = b1 (b2 x2 + b3 x3 ) + (a3 b2 a2 b3 ).
Therefore, setting b21 + b22 + b23 = b2 , we have that
b2 x1 = b1 (b1 x1 + b2 x2 + b3 x3 ) + (a3 b2 a2 b3 ).
Similarly
b2 x2 = b2 (b1 x1 + b2 x2 + b3 x3 ) + (a1 b3 a3 b1 ),
b2 x3 = b3 (b1 x1 + b2 x2 + b3 x3 ) + (a2 b1 a1 b2 ).
Observing that b1 x1 + b2 x2 + b3 x3 vanishes when
(x1 , x2 , x3 ) = (a3 b2 a2 b3 , a1 b3 a3 b1 , a2 b1 a1 b2 ),
 7. LINEAR ALGEBRA 125

we obtain a particular solution to the system:

(x1 , x2 , x3 ) = b2 (a3 b2 a2 b3 , a1 b3 a3 b1 , a2 b1 a1 b2 ).
Adding to this the general solution of the homogeneous system yields the
solution of the nonhomogeneous system.
2010:8. Let A be an invertible symmetric n n matrix with entries
{aij } in Z2 . Prove that there is an n n matrix M with entries in Z2 such
that A = M t M only if ai,i = 0 for some i.
Solution 1. [A. Kim] Let the entries of M be mi,j . Then
  
aii = mji mji = m2ji = mji .
j j j

Suppose that aii = 0 for each i. Then each column must have evenly many
entries equal to 1. But then the sum of the row vectors of M must be the
zero vector, and so the rows are linearly dependent. Hence the rank of M is
less than n, and so M is not invertible.
Solution 2. Let x = (x1 , x2 , . . . , xn )t be a column vector over Z2 and
observe that
n n
t 2
xAx = aii xi = ( aii xi )2 .
i=1 i=1
t
naij + aji = 2aij = 0.) Dene N = {x : xAx = 0}. Then
(Note that
N = {x : i=1 aii xi = 0}.
Suppose that A = M t M . Then M is invertible and so x M x is a
surjection (onto). Therefore the equation
0 = xAxt = xM t M xt = (M xt )t (M xt )
is not satised for each x, so that N is a proper subspace of (Z2 )n . Therefore,
there must exist i for which ai,i = 0.
[Problem 2010:8 was contributed by Franklin Vera Pachebo.]
2011:7. Suppose that there are 2011 students in a school and that each
student has a certain number of friends among his schoolmates. It is assumed
that if A is a friend of B, then B is a friend of A, and also that there may
exist certain pairs that are not friends. Prove that there is a nonvoid subset
S of these students for which every student in the school has an even number
of friends in S.

Solution. Dene a 2011 2011 matrix with entries aij in Z2 {0, 1},
the eld of integers modulo 2, as follows: aij = 1 if and only if i and j are
distinct and persons i and j are friends, and aij = 0 otherwise.

The determinant of this matrix is equal to


a1,(1) a2,(2) a2011,(2011) ,
126 7. LINEAR ALGEBRA

where the sum has n! terms and is taken over all permutations of the
symmetric group S2011 . Since 2011 is odd, any permutation of period 2,
consisting of a product of independent transpositions, must leave at least
one element xed so that its contribution to the determinant sum is 0. The
remaining permutations come in inverse pairs that yield the same product
value in the sum. It follows that the determinant of the matrix has the
value 0.
Therefore there is a linearly dependent subset of rows, and therefore a set
of rows whose modulo 2 sum is 0. The portion of each column belonging to
each of these rows must contain an even number of 1s. This means that the
person corresponding to that column must have evenly many friends among
the persons corresponding to the rows.
[Problem 2011:7 was contributed by Ali Feiz Mohammadi.]
2011:9. Suppose that A and B are two square matrices of the same
order for which the indicated inverses exist. Prove that

(A + AB 1 A)1 + (A + B)1 = A1 .

Solution 1. [V. Baydina]

(A + AB 1 A)1 + (A + B)1 = [A(I + B 1 A)]1 + (A + B)1

= (B 1 B + B 1 A)1 A1 + (B + A)1
= [B 1 (B + A)]1 A1 + (B + A)1
= (B + A)1 (BA1 + I)
= (B + A)1 (BA1 + AA1 )
= (B + A)1 (B + A)A1 = A1 .

Solution 2. [K. Ng]

[A + AB 1 A][A1 (A + B)1 ] = I A(A + B)1 + AB 1 AB 1 A(A + B)1

= I + AB 1 A(I + B 1 A)(A + B)1
= I + AB 1 A(I + B 1 A)[B(I + B 1 A)]1
= I + AB 1 A(I + B 1 A)(I + B 1 A)1 B 1
= I + AB 1 AB 1 = I.

Hence A1 (A + B)1 = (A + AB 1 A)1 as desired.

Solution 3. Fix the matrix A and dene a new multiplication of the set
of square matrices by X Y = XA1 Y . This multiplication is associative
and distributive with respect to ordinary matrix addition. For any matrix
X, we have that X A = X = A X so that A is the new identity matrix;
the inverse of a nonsingular matrix X is equal to X AX 1 A.
 7. LINEAR ALGEBRA 127

Observe that
(A + B) (A + B ) = (A + B)A1 (A + B ) = (I + BA1 )(A + B )
= (A + B) + (B + BA1 B ) = (A + B) + (B + A)
= (A + B) + (A + B ).

multiplying this equality on the left by (A + B) and on the right by

(A + B ) yields that
A = (A + B ) + (A + B) = A(A + AB 1 A)1 A + A(A + B)1 A.
Multiplying this equation normally on the left and on the right by A1 yields
the desired result.
2012:10. Let A be a square matrix whose entries are complex numbers.
Prove that A = A if and only if AA = A2 .
Comment. If A = A , then clearly AA = A2 . Henceforth, we assume
that AA = A2 and that A is a n n matrix.
Solution 1. Let B = i(A A ), so that B is hermitian and AB = O.
Then B is unitarily equivalent to a diagonal matrix, so that it suces to
show that all the eigenvalues of B are 0.
Suppose, if possible, that B has a nonzero eigenvalue with nontrivial
eigenvector x. Then O = ABx = Ax, so that
O = Ax = A x ix.
Therefore
0 = x Ax = x A x ix x.
Since
x A x =
x, A x =
A x, x =
x, Ax = x Ax = 0,
we have that ix2 = ix x = 0, which yields a contradiction. Therefore
B = O and A = A .
Solution 2. There exists a unitary matrix U (with U U = U U = I) for
which A = U T U for some upper triangular matrix T . Let the diagonal of
T be (1 , 2 , . . . , n ) and its other entries be tij for 1 i < j n.
We have that
T T = U AU U A U = U AA U = U A2 U = U AU U AU = T 2 ,
so that the trace of T T equals the trace of T 2 . Thus,
n  n

|k |2 + |tij |2 = 2k .
k=1 1i<jn k=1

The right side of this equation must be real, and

n n
 n
k |
2
k |
2
|k |2 ,
k=1 k=1 k=1
128 7. LINEAR ALGEBRA

so that each tij is zero and 2k = |k |2 for each k. Thus T is a diagonal matrix
with real eigenvalues, and so hermitian. Therefore A is hermitian.
Solution 3. Observe that
A 2 = (A2 ) = (AA ) = AA = A2 .
Suppose that it has been established that A k = Ak for some integer k not
less than 2. Then
Ak+1 = A(A k ) = (AA )(A k1 ) = A 2 (A k1 ) = A k+1
so that A m = Am for all integers m 2.
Let p be the minimal polynomial for A. Since every eigenvalue of A2 =

AA is nonnegative, it follows that every eigenvalue of A is real, so that p has
real coecients and p(A ) = (p(A)) = O. Thus p is the minimal polynomial
for A .
Suppose, if possible, that p(t) = t2 q(t), for some polynomial q(t). Let x
be any vector, and let y = q(A )x. Then
AA y = A2 y = A 2 y = p(A )x = 0
so that

A y, A y =
y, AA y = 0
from which A q(A )x = A y = 0. But this contradicts the fact that p is the

minimal polynomial for A . Thus p(t) = a0 + a1 t + a2 t2 + where at least

one of a0 and a1 is nonzero.
Suppose that a0 = 0. Then
O = p(A) p(A ) = a1 (A A )
so that A = A . On the other hand, if a0 is nonzero, then
O = Ap(A) A p(A ) = a0 (A A ) + a1 A2 a1 A 2 = a0 (A A )
and again A = A . Thus, the desired result holds.
Solution 4. Suppose that M = (mij ) is an arbitrary n n matrix. Then
tr(M ), the trace of M (the sum of its diagonal elements), is equal to the
complex conjugate of tr(M ). Since the ith diagonal
element of M M is

equal to j |mij | , it follows that tr(M M ) = i,j |mij |2 and we have that
2

tr(M M ) = tr(M M ).
It follows from this that M = O if and only if tr(M M ) = 0. If M = M ,
then its trace is real and tr(M ) = tr(M ).
Consider the situation of the problem where AA = A2 . We have that

(A A )(A A ) = (A A )(A A) = (AA A2 ) + (A A A2 ) = A A A2 ,

from which we nd that

tr((A A )(A A ) ) = tr(A A) tr(A2 ) = tr(AA ) tr(A2 )
= tr(AA ) tr(AA ) = 0
since the trace of AA is real. It follows that A A = O so that A = A .
 7. LINEAR ALGEBRA 129

[Problem 2012:10 and its solutions are drawn from the American Math-
ematical Monthly 101:4 (April, 1994), 362; 104:3 (1997), 277278.]
2013:7. Let (V,
) be a two-dimensional inner product space over the
complex eld C and let z1 and z2 be unit vectors in V . Prove that
1
sup{|
z, z1 
z, z2 | : z = 1}
2
with equality if and only if
z1 , z2  = 0.
Solution. Let
z1 , z2  = a + bi for real a and b. Since |
z, z1 
z, z2 | =
|
z, z1 
z, z2 |, there is no loss of generality in assuming that a = Re
z1 , z2 
0. Suppose that w = (z1 + z2 )/(z1 + z2 ). Then w = 1 and the supre-
mum in the problem is not less than 
w, z1 
w, z2 . Note that |
w, z1 | =
|1 +
z2 , z1 |/(z1 + z2 ) and that |
w, z2 | = |1 +
z1 , z2 |/(z1 + z2 ). Also
z1 + z2  =
z1 + z2 , z1 + z2  =
z1 , z1  +
z2 , z2  +
z1 , z2  +
z2 , z1  =
2 + 2Re
z1 , z2  = 2 + 2a.
|1 +
z1 , z2 |2
|
w, z1 
w, z2 | =
2 + 2a
(1 + a)2 + b2
=
2(1 + a)
 
1 b2 1
= 1+a+ ,
2 1+a 2
with equality if and only if a = b = 0, i.e.
z1 , z2  = 0. Thus, if the supremum
is 12 , then
z1 , z2  = 0.
On the other hand, suppose that
z1 , z2  = 0. Then {z1 , z2 } is an or-
thonormal basis, and we can write z = uz1 + vz2 where |u|2 + |v|2 = 1 for
z = 1. Then
1 1
|
z, z1 
z, z2 | = |u||v| (|u|2 + |v|2 ) =
2 2
with equality if and only if |u| = |v| = 21/2 . Therefore, the supremum is 12
in this case.
Comment. If V is a real, rather than complex, vector space, then a
trigonometric solution is possible. Wolog, we can assume that a basis has
been selected so that z1 = (1, 0) and z2 = (cos , sin ). Suppose that
z = (cos , sin ), then

z, z1 
z, z2  = cos (cos cos + sin sin ) = cos (cos( ))
1
= [cos + cos( 2)].
2
If cos = 0 and sin = 1, then the value of this expression is | 12 cos(
2)| 12 , so that the supremum is equal to 1/2, attainable when 2 = .
130 7. LINEAR ALGEBRA

On the other hand, when cos > 0 and 2 = , the expression is greater
than 1/2 and the supremum exceeds 1/2. Similarly, when cos < 0, and 2 =
+ , the expression is less than 1/2 and again the supremum exceeds 1/2.
2013:8. For any real square matrix A, the adjugate matrix, adj A, has
as its elements the cofactors of the transpose of A, so that
A adj A = adj A A = (det A)I.

(a) Suppose that A is an invertible square matrix. Show that

(adj (At ))1 = (adj (A1 ))t .
(b) Suppose that adj (At ) is orthogonal (i.e., its inverse is its transpose).
Prove that A is invertible.
(c) Let A be an invertible n n square matrix and let det (tI A) =
tn + c1 tn1 + + cn1 t + cn be the characteristic polynomial of
the matrix A. Determine the characteristic polynomial of adj A.
Solution. (a) Since At adj (At ) = (det A)I and A1 (adj (A1 )) =
(det A)1 I, it follows that
1
(adj (At ))1 = At = (adj (A1 ))t .
det A
(b) Since adj (At ) (adj (At ))t = I, then
At = At adj (At )(adj At )t = (det A)(adj At )t .
Since A = O, det A = 0.
(c)
1 1
det(tI adj A) = det(tA A adj A) = det(tA (det A)I)
det A   det A
(t)n det A
= det I A
det A t
 n  
(t)n  (1)n cn nk
= ck
(1)n cn t
k=0
n

= (1)n(nk) ck cnnk1 tk ,
k=0
with c0 = 1.
2014:1. The permanent, per A, of an n n matrix A = (ai,j ), is equal
to the sum of all possible products of the form a1,(1) a2,(2) an,(n) , where
runs over all the permutations on the set {1, 2, . . . , n}. (This is similar to
the denition of determinant, but there is no sign factor.) Show that, for any
n n matrix A = (ai,j ) with positive real terms,
n1

per A n! ai,j .
1i,jn
 7. LINEAR ALGEBRA 131

Solution. By the arithmetic-geometric means inequality, the sum den-

ing the permanent (having n! terms) is not less than n! times the product of
all these terms raised to the power 1/n!. Each term in the denition of the
permanent has n factors, so that product of all the terms has n n! factors.
There are n2 entries ai,j ; each appears the same number nn!n2 = (n1)!
times. The result follows.

2015:9. What is the dimension of the vector subspace of Rn generated

by the set of vectors
((1), (2), (3), . . . , (n))

where runs through all n! of the permutations of the rst n natural numbers.
Solution 1. [J. Love] The dimension cannot exceed n. Taking the dif-
ference of two permutations with identical outcomes except in the ith and
nth positions where takes the values 1 and 2, we nd that the vector space
contains the vectors

(0, 0, . . . , 0, 1, 0, . . . , 0, 1) = (3, 4, . . . , i + 1, 2, i + 2, . . . , n, 1)
(3, 4, . . . , i + 1, 1, i + 2, . . . , n, 2).

This set S of n 1 vectors is linearly independent and generates the (n 1)-

dimensional subspace {(x1 , x2 , . . . , xn ) : x1 + x2 + + xn = 0}. However,
the sum of the entries of any element in the generating set is 12 n(n + 1), so
that any generators along with S is a basis of n elements for the whole space.
Thus the required dimension is n.
Solution 2. [C. Wang] As in the rst solution, we can show that the span
of the generating set contains all vectors of the form (1, 0, . . . , 1, . . . , 0) and
therefore the vector
1
n(n + 1)(1, 0, 0, . . . , 0, 0) = (1, 2, 3, . . . , n 1, n) + (2, 2, 0, . . . , 0, 0)
2
+ (3, 0, 3, . . . , 0, 0) + (n 1, 0, 0, . . . , (n 1), 0)
+ (n, 0, 0, . . . , 0, n).

Similarly, it can be shown that this span contains all of the basis vectors
(0, 0, . . . , 1, . . . , 0). Hence the required dimension is n.
Solution 3. The dimension cannot exceed n, and is in fact equal to n.
We prove that

{(1, 2, 3, 4, , n 1, n), (2, 3, 4, 5, , n, 1), (3, 4, 5, 6, , 1, 2), ,

(n, 1, 2, 3, , n 2, n 1)}

is linearly independent by showing that the determinant of the matrix whose

rows are these vectors is nonzero.
132 7. LINEAR ALGEBRA

Using the fact that the absolute value of the determinant remains unc-
hanged if one column or row is subtracted from another, we have that
 
1 2 3 4 n2 n1 n 
 
2 3 4 5 n1 n 1 
 
3 4 5 6 n 1 2 
 
 
n 1 2 3 n3 n2 n 1
 
1 1 1 1 1 1 1 
 
2 1 1 1 1 1 (n 1)
 
3 1 1 1 1 (n 1) 1 
= 
4 1 1 1 (n 1) 1 1 
 
 
n (n 1) 1 1 1 1 1 
 
 1 1 1 1 1 1 1 

 1 0 0 0 0 0 n
 
 2 0 0 0 0 n 0 
= 
 3 0 0 0 n 0 0 
 
 
n 1 n 0 0 0 0 0 
 
 n 1 1 1 1 1 1 

 1 0 0 0 0 0 1
 
n2  2 0 0 0 0 1 0 
=n  3 1

 0 0 0 0 0 
 
 
n 1 1 0 0 0 0 0 
 
n(n + 1)/2 1 1 1 1 1 1 

 0 0 0 0 0 0 1
 
n1  0 0 0 0 0 1 0 
=n  1

 0 0 0 0 0 0 
 
 
 0 1 0 0 0 0 0 
n(n + 1)
= nn2 (1) = 0.
2
 CHAPTER 8

Geometry

2001:2. Let O = (0, 0) and Q = (1, 0). Find the point P on the line with
equation y = x + 1 for which the angle OP Q is a maximum.
The solution to this problem appears in Chap. 5.
2001:8. A regular heptagon (polygon with seven equal sides and seven
equal angles) has diagonals of two dierent lengths. Let a be the length of
a side, b be the length of a shorter diagonal and c be the length of a longer
diagonal of a regular heptagon (so that a < b < c). Prove ONE of the
following relationships:
a2 b2 c2
2
+ 2 + 2 =6
b c a
or
b2 c2 a2
+ + = 5.
a2 b2 c2
The solution to this problem appears in Chap. 2.
2002:1. Let A, B, C be three pairwise orthogonal faces of a tetrahedron
meeting at one of its vertices and having respective areas a, b, c. Let the face
D opposite this vertex have area d. Prove that
d 2 = a 2 + b2 + c 2 .

Solution 1. Let the tetrahedron be bounded by the three coordinate

planes in R3 and the plane with equation ux + yv + wz = 1, where u, v, w
are positive. The vertices of the tetrahedron are (0, 0, 0), (u, 0, 0), (0, v, 0),
(0, 0, w). Let d, a, b, c be the areas of the faces opposite these respective
vertices. Then the volume V of the tetrahedron is equal to
1 1 1 1
au = bv = cw = dk,
3 3 3 3

Springer International Publishing Switzerland 2016 133

E.J. Barbeau, University of Toronto Mathematics Competition
(20012015), Problem Books in Mathematics,
DOI 10.1007/978-3-319-28106-3 8
134 8. GEOMETRY

where k is the distance from the origin to its opposite face. The foot of
the perpendicular from the origin to this face is located at ((um)1 , (vm)1 ,
(wm)1 ), where m = u2 + v 2 + w2 , and its distance from the origin is
m1/2 . Since a = 3V u1 , b = 3V v 1 , c = 3V w1 and d = 3V m1/2 , the
result follows.
Solution 2. [J. Chui] Let edges of lengths x, y, z be common to the
respective pairs of faces of areas (b, c), (c, a), (a, b). Then 2a = yz,2b = zx
and 2c = xy. The fourth  face is bounded by sides of length u = y2 + z2 ,

v = z 2 + x2 and w = x2 + y 2 . By Herons formula, its area d is given by
the relation
16d2 = (u + v + w)(u + v w)(u v + w)(u + v + w)
= [(u + v)2 w2 ][w2 (u v)2 ]
= [2uv + (u2 + v 2 w2 )][2uv (u2 + v 2 w2 )]
= 2u2 v 2 + 2v 2 w2 + 2w2 u2 u4 v 4 w4
= 2(y 2 + z 2 )(x2 + z 2 ) + 2(x2 + z 2 )(x2 + y 2 ) + 2(x2 + y 2 )(y 2 + z 2 )
(y 2 + z 2 )2 (x2 + z 2 )2 (x2 + y 2 )2
= 4x2 y 2 + 4x2 z 2 + 4y 2 z 2 = 16a2 + 16b2 + 16c2 ,
whence the result follows.
Solution 3. Use the notation of Solution 2. There is a plane containing the
edge bounding the faces of areas a and b perpendicular to the edge bounding
the faces of areas c and d. Suppose it cuts  the latter faces in altitudes of
respective lengths u and v. Then 2c = u x2 + y 2 , whence u2 (x2 + y 2 ) =
x2 y 2 . Hence
x2 y 2 + x2 z 2 + y 2 z 2 4(a2 + b2 + c2 )
v 2 = z 2 + u2 = = ,
x2 + y 2 x2 + y 2
so that

2d = v x2 + y 2 = 4d2 = 4(a2 + b2 + c2 ),
as desired.
Solution 4. [R. Ziman] Let a, b, c, d be vectors orthogonal to the
respective faces of areas a, b, c, d that point inwards from these faces and
have respective magnitudes a, b, c, d. If the vertices opposite the respective
faces are x, y, z, O, then the rst three are pairwise orthogonal and
2a = y z, 2b = z x, 2c = x y,
and
2d = (z y) (z x) = (z x) (y z) (x y).
Hence d = (a + b + c), so that
d2 = d d = (a + b + c) (a + b + c) = a2 + b2 + c2 .
 8. GEOMETRY 135

2004:9. Let ABCD be a convex quadrilateral for which all sides and
diagonals have rational length and AC and BD intersect at P . Prove that
AP , BP , CP , DP all have rational length.
Solution 1. Because of the symmetry, it is enough to show that the length
of AP is rational. The rationality of the lengths of the remaining segments
can be shown similarly. Coordinatize the situation by taking A (0, 0),
B (p, q), C (c, 0), D (r, s) and P (u, 0). Then, equating slopes, we
nd that
s sq
=
ru rp
so that
sr ps
=ru
sq
whence u = r srps psqr
sq = sq .
Note that |AB|2 = p2 + q 2 , |AC|2 = c2 , |BC|2 = (p2 2pc + c2 ) + q 2 ,
|CD|2 = (c2 2cr + r2 ) + s2 and |AD|2 = r2 + s2 , we have that
2rc = |AC|2 + |AD|2 |CD|2
so that, since c is rational, r is rational. Hence s2 is rational.
Similarly
2pc = |AC|2 + |AB|2 |BC|2 .
Thus, p is rational, so that q 2 is rational.
2qs = q 2 + s2 (q s)2 = q 2 + s2 [(p r)2 + (q s)2 ] + p2 2pr + r2
is rational, so that both qs and q/s = (qs)/s2 are rational. Hence
p r(q/s)
u=
1 (q/s)
is rational.
Solution 2. By the cosine law, the cosines of all of the angles of the
triangles ACD, BCD, ABC and ABD are rational. Now
AP sin ABP
=
AB sin AP B
and
CP sin P BC
= .
BC sin CP B

Since AP B + BP C = 180 , therefore sin AP B = sin CP B and
AP AB sin ABP AB sin ABP sin P BC
= =
CP BC sin P BC BC sin2 P BC
AB(cos ABP cos P BC cos(ABP + P BC))
=
BC(1 cos2 P BC)
AB(cos ABD cos DBC cos ABC)
=
BC(1 cos2 DBC)
is rational. Also AP + CP is rational, so that (AP/CP )(AP + CP ) =
((AP/CP ) + 1)AP is rational. Hence AP is rational.
136 8. GEOMETRY

2006:4. Two parabolas have parallel axes and intersect in two points.
Prove that their common chord bisects the segments whose endpoints are the
points of contact of their common tangent.
Solution 1. Wolog, we may assume that the parabolas have the equations
y = ax2 and y = b(x 1)2 + c. The common chord has equation
a[b(x 1)2 + c y] b[ax2 y] = 0,
or
(8.1) (a b)y + 2abx a(b + c) = 0.

Consider a point (u, au2 ) on the rst parabola. The tangent at this point
has equation y = 2aux au2 . The abscissa of the intersection point of this
tangent with the parabola of equation y = b(x 1)2 + c is given by the
equation
bx2 2(b + au)x + (au2 + b + c) = 0.
This has coincident roots if and only if
(8.2) (b + au)2 = b(au2 + b + c) a(a b)u2 + 2abu bc = 0.
In this situation, the coincident roots are x = 1 + (au)/b and the point of
contact of the common tangent with the second parabola is
 
au a2 u2
1+ , +c .
b b
The midpoint of the segment joining the two contact points is
 
b + au + bu abu2 + a2 u2 + bc
, .
2b 2b

Plugging this into the left side of (8.1) and using (8.2) yields that
[1/(2b)][(a b)a(a + b)u2 + (a b)bc + 2ab2 + 2ab(a + b)u 2ab(b + c)]
= [(a + b)/(2b)][a(a b)u2 + 2abu bc] = 0.
Thus, the coordinates of the midpoint of the segment satisfy (8.1) and the
result follows.
Solution 2. [A. Feizmohammadi] Let the two parabolas have equations
y = ax(x u) and y = bx(x v) with au = bv. Since the two parabolas
must open the same way for the situation to occur, wolog, we may suppose
that a > b > 0. The parabolas intersect at the points (0, 0) and ((au
bv)/(ab), (ab(aubv)(uv))/(ab)2 ), and the common chord has equation
(a b)y ab(u v)x = 0.
Let y = mx + k be the equation of the common tangent. Then both of
the equations ax2 (au + m)x k = 0 and bx2 (bv + m)x k = 0 have
coincident solutions. Therefore (au + m)2 + 4ak = (bv + m)2 + 4bk = 0, from
which (by eliminating k),
ab(au2 bv 2 ) + 2ab(u v)m + (b a)m2 = 0.
 8. GEOMETRY 137

The common tangent of equation y = mx + k touches the rst parabola at

 
au + m m2 a2 u2
,
2a 4a
and the second parabola at
 
bv + m m2 b2 v 2
, .
2b 4b
The midpoint of the segment joining these two points is
 
ab(u + v) + (a + b)m (a + b)m2 ab(au2 + bv 2 )
, .
4ab 8ab
Using these coordinates as the values of x and y, we nd that
8ab[(a b)y ab(u v)x] = (a b)[(a + b)m2 ab(au2 + bv 2 )]
[2a2 b2 (u v)(u + v) + 2ab(a + b)(u v)m]
= (a + b)[(a b)m2 2ab(u v)m]
[ab(a b)(au2 + bv 2 ) + 2a2 b2 (u2 v 2 )]
= (a + b)ab(au2 bv 2 )
ab(a + b)(au2 bv 2 ) = 0.
Comment. There are two possible common tangents, and the common
chord bisects the segment on each of them (Fig. 8.1).

y = x(1x)

y =2x(x3)

Fig. 8.1. Two parabolas y = 2x(x 3) and y = x(x 1) and

their common tangents
138 8. GEOMETRY

2006:10. Let P be a planar polygon that is not convex. The vertices

can be classied as either convex or concave according as to whether the
angle at the vertex is less than or greater than 180 respectively. There must
be at least two convex vertices. Select two consecutive convex vertices (i.e.,
two interior angles less than 180 for which all interior angles in between
exceed 180 ) and join them by a segment. Reect the edges between these
two convex angles in the segment to form along with the other edges of P a
polygon P1 . If P1 is not convex, repeat the process, reecting some of the
edges of P1 in a segment joining two consecutive convex vertices, to form a
polygon P2 . Repeat the process. Prove that, after a nite number of steps,
we arrive at a polygon Pn that is convex.
Solution. Note that for the problem to work, we need to assume that, at
each stage, we create a polygon that does not cross itself. Suppose that P is
a m-gon. Then each Pi has at most m sides, as some angles of 180 may be
created at the vertices. The perimeter of Pi is the same as that of P for each
index i. Also P P1 P2 Pi . Therefore there is a disc that
contains all of the polygons Pi (the perimeter is at least as long as any line
segment contained within the polygon).
Let a be a vertex of P , and let ai be the corresponding vertex obtained
from a in Pi , each ai being equal to its predecessor or obtained from it by a
reection that takes it outside of Pi1 . In the latter case, ai must lie outside
of Pj when j < i. Because the sequence {ai } is bounded, it must have an
accumulation point a .
Let  > 0 be small enough that an -neighbourhood about a excludes all
other vertices of P . Consider an -neighbourhood U of a . It contains a point
al . If al+1 U , then al+1 is the reection of al in a line joining two other
vertices of Pl and lying on the opposite side of the centre of U to al , and so
U Pl+1 . This entails that ak U for k l + 1, contradicting that a is an
accumulation point. Hence ak U for k l. It follows that limi ai = a .
We show, in fact, that ai = a for suciently large i.
Let P  be the limiting polygon with vertices a . Suppose, if possible that

a is a concave vertex. We can nd a neighbourhood Vx around each vertex
x of P  such that, selecting any point from each of these neighbourhoods Vx
will create a polygon with a concave vertex in Va . Thus, there is an index r
such that the polygon Pr has a vertex inside each of these neighbourhoods.
If the concave vertex ar is reected about a line joining two other vertices
in Pr to ar+1 , then ar will be an interior point of Pr+1 and each subsequent
ai will lie outside of the interior of Pr+1 , and so outside of Ua , yielding a
contradiction.
Hence P  has acute angles at all of its vertices. But then for suciently
large n, so will Pn and so ai = an for i n. The result follows.
 8. GEOMETRY 139

2007:8. Suppose that n points are given in the plane, not all collinear.
Prove that there are at least n distinct straight lines that can be drawn
through pairs of the points.
Solution. This can be obtained as a corollary to Sylvesters Theorem:
Suppose that n points are given in the plane, not all collinear. Then there is
exists a line that contains exactly two of them. The proof of this is given in
Appendix A.
We can now solve the problem by an induction argument. It is clearly
true for n = 3; suppose it holds for n 1 3 and that n points are given.
Pick a line that passes through exactly two points P and Q of the set. At
least one of these points, say P , is not collinear with the rest. Remove this
line and the point Q. We can nd n 1 distinct lines determined by pairs of
the other n 1 points, and restoring the line through P Q yields the nth line.
2008:1. Three angles of a heptagon (7-sided polygon) inscribed in a
circle are equal to 120 . Prove that at least two of its sides are equal.
Solution. Consider two adjacent sides of the heptagon for which the angle
between them is 120 . The chord of the circle joining the endpoints that are
not common to the chords subtends an angle of 60 at the circumference of
the circle and therefore an angle of 120 at the centre of the circle. If the
three pairs of adjacent sides forming angles of 120 are mutually disjoint from
each other, then they constitute six sides of the heptagon which subtend in
total angles totalling 360 at the centre of the circle, leaving no positive angle
for the seventh side to subtend. Hence, two of the pairs of sides must have
an edge in common. However, since each pair subtends the same angle at
the centre, the edges that the pairs do not have in common must be of equal
length and the result follows.
2008:6. 2008 circular coins, possibly of dierent diameters, are placed
on the surface of a at table in such a way that no coin is on top of another
coin. What is the largest number of points at which two of the coins could
be touching?
The solution to this problem appears in Chap. 10.
2008:10. A point is chosen at random (with the uniform distribution)
on each side of a unit square. What is the probability that the four points
are the vertices of a quadrilateral with area exceeding 12 ?
Solution 1. The desired probability is 12 .
Suppose that the points are located in a counterclockwise direction dis-
tance x, y, z, w from the four vertices in order. Then the area of that part
of the square lying outside of the inner quadrilateral is
1 1
[x(1w)+y(1x)+z(1y)+w(1z)] = [(x+y+z+w)(xw+yx+zy+wz)].
2 2
140 8. GEOMETRY

1
The condition that the area of the inner quadrilateral exceeds 2 is that
(x + y + z + w) (xw + yx + zy + wz) < 1.
The area is equal to 12 if and only if x + z = 1 or y + w = 1, both of which
occur with probability 0.
On the other hand, note that the points are located at distances x, w =
1 w from one vertex of the square and distance z, y  = 1 y from the
diagonally opposite vertex. Then the area of that part of the square lying
outside the inner quadrilateral is
1 1
[xw +(1x)(1y  )+zy  +(1w )(1z)] = [(xw +xy  +zy  +zw )(x+y  +z+w )+2],
2 2

so that the condition that the area of the inner quadrilateral exceeds 12 is
that
(x + y  + z + w ) (xw + y  x + zy  + w z) > 1.
Since x, y, z, w and so x, y  , z, w , are chosen independently and uniformly in
each case, the probabilities of either inequality occurring is the same as the
probability of the other. Therefore, the desired probability is 12 .
Solution 2. Let the square have vertices (0, 0), (1, 0), (1, 1), (0, 1) in the
cartesian plane and let the four selected points by (a, 0), (1, b), (c, 1) and
(0, d), where 0 a, b, c, d 1. Then the area of the inner quadrilateral is
equal to
1 1 1 1 1
1 (1 a)b (1 c)(1 b) (1 d)c ad = [1 + (d b)(c a)].
2 2 2 2 2
The area of the quadrilateral is exactly 12 when b = d or a = c, which
occurs with probability 0. The area exceeds 12 if and only if, either b < d
and a < c, or b > d and a > c. The area is less than 12 if and only if, either
b < d and c < a, or b > d and c > a. Since a, b, c, d are selected randomly
and independently from [0, 1], by symmetry, the two latter events have equal
probability of 12 .
2010:1. Let F1 and F2 be the foci of an ellipse and P be a point in
the plane of the ellipse. Suppose that G1 and G2 are points on the ellipse
for which P G1 and P G2 are tangents to the ellipse. Prove that F1 P G1 =
F2 P G2 .
Solution. Let H1 be the reection of F1 in the tangent P G1 , and H2
be the reection of F2 in the tangent P G2 . We have that P H1 = P F1 and
P F2 = P H2 . By the reection property, P G1 F2 = F1 G1 Q = H1 G1 Q,
where Q is a point on P G1 produced. Therefore, H1 F2 intersects the ellipse
in G1 . Similarly, H2 F1 intersects the ellipse in G2 . Therefore
H1 F2 = H1 G1 + G1 F2 = F1 G1 + G1 F2
= F1 G2 + G2 F2 = F1 G2 + G2 H2 = H2 F1 .
 8. GEOMETRY 141

Therefore, triangle P H1 F2 and P F1 H2 are congruent (SSS), so that

H1 P F2 = H2 P F1 . It follows that
2F1 P G1 = H1 P F1 = H1 P F2 F2 P F1
= H2 P F1 F2 P F1 = H2 P F2 = 2F2 P G2
and the desired result follows.
2012:1. An equilateral triangle of side length 1 can be covered by ve
equilateral triangles of side length u. Prove that it can be covered by four
equilateral triangles of side length u. (A triangle is a closed convex set that
contains its three sides along with its interior.)
Solution. Consider the set of six points consisting of the vertices of the
triangle and the midpoints of the three sides. By the Pigeonhole Principle,
one of the ve covering equilateral triangles must cover two of these six points.
Since the distance between any two of the six points is at least 12 and any
two points within an equilateral triangle cannot be further apart than two of
its vertices, we must have that u 12 .
However, it is clear that the given equilateral triangle can be covered by
four such triangles of side-length 12 determined by joining pairs of midpoints
of the sides. Each of these four triangles can be included within a triangle of
side-length u.
2012:5. Let C be a circle and Q a point in the plane. Determine the locus
of the centres of those circles that are tangent to C and whose circumference
passes through Q.
Solution 1. There are several cases to consider. Let O be the centre and
r the radius of the given circle. Let P be the centre of the variable circle.
(i) Q lies on C. In this case, the variable circle is tangent to C at Q
and the locus is the set of points on the line OQ with the exception
of Q itself.
(ii) Q lies in the interior of C. Suppose that the variable circle is tangent
to C at T . Then
|OP | + |P O| = |OP | + |P T | = |OT | = r
so that P lies on the ellipse with foci O and Q and major axis of
length r. (If Q = O, this becomes a circle with diameter r.)
Conversely, if P is a point on the ellipse just described, let OP produced
meet C at T and construct a circle with centre P and radius |P T |. This
circle will be tangent to C and |QP | = r |OP | = |T P |, so that it will pass
through Q.
(iii) Q lies outside of C and the variable circle touches C at T , say, with
the two circles external to each other. Let a be the radius of the
variable circle. The segment OP contains T and
|OP | = r + a = r + |P Q|
142 8. GEOMETRY

so that |OP | |P Q| = r, a constant. Hence P lies on one branch

of a hyperbola with foci O, Q whose asymptotes are the lines that
pass through O and the contact points of the tangents drawn to C
from Q.
(iv) Q lies outside of C and the variable circle touches C at T , say, with
C contained inside the variable circle. Then, with a the radius of
the variable circle,
|QP | |OP | = a (a r) = r
so that P lines on a branch of the hyperbola with foci O, Q.
Solution 2. Let C have equation x2 + y 2 = 1 and Q (q, 0), with q 0.
Suppose that (X, Y ) is the centre of a circle that passes through Q and
touches C. Then
 
(X q)2 + Y 2 = | X 2 + Y 2 1|

= X 2 2qX + q 2 + Y 2 = X 2 + Y 2 + 1 2 X 2 + Y 2

= 2 X 2 + Y 2 = (1 q 2 ) + 2qX
= 4X + 4Y 2 = (1 q 2 )2 + 4q(1 q 2 )X + 4q 2 X 2
2

= (1 q 2 )(2X q)2 + 4Y 2 = (1 q 2 ).
When q = 0, the locus is a circle concentric with the given circle with half
the radius. When 0 < q < 1, the locus isan ellipse with centre at (q/2, 0),
semi-major axis 1/2 and semi-minor axis 1 q 2 /2. When q = 1, the locus
is the line Y = 0.
Suppose that q > 1. The equation of the locus can be written
4Y 2
(2X q)2 = 1,
q2 1
from which it can be seen that the locus is a hyperbola with asymptotes given
by the equations 
q 2 1(2X q) = 2Y.

These are lines passing through the point (q/2, 0) with slopes q 2 1.
Thetangents to the circle C from (q, 0) touch the circle at the points
(1/q, q 2 1/q). When the centre of the variable circle is on the right
branch of the hyperbola, the two circles touch externally at a point that lies
on the minor arc between these points. When the centre is on the left branch
of the hyperbola, C touches the variable circle internally at a point on the
major arc joining these two points. Note that the vertex of the left branch
of the hyperbola is ((q 1)/2, 0) which lies on the positive x-axis; this is the
centre of a circle of radius (q + 1)/2 > 1 that touches C at the point (1, 0).
 8. GEOMETRY 143

2012:8. Determine the area of the set of points (x, y) in the plane that
satisfy the two inequalities:
x2 + y 2 2
x4 + x3 y 3 xy + y 4 .
Solution. The second inequality can be rewritten as
0 y 4 + xy x4 x3 y 3 = (y 3 + x)(y x3 ).
The locus of the equation y 3 + x = 0 can be obtained from the locus of the
equation y x3 = 0 by a 90 rotation about the origin that relates the points
(u, v) and (v, u). Thus the two curves partition the disc x2 + y 2 2 into
four regions of equal area, and the inequality is satised by the two regions
that cover the y-axis. Thus the area of the set is half the area of the disc,
namely .
2013:2. ABCD is a square; points U and V are situated on the respective
sides BC and CD. Prove that the perimeter of triangle CU V is equal to twice
the sidelength of the square if and only if U AV = 45 .
Solution 1. [Y. Babich; J. Song] Produce CD to W so that DW = BU .
Then triangles ABU and ADW are congruent (SAS) so that AU = AW .
First, assume that U AV = 45 . Then
V AW = V AD + DAW = V AD + BAU
= 90 U AV = 45 .
Since also AU = AW and AV is common, triangles U AV and W AV are
congruent (SAS) and U V = V W = DV + BU .
Therefore
CU + CV + U V = (CU + BU ) + (CV + DV ) = BC + CD
as desired.
On the other hand, assume that CU + CV + U V = BC + CD. Then
U V = (BC CU ) + (CD CV ) = BU + V D = V W,
so that triangles AU V and AW V are congruent (SSS) and U AV =
W AV = V AD + U AB. Since U AB + U AV + V AD = 90 , it
follows that U AV = 45 .
Solution 2. Let the side length of the square be 1, |BU | = u, |DV | = v,
|U V | = w. Then |CU | = 1 u and |CV | = 1 v.
Suppose that U AV = 45 . Since 45 = BAU + DAV ,
u+v
1 = tan 45 = ,
1 uv
144 8. GEOMETRY

whereupon 1 uv = u + v and 1 + u2 v 2 = u2 + 4uv + v 2 . By the Law of

Cosines,

w2 = (1 + u2 ) + (1 + v 2 ) 2(1 + u2 )(1 + v 2 )

= u2 + v 2 + 2 2(1 + u2 v 2 + u2 + v 2 ) = u2 + v 2 + (2 4(u2 + 2uv + v 2 ))
= u2 + v 2 + 2(1 (u + v)) = u2 + v 2 + 2uv = (u + v)2 .

Hence w = u + v and the perimeter of CU V is equal to (1 u) + (1 v) +

(u + v) = 2.
On the other hand, suppose that |U V | = u + v. Then, by Pythagoras
Theorem,
u2 + 2uv + v 2 = (1 2u + u2 ) + (1 2v + v 2 )
so that u + v = 1 uv. Therefore
u+v
tan(U AB + V AD) = = 1,
1 uv
so that U AB + V AD = 45 and U AV = 45 .
Solution 3. [Z. Liu] Let A (0, 0), B (1, 0), C (1, 1), D (0, 1),
U (1, u), and V (v, 1), where 0 < u, v < 1. Since

cos U AV |AU | |AV | = AU AV ,

U AV = 45 2(u + v) = 1 + u2 1 + v2
(u + v) = u + 2uv + v = 1 2uv + u2 v 2 = (1 uv)2
2 2 2

u + v = 1 uv (1 u)2 + (1 v)2 = (u + v)2

|U V | = |BU | + |V D|
|CU | + |CV | + |U V | = |BU | + |CU | + |CV | + |V D| = |BC| + |CD|.

The result follows.

Solution 4. [J. Zung] Let U be xed on BC. Then as V moves from
C to D, the lengths of CV and U V strictly increase, so that the perimeter
of triangle CU V strictly increases from 2CU < BC + CD to CU + U D +
CD > BC + CD. Therefore, there is a unique position of V such that
CU + CV + U V = BC + CD.
Also, as V moves from C to D, the angle U AV strictly increases from an
angle less than 45 to an angle greater than 45 . Therefore, there is a unique
position of V such that U AV = 45 . We need to show that the position of
V is the same in both situations.
Select V so that U V is tangent to the circle with centre A that passes
through B and D. Let T be the point of tangency. Then
CU +CV +V U = CU +CV +V T +T U = CU +CV +V D +U B = CD +CB.
Also, V AT = 12 DAT and U AT = 12 BAT , so that U AV = 45 . Thus
the point V is the unique point for both of the foregoing situations, and the
desired result follows.
 8. GEOMETRY 145

Solution 5 (of if part). [S. Rumsey] Suppose U AV = 45 . Let the

image of B reected in AU be E and the image of D reected in AV be F .
Then
EAU + F AV = BAU + DAV = 45 = U AV,
so that E and F must fall on the same line through A. Since AE = AB =
AD = AF , then E = F .
Also,
AEU = ABU = 90 = ADV = AEV,
so that U, E, V are collinear and E lies on U V . Therefore
CU +CV +U V = CU +CV +U E +EV = CU +CV +BU +DV = BC +CD,
as desired.
2013:5. A point on an ellipse is joined to the ends of its major axis.
Prove that the portion of a directrix intercepted by the two joining lines
subtends a right angle at the corresponding focus.
Solution 1. Suppose that the focus is at the origin of the Cartesian plane,
the directrix is the line x = 1and the eccentricity of the ellipse is e. Then
the equation of the ellipse is x2 + y 2 = e(1 x), or
y 2 = (1 e2 )x2 2e2 x + e2 .
The major axis is the x-axis and the ellipse intersects this axis at the points
(e/(1 + e), 0) and (e/(1 e), 0). Let (u, v) be an arbitrary point on the
ellipse. Then the lines determined by the point (u, v) and these endpoints
respectively intersect the directrix at
   
v v
1, and 1, .
(1 + e)u e (1 e)u + e
The product of the slopes of the segments joining these to the origin is
v2 (1 e2 )u2 2e2 u + e2
= = 1.
(1 e2 )u2 + [e(1 + e) e(1 e)]u e 2 (1 e2 )u2 + 2e2 u e2

Solution 2. [J. Song; T. Xiao] Let the equation of the ellipse be x2 /a2 +
y /b2 = 1 and the point P (u, v). The right directrix of the ellipse has
2

equation x = a2 /c and the right focus F is at (c, 0), where a2 = b2 + c2 .

The equation of the line through (a, 0) and P is (u + a)y = v(x + a)
and this meets the directrix at the point
 2 
a av(a + c)
G , .
c c(u + a)
The equation of the line through (a, 0) and P is (u a)y = v(x a) and this
meets the directrix at the point
 2 
a av(a c)
H , .
c c(u a)
146 8. GEOMETRY

Let K be the point (a2 /c, 0) where the directrix and major axis intersect.
Because P lies on the ellipse, we have that a2 v 2 = b2 (a2 u2 ).
There are two possible ways to proceed. Taking the dot product of the

vectors F G and F H, we obtain that
   2   
a2 av(a + c) a av(a c) 1 a2 v 2 (a2 c2 )
c, c, = 2 (a2 c2 )2
c c(u + a) c c(u a) c a2 u 2
a2 c 2 2
= [a c2 b2 ] = 0,
c2
so that GF H = 90 .
Alternatively, we can use the Law of Cosines to obtain that
2|GF ||HF | cos GF H = |F G|2 + |F H|2 |GH|2
= 2|F K|2 + |GK|2 + |HK|2 (|GK| + |HK|)2
= 2[|F K|2 |GK||HK|]
 2 
a2 a2 v 2 (a2 c2 )
=2 c 2 2
c c (a u2 )
2 2(a2 c2 ) 2
= 2
[(a2 c2 )2 b2 (a2 c2 )] = [a c2 b2 ] = 0,
c c2

from which the result follows. (Note that |u a| = a u since u < a.)
2014:8. The hyperbola with equation x2 y 2 = 1 has two branches,
as does the hyperbola with equation y 2 x2 = 1. Choose one point from
each of the four branches of the locus of (x2 y 2 )2 = 1 such that area of the
quadrilateral with these four vertices is minimized.

Solution. Let N, W, S, E be arbitrary points on the four branches that are

symmetric respectively about the positive y-axis (northern), negative x-axis
(western), negative y-axis (southern) and positive x-axis (eastern). Consider
the diagonal N S of N W SE. If the secant through W that is parallel to
N S passes through two distinct points of the western branch, then the area
[N W S] of triangle N SW can be made smaller by replacing W by the point
of tangency to the western branch by a tangent parallel to N S. Similarly,
the area [N SE] can be made smaller when the tangent at E is parallel to
N S. We can follow a similar argument for the diagonal W E. Thus, for any
quadrilateral N W SE, we can nd a quadrilateral of no larger area where the
tangents to the hyperbolae at N and S are parallel to W E and the tangents
to the hyperbolae at E and W are parallel to N S. Thus we may assume that
N W SE is a parallelogram.

Observe that the slopes of the diagonals W E as well as the tangents to

the northern and southern branches range over the open interval (1, 1), and
that the slopes of the diagonals N S and the tangents to the western and
eastern branches are similarly related. Each allowable tangent slope occurs
exactly twice, the tangents being related by a 180 rotation about the origin.
 8. GEOMETRY 147

Thus, in the minimization problem, we need consider only parallelograms

centred at the origin with the foregoing tangent and diagonal relationship.
Let N (a, b) be any point on the northern branch, where, wolog, we
may assume that 0 a < b (and b2 a2 = 1). The slope of the tangent at N
is a/b, so the point E, located on the line with this slope through the origin
must be at (b, a). Similarly, S (a, b) and W (b, a). Thus, N W SE
is in fact a rectangle whose sides are perpendicular to the asymptotes of the
hyperbola and whose area is

|N E| |ES| = ( 2)(b a)( 2)(b + a) = 2(b2 a2 ) = 2.
Therefore, the required minimum area is 2.
[Problem 2014:8 was contributed by Robert McCann.]
2015:2. Given 2n distinct points in space, the sum S of the lengths of
all the segments joining pairs of them is calculated. Then n of the points
are removed along with all the segments having at least one endpoint from
among them. Prove that the sum of the lengths of all the remaining segments
is less than 12 S.
Solution 1. [J. Love] Let P1 , P2 , . . . , Pn be the points that remain after the
points Q1 , Q2 , . . . , Qn are removed.
 By repeated application of the triangle
inequality, for each of the n2 pairs (i, j), we have that
|Pi Pj | |Pi Qj | + |Qj Qi | + |Qi Pj |.
Note that, if (i, j) = (r, s), then Pi Pj = Pr Ps , Pi Qj = Pr Qs and Qi Qj =
Qr Qs . For each retained segment, there is a distinct set of three deleted
intervals whose total length exceeds that of the retained segment. Hence, the
total length of the deleted segments exceeds that of the retained segments.
(The inequality is strict because we have not included among the deleted
segments those of the form Pi Qi .) The result follows.
Solution 2. Dene Pi and Qi as before, For each triple i, j, k with i = j
and 1 i, j, k n, write the triangle inequality

|Pi Pj | |Pi Qk | + |Pj Qk |.

n  
There are n 2 inequalities in all; for each of the n2 choices of Pi Pj
we have an inequality for each of the n choices
 of Qk . Each |Pi Pj | appears
in n of the inequalities. There are 2n n2 = n2 (n 1) terms on the right
side of the inequalities and each of the n2 terms of the form |Pi Qk | appears
n 1 times. Let T be the sum of all the lengths |Pi Pj |. Then the sum of
the lengths of all the intervals involving at least one Qk is S T , and this
includes all the intervals of the form Pi Qk . Adding all the inequalities yields
that nT (n 1)(S T ), from which we nd that
(2n 1)T (n 1)S.
The desired result follows.
148 8. GEOMETRY

Solution 3. [S. Xiu] We begin with 2n points X1 , X2 , . . . , X2n and remove

the points one by one beginning with X2n . Let Sk be the sum of the lengths of
all the segments joining pairs of points in {X1 , X2 , . . . , Xk } for n k 2n1.
 
When 1 i < j k, |Xi Xj | |Xi Xk+1 | + |Xj Xk+1 |. There are k2
inequalities of this type in which each |Xi Xk+1 | appears k 1 times, once for
k
each Xi Xj with i = j, so that Sk (k 1) i=1 |Xi Xk+1 |. Therefore
k

(k 1)Sk+1 = (k 1)(Sk + |Xi Xk+1 |) kSk .
i=1
Thus Sk [(k 1)/k]Sk+1 . Hence
n1 n 2n 2 n1 1
Sn S2n = S < S.
n n+1 2n 1 2n 1 2
 CHAPTER 9

Group Theory

2003:10. Let G be a nite group of order n. Show that n is odd if and only
if each element of G is a square.
Solution. Suppose that n is odd. Then, by Lagranges Theorem, each
element of G is of odd order, so that, if a G, then a2k+1 = e (the identity)
for some nonnegative integer k. Hence a = (ak+1 )2 is a square. On the other
hand, suppose that n is even. Pair o each element of G with its inverse.
Some elements get paired o with a distinct element, and others get paired
o with themselves. Since n is even and an even number of elements get
paired o with a distinct element, there must be an even number of elements
that get paired o with themselves. Since the identity gets paired o with
itself, there must be some other element v that also is its own inverse, i.e.
v 2 = e. Consider the mapping x x2 dened from G to itself. Since e and
v have the same image and since G is nite, this mapping cannot be onto.
Hence, there must be an element of G which is not the square of another
element. The result follows.
2005:6. Let G be a subgroup of index 2 contained in Sn , the group of
all permutations of n elements. Prove that G = An , the alternating group of
all even permutations.
Solution 1. Let u Sn \G. Since G is of index 2, uG = Gu whence
u1 Gu = G. Thus, x1 gx G for each element g in G and x in Sn .
If t is a transposition, than x1 tx runs through all transpositions as x runs
through G. Therefore, if G contains a transposition, then G must contain
every transposition, and so must be all of Sn , giving a contradiction. Thus,
every transposition must belong to Gu. Hence, if t1 and t2 are two trans-
positions, then t1 = g1 u and t2 = t1 2 = g2 u for g1 and g2 in G. Hence,
t1 t2 = (g1 u)(u1 g21 ) = g1 g21 G. Thus, G contains every even permuta-
tion, and so contains An . The result follows.

Springer International Publishing Switzerland 2016 149

E.J. Barbeau, University of Toronto Mathematics Competition
(20012015), Problem Books in Mathematics,
DOI 10.1007/978-3-319-28106-3 9
150 9. GROUP THEORY

Solution 2. As in Solution 1, we see that G must be normal, and so

must be the kernel of a homomorphism from Sn into the multiplicative
group {1, 1}. Suppose that x Sn and x has order m, then 1 = (xm ) =
((x))m , whence m is even or x G. It follows that every element of odd
order must belong to G, and so in particular, G must contain all 3-cycles.
Since (ab)(ac) = (abc) and (ab)(cd) = (abc)(adc) (multiplying from the left),
every product of a pair of transpositions and hence every even permutation
is a product of 3-cycles. Hence G contains all even permutations and so must
coincide with An .
2008:7. Let G be a group of nite order and identity e. Suppose that is
an automorphism of G onto itself with the following properties: (1) (x) = x
if and only if x = e; (2) ((x)) = x for each element x of G.
(a) Give an example of a group and automorphism for which these
conditions are satised.
(b) Prove that G is commutative (i.e., xy = yx for each pair x, y of
elements in G).
Solution. (a) An example is Z3 {0, 1, 2}, the integers modulo 3 with
mod 3 addition as the group operation, with the automorphism that switches
1 and 2.
(b) Suppose for each element x, we denote (x) by x . We dene the
mapping : G G by (x) = x x1 . We show that is one-one.
Suppose that (x) = (y). Then x x1 = y  y 1 , so that
x1 y = x1 y  = (x1 ) y  = (x1 y) .
Hence x1 y = e, so that y = x. Since is one-one and G is nite, is onto.
For any element z G, select x so that z = x x1 . Then z  = x(x1 ) ,
so that
zz  = x x1 x(x )1 = e = z  z,
whence z  = z 1 . Therefore
(z) = z 1
for each z G.
Let x, y be any two elements of G. Then
xy = [(xy) ] = (x y  ) = (x1 y 1 )1 = yx.
The desired result follows.
2011:8. The set of transpositions of the symmetric group S5 on
{1, 2, 3, 4, 5} is
{(12), (13), (14), (15), (23), (24), (25), (34), (35), (45)}
 9. GROUP THEORY 151

where (ab) denotes the permutation that interchanges a and b and leaves
every other element xed. Determine a product of all transpositions, each
occuring exactly once, that is equal to the identity permutation , which
leaves every element xed.
Solution. Multiplication proceeds from left to right. Observe that
(ab)(ac)(bc) = (ac) for any a, b and c. We observe that (13)(23) carries 1
to 2 and that (24)(14) carries 2 to 1, so that (13)(24)(14)(23) = (12)(34).
Therefore
(12)(13)(24)(14)(23)(34) = .
Therefore
(15)(12)(25)(13)(24)(14)(23)(35)(34)(45) = .
Comment. Other answers were
(34)(45)(35)(23)(24)(15)(25)(14)(13)(12) (S. Sagatov),
(12)(23)(34)(45)(14)(13)(15)(25)(24)(35) (S.E. Rumsey).
2013:9. Let S be a set upon whose elements there is a binary operation
(x, y) xy which is associative (i.e. x(yz) = (xy)z). Suppose that there
exists an element e S for which e2 = e and that for each a S, there is
at least one element b for which ba = e and at most one element c for which
ac = e. Prove that S is a group with this binary operation.
Solution 1. We show that e is the identity element of S and that each
element of S has a two-sided inverse. Let t be an arbitrary element of S.
There is an element s S for which st = e. Hence e = e2 = s(tst). Since
also st = e, we must have that t = tst = te. Therefore e is a right identity
in S.
Suppose that r S is such that rs = e. Then et = rst = re = r (since e
is a right identity). Therefore ets = rs = e. Since ee = e as well, e = ts and
so et = tst = t.
Solution 2. [J. Love] Let a S. There exists b such that ba = e.
Therefore, b(ae) = (ba)e = e2 = e, so that a = ae (a S). We have that
e = ba = (be)a = b(ea) whereupon a = ea. Hence e is a two-sided identity.
Suppose that ba = e. Select d S so that db = e. Then d = de =
d(ba) = (db)a = ea = a, so that a is a two-sided inverse of b.
[Problem 2013:9 is #4504 from the American Mathematical Monthly
59:8 (October, 1952), 554; 61:1 (January, 1954), 5455.]
 CHAPTER 10

Combinatorics and Finite Mathematics

2001:3.
(a) Consider the innite integer lattice in the plane (i.e., the set of
points with integer coordinates) as a graph, with the edges being the
lines of unit length connecting nearby points. What is the minimum
number of colours that can be used to colour all the vertices and
edges of this graph, so that
(i) each pair of adjacent vertices gets two distinct colours;
(ii) each pair of edges that meet at a vertex gets two distinct
colours; and
(iii) an edge is coloured dierently than either of the two vertices
at the ends?
(b) Extend this result to lattices in real n-dimensional space.

Solution. Consider an n-dimensional lattice. Each vertex has 2n edges

emanating from it (consisting of those points with exactly one coordinate
diering by +1 or 1), so at least 2n + 1 colours are needed. We create
a colouring with this number of colours by building up one dimension at a
time.
The integer points on the line and the edges between them can be
coloured 1 (3) 2 (1) 3 (2) 1 and so on, where the
edge colouring is in parentheses. Form a plane by stacking these lines unit
distance apart, making sure that each vertex has a dierent coloured vertex
above and below it; use colours 4 and 5 judiciously to colour the vertical
edges. Now go to three dimensions; stack up planar lattices and struts unit
distance apart, colouring each with the colours 1, 2, 3, 4, 5, while making sure
that vertically adjacent vertices have separate colours, and use the colours 6
and 7 for vertical struts. Continue on.
Comment. Here is a another colouring that will work for the plane: Let
the colours be numbered 0, 1, 2, 3, 4. Colour the point (x, 0) with the colour

Springer International Publishing Switzerland 2016 153

E.J. Barbeau, University of Toronto Mathematics Competition
(20012015), Problem Books in Mathematics,
DOI 10.1007/978-3-319-28106-3 10
154 10. COMBINATORICS AND FINITE MATHEMATICS

x (mod 5); colour the point (0, y) with the colour 2y (mod 5); colour the
points along each horizontal line parallel to the x-axis consecutively; colour
the vertical edge whose lower vertex has colour m (mod 5) with the colour
m + 1 (mod 5); colour the horizontal edge whose left vertex has the colour n
(mod 5) with the colour n + 3 (mod 5).
2002:9. A sequence whose entries are 0 and 1 has the property that,
if each 0 is replaced by 01 and each 1 by 001, then the sequence remains
unchanged. Thus, it starts out as 010010101001 . . . . What is the 2002th
term of the sequence?
Solution. Let us dene nite sequences as follows. Suppose that S1 = 0.
Then, for each k 2, Sk is obtained by replacing each 0 in Sk1 by 01 and
each 1 in Sk1 by 001. Thus,
S1 = 0; S2 = 01; S3 = 01001; S4 = 010010101001;
S5 = 01001010100101001010010101001;
Each Sk1 is a prex of Sk ; in fact, it can be shown that, for each k 3,
Sk = Sk1 Sk2 Sk1 ,
where indicates juxtaposition. The respective number of symbols in Sk for
k = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 is equal to 1, 2, 5, 12, 29, 70, 169, 408, 985,
2378.
The 2002th entry in the given innite sequence is equal to the 2002th
entry in S10 , which is equal to the (2002 985 408)th = (609)th entry in
S9 . This in turn is equal to the (609 408 169)th = (32)th entry in S8 ,
which is equal to the (32)th entry of S6 , or the third entry of S3 . Hence, the
desired entry is 0.
2003:6. A set of n lightbulbs, each with an on-o switch, numbered
1, 2, . . . , n are arranged in a line. All are initially o. Switch 1 can be
operated at any time to turn its bulb on or o. Switch 2 can turn bulb 2 on
or o if and only if bulb 1 is o; otherwise, it does not function. For k 3,
switch k can turn bulb k on or o if and only if bulb k 1 is o and bulbs
1, 2, . . . , k 2 are all on; otherwise it does not function.
(a) Prove that there is an algorithm that will turn all of the bulbs on.
(b) If xn is the length of the shortest algorithm that will turn on all n
bulbs when they are initially o, determine the largest prime divisor
of 3xn + 1 when n is odd.
Solution. (a) Clearly x1 = 1 and x2 = 2. Let n 3. The only way that
bulb n can be turned on is for bulb n 1 to be o and for bulbs 1, 2, . . . , n 2
to be turned on. Once bulb n is turned on, then we need get bulb n 1
turned on. The only way to do this is to turn o bulb n 2; but for switch
n 2 to work, we need to have bulb n 3 turned o. So before we can think
about dealing with bulb n 1, we need to get the rst n 2 bulbs turned
o. Then we will be in the same situation as the outset with n 1 rather
than n bulbs. Thus the process has the following steps: (1) Turn on bulbs
 10. COMBINATORICS AND FINITE MATHEMATICS 155

1, . . . , n 2; (2) Turn on bulb n; (3) Turn o bulbs n 2, . . . , 1; (3) Turn on

bulbs 1, 2, . . . , n 1. So if, for each positive integer k, yk is the length of the
shortest algorithm to turn o the rst k bulbs after all are lit, then
xn = xn2 + 1 + yn2 + xn1 .
Since we can clearly nd an algorithm for turning on or o bulb 1, an induc-
tion argument will provide an algorithm for turning any number of bulbs on
or o.

We show that xn = yn for n = 1, 2, . . . . Suppose that we have an

algorithm that turns all the bulbs on. We prove by induction that at each
step we can legitimately reverse the whole sequence to get all the bulbs o
again. Clearly, the rst step is to turn either bulb 1 or bulb 2 on; since
the switch is functioning, we can turn the bulb o again. Suppose that we
can reverse the rst k 1 steps and are at the kth step. Then the switch
that operates the bulb at that step is functioning and can restore us to the
situation at the end of the (k 1)th step. By the induction hypothesis, we
can go back to having all the bulbs o. Hence, given the bulbs all on, we
can reverse the steps of the algorithm to get the bulbs o again. A similar
argument allows us to reverse the algorithm that turns the bulbs o. Thus,
for each turning-on algorithm there is a turning-o algorithm of equal length,
and vice versa. Thus xn = yn .
We have that xn = xn1 + 2xn2 + 1 for n 3. By, induction, we show
that, for m = 1, 2, . . . ,
x2m = 2x2m1 and x2m+1 = 2x2m + 1 = 4x2m1 + 1.
This is true for m = 1. Suppose it is true for m 1. Then
x2(m+1) = x2m+1 + 2x2m + 1 = 2(x2m + 1) + 4x2m1
= 2(x2m + 2x2m1 + 1) = 2x2m+1 ,
and
x2(m+1)+1 = x2(m+1) + 2x2m+1 + 1 = 2x2m+1 + 4x2m + 3
= 2(x2m+1 + 2x2m + 1) + 1 = 2x2(m+1) + 1.
Hence, for m 1,
3x2m+1 + 1 = 4(3x2m1 + 1) = = 4m (3x1 + 1) = 4m+1 = 22(m+1) .
Thus, the largest prime divisor is 2.
2004:4. Let n be a positive integer exceeding 1. How many permutations
{a1 , a2 , . . . , an } of {1, 2, . . . , n} are there which maximize the value of the sum
|a2 a1 | + |a3 a2 | + + |ai+1 ai | + + |an an1 |
over all permutations? What is the value of this maximum sum?
The solution to this problem appears in Chap. 3.
156 10. COMBINATORICS AND FINITE MATHEMATICS

2005:10. Let n be a positive integer exceeding 1. Prove that, if a graph

with 2n+1 vertices has at least 3n+1 edges, then the graph contains a circuit
(i.e., a closed non-self-intersecting chain of edges whose terminal point is its
initial point) with an even number of edges. Prove that this statement does
not hold if the number of edges is only 3n.
Solution 1. If there are two vertices joined by two separate edges, then
the two edges together constitute a chain with two edges. If there are two
vertices joined by three distinct chains of edges, no pair sharing a vertex other
than the endpoints, then the number of edges in two of the chains have the
same parity, and these two chains together constitute a circuit with evenly
many edges. Henceforth, we assume that neither of these situations occurs.
We establish the general result by induction.
The result can be checked for n = 2. Then we have 5 vertices and at
least 7 edges (but not more than 10). If we can nd an even circuit among
the seven edges, then a fortiori we can nd one among more edges. Suppose
that the edges P Q and P R and one other edge are missing. Then
one of the circuits P S R T P , Q S T R Q,
P S T Q P will serve.
Suppose that the result holds for 2 n m 1. We may assume
that we have a graph G that contains no instances where two separate edges
join the same pair of vertices and no two vertices are connected by more
than two distinct chains. Since 3n + 1 > 2n, the graph is not a tree, and
therefore must contain at least one circuit. Consider one of these circuits,
L. If it has evenly many edges, the result holds. Suppose that it has oddly
many edges, say 2k + 1 with k 1. Since any two vertices in the circuit
are joined by at most two chains (the two chains that make up the circuit),
there are exactly 2k + 1 edges joining pairs of vertices in the circuit. Apart
from the circuit, there are (2m + 1) (2k + 1) = 2(m k) vertices and
(3m + 1) (2k + 1) = 3(m k) + k 3(m k) + 1 edges.
We now create a new graph G , by coalescing all the vertices and edges
of L into a single vertex v and retaining all the other edges and vertices of G.
This graph G contains 2(m k) + 1 vertices and at least 3(m k) + 1 edges,
and so by the induction hypothesis, it contains a circuit M with an even
number of edges. If this circuit does not contain v, then it is a circuit in the
original graph G, which thus has a circuit with evenly many edges. If the
circuit does contain v, it can be lifted to a chain in G joining two vertices of
L by a chain of edges in G . But these two vertices of L must coincide, for
otherwise there would be three chains joining these vertices. Hence we get a
circuit, all of whose edges lie in G ; this circuit has evenly many edges. The
result now follows by induction.
Here is a counterexample with 3n edges. Consider 2n + 1 vertices parti-
tioned into a singleton and n pairs. Join each pair with an edge and join the
singleton to each of the other vertices with a single edge to obtain a graph
with 2n + 1 vertices, 3n edges whose only circuits are triangles.
 10. COMBINATORICS AND FINITE MATHEMATICS 157

Solution 2. [J. Tsimerman] We can assume wolog that G has no double

edges. For any graph H, let k(H) be the number of circuits minus the
number of components (two vertices being in the same component if and
only if they are connected by a chain of edges). Let G0 be the graph with
2n + 1 vertices and no edges. Then k(G0 ) = (2n + 1). Suppose that edges
are added one at a time to obtain a succession Gi of graphs culminating in
the graph G = G3n+1 with 2n + 1 vertices and 3n + 1 edges. Since adding
an edge either reduces the number of components (when it connects two
vertices of separate components) or increases the number of circuits (when it
connects two vertices in the same component), k(Gi+1 ) k(Gi ) + 1. Hence
k(G) (2n + 1) + (3n + 1) = n. Thus, the number of circuits in G is at
least equal to the number of components in G plus n, which is at least n + 1.
Thus, G has at least n + 1 circuits.
Since 3(n + 1) > 3n + 1, there must be two circuits that share an edge.
If either circuit has an even number of edges, then the problem is solved.
Otherwise, each has an odd number of edges. Suppose they share exactly
one edge C0 D0 . Indicating the two circuits by the numerals 1 and 2 and
letting k denote a path not containing a repeated edge in the circuit
with numeral k, we can form a circuit with evenly many edges by cutting
out the common edge:
C0 1 D0 2 C0 .
Since the deleted edge was counted twice in summing the edges of the two
circuits, the new circuit has evenly many edges.
However, the two circuits could have more than one edge in common. We
can still perform the above process to delete a single edge to get a single closed
path that will have some edges traversed twice. This process of eliminating a
single edge may present parity problems, so we present a method of dealing
with two repeated edges C0 D0 and C1 D1 simultaneously. If there
are two edges in common, the closed path can be a double circuit of one of
two types.
The rst type of closed path is:
C0 D0 1C1 D1 1C0 D0 2D1 C1 2C0 .
We can remove both edges (C0 , D0 ) and (C1 , D1 ) to get a closed path with
evenly many edges:
C0 2 C1 1 D0 2 D1 1 C0 .

The second type of closed path is:

C0 D0 1C1 D1 1C0 D0 2C1 D1 2C0 .
Let there be p edges in the path D0 1 C1 , q edges in D1 1 C0 ,
r edges in D0 2 C1 and s edges in D1 2 C0 . Note that p + q
and r + s are both odd. If p and r have the opposite parity, then p and s
have the same parity and C0 D0 1 C1 D1 2 C0 is a
158 10. COMBINATORICS AND FINITE MATHEMATICS

closed path with evenly many edges. If p and r have the same parity, then
D0 1 C1 2 D0 is a closed path with evenly many edges.
Thus, we can take our two circuits and by bifurcating at a common edge
obtain a single closed path with some nite number of edges traversed twice.
By the foregoing process, we can eliminate the repeated edges one or two at
a time until we end up with the required type of circuit.
A counterexample can be obtained by taking a graph with vertices A1 ,
. . . , An , B0 , B1 , . . . , Bn , with edges joining the vertex pairs (Ai , Bi1 ),
(Ai , Bi ) and (Bi1 , Bi ) for 1 i n.
2006:1.
(a) Suppose that a 66 square grid of unit squares (chessboard) is tiled
by 1 2 rectangles (dominoes). Prove that it can be decomposed
into two rectangles, tiled by disjoint subsets of the dominoes.
(b) Is the same thing true for an 8 8 array?
Solution. (a) There are 18 dominoes and 10 interior lines in the grid.
Suppose, if possible, that the decomposition does not occur for some dispo-
sition of dominoes. Then each of the lines must be straddled by at least one
domino. We argue that, in fact, at least two dominoes must straddle each
line. Since no domino can straddle more than one line, this would require 20
dominoes and so yield a contradiction.
Each interior line has six segments. For a line next to the side of the
square grid, an adjacent domino between it and the side must either cross
one segment or be adjacent to two segments. Since the number of segments
is even, evenly many dominoes (at least two) must cross a segment. For
the next line in, an adjacent domino must be adjacent to two segments, be
adjacent to one segment and cross the previous line, or cross one segment.
Since the number of dominoes straddling the previous line is even, there must
be evenly many that cross the segment. In this way, we can work our way
from one line to the next.
(b) Figure 10.1 shows a possible arrangement of the dominoes.

2006:5. Suppose that you have a 3 3 grid of squares. A line is a set of

three squares in the same row, the same column or the same diagonal; thus,
there are eight lines.
Two players A and B play a game. They take alternate turns, A putting
a 0 in any unoccupied square of the grid and B putting a 1. The rst player
is A, and the game cannot go on for more than nine moves. (The play is
similar to noughts and crosses, or tic-tac-toe.) A move is legitimate if it does
not result in two lines of squares being lled in with dierent sums. The
winner is the last player to make a legitimate move.
(For example, if there are three 0s down the diagonal, then B can place
a 1 in any square provided it completes no other line, for then the sum would
dier from the diagonal sum. If there are two zeros at the top of the main
 10. COMBINATORICS AND FINITE MATHEMATICS 159

Fig. 10.1. A nondecomposable tiling of dominoes

diagonal and two ones at the left of the bottom line, then the lower right
square cannot be lled by either player, as it would result in two lines with
dierent sums.)
(a) What is the maximum number of legitimate moves possible in a
game?
(b) What is the minimum number of legitimate moves possible in a
game that would not leave a legitimate move available for the next
player?
(c) Which player has a winning strategy? Explain.
Solution. (a) A game cannot continue to nine moves. Otherwise, each
line sum must be three times the value on the centre square of the grid
(why?) and so must be 0 or 3. But some line must contain both zeros and
ones, yielding a contradiction. [An alternative argument is that, if the array is
lled, not all the rows can have the same numbers of 0s and 1s, and therefore
cannot have the same sums.] However, an eight-move grid is possible, in
which one player selects the corner squares and the other the squares in the
middle of the edges.
(b) Consider any game after four moves have occurred and it is As turn to
play a zero. Suppose, rst of all, that no lines have been lled with numbers.
The only way an inaccessible square can occur is if it is the intersection of
two lines each having the other two squares lled in. This can happen in at
most one way. So A would have at least four possible squares to ll in. On
160 10. COMBINATORICS AND FINITE MATHEMATICS

the other hand, if three of the rst four moves complete a line, the fourth
number can bar at most three squares for A in the three lines determined by
the fourth number and one of the other three numbers. Thus, A would have
at least two possible positions to ll. Thus, a game must go to at least ve
moves.

A ve-move game can be obtained when A has placed three 0s down

the left column and B has 1 in the centre square and another square of the
middle column. Each remaining position is closed to B as it would complete
a line whose sum is not 0.
(c) A has a winning strategy. Let A begin by putting 0 in the centre
square. After three moves, A can guarantee either a row, column or diagonal
with two 0s and one 1. Suppose that we have 0, 0, 1 down a diagonal.
Wherever B puts his 1, A can complete the line through the centre containing
it. For each of Bs responses, A can make a move that stymies B. Suppose
that we have 1, 0, 0 down a column through the centre parallel to a side of
the grid. If B puts 1 in a row with a 0, then A can complete that row. B has
one possible move, after which A can still move. If B puts 1 in a row with
another 1, A puts a 0 at the other end of the diagonal. B has only one move,
to which A can respond.
2007:1. A m n rectangular array of distinct real numbers has the
property that the numbers in each row increase from left to right. The
entries in each column, individually, are rearranged so that the numbers in
each column increase from top to bottom. Prove that in the nal array, the
numbers in each row will increase from left to right.
Solution 1. We prove this by induction. Note that the permutation
yielding the nal arrangement of each column is uniquely determined, so
that if we can perform a sequence of transposition (switches) resulting in the
entries increasing from top to bottom, the composite of these transpositions
is the required permutation.
We can arrange the rows so that the rst column is increasing from top
to bottom; all the rows will still be increasing from left to right. Suppose
that we have performed a sequence of rearrangements of sets of columns so
that (a) each row is increasing, and (b) the rst k columns are increasing for
1 k n 1. Let the entries in the kth column be a1 , a2 , . . . , am and in the
(k + 1)th column be b1 , b2 , . . . , bm . We have that a1 a2 a3 am
and ai bi for 1 i m.
Suppose that br is the minimum of all the bi (1 i m). We interchange
the elements in the rst and rth rows of the jth column for k + 1 j n.
Since a1 ar br and ar br b1 , the rst and rth rows are still increasing.
Let bs be the minimum of all the bi except for br ; interchange the elements
in the second and sth rows of the jth column for k + 1 j n. Since
a2 as bs and as bs b2 , the new second and sth rows are increasing.
 10. COMBINATORICS AND FINITE MATHEMATICS 161

Observe that br bs bi for i = r, s, so that the (k + 1)th column is

increasing down to the third entry.
We can continue in this way, moving the third smallest bi to the third
row, and so on, ending up with changing the order of the columns from the
(k + 1)th to the nth and keeping the rows increasing. The result follows by
induction on k.
Solution 2. [M. Cai] Let ai,j be the (i, j)th entry in the array before the
rearrangement and bi,j the (i, j)th entry after the rearrangement. Then
b1,j b2,j bm,j
for 1 j n. We need to show that, for each i with 1 i m,
bi,1 bi,2 bi,n .

Let 1 j n 1. For 1 r m, we have that

bm,j+1 ar,j+1 and ar,j ar,j+1 .
There exist s with 1 s m for which bm,j = as,j . Hence, for 1 j n 1,
bm,j = as,j as,j+1 bm,j+1 .

Suppose, as an induction hypothesis, it has been established that bi,j

bi,j+1 for 2 k + 1 i m and 1 j n 1. Then bk,j = at,j at,j+1 for
some t with 1 t m. Since the set of numbers {bi,j+1 : k i m} is the
set of the largest m k + 1 numbers of the set {ai,j+1 : 1 i m}, we must
have that at,j+1 bt,j+1 . The result follows.
Solution 3. There is nothing to prove when n = 1. Suppose that n 2,
that a given column, not the last, is A = (a1 , a2 , . . . , am )t to begin with and
B = (b1 , b2 , . . . , bm )t after the rearrangement, so that b1 b2 bm .
Let the elements in a column to the right of this be C = (c1 , c2 , . . . , cm )t to
begin with and D = (d1 , d2 , . . . , dm )t after the rearrangement. We have that
d1 d2 dm and ai ci for 1 i m. We need to show that bk dk
for 1 k m.
Since bk is the kth largest number in the given column A to the left,
there are m k + 1 elements in that column not less than it. Hence there
are at least m k + 1 elements in the column C to the right that are not less
than bk , and these elements include dk , dk+1 , . . . , dm . Hence bk dk .
2007:3. Prove that the set {1, 2, . . . , n} can be partitioned into k subsets
with the same sum if and only if k divides 12 n(n + 1) and n 2k 1.
Solution. The necessity of the conditions follows from the fact that the
sum of all the numbers in the set is 12 n(n + 1) = ks, where s is the sum of the
numbers in each subset, and there is at most one subset with a single element;
hence the number k of subsets is at most 12 (n 1) + 1. (Alternatively: Since
n must lie in one of the subsets, the sum of each subset is at least n, and so
kn 12 n(n + 1). Therefore n 2k 1.)
162 10. COMBINATORICS AND FINITE MATHEMATICS

On the other hand, let k, s and n be positive integers for which ks =

2 n(n + 1) and 2k 1 n. Then ks 2 (2k 1)(2k) from which s 2k 1,
1 1

k 2 (s + 1) and 2 n(n + 1) = ks 2 (s + 1)s. Thus both n and k

1 1 1
do not
exceed s.
For k = 1 and each value of s, the conditions holds with n = s, which
corresponds to the partition given by the single set (1, 2, . . . , s) = (1, 2, . . . , n).
The only triples (k, n, s) satisfying the conditions with s 3 are (k, n, s) =
(1, 1, 1), (1, 2, 3) and (2, 3, 3), corresponding respectively to the partitions
({1}), ({1, 2}) and ({1, 2}, {3}).
We establish the result by double induction on s and k. Suppose that
the result has been established for all possible subset sums up to s 1 3, as
well as for s and possible number of subsets in the partition up to k 1 1.
Consider the case where the conditions are satised for (k, n, s), and we have
to nd a suitable partition. Let n = 2k1, so that s = n(n+1)/(2k) = 2k1.
Then a partition of the required type consists of the singleton {2k 1} and
the k 1 pairs {i, 2k 1 i} with 1 i k 1. If n = 2k, then s =
n(n + 1)/(2k) = 2k + 1 and we can use the k pairs {i, 2k + 1 i} with
1 i k. We use induction on n to establish the result for n > 2k.
Suppose that 2k < n < 4k 1 and that s = n(n + 1)/(2k) is an integer.
We have that 2 < n/k < (n + 1)/k < 4 so that
 
2ks n+1
n+1= < s = 2n < 2n.
n 4k
Suppose that n = s n 1, so that 0 < n < n 1.
First, let s be odd. The sum of the numbers from 1 to s n 1 inclusive
is equal to
 
(s n)(s n 1) s2 (2n + 1)s n(n + 1) s1
= + =s n+k .
2 2 2 2
Let k  = 12 (s 1) n + k; note that k  < k and that
n (2k  1) = (s n 1) (s 2 2n + 2k) = n 2k + 1 > 0,
so that n > 2k  1. By the induction hypothesis, we can partition {1, 2,
. . . , s n 1} into k  subsets, each of whose sum is s. Let r = 12 (s 1), and
adjoin the nr = kk  doubletons {sn, n}, {sn+1, n1}, . . . , {r, r+1} to
get the desired partition of {1, 2, . . . , n} into k subsets with the same sum s.
Now let s be even. Then
(s n)(s n 1) s
= [s 1 2n + 2k].
2 2
Let k  = 2k 2n + s 1 = (n 2k)(n + 1 2k)/(2k) and s = s/2. Note
that k  need not be less than k, but in any case the new purported subset
sum s is less than s. Since s < 2(n 1), 2s = s > 2(s n 1) = 2n . By
the induction hypothesis, we can partition the set {1, 2, . . . , s n 1} into
k  subsets whose sums are all s . Since k  is odd and since s > s n 1, we
 10. COMBINATORICS AND FINITE MATHEMATICS 163

can augment this family of subsets by the singleton {s } to get evenly many
subsets with sum s . Pair them o to get 12 (k  + 1) = k n + s subsets with
sum s, and further augment the family with the n s doubletons {s n, n},
{s n + 1, n 1} . . . , {s 1, s + 1} to get a partition of {1, 2, . . . , n} into
k sets with sum s.
Finally, suppose that n 4k 1. Let n = n 2k, k  = k. Then, if
s = n(n + 1)/(2k), then the sum s = n (n + 1)/2k is an integer equal to
(n 2k)(n 2k + 1) n(n + 1)
= (2n + 1) + 2k,
2k 2k
that is not less than 2k 1 > 0. Note that n = n 2k 2k 1 = 2k  1
and that
(n n )(n + n + 1)
s s = = 2n 2k + 1.
2k
Determine a partition of {1, 2, . . . , n 2k} into k subsets with sum s , and
adjoin to these subsets the k doubletons

{n 2k + 1, n}, {n 2k + 2, n 1}, . . . , {n k, n k + 1}.

The whole result is now established.

Comment. The case 2k < n < 4k 1 with k a divisor of 12 n(n + 1) is
not realizable for k = 2 and k = 4. For k 1 (mod 4) and s odd, a possible
value of n is 3k 1, and following the notation in the solution, we get the
possibilities  
3(3k 1)
(k, n, s) = k, 3k 1,
2
 
k 1 3(k 1) 3(3k 1)
(k  , n , s ) = , , .
4 2 2
For k 1 (mod 4) and s even, we have the possibilities
 
3(3k + 1)
(k, n, s) = k, 3k,
2
 
k + 1 3(k + 1) 3(3k + 1)
(k  , n , s ) = , , .
2 2 4
For k 3 (mod 4) and s odd, we have the possibilities
 
3(3k + 1)
(k, n, s) = k, 3k,
2
 
k + 1 3(k + 1) 3(3k + 1)
(k  , n , s ) = , , .
4 2 2
For k 3 (mod 4) and s even, we have the possibilities
 
3(3k 1)
(k, n, s) = k, 3k 1,
2
164 10. COMBINATORICS AND FINITE MATHEMATICS

 
k 1 3(k 1) 3(3k 1)
(k  , n , s ) =, , .
2 2 4
This exhausts the possibilities for k 5. For k = 6, there is one possibility:
(k, n, s) = (6, 15, 20); (k  , n , s ) = (1, 4, 10)
with the partition
{1, 2, 3, 4, 10}, {5, 15}, {6, 14}, {7, 13}, {8, 12}, {9, 11}.

In dealing with the case of even s, it can indeed occur that k  > k (which
is why we have to extend the induction to s as well as k). Applying the
reduction to (k, n, s) = (14, 48, 84) leads to (k  , n , s ) = (15, 35, 42) which in
turn reduces to (k  , n , s ) = (1, 6, 21). The partition that corresponds to
(14, 48, 84) turns out to be
{1, 2, 3, 4, 5, 6, 7, 21, 42}, {7, 8, 34, 35}, {9, 10, 32, 33}, {11, 12, 30, 31},
{13, 14, 28, 29}, {15, 16, 26, 27}, {17, 18, 24, 25}, {19, 20, 22, 23}, {36, 48}, {37, 47},

{38, 46}, {39, 45}, {40, 44}, {41, 43}.

2008:6. 2008 circular coins, possibly of dierent diameters, are placed

on the surface of a at table in such a way that no coin is on top of another
coin. What is the largest number of points at which two of the coins could
be touching?

Solution. Suppose that there are n coins where n 3. We show that

the answer is 3n 6. When n = 2008, this number is equal to 6018. We
can achieve this as follows. Start with three mutually touching coins. Insert
a coin of suitable diameter in the middle so that it touches all three coins.
Continue adding coins, each in the middle of three mutually touching coins
that have already been placed.

To show that this is an upper bound, construct a graph whose vertices

are the centres of the coins, with two vertices connected by an edge if and
only if the their two coins touch. This is a planar graph with n vertices, with
the number of edges equal to the number of points where two coins touch.
We use the following result from graph theory: Any planar graph with V
vertices, E edges and F faces (including the unbounded face) has at most
3V 6 edges.
Since each face has at least three edges, and each edge belongs to two
faces, we must have 2E 3F . Hence, using Eulers equation,
2 1
2 = V E + F V E + E = V E,
3 3
whence E 3V 6.
[Problem 2008:6 was contributed by David Arthur.]
 10. COMBINATORICS AND FINITE MATHEMATICS 165

2008:10. A point is chosen at random (with the uniform distribution)

on each side of a unit square. What is the probability that the four points
are the vertices of a quadrilateral with area exceeding 12 ?
The solution to this problem appears in Chap. 8.
2009:10. Suppose that a path on an m n grid consisting of the lattice
points {(x, y) : 1 x m, 1 y n} (x and y both integers) consisting of
mn 1 unit segments begins at the point (1, 1), passes through each point
of the grid exactly once, does not intersect itself and nishes at the point
(m, n). Show that the path partitions the rectangle bounded by the lines
x = 1, x = m, y = 1, y = n into two subsets of equal area, the rst consisting
of regions opening to the left or up, and the second consisting of regions
opening to the right or down.
Solution. Embed the grid into a rectangle composed of mn unit square
cells, each with a grid point as centre. Extend the path horizontally to the left
and right end of this rectangle from the respective points (1, 1) and (m, n).
It is equivalent to show that this rectangle is decomposed by this extended
path into two subsets of equal area. As the path proceeds, at any grid point,
it either turns left through a right angle, or turns right through a right angle
or proceeds straight ahead. Since it ends up going in the same direction as
it started, it makes an equal number of left and right turns. The path keeps
the region opening left or up on its left and the region opening right or down
on its right. At each left turn, the path splits the area of the cell it is in into
two subsets of areas 1/4 on the left and 3/4 on the right; at each right turn,
it splits the cell in the opposite way. If the path goes straight ahead, it splits
the cell area in half. It follows from this that the area on the left side of the
path equals the area on the right side of the path, and the result holds.
2010:2. Let u0 = 1, u1 = 2 and un+1 = 2un + un1 for n 1. Prove
that, for every nonnegative integer n,
  (i + j + k)! 
un = : i, j, k 0, i + j + 2k = n .
i!j!k!
Solution 1. Suppose that we have a supply of white and of blue coaches,
each of length 1, and of red coaches, each of length 2; the coaches of each
colour are indistinguishable. Let vn be the number of trains of total length n
that can be made up of red, white and blue coaches of total length n. Then
v0 = 1, v1 = 2 and v2 = 5 (R, WW, WB, BW, BB). In general, for n 1,
we can get a train of length n + 1 by appending either a white or a blue
coach to a train of length n or a red coach to a train of length n 1, so that
vn+1 = 2vn + vn1 . Therefore vn = un for n 0.
We can count vn in another way. Suppose that the train consists of
i white coaches, j blue coaches and k red coaches, so that i + j + 2k = n.
There are (i+j +k)! ways of arranging the coaches in order; any permutation
of the i white coaches among themselves, the j blue coaches among themselves
166 10. COMBINATORICS AND FINITE MATHEMATICS

and k red coaches among themselves does not change the train. Therefore
  (i + j + k)! 
un = : i, j, k 0, i + j + 2k = n .
i!j!k!

Solution 2. Let f (t) = n=0 un tn . Then

f (t) = u0 + u1 t + (2u1 + u0 )t2 + (2u2 + u1 )t3 +

= u0 + u1 t + 2t(f (t) u0 ) + t2 f (t)
= u0 + (u1 2u0 )t + (2t + t2 )f (t)
= 1 + (2t + t2 )f (t),
whence

1 1
f (t) = = = (t + t + t2 )n
1 2t t2 1 t t t2 n=0
  
(i + j + k)! i+j+2k
= t : i, j, k 0, i + j + 2k = n .
n=0
i!j!k!

For a given exponent m 0, we need to determine the coecient of tm in

the right sum. The values of k range between 0 and m/2, and for each we
get a term in tm when i + j + k = m k, so the coecient of tm is
m/2  
 (i + j + k)!
; i, j 0, i + j + k = m k
i!j!k!
k=0
  (i + j + k)! 
; i, j, k 0, i + j + k = m k
i!j!k!
Replacing m by n, we nd that
  
(i + j + k)!
f (t) = : i, j, k 0, i + j + 2k = n ,
i!j!k!
from which the desired result follows.
Solution 3. Let wn be the sum in the problem. It is straightforward
to check that u0 = w0 and u1 = w1 . We show that, for n 1, wn+1 =
2wn + wn1 from which it follows by induction that un = wn for each n. By
convention, let (1)! = . Then, for i, j, k 0 and i + j + 2k = n + 1, we
have that
(i + j + k)! (i + j + k)(i + j + k 1)!
=
i!j!k! i!j!k!
(i + j + k 1)! (i + j + k 1)! (i + j + k 1)!
= + + ,
(i 1)!j!k! i!(j 1)!k! i!j!(k 1)!
 10. COMBINATORICS AND FINITE MATHEMATICS 167

whence
  (i + j + k 1)! 
wn+1 = : i, j, k 0, (i 1) + j + 2k = n
(i 1)!j!k!
  (i + j + k 1)! 
+ : i, j, k 0, i + (j 1) + 2k = n
i!(j 1)!k!
  (i + j + k 1)! 
+ : i, j, k 0, i + j + 2(k 1) = n 1
i!j!(k 1)!
= wn + wn + wn1 = 2wn + wn1
as desired.
[Problem 2010:2 is #10663 in American Mathematical Monthly 105:5
(May, 1998), 464; 107:4 (April, 2000), 370371.]
2010:4. The plane is partitioned into n regions by three families of
parallel lines. What is the least number of lines to ensure that n 2010?

Solution. Suppose that there are x, y and z lines in the three families.
Assume that no point is common to three distinct lines. The x + y lines of
the rst two families partition the plane into (x + 1)(y + 1) regions. Let be
one of the lines of the third family. It is cut into x + y + 1 parts by the lines in
the rst two families, so the number of regions is increased by x + y + 1. Since
this happens z times, the number of regions that the plane is partitioned into
by the three families of lines is
n = (x + 1)(y + 1) + z(x + y + 1) = (x + y + z) + (xy + yz + zx) + 1.

Let u = x + y + z and v = xy + yz + zx. Then (by the Cauchy-Schwarz

Inequality for example), v x2 + y 2 + z 2 , so that u2 = x2 + y 2 + z 2 + 2v 3v.
Therefore, n u + 13 u2 + 1. This is a value less than 2003 when u = 76.
However, when (x, y, z) = (26, 26, 25), then u = 77, v = 1976 and n = 2044.
Therefore, we need at least 77 lines, but a suitably chosen set of 77 lines will
suce.
2011:3. Suppose that S is a set of n nonzero real numbers such that
exactly p of them are positive and exactly q are negative. Determine all the
pairs (n, p) such that exactly half of the threefold products abc of distinct
elements a, b, c of S are positive.
Solution 1. A threefold product is positive if and only if all factors are
positive or exactly one factor is positive; it is negative if and only if all factors
are negative or exactly one factor is negative. This leads to the equation
       
p q q p
+p = +q .
3 2 3 2
This simplies to
(p q)[(p q)2 3(p + q) + 2] = 0.
168 10. COMBINATORICS AND FINITE MATHEMATICS

We have that n = p + q; let m = |p q|. Then the condition is satised if and

only if either p = q and n = 2p is even, or n = 13 (m2 + 2), m is not divisible
by 3, and p and q are equal to (m + 1)(m + 2)/6 and (m 1)(m 2)/6 in
some order.
Solution 2. Since half of the total number of products are to be positive,
we obtain the condition
     
p np 1 n
+p = .
3 2 2 3
This reduces to
2p(p 1)(p 2) + 6p(n p)(n p 1) = n(n 1)(n 2),
and ultimately to
0 = (8p3 12p2 n + 6pn2 n3 ) + (6pn + 3n2 ) + (4p 2n)
= (2p n)[(2p n)2 3n + 2].
Therefore, we must have either that 2p = n or n = 13 (m2 + 2) where m =
|2p n|. The argument can be completed as in Solution 1.
Solution 3. There is no loss of generality in assuming that all the positive
numbers are +1 and the negative numbers are 1. Then, when exactly half
the threefold products are positive, the sum of all the threefold products
vanishes. Let m = p q, so that m is the sum of the numbers in S. Then
  
m3 = {a3 : a S} + 3 {a2 b : a, b S} + 6 {abc : a, b, c S}
= m + 3(n 1)m + 0
whence either m = 0 or m2 = 3n 2. This leads to the same conclusion as
in Solution 1.
[Problem 2011:3 is #10683 in the American Mathematical Monthly 105:8
(October, 1998), 768; 107:8 (October, 2000), 754.]
2011:6. [The problem posed on the competition is not available for
publication, as it appeared as Enigma problem 1610 in the August 25, 2010
issue of the New Scientist. It concerned the determination of the nal score
of a three game badminton match where the score of each competitor was in
arithmetic progression. The full statement of the problem is available on the
internet.]
The solution to this problem appears in Chap. 2.
2011:7. Suppose that there are 2011 students in a school and that each
student has a certain number of friends among his schoolmates. It is assumed
that if A is a friend of B, then B is a friend of A, and also that there may
exist certain pairs that are not friends. Prove that there is a nonvoid subset
S of these students for which every student in the school has an even number
of friends in S.
 10. COMBINATORICS AND FINITE MATHEMATICS 169

The solution to this problem appears in Chap. 7.

2011:10. Suppose that p is an odd prime. Determine the number of
subsets S contained in {1, 2, . . . , 2p 1, 2p} for which (a) S has exactly p
elements, and (b) the sum of the elements of S is a multiple of p. [3]
The solution to this problem appears in Chap. 2.
2012:9. In a round-robin tournament of n 2 teams, each pair of teams
plays exactly one game that results in a win for one team and a loss for the
other (there are no ties).
(a) Prove that the teams can be labelled t1 , t2 , . . . , tn , so that, for each
i with 1 i n 1, team ti beats ti+1 .
(b) Suppose that a team t has the property that, for each other team
u, one can nd a chain u1 , u2 , . . . , um of (possibly zero) distinct
teams for which t beats u1 , ui beats ui+1 for 1 i m 1 and um
beats u. Prove that all of the n teams can be ordered as in (a) so
that t = t1 and each ti beats ti+1 for 1 i n 1.
(c) Let T denote the set of teams that can be labelled as t1 in an
ordering of teams as in (a). Prove that, in any ordering of teams
as in (a), all the teams in T occur before all the teams that are not
in T .
Preliminaries. Let a > b denote that team a beats team b. The result is
clear when n = 2. Assume that the result holds for all numbers of teams up
to n 1 2.
(a) Solution 1. By the induction hypothesis, we can order n 1 of the
teams so that t1 > t2 > > tn1 . Suppose that t is the nth team not
included in this list. If t either beats t1 or is beaten by tn1 then it can be
appended at one end of the list. Suppose therefore that t1 > t > tn1 . Let
i be the largest index for which ti > t. Then i < n 1 and we can insert t
between ti and ti+1 .
Solution 2. Let t be any team, A be the set of teams that beat t and B
be the set of teams that are beaten by t. We use induction. If A is nonvoid,
then the teams in A can be ordered as in (a) so that each team beats the
next; similarly, if B is nonvoid, its teams can be ordered in the same way.
The orderings in A and B with t interpolated give the required ordering of
all the teams.
(b) Solution 1. Suppose that t > u1 > u2 > > ur is a list of teams,
each beating the next, of maximum length. Suppose, if possible, that a team
v is not included in the list. Then v > ur . If there is a team ui with ui > v,
then by taking such a team with the maximum value of i, we can insert v
between ui and ui+1 , contradicting the maximality of the list. Therefore
v > ui for each i.
By the hypothesis on t, we can nd teams v1 , v2 , . . . , vs for which t >
v1 > v2 > v3 > > vs = v. Consider team vs1 . This team is not one
of the ui since it beats vs = v. Therefore, arguing as for v, we nd that
170 10. COMBINATORICS AND FINITE MATHEMATICS

vs1 > ur . If, for some i, ui > vs1 , then we could insert > vs1 > vs >
between ui and ui+1 in the list, contradicting maximality. Hence vs1 > ui
for each i. We can continue on in this way to argue in turn that vs2 , . . . ,
v1 are all distinct from the ui and beat each of the ui .
But then, we could form a chain
t > v 1 > v 2 > > v s > u1 > > ur = u
again contradicting the maximality. Hence any maximum chain contains all
the teams.
Solution 2. [J. Zung] We can consider the set T of teams as a directed
graph with root t where each pair of vertices is connected by a directed edge
a b, corresponding to team a beating team b. Let T1 be a minimum
spanning tree of T with root t, (i.e., T and T1 have the same vertices, T1
is connected, but T1 has no loops, and there is a path from t to any other
vertex). If every vertex in T1 has at most one exiting edge, then T1 is a path
of all the vertices. Otherwise, there are vertices u, v, w for which u v,
u w. Suppose, wolog, in T , that v w. We convert T1 to a new tree
T2 with one fewer vertex having more than one exiting edge by removing the
edge u w and inserting the edge v w.
We can continue the same process to get a succession of spanning trees
from which the sum over all vertices of the number of edges required to go
from t to those nodes increases. The process must terminate with a spanning
path.
(c) Solution. First note that any team x that beats a team t in T must
itself belong to T . For let an ordering of the all the teams be given and
append x at the beginning. We must then have an ordering
x > t > > x >
in which x appears twice and every other team once. It follows from this that
any team is at the end of a chain that begins with one of the x and consists
of distinct teams. Thus, by (b), any team beating a member of T must lie in
T . The desired result follows from this.
2015:2. Given 2n distinct points in space, the sum S of the lengths of
all the segments joining pairs of them is calculated. Then n of the points
are removed along with all the segments having at least one endpoint from
among them. Prove that the sum of the lengths of all the remaining segments
is less than 12 S.
The solution to this problem appears in Chap. 8.
 CHAPTER 11

Number Theory

2002:3. In how many ways can the rational 2002/2001 be written as the
product of two rationals of the form (n + 1)/n, where n is a positive integer?
Solution. We begin by proving a more general result. Let m be a positive
integer, and denote by (m) and (m + 1), the number of positive divisors
of m and m + 1 respectively. Suppose that
m+1 p+1 q+1
= ,
m p q
where p and q are positive integers exceeding m. Then (m + 1)pq = m(p + 1)
(q + 1), which reduces to (p m)(q m) = m(m + 1). It follows that
p = m + u and q = m + v, where uv = m(m + 1). Hence, every representation
of (m + 1)/m corresponds to a factorization of m(m + 1).
On the other hand, observe that, if uv = m(m + 1), then
m+u+1 m+v+1 m2 + m(u + v + 2) + uv + (u + v) + 1
=
m+u m+v m2 + m(u + v) + uv
m2 + (m + 1)(u + v) + m(m + 1) + 2m + 1
=
m2 + m(u + v) + m(m + 1)
(m + 1)2 + (m + 1)(u + v) + m(m + 1)
=
m2 + m(u + v) + m(m + 1)
(m + 1)[(m + 1) + (u + v) + m] m+1
= = .
m[m + (u + v) + m + 1] m
Hence, there is a one-one correspondence between representations and un-
ordered pairs (u, v) of complementary factors of m(m + 1). (Since m(m + 1)
is not square, u and v are always distinct.) Since m and m + 1 are coprime,
the number of factors of m(m + 1) is equal to (m) (m + 1), and so the
number of representations is equal to 12 (m) (m + 1).

Springer International Publishing Switzerland 2016 171

E.J. Barbeau, University of Toronto Mathematics Competition
(20012015), Problem Books in Mathematics,
DOI 10.1007/978-3-319-28106-3 11
172 11. NUMBER THEORY

Now consider the case that m = 2001. Since 2001 = 3 23 29,

(2001)=8; since 2002 = 2 7 11 13, (2002) = 16. Hence, the de-
sired number of representations is 64.
2003:4. Show that n divides the integer nearest to
(n + 1)!
.
e
The solution to this problem appears in Chap. 4.
2004:3. Suppose that u and v are positive integer divisors of the positive
integer n and that uv < n. Is it necessarily so that the greatest common
divisor of n/u and n/v exceeds 1?
Solution 1. Let n = ur = vs. Then uv < n v < r, u < s, so that
n2 = uvrs rs > n. Let the greatest common divisor of r and s be g and
the least common multiple of r and s be m. Then m n < rs = gm, so
that g > 1.
Solution 2. Let g = gcd (u, v), u = gs and v = gt. Then gst g 2 st < n
so that st < n/g. Now s and t are a coprime pair of integers, each of which
divides n/g. Therefore, n/g = dst for some d > 1. Therefore n/u = n/(gs) =
dt and n/v = n/(gt) = ds, so that n/u and n/v are divisible by d, and so
their greatest common divisor exceeds 1.
Solution 3. uv < n = nuv < n2 = n < (n/u)(n/v). Suppose, if
possible, that n/u and n/v have greatest common divisor 1. Then the least
common multiple of n/u and n/v must equal (n/u)(n/v). But n is a common
multiple of n/u and n/v, so that (n/u)(n/v) n, a contradiction. Hence the
greatest common divisor of n/u and n/v exceeds 1.
Solution 4. [D. Shiroko] If n/u and n/v be coprime, then there are
integers x and y for which (n/u)x + (n/v)y = 1, whence n(xv + yu) = uv.
Since n and uv are positive, then so is the integer xv + yu. But uv < n =
0 < xv + yu < 1, an impossibility. Hence the greatest common divisor of n/u
and n/v exceeds 1.
2004:9. Let ABCD be a convex quadrilateral for which all sides and
diagonals have rational length and AC and BD intersect at P . Prove that
AP , BP , CP , DP all have rational length.
The solution to this problem appears in Chap. 8.
2005:7. Let f (x) be a nonconstant polynomial that takes only integer
values when x is an integer, and let P be the set of all primes that divide
f (m) for at least one integer m. Prove that P is an innite set.
Solution 1. Let r be the greatest integer that divides all of the values
f (m), where m Z, and let g(x) = f (x)/m. Then g(x) is a nonconstant
polynomial that satises the condition of the problem, and the set of all
primes dividing at least one f (m) is the union of the set of primes dividing at
least one of g(m) and the nites et of primes dividing r. Thus, it is enough
to deal with g(x).
 11. NUMBER THEORY 173

Let {p1 , p2 , . . . , pk } be a nite set of distinct primes. Since the greatest

common divisor of the numbers g(m) for m Z is 1, there is a set S1 of
numbers m for which g(m) is not a multiple of p1 . Suppose, as an induction
hypotheses, 1 i k 1 and we have established that there is a set Si of
integers m for which g(m) is divisible by none of p1 , p2 , . . . , pi . For any
integers n and s, f (n) f (n + sp1 p2 pi ) mod p1 p2 pi . Let a Si ; then
a + sp1 p2 pi also belongs to Si for all s.
There is an integer b for which g(b) is not a multiple of pi+1 . Then
neither is g(n) a multiple of pi+1 when n b (mod pi+1 ). Since p1 p2 pi
and pi+1 are coprime, a + sp1 p2 pi runs through a complete set of residues
as s varies, and in particular is congruent to b for some values of s. Thus, Si
contains some number n for which g(n) is not a multiple of pi+1 . Hence the
set Si+1 = {m : g(m) 0pj (1 j i + 1)} is nonvoid.
Therefore there exists an integer m for which g(m) is divisible by none
of p1 , p2 , . . . , pk , for an arbitrary set of primes, the result follows.
n
Solution 2. Let f (x) = k ak xn . The number a0 = f (0) is rational.
Indeed, each of the numbers f (0), f (1), , f (n) is an integer; writing these
conditions out yields a system of n+1 linear equations with integer coecients
for the coecients a0 , a1 , . . . , an whose determinant is nonzero. The solution
of this equation consists of rational values. Hence all the coecients of f (x)
are rational. Multiply f (x) by the least common multiple of its denominators
to get a polynomial g(x) which takes integer values whenever x is an integer.
Suppose, if possible, that values of f (x) for integral x are divisible only by
primes p from a nite set Q. Then the same is true of g(x) for primes from
a nite set P consisting of the primes in Q along with the prime divisors of
the least common multiple. For each prime p  P , select a positive integer
ap such that pap does not divide g(0). Let N = {pap : p P }. Then, for
 bpinteger u, g(N u) g(0) 0 (mod N ). However, for all u, g(N u) =
each
p , where 0 bp ap . Since there are only nitely many numbers of
this type, some number must be assumed by g innitely often, yielding a
contradiction. (Alternatively: one could deduce that g(N u) N for all u
and get a contradiction of the fact that |g(N u)| tends to innity with u.)
Comment. There is another approach to showing that the coecients of
f (x) are rational. The proof is by induction on the degree. Suppose that
pk (x) is a polynomial of degree k assuming integer values at x = n, n +
1, . . . , n + k. Then, there are integers ci for which
     
x x x
pk (x) = ck,0 + ck,1 + + ck,k .
k k1 0
   x   
To see this, rst observe that xk , k1 , . . . , x0 constitute a basis for the
vector space of polynomials of degree not exceeding k. So there exist real ck,i
174 11. NUMBER THEORY

as specied. We prove by induction on k that the ck,i must in fact be integers.

The result is trivial when k = 0. Assume its truth for k 0. Suppose that
 
x
pk+1 (x) = ck+1,0 + + ck+1,k+1
k+1
takes integer values at x = n, n + 1, . . . , n + k + 1. Then
 
x
pk+1 (x + 1) pk+1 (x) = ck+1,0 + + ck+1,k
k
is a polynomial of degree k which takes integer values at n, n + 1, . . . , n + k,
and so ck+1,0 , . . . , ck+1,k are all integers. Hence,
   
n n
ck+1,k+1 = pk+1 (n) ck+1,0 ck+1,k
k+1 1
is also an integer.
Solution 3. [R. Barrington Leigh] Let n be the degree of f . It can
be established that if p be a prime and k a positive integer, then f (x)
f (x + pnk ) (mod pk ). The result of the problem holds for n = 0. Assume that
it holds for n = m1 and that f (x) has degree m. Let g(x) = f (x)f (x1),
so that the degree of g(x) is m 1. Then, modulo pk ,

nk
p

f (x + pnk ) f (x) = g(x + i)
i=1

p(n1)k

= (g(x + i) + g(x + i + p(n1)k ) + + g(x + i + (pk 1)p(n1)k ))
i=1

p(n1)k

pk g(x + i) 0.
i=1

(Note that this does not require the coecients to be integers.)

Suppose, if possible, that the set P of primes p that divide at least one
value of f (x) for integer x is nite, and that, for each p  P , the positive
integer a is chosen so that pa does not divide f (0). Let q = {pa : p P }.
Then pa does not divide f (0), nor any of the values f (q n ) for positive integer
n, as these are all congruent modulo pa . Since any prime divisor of f (q n )
belongs to P , it must be that f (q n ) is a divisor of q. But this contradicts the
fact that |f (q n )| becomes arbitrarily large with n.
2007:2. Determine distinct positive integers a, b, c, d, e such that the
ve numbers a, b2 , c3 , d4 , e5 constitute an arithmetic progression. (The
dierence between adjacent pairs is the same.)
Solution 1. One example is obtained by taking the arithmetic progression
(1, 9, 17, 25, 33) and multiplying by 324 530 1124 1720 to obtain

(a, b, c, d, e) = (324 530 1124 1720 , 313 515 1112 1710 , 38 510 118 177 , 36 58 116 175 , 35 56 115 174 ).
 11. NUMBER THEORY 175

Solution 2. [G. Siu] Let

(a, b, c, d, e) = (339724 6520 , 79712 6510 , 657 978 , 3976 655 , 975 654 ).
Then
(a, b2 , c3 , d4 , e5 ) = (33k, 49k, 65k, 81k, 97k)
where k = 6520 9724 .
Comment. Two solvers found the loophole that the arithmetic
progression itself could be constant, and gave the example (a, b, c, d, e) =
(n60 , n30 , n20 , n15 , n12 ) for an integer n 2.
2007:6. Let h(n) denote the number of nite sequences {a1 , a2 , . . . , ak }
of positive integers exceeding 1 for which k 1, a1 a2 ak and
n = a1 a2 ak . (For example, if n = 20, there are four such sequences {20},
{10, 2}, {5, 4} and {5, 2, 2} and h(20) = 4.)
Prove that

h(n)
= 1.
n=1
n2

The solution to this problem appears in Chap. 4.

2007:9. Which integers can be written in the form
(x + y + z)2
xyz
where x, y, z are positive integers?
Solution. Let F (x, y, z) be the expression in question. Wolog, suppose
that x y z. Suppose that F (x, y, z) = n. Then
nxyz = (x + y + z)2 = (x + y)2 + 2z(x + y) + z 2
from which z divides (x + y)2 . Let w = (x + y)2 /z. Then
(zx + zy + (x + y)2 )2 (z + x + y)2
F (x, y, w) = = = F (x, y, z).
zxy(x + y)2 zxy
If x + y z, then w x + y. So, if there is a representation of n, we can nd
one for which z x + y. Then
 
x y z 1 1 1
n= + + +2 + +
yz xz xy x y z
 
1 1 1 1 2 2 2
+ + + + + +
z x x y x y z
7 3
+ .
x z
If (x, y, z) = (1, 1, 1), then n = 9. Otherwise n < 9. We have that F (9, 9, 9) =
1, F (4, 4, 8) = 2, F (3, 3, 3) = 3, F (2, 2, 4) = 4, F (1, 4, 5) = 5, F (1, 2, 3) = 6,
F (1, 1, 2) = 8.
176 11. NUMBER THEORY

However, F (x, y, z) cannot equal 7. Supposing that 2 x y z

x + y 2y, we have that
(x + y + z)2 (x2 + y 2 ) + 2x(y + z) + z 2 + 2yz

xyz 2yz
2
x y x(y + z) z2
+ + + +2
2yz 2z yz 2yz
1 1 y(2z) z
+ + + +2
2 2 yz 2y
= 1 + 2 + 1 + 2 = 6.
Now let x = 1 and y z 1 + y. Since F (1, 1, 1) and F (1, 1, 2) dier from
7, y 2. But then
(2y + 1)2 4 1
F (1, y, y) = = 4 + + 2 < 7,
y2 y y
and
(2y + 2)2 4(y + 1) 4
F (1, y, y + 1) = = = 4 + < 7.
y(y + 1) y y
Hence F (x, y, z) = 7 is not possible. Therefore, only the integers 1, 2, 3, 4,
5, 6, 8 can be represented.
2008:8. Let b 2 be an integer base of numeration and let 1 r b1.
Determine the sum of all r-digit numbers of the form
ar1 ar2 . . . a2 a1 a0 ar1 br1 + ar2 br2 + + a1 r + a0
whose digits increase strictly from left to right: 1 ar1 < ar2 < <
a1 < a0 b 1.
 b 
Solution. The answer is 12 . . . r r+1 .

Let ns denote the s-digit number 12 . . . s for each positive integer s

b 1. We use a double induction and show that for b  1 k r, the sum
of all the r-digit numbers whose units digit is k is nr kr .
When r = 1, there is only
  one number ending in k, namely k itself, and
the desired sum is k = 1 k1 . For arbitrary r and k = r, the only eligible
number is 12 . . . r = nr .
Suppose that the result holds up to r 1 and up to k. The r-digit
numbers that end in k + 1 fall into two classes: (a) those whose second last
 k 1 and (b) those whosek1
digit does not exceed last two digits are k and
k + 1. There are k1
r1 numbers in class (a) and r2 numbers in class (b).
Since the numbers in class (a) can be formed by taking all the numbers
with r-digits and last digit k and adding 1 to each such number, the sum of
the numbers in class (a) is equal to
   
k k1
nr + .
r r1
 11. NUMBER THEORY 177

The numbers in class (b) can be found by taking all the (r 1)-digit numbers
ending in k, multiplying them by b and adding k + 1 to each such number.
The sum of these numbers is
       
k k1 k k1
bnr1 + (k + 1) = (nr r) + (k + 1).
r1 r2 r1 r2
Therefore the sum of all the numbers in classes (a) and (b) is
         
k k1 k k k1
nr + + nr r + (k + 1)
r r1 r1 r1 r2
         
k k k1 k1 k
= nr + + + (k + 1) r
r r1 r1 r2 r1
   
k+1 k+1
= nr + 0 = nr .
r r
Therefore the sum of all the r-digit numbers with increasing digits is
b1  
  
k b
nr = nr .
r r+1
k=r

2009:3. For each positive integer n, let p(n) be the product of all positive
integral divisors of n. Is it possible to nd two distinct positive integers m
and n for which p(m) = p(n)?

Solution. The answer is no. The function p(n) is one-one. Observe that,
since each divisor d can be paired with n/d (which are distinct except when n
is square and d is its square root), p(n) = n (n)/2 , where (n) is the number
of divisors of n.

Suppose that p(m) = p(n) and that q is any prime. Let the highest
powers of q dividing m and n be q u and q v respectively. The exponent of
the highest power of q that divides p(m) is (u (m))/2 and the highest power
of q that divides p(n) has exponent (v (n))/2. (The exponents are indeed
integers, since (n) is odd if and only if all u are even.)
Since p(m) = p(n), it follows that u (m) = v (n). If u < v for some
prime q, it is so for all primes q, so that m < n and (m) < (n). Thus,
u < v is impossible, as is u > v. Therefore u = v for each prime q. But then
this means that m = n.
2009:6. Determine all solutions in nonnegative integers (x, y, z, w) to
the equation
2x 3y 5z 7w = 1.

Solution. By parity considerations, we must have that x 1. Suppose,

rst of all, that y 1, z 1 and w 1. Since 2x 3y = 6y 2xy 2xy 1
(mod 5), it follows that x y (mod 4), so that x and y have the same parity.
178 11. NUMBER THEORY

Since 2x 3y 1 (mod 7), it follows that (x, y) (0, 0), (1, 4), (2, 2), (3, 0), (4, 4),
(5, 2) (mod 6). Therefore, x and y are both even. Let x = 2u and y = 2v.
Then
5z 7w = 22u 32v 1 = (2u 3v 1)(2u 3v + 1).
The two factors on the right must be coprime (being consecutive odd num-
bers), so that one of them is a power of 5 and the other is a power of 7. Thus
2u 3v 1 = 5z and 2u 3v 1 = 7w , whence
7w 5z = 2.

Observe that 7w is congruent, modulo 25, to one of 1, 7, 1, 7, so that

7w 2 is never divisible by 25. Therefore, z = 1 and so w = 1 and we obtain
the solution (x, y, z, w) = (2, 2, 1, 1).
Now we consider the cases where at least one of y, z, w vanishes.
Suppose that y = 0. Then 2x 1 = 5z 7w . If x is even, then the left
side is divisible by 3. Therefore x is odd. But if x is odd, then the left side
is not divisible by 5. Therefore z = 0 and 2x 7w = 1. This implies either
that w = 0 or that x is divisible by 3. There are two immediate possibilities:
(x, w) = (1, 0) and (x, w) = (3, 1). Suppose that x = 3r where r 2. Then
7w = (2r )3 1 = (2r 1)(22r + 2r + 1),
so that both factors on the right are nontrivial powers of 7. But it is not
possible for both factors of the right side to be divisible by 7, and we obtain
no further solutions.
Therefore, when y = 0, we have only the solutions (x, y, z, w) = (1, 0, 0, 0),
(3, 0, 0, 1).
It is not possible for x and y to both exceed 0 while z = 0, since otherwise
2x 3y 5z 7w would be congruent to 1 (modulo 3).
Finally, let x and y both exceed 0 and let w = 0. Since 5z + 1 2
(mod 4), we must have that x = 1. Observe that 5z + 1 0 (mod 9) only
when z 3 (mod 6). However, when z 3 (mod 6), 5z + 1 0 (mod 7).
Therefore, either 5z + 1 is not divisible by 9, or it is divisible by 7. In either
case, it cannot be of the form 2x 3y when y 2. Therefore, y = 1. This leads
to the sole solution (x, y, z, w) = (1, 1, 1, 0).
In all, there are four integral solutions to the equation, namely
(x, y, z, w) = (2, 2, 1, 1), (1, 1, 1, 0), (3, 0, 0, 1), (1, 0, 0, 0).

2009:9. Let p be a prime congruent to 1 modulo 4. For each real number

x, let {x} = x x denote the fractional part of x. Determine
  k 2  1

: 1 k (p 1) .
p 2
 11. NUMBER THEORY 179

Solution. When x is not divisible by p, then x x (mod p), since p

is odd. Also, x2 y 2 (mod p) if and only if x y or x y (mod p), so
that the mapping x x2 is a two-one mapping on Zp (the integers modulo p
that are coprime with p); thus, precisely 12 (p 1) elements of Zp are squares.
We rst show that m is a square modulo p if and only if p m is a square
modulo p.

If s is a square in Zp with r2 = s, then s1 = (r1 )2 is also a square,

so we can split the squares into disjoint sets of inverses. s = s1 can occur
only if s2 = 1 or s = 1 in Zp . Thus, these disjoint sets contain precisely
two elements when the square is not 1 nor 1, if applicable. Since there are
an even number 12 (p 1) of squares and 1 is a square, there must be another
singleton set of inverse squares, and this can consist only of 1. Hence 1 is
a square in Zp . Therefore, if m is a square in Zp , so also is m = (1)m. In
other words, m is a square modulo p if and only if pm is a square modulo p.
The set {k 2 : 1 k 12 (p 1)} contains each nonzero square exactly
once, and so there are 14 (p 1) pairs of squares of the form {m, p m} where
1 m 12 (p1). If k 2 r with 1 r p1, then {k 2 /p} = r/p. Therefore

  k2    m  
1 (p1)
2
pm 1
= + : 1 m (p 1), m a square (mod p)
p p p 2
k=1
p1
= .
4

2011:10. Suppose that p is an odd prime. Determine the number of

subsets S contained in {1, 2, . . . , 2p 1, 2p} for which (a) S has exactly p
elements, and (b) the sum of the elements of S is a multiple of p.
The solution to this problem appears in Chap. 2.
2012:2. Suppose that f is a function dened on the set Z of integers that
takes integer values and satises the condition that f (b) f (a) is a multiple
of b a for every pair a, b, of integers. Suppose also that p is a polynomial
with integer coecients such that p(n) = f (n) for innitely many integers n.
Prove that p(x) = f (x) for every positive integer x.

Solution. Let q(x) = f (x) p(x) for each integer x. The function q(x)
vanishes on an innite subset M of integers and q(a)q(b) is divisible by ab
for every pair a, b of
integers, since the same
is true of p(x). (If p(x) = ci x i ,
i i i1 i2 i2 i1
then p(a)p(b) = ci (a b ) = (ab) ci (a +a b+ +ab +b ).)

Let n be any integer. Then q(n) = q(n) q(m) is divisible by n m for

each m M . But then either q(n) = 0 or |q(n)| |n m| for innitely many
integers m. However, the latter case cannot occur.

Comment. The function f (x) itself need not be a polynomial. For ex-
ample, we might have f (x) = x + sin x.
180 11. NUMBER THEORY

2012:4. (a) Let n and k be positive integers. Prove that the least
common multiple of {n, n + 1, n + 2, . . . , n + k} is equal to
 
n+k
rn
k
for some positive integer r.
(b) For each positive integer k, prove that there exist innitely many
positive integers n, for which the number r dened in part (a) is equal to 1.
Solution 1. (a) Observe that
 
n+k n(n + 1)(n + 2) (n + k)
n = .
k (1)(2)(3) (k)
Let p be an arbitrary prime and suppose that pb is the highest power of p
that divides any of n, n + 1, . . . , n + k; thus, pb is the power of p that occurs
in the prime factorization of the least common
 multiple of n, n + 1, . . . , n + k.
We need to show that p divides n n+k k to no greater a power than pb .
 
Suppose that n + u is divisible by pb where 0 u k. Write n n+k k =
(n + u) A B where
(n + u + 1)(n + u + 2) (n + k)
A=
(1)(2)(3) (k u)
and
(n + u 1)(n + u 2) (n)
B= .
(k u + 1)(k u + 2) (k)
Since n + u + i i (mod pb ) it follows that p divides n + u + i and i to the
same power, for 1 i k u. Therefore A, written in lowest terms, has
a numerator that is not divisible by p. Since n + u i i (mod pb ) for
1 i u and k u + 1, k u + 2, . . . , k constitutes a set of u consecutive
integers, for each power pa with a b, pa divides a number among k u + 1,
. . . , k at most once more (either u/pa  or u/pa  + 1) than it divides one of
the numbers 1, 2, . . . , u or one of the numbers n + u 1, n + u 2, . . . ,
n. It follows that the power of p that divides the denominator of B cannot
exceed the power that divides the numerator of B by more than b. Since
b
n+k divisible by p , it follows that the highest power of p that
n + u is exactly
divides n k is no greater than b. The desired result follows.
 
(b) Let n = r k! for any positive integer r. Then n n+k k = rn(n +
1)(n + 2) (n + k) which is clearly a common multiple of n, n + 1, . . . , n + k.
Since this number is a divisor of the least common multiple of n, n + 1, . . . ,
n + k, it must be equal to the least common multiple.
 11. NUMBER THEORY 181

Solution 2. [J. Zung] (a) For a prime p, let fp (m) denote the exponent
of the highest power of p that is a divisor of the integer m. It is required to
prove that
  
n+k
fp (n(n + 1) (n + k)) fp (k!) = fp n
k
fp (lcm (n, n + 1, . . . , n + k))
= max{fp (m) : n m n + k}.

Suppose that M is the maximum of fp (m) for n m n + k. Then

fp (n(n+1) (n+k)) is the sum over i from 1 to M of the number of multiples
of pi that occur in the set {n, n + 1, . . . , n + k}. This is at most once more
than the number of multiples of pi that occur in the set {1, 2, . . . , k} with
equality occurring when n is a multiple of pi . Therefore
 * k + 
fp (n(n + 1) (n + k)) + 1 = fp (k!) + M,
pi
1iM

which yields the desired result.

(b) Equality occurs when n is a multiple of pfp (k!) for every prime p.
2013:1. Let a be an odd positive integer exceeding 3, and let n be a
positive integer. Prove that
n
a2 1
has at least n + 1 distinct prime divisors.
(b) When a = 3, determine all the positive integers n for which the
assertion in (a) is false.
Solution 1. (a) For n = 1, note that a2 1 = uv, where u = a 1 and
v = a + 1. Since u and v are consecutive even integers, one of them is the
product of 2 and an odd integer. Thus, a2 1 has at least two prime divisors.
We complete the proof by induction. Suppose it holds for the exponent n.
Observe that
n n1 n1
a2 1 = (a2 + 1)(a2 1).
The second factor on the right, by the induction hypothesis, has at least n
distinct prime divisors. The rst factor, being congruent to 2 modulo 4, must
have an odd divisor exceeding 1. Any odd prime that divides the rst factor
cannot divide the second, and so there must be at least n + 1 prime factors
in all.
2
(b) Observe that 32 1 = 23 , 32 1 = 34 1 = 24 5,
3
32 1 = (34 1)(34 + 1) = (24 5) (2 41) = 25 5 41
and
4
32 1 = (38 1)(38 +1) = (25 541)(217193) = 26 51741193.
n
By induction, as in (a), it can be shown that 32 1 has n + 1 distinct prime
factors when n 4. It has exactly n distinct prime factors when n = 1, 2, 3.
182 11. NUMBER THEORY

Solution 2. [J. Love] (a) Observe that, for each positive integer n,

n n1
a2 1 = (a 1)(a + 1)(a2 + 1) . . . (a2 + 1).

This quantity is divisible by 2. Since a is odd, a2 1 mod 4, so that the

last n 1 terms of the product on the right is equal to twice an odd integer,
and therefore has at least one odd prime divisor. Since a > 3, at least one of
a 1 and a + 1 is equal to twice an odd prime.
It remains to show that no two of the factors on the right side can be
divisible by the same odd prime. If p is an odd prime divisor of a 1, then
each of the remaining factors is congruent to 2 modulo p, If i < j and p
i j i j
divides a2 + 1 and a2 + 1, then, since a2 + 1 divides a2 1, then so does p.
j j n
But then p divides 2 = (a2 + 1) (a2 1), which is false. Therefore, a2 1
i
is divisible by 2 and n odd primes, distinct divisors of a2 1 and a2 + 1 for
1 i n 1.
n
(b) As in (a), we can prove that 2 divides N 32 1 and that distinct
n1
odd primes divide each of the n 1 factors 32 + 1, 34 + 1, . . . , 32 +1
of N . Thus, there are at least n distinct primes dividing N . However,
3
32 + 1 = 6562 = 2 17 193 is divisible by 2 odd primes, so that, when
n 4, N is divisible by n + 1 distinct primes. However, since 32 1 = 23 ,
34 1 = 24 5 and 38 1 = 6560 = 25 5 41, the assertion in (a) fails for
a = 3 and n = 1, 2, 3.
2013:4. Let S be the set of integers of the form x2 + xy + y 2 , where x
and y are integers.

(a) Prove that any prime p in S is either equal to 3 or is congruent to

1 modulo 6.
(b) Prove that S includes all squares.
(c) Prove that S is closed under multiplication.

Solution. (a) Let f (x, y) = x2 +xy+y 2 . Then 3 = f (1, 1) is representable.

It is straightforward to establish that 2 is not representable. (If x2 +xy +y 2 =
2, then (2x+y)2 +3y 2 = 8.) Let p be a prime exceeding 3 for which p = f (u, v)
for some integers u and v. Then, modulo 3,

0 p 4p = 4(u2 + uv + v 2 ) = (2u + v)2 + 3v 2 (2u + v)2 1.

Since p is odd, the result follows. (Alternatively, modulo 3, we have that

0 p = (u v)2 + 3uv (u v)2 1.)
(b) Observe that a2 = f (a, 0) = f (a, a) for each integer a.
 11. NUMBER THEORY 183

(c) Observe that x2 + xy + y 2 = (x y)(x y 2 ), where is an

imaginary cube root of unity, i.e., 1 + + 2 = 0 (this can be deduced
from the factorization of x3 y 3 into linear factors over C). Then, since
2 = 1 and 3 = 1.
(x2 + xy + y 2 )(u2 + uv + v 2 )
= [(x y)(x y 2 )][(u v)(u v 2 )]
= [(x y)(u v)][(x y 2 )(u v 2 )]
= [xu (xv + yu) + yv 2 ][xu (xv + yu) 2 + yv]
= [(xu yv) (xv + yu + yv)(xu yv) (xv + yu + yv) 2 ].
It is readily checked that
f (x, y)f (u, v) = f (xu yv, xv + yu + yv).

2014:2. For a positive integer N written in base 10 numeration, N 

denotes the integer with the digits of N written in reverse order. There are
pairs of integers (A, B) for which A, A , B, B  are all distinct and A B =
B  A . For example,
3516 8274 = 4728 6153.
(a) Determine a pair (A, B) as described above for which both A and
B have two digits, and all four digits involved are distinct.
(b) Are there any pairs (A, B) as described above for which A has two
and B has three digits?

Solution. (a) Let A = 10a+b and B = 10u+v where 1a, b, u, v9. The
condition AB = B  A leads to 99(au bv) = 0, so that a necessary condition
for the property to hold is au = bv. There are many possibilities, including
(a, b; u, v) = (1, 3; 6, 2). Thus, for example, 13 62 = 26 31. (The pos-
sibilities include (A, B) = (12, 42), (12, 63), (13, 62), (12, 84), (14, 82), (23, 64),
(24, 63), (24, 84), (23, 96), (26, 93), (34, 86), (46, 96), (48, 63).)
(b) Let A = 10a + b and B = 100u + 10v + w, where abuw = 0 and
0 a, b, u, v, w 9. The condition AB = B  A leads to
111(au bw) + 10[a(v w) + b(u v)] = 0.
Therefore a(v w) + b(u v) must be a multiple of 111. If the two terms
of this sum dier in sign, then the absolute value of the sum cannot exceed
9 9 = 81. If the two terms of the sum have the same sign, then the signs of
v w and u v must be the same, and the absolute value of the sum cannot
exceed
|a(v w) + b(u v)| 9|(v w) + (u v)| = 9|u w| = 81.
Therefore a(v w) + b(u v) = 0, so that au = bw. If v w and u v do not
vanish, they must have opposite signs. These conditions are necessary and
184 11. NUMBER THEORY

sucient for an example. Trial and error leads to (a, b; u, v, w) = (1, 2; 2, 3, 1).
Thus, we obtain
12 231 = 2772 = 132 21.
In fact, we can replace these numbers by multiples that involve no carries
to obtain the examples (12s, 231t) where 1 s 4 and 1 t 3. Other
examples of pairs are (13, 682), (28, 451).
Comment. At a higher level, we have that
992 483 156 = 651 384 299 = 27 32 7 13 23 31.
2014:3. Let n be a positive integer. A nite sequence {a1 , a2 , . . . , an } of
positive
n integers ai is said to be tight if and only if 1 a1 < a2 < < an , all
2 dierences aj ai with i < j are distinct, and an is as small as possible.
(a) Determine a tight sequence for n = 5.
(b) Prove that there is a polynomial p(n) of degree not exceeding 3 such
that an p(n) for every tight sequence {ai } with n entries.
Solution. (a) A sequence having all dierences dierent remains so if the
same integer is added or subtracted from
 each term. Therefore a1 = 1 for
any tight sequence. Since there are 52 = 10 dierences for a tight sequence
with ve entries, a5 11. We show that a5 = 11 cannot occur.
Suppose, if possible, that a1 = 1 and a5 = 11. Since each of the dif-
ferences from 1 to 10, inclusive, must occur, and since the largest dierence
10 is the sum of the four dierences of adjacent entries, the dierences of
adjacent entries must be 1, 2, 3, 4 in some order. The dierence 1 cannot be
next to either of the dierences 2 or 3, for then there a dierence between
nonadjacent entries would be 3 or 4, respectively. Thus 1 is the dierence of
either the rst two or the last two entries and the adjacent dierence would
be 4. But then a3 a1 = a5 a3 = 5, which is not allowed. Since there is no
other possibility, a5 12.
Trying a5 = 12, we nd that the sequence {1, 2, 5, 10, 12} is tight.
(b) The strategy is to construct a sequence that satises the dierence
condition such that the dierences between pairs of terms have dierent con-
gruence classes with respect to some modulus, dependent only on the number
of terms between them. Thus, we make the dierence between adjacent terms
congruent to 1, between terms separated by one entry congruent to 2, etc.
To this end, let a1 = 1 and
ai+1 = ai + 1 + (i 1)(n 1)
for 1 i n 1. Then,
1
ai = i + (i 1)(i 2)(n 1).
2
In particular,
1 1
an = n + (n 1)2 (n 2) = (n3 4n2 + 5n 2).
2 2
 11. NUMBER THEORY 185

When 1 i < j n, we have that

aj ai = (aj aj1 ) + (aj1 aj2 ) + + (ai+1 ai ) j i
modulo (n 1). Furthermore,
(aj+1 ai+1 ) (aj ai ) = (aj+1 aj ) (ai+1 ai ) = (j i)(n 1) > 0.
It follows that all the dierences are distinct.
Any tight sequence with n terms must have its largest term no greater
than p(n) = 12 (n3 4n2 + 5n 2).
Comment. This upper bound is undoubtedly too generous and it may
be that there is an upper bound quadratic in n.
2014:5. Let n be a positive integer. Prove that
    n 
1 
n n n
1 1 1 2k1 k
 n = = = n1 : k odd , 1 k n .
k k
k2nk 2n1 k 2 k
k=1 k=1 k=1

The solution to this problem appears in Chap. 4.

2015:6. Using the digits 1, 2, 3, 4, 5, 6, 7, 8, each exactly once, create two
numbers and form their product. For example, 472 83,156 = 39,249,632.
What are the smallest and the largest values such a product can have?
Solution. Observe that, when a > b > 0, c > 0, then (10b + c)a
(10a + c)b = c(a b) > 0. If a has as many digits as 10b + c (where c 9)
and a > 10b + c > b, then taking the last digits from 10b + c and appending
it to a gives the product (10a + c)b < (10b + c)a. From this, we see that if the
two numbers have unequal numbers of digits or the same number of digits,
removing the last digit from the smaller and appending it to the larger, will
result in a smaller product of the pair. Therefore, the smallest product will
occur when one number has a single digit and the other seven digits. For
both factors, the digits will appear in increasing order. It is straightforward
to see that the smallest product is 1 2,345,678 = 2,345,678.
Similarly, if the two factors have unequal numbers of digits, removing the
last digit from the larger factor and appending it to the smaller will result
in a larger product for the pair. Therefore, the largest product will occur
when both factors have four digits and the digits appear in decreasing order.
Observe the (a + c)(b + d) (a + d)(b + c) = (a b)(d c). If a > b, this
will be positive if and only if c < d. One number begins with 8 and the other
with either 7 or 6. But since 87pq 6mcd < 8mrs 67pq < 8mrs 76pq, one
number begins with 8 and the other with 7. Applying the result further, we
deduce that each digit after the rst in the number beginning with 8 is less
than the corresponding digit after the rst in the number beginning with 7.
Therefore, we deduce that the largest product is 8531 7642 = 65,193,902.
 APPENDIX A

Definitions, Conventions, Notation and Basics

1. General

1.1. Wolog = Without loss of generality

1.2. Kronecker delta. For indices (i, j), the Kronecker delta function
is dened by

1 if i = j;
ij =
0 if i = j.

1.3. Logarithm. The symbol log refers to the natural logarithm, or

logarithm to base e. The logarithm to base b will be indicated as logb .

1.4. Floor and ceiling. For any real number x, the oor of x, denoted
by x, is the largest integer that does not exceed x. The ceiling of x, denoted
by x, is the smallest integer that is not less than x.

2. Algebra

2.1. Binomial symbol. Let x be any real number and n be a positive

integer. The binomial symbol is dened as
 
x x(x 1) (x n + 1)
= .
n n!
A fundamental identity is:
     
x+1 x x
= +
n n n1
for n 1.
Springer International Publishing Switzerland 2016 187
E.J. Barbeau, University of Toronto Mathematics Competition
(20012015), Problem Books in Mathematics,
DOI 10.1007/978-3-319-28106-3
188 A. DEFINITIONS, CONVENTIONS, NOTATION AND BASICS

2.2. Roots of unity. The solutions of the equation z n 1 = 0, known as

the nth roots of unity, are the complex numbers cos(2k/n) + i sin(2k/n) =
e(2k/n)i where 0 k n 1. This is the set {1, , 2 , . . . , n1 } where
= cos(2/n) + i sin(2/n) is a fundamental nth root of unity. Every nth
root of 1 except 1 itself satises the equation 1 + z + z 2 + + z n1 = 0.

2.3. Fundamental theorem of algebra. Every polynomial p(z) with

complex coecients of positive degree n can be written as a product
a(z r1 )m1 (z r2 )m2 (z rk )mk ,
where m1 + m2 + + mk = n. For each i, the number ri is a root of p(z)
with multiplicity mi .

2.4. Symmetric function of the roots. Let r1 , r2 , . . . , rn be the n

roots of the polynomial z n + an1 z n1 + an2 z n2 + + a1 z + a0 . Then
r1 + r2 +  + rn = an1 , r1 r2 rn = (1)n a0 , and, for 1 k n, the
sum of all nk k-fold products of the roots is (1)k ank .

2.5. Sum of powers of the roots. With the same notation as before,
let pk = r1k + r2k + + rnk be the sum of the kth powers of the roots for
nonnegative integers k. Thus p0 = n, p1 = an1 , p2 = a2n1 2an2 . More
generally, when 1 k n 1,
pk + an1 pk1 + an2 pk2 + + ank+1 p1 + kank = 0.
For k = n + r n, we have that
pn+r + an1 pn+r1 + an2 pn+r2 + + a1 pr+1 + a0 pr = 0.
2.6. Rational function. A rational function is a ratio of two polyno-
mials. The roots of the numerator are the roots of the rational function, and
the roots of the denominator are the poles of the rational function.

2.7. Greatest common divisor and Euclidean algorithm. A great-

est common divisor of two polynomials is a polynomial of maximum degree
which divides both the polynomials. The polynomials are coprime i their
greatest common divisors are constants. Given any two polynomials f (z) and
g(z), we can determine polynomials q(z) and r(z), with the degree of r less
than that of q for which f (z) = q(z)g(z) + r(z). A greatest common divisor
of the pair (f (z), g(z)) is equal to a greatest common divisor of (g(z), r(z)).
By dividing g(z) by r(z), and continuing the process, we get a succession of
equations
f (z) = q(z)g(z) + r(z)
g(z) = q1 (z)r(z) + r1 (z)
r(z) = q2 (z)r1 (z) + r2 (z)

ri1 (z) = qi+1 (z)ri (z) + ri+1 (z)
 3. INEQUALITIES 189

rk1 (z) = qk+1 (z)rk (z),
where the degrees of the remainders strictly decrease and the process even-
tually terminates with an exact division. The polynomial rk (z) is a greatest
common divisor of f (z) and g(z). This method of obtaining a greatest com-
mon divisor is the Euclidean algorithm.

2.8. Trinomial theorem. The trinomial theorem states that

  (i + j + k)! 
(a + b + c)n = ai bj ck : i, j, k 0, i + j + k = n .
i!j!k!

3. Inequalities

3.1. Arithmetic-geometric means inequality. Let a1 , a2 , , an be

nonnegative real numbers. Then
a1 + a2 + + an
(a1 a2 an )1/n ,
n
with equality if an only if a1 = a2 = = an . This can be generalized to a
weighted version. Suppose that 1 , 2 , . . . , n are positive real numbers for
which 1 + 2 + + n = 1. Then
a1 1 a2 2 ann = 1 a1 + 2 a2 + + n an .

3.2. Power means inequalities. Let a1 , a2 , . . . , an be nonnegative

real numbers, and let m be a nonzero integer. The mth power mean Mm is
dened as
Mm = (am m m 1/m
1 + a2 + + an ) .
 1/n
We also dene M0 to be the geometric mean ( (ai )) of the ai . Then for
r < s, we have that
min(ai ) Mr Ms max ai .
M2 is the root-mean-square; M1 is the arithmetic mean; M0 is the geometric
mean; and M1 is the harmonic mean of the ai .

3.3. Cauchy-Schwarz Inequality. Let ai and bi be real numbers, for

1 i n. Then
(a1 b1 + a2 b2 + + an bn )2 (a21 + a22 + + a2n )(b21 + b22 + + b2n ),
with equality if and only if ai = bi for each i. There is also an integral version:
  2    
b b b
f (x)g(x)dx f (x)2 dx g(x)2 dx .
a a a
190 A. DEFINITIONS, CONVENTIONS, NOTATION AND BASICS

3.4. Jensen Inequality. There are two versions, one for convex and
one for concave functions.
A real-valued function dened on R is convex if and only if, for any real
x and y and for 0 t 1, then f ((1 t)x + ty) (1 t)f (x) + tf (y).
Then for any set {x1 , x2 , . . . , xn } of real numbers and any set of positive
reals {1 , 2 , . . . , n } with 1 + 2 + + n = 1, we have that
f (1 x1 + 2 x2 + + n xn ) 1 f (x1 ) + 2 f (x2 ) + + n f (xn ).
A twice-dierential function on an open interval whose second derivative is
nonnegative is convex.
A real-valued function dened on R is concave if and only if, for any
real x and y and for 0 t 1, then f ((1 t)x + ty) (1 t)f (x) + tf (y).
Then for any set {x1 , x2 , . . . , xn } of real numbers and any set of positive reals
{1 , 2 , . . . , n } with 1 + 2 + + n = 1, we have that
f (1 x1 + 2 x2 + + n xn ) 1 f (x1 ) + 2 f (x2 ) + + n f (xn ).
A twice dierentiable function on an open interval whose second derivative
is nonpositive is concave.

4. Sequences and Series

4.1. Convergence of monotone sequences. A bounded monotone

(either increasing or decreasing) sequence converges.

4.2. Liminf and Limsup. Suppose that {xn } is a real innite se-
quence. Then a = lim supn xn if and only if, given  > 0, there are
innitely many values of n for which xn > a  and there is an index N for
which xn < a +  for n N .
Also, a = lim inf n xn if and only if, given  > 0, there are innitely
many values of n for which xn < a +  and there is an index N for which
xn > a  for n N .

4.3. Abels Partial Summation Formula. :

n
 n

xk yk = (x1 + x2 + + xn )yn+1 + (x1 + x2 + + xk )(yk yk+1 ).
k=1 k=1

5. Analysis

5.1. Basic theorems. Let f (x) be a continuous real-valued function

on a real closed integer [a, b]. The Intermediate Value Theorem asserts that,
within the interval [a, b], the function assumes every value between f (a) and
f (b). The Extreme Value Theorem asserts that f (x) is bounded on [a, b] and
assumes both its least upper bound and greatest lower bound in the interval.
 6. LINEAR ALGEBRA 191

Suppose, further, that f (x) is dierentiable on the open interval (a, b). The
Mean Value Theorem states that there is a number c (a, b) for which
f (b) f (a)
f  (c) = .
ba
A special case is Rolles Theorem, which provides that if f (a) = f (b), then
there is a number c (a, b) for which f  (c) = 0. If f (x) is n times dieren-
tiable on an open interval containing a, then Taylors Theorem asserts that,
if a + u belongs to the interval, there is a number v between a and a + u for
which
n+1
 f (k) (a) f (n+1) (v) k+1
f (a + u) = uk + u .
n! (n + 1)!
k=0
The last term, called the Remainder, can be written in a number of alternative
ways.
If two real-valued functions f (x) and g(x) are dened and dierentiable
on a neighbourhood of a, if limxa f (x) = limxa g(x) = 0 and limxa
(f  (x)/g  (x)) = c, then, according to lH
opitals Rule, limxa (f (x)/g(x))=0.
If f (x) is a continuous real-valued functions dened on a closed interval
[a, b], then the Integral Mean Value  n Theorem asserts that there is a point
c [a, b] for which f (c)(b a) = a f (x) dx.

5.2. Fixpoint. Let f : S S be a function taking a set into itself. A

xpoint of f is an element s S for which f (s) = s.

5.3. Continuity of monotone functions. Let f be a monotonic (al-

ways increasing or always decreasing) real-valued function dened on an open
interval I. Then for each a I, both the one-sided limits limxa f (x) and
limxa f (x) exist. If they are equal, f (x) is continuous at a. If they are
unequal, the f (x) has a jump discontinuity at a. Every discontinuity of a
monotonic function is of this type.

5.4. Limits involving logarithms.

lim t log t = 0
t0

log(1 + t)
lim =1
t0 t

6. Linear Algebra

6.1. Linear independence; basis. A set {x1 , x2 , . . . , xn } of vectors is

linearly independent if and only if the condition c1 v1 + c2 v2 + + cn vn = 0
for scalars c1 , c2 , , cn implies that c1 = c2 = = cn = 0. The span of a
set {v1 , v2 , , vn } of a set of vectors, denoted by
v1 , v2 , , vn  is the set
of all linear combinations c1 v1 + c2 v2 + + cn vn = 0 for scalars ci . A basis
192 A. DEFINITIONS, CONVENTIONS, NOTATION AND BASICS

of a vector space is a linearly independent set whose span is the entire space.
The dimension of a vector space is the number of elements in any basis.

6.2. Matrix and vector conventions. The action of a matrix on a

vector is indicated by the vector in a column appearing to the right of a
matrix. Multiplication of matrices proceeds from left to right.

6.3. Linear transformations. A linear transformation T from a vec-

tor space V to a vector space W is a mapping that satises the condition
T (ax + by) = aT (x) + bT (y) for any vectors x and y in V and any scalars
a and b. The kernel of a linear transformation on a vector space is the set
of vectors carried to 0 by the transformation. The dimension of the kernel
is the nullity of the transformation. The rank of the transformation is the
dimension of the space of vectors onto which it maps.

6.4. Transpose. The transpose of a m n matrix M = (aij ) is the

n m matrix whose (i, j)th element is aji , where 1 i m, 1 j n. The
transpose of M is denoted by M t . A symmetric matrix is one that is equal
to its transpose.

6.5. Determinant and adjugate matrix. Let M = (mi,j ) be a n n

square matrix. The determinant of M , denoted by det M or |M |, is equal to
the sum of n! terms

(1) () m1,(1) m2,2 ai.(i) an,(n)
where the sum is over all permutations of the set {1, 2, . . . , n} and ()
equals 1 when can be written as the product of an even number of trans-
positions and 1 when can be written as the product of an odd number
of transpositions. For each i with 1 i n, the determinant of M can be
written in the form
n n
det M = mij Mij = mji Mji ,
j=1 j=1

where the number Mij is the cofactor of Mij . The adjugate matrix, denoted
adjM , for M has as its (i, j) entry Mji ; i.e., it is the transpose of the matrix
of cofactors of M . We note that
M adjM = adjM M = (det M )I.

6.6. Invertible and noninvertible matrices. A square matrix M is

nonsingular or invertible if there exists an inverse matrix, denoted by M 1
for which M M 1 = M 1 M = I, where I is the identity matrix with ones
down the main diagonal and zeros elsewhere. Otherwise, it is singular or
noninvertible. The matrix M is nonsingular if and only if its determinant
det M = |M | does not vanish. The nullity of a nonsingular n n matrix is
0, and its rank is n.
 6. LINEAR ALGEBRA 193

6.7. Orthogonal matrix. A real square matrix M is orthogonal if and

only if the product M M t = I with its transpose is the identity matrix.

6.8. Eigenvectors and eigenvalues. Let M be a n n square matrix.

An eigenvector of M is a nonzero column vector x for which M x = x for
some scalar . Such a scalar is called an eigenvalue. The eigenvalues are
the roots of the characteristic polynomial det(I M ) of the matrix. The
(algebraic) multiplicity of an eigenvalue is equal to its multiplicity as a root
of the characteristic polynomial. The trace of M is the sum of its eigenvalues
(counting multiplicity) and is equal to the sum of its main diagonal elements.
The determinant of M is equal to the product of its eigenvalues.

6.9. Minimal polynomial of a matrix. The minimal polynomial p(t)

of a matrix M is the monic nontrivial polynomial of lowest degree that satis-
es p(M ) = O. The minimal polynomial of a matrix divides its characteristic
polynomial.
The companion matrix of a monic polynomial p(t) = tn + an1 tn1 +
+ a1 t + a0 is the n n matrix

0 1 0 0 0
0 0 1 0 0

0 0 0 1 0


a0 a1 a2 a3 an1
whose characteristic and minimal polynomials are both equal to p(t).

6.10. Jordan canonical form. For any n n matrix n, there exists

a nonsingular matrix S for which S 1 M S = J, where the upper triangu-
lar matrix J is its Jordan canonical form, where the eigenvalues, repeated
according to their multiplicities, occur down the main diagonal. The Jordan
canonical form consists of square blocks Bi down its diagonal

B1 O O O O
O B2 O O O

O O B3 O O


O O O O Bk
where each block has the form

1 0 0 0 0 0
0 1 0 0 0 0

0 0 1 0 0 0
.


0 0 0 0 0 1
0 0 0 0 0 0
194 A. DEFINITIONS, CONVENTIONS, NOTATION AND BASICS

6.11. Hermitian transpose. For any m n matrix M with complex

entries mij , the hermitian transpose is the n m matrix M obtained by
taking the complex conjugate of the entries of M and transposing; thus,
M = M t and the (i, j)the element of M is mji . A matrix A for which
A = A is said to be hermitian.
For the complex column vector x with ith entry xi , x is a row vector
whose
ith entry is xi . The inner product,
x, y, of two column vectors is
xi xj = x y. For matrix A and column vectors x, y, we have that
x, Ay =

A x, y.

6.12. Three-dimensional vectors. For any vectors a, b of the real

3-dimensional linear space R3 , we have the scalar product
a b = a1 b1 + a2 b2 + c1 c2 = |a| |b| cos
and the vector product
a b = (a2 b3 a3 b2 )i + (a3 b1 a1 b3 )j + (a1 b2 a2 b1 )k
= |a| |b| (sin )c
where is the angle between the vectors a, and b, i = (1, 0, 0), j = (0, 1, 0),
k = (0, 0, 1), and c is a vector orthogonal to both a and b. We have the
following properties:
(1) i j = k; j k = i; k i = j;
(2) a b = b a;
(3) a (b c) = (a c)b (a b)c;
(4) a (b c) = b (c a) = c (a b)

a 1 b1 c 1
= det a2 b2 c2 .
a 3 b3 c 3

7. Geometry

7.1. Ptolemys Theorem. The sum of the products of the lengths of

pairs of opposite sides of a concyclic quadrilateral is equal to the product of
the lengths of the diagonals.

7.2. Sylvesters Theorem. Suppose that n 2 points are given in the

plane, not all collinear. Then there exists a line that contains exactly two of
them. To see this, suppose that n points are given. Pick a point P and line
m through two points of the set that does not contain P and for which the
distance from P to the line is minimum. If m contains exactly two points
of the set, then the result is established. Otherwise, let T be the foot of the
perpendicular from P to m and select two points Q and R on m on the same
side of T with Q between T and R (Q could coincide with T ). Let S be the
foot of the perpendicular from Q to the line P R. Since the right triangles
 7. GEOMETRY 195

T Q R

Fig. A.1. Sylvesters Theorem

P T R and QSR are similar and the hypotenuse P S of the rst triangle is
longer than T R, and so larger than QR, P T must be longer than QS. Thus
the distance between Q and P R is smaller than the distance between P and
m, contradicting our choice of P and m. (See Fig. A.1.) The result follows.

7.3. Convex set. A convex set is any set in the plane or in space that
contains along with any two of its points, all points within the line segment
that connects them.

7.4. Herons formula for the area of a triangle. Let the sides of a
triangle be a, b, c and let its semiperimeter 12 (a + b + c) be s. Then the area
F of the triangle is given by

F = s(s a)(s b)(s c).
We have that
16F 2 = 2(a2 b2 + b2 c2 + c2 a2 ) (a4 + b4 + c4 ).

7.5. Parabolas. Let a point F (the focus) and a line d (the directrix )
not containing F be given in the plane. The locus of points in the plane
that are equidistant from the point F and the line d is called a parabola. All
parabolas in the plane are similar. The graph of any equation of the form
y = ax2 + bx + c with a = 0 is a parabola. One characteristic of a parabola
is its reection property. For any point P on the parabola, the reection of
the line F P about the normal to the parabola at P is a line parallel to its
axis.
196 A. DEFINITIONS, CONVENTIONS, NOTATION AND BASICS

7.6. Ellipses. An ellipse is the locus of points P in the plane for which
the distance P F from P to a xed point F (the focus) divided by the distance
from P to a xed line d (the directrix ) is a constant e (0, 1). This constant
is called the eccentricity. Alternatively, an ellipse can be described as the
locus of points P for which the sum of the distances from P to two xed
points F1 and F2 (the foci ) is constant. The ellipse has a reection property
in that, for any point P on the ellipse, the line P F1 and P F2 are reected
images of each other with respect to the normal to the ellipse at P .
A standard equation for an ellipse in the plane is
x2 y2
+ = 1,
a2 b2
where
a > b > 0. The foci of this ellipse are at the points (c, 0), where
c = a2 b2 . The chord of the ellipse containing the foci is called the major
axis, and its length is 2a. The chord perpendicular to the major axis is called
the minor axis, with length 2b. The eccentricity of this ellipse is e = c/a and
the directrices for the two foci are the vertical lines with equations x = a2 /c.

7.7. Hyperbolas. A hyperbola is the locus of points P in the plane

for which the distance P F from P to a xed point F (the focus) divided
by the distance from P to a xed line d (the directrix ) is a constant e (the
eccentricity) exceeding 1. Alternatively, a hyperbola can be described as the
locus of points P for which the dierence of the distances from P to two xed
points F1 and F2 is constant. A standard equation for a hyperbola in the
plane is
x2 y2
= 1,
a2 b2
where a > b > 0. This curve is symmetric about each of the coordinate axes,
is unbounded and, for large values of |x|, approaches its asymptotes, the two
lines y = (b/a)x.

8. Group Theory

8.1. Definition of group. A group G is a set with an associative binary

operation (x, y) xy that contains an identity element e for which e2 = e
and ue = eu = u for every element u G. For each element a G, there
exists an inverse element v for which uv = vu = e.

8.2. Subgroups and cosets. A subgroup H of a group G is a group

with the same binary operation as in G. A left coset of H in G is a set of the
form xH = {xh : h H}; a right coset has the form Hx = {hx : h H}.
A subgroup H of G is normal i xH = Hx for all x G, or, equivalently,
x1 hx H whenever h H.
 9. COMBINATORICS 197

8.3. Order. The order of a subgroup H of G is the number of elements

in H and the order of an element x G is the smallest positive integer m for
which xm = e. Lagranges Theorem provides that the order of each subgroup
and of each element divides the order of the group. The index of a subgroup
of a group is equal to the order of the group divided by the order of the
subgroup.

8.4. Homomorphism. A homomorphism of a group G to a group K

is a mapping : G K for which (xy) = (x)(y) and (x1 ) = ((x))1
for all x, y G. The kernel of the homomorphism is the set of elements
of G that get mapped to the identity of K. The kernel of a homomorphism
is a normal subgroup of G. Conversely, every normal subgroup H of G is
the kernel of a homomorphism of G onto some other group, in particular the
factor group G/H whose elements are the cosets of H.

8.5. Automorphism. An automorphism of a group G is a one-one

mapping of G onto itself for which (e) = e, (x1 ) = ((x))1 and
(xy) = (x)(y) for all x, y G.

8.6. Permutations. A permutation is a one-one mapping of the set

{1, 2, . . . , n} onto itself. The symmetric group Sn is the group of all n! per-
mutations on the rst n positive integers with the operation of functional
composition. A transposition is a permutation that switches precisely two
numbers and leaves the rest xed; the transposition that takes a to b and b
to a is denoted by (ab). If x is any permutation that takes a to c and b to
d, then x1 (ab)x = (cd). A permutation is even i it can be written as the
product of an even number of transpositions; it is odd, otherwise. The set
of all even permutations constitutes the alternating group An . A cycle is a
permutation of the form (abc . . . z) that takes a to b, b to c, and so on until
nally z to a. The order of a cycle is equal to the number of elements aected
by the cycle. Each permutation can be written as a product of cycles, with
multiplication proceeding from left to right.

9. Combinatorics

9.1. Pigeonhole principle. The Pigeonhole principle states that if a

nite set of objects is partitioned into a number of non-overlapping categories,
and there are more objects than categories, then some category must be
assigned at least two objects.

9.2. Graphs. A graph consists of a set of points, called vertices or

nodes, along with a set of edges that connect certain pairs of vertices. A
vertex may be connected to itself, and two vertices may be connected by
more than one edge. If P and Q are two vertices, then the edge joining them
is denoted by P Q. A vertex is incident with an edge if it is one of its
198 A. DEFINITIONS, CONVENTIONS, NOTATION AND BASICS

endpoints. A path is a nite sequence of vertices, each consecutive pair of

which is joined by an edge. A closed path is a path that includes an edge
joining the nal and initial vertices. A path is a chain if it does not contain
any vertex more than once. A closed path is a circuit or loop if it does not
encounter any vertex more than once.

9.3. Trees. A tree is a graph that contains no loop and for which there
is exactly one path between any two vertices. The number of vertices in a
tree exceeds the number of edges by 1. A root of a tree is any vertex on the
tree that is incident with exactly one edge. A spanning tree or path is one
that contains every vertex of the graph.

9.4. Planar graphs. A planar graph is a graph that can be represented

by vertices and edges drawn on a plane or spherical surface without any edges
intersecting. Faces are regions surround by loops that contain no vertices. If
V , E, F are respectively the number of vertices, edges and faces in a planar
graph or a three-dimensional polyhedron, then Eulers equation V E+F = 2
is valid.

9.5. Directed graphs. A directed graph is one for which each edge is
assigned a direction so that it is conceived as emanating from its rst endpoint
to its second: P Q. A (directed) path or tree in a directed graph is such
that the exit vertex of each edge is the entry vertex of the next: P Q R.
A root of a spanning path or tree is a vertex from which each other vertex
can be reach along a directed path.

10. Number Theory

10.1. Base of numeration. Let b be a positive integer exceeding 1.

Every positive integer a can be written in the form
a = ar bn + an1 bn1 + + a1 b + a0
where r is a positive integer and the coecients ak are integers for which
0 ak b 1. Such a representation, denoted by
ar ar1 . . . a1 a0 ,
is the base b numeration of a.

10.2. Greatest common divisor and least common multiple. The

greatest common divisor of two integers a, b, denoted by gcd (a, b), is the
largest positive integer that divides both of them. Given any pair (a, b)
of integers, there exist integers x and y for which gcd (a, b) = ax + by.
Two integers are coprime if their greatest common divisor is 1. The least
common multiple of a and b, denoted by l cm (a, b), is the smallest positive
integer that both of them divide into. For any pair (a, b) of integers, ab =
gcd(a, b) lcm(a, b).
 10. NUMBER THEORY 199

Let a and b be integers whose greatest common divisor is g. Then there

exist integers x and y for which ax + by = g. As x and y vary, ax + by runs
through all the multiples of g.

10.3. Modularity and residues. Let m be a positive integer exceed-

ing 1, and let a and b be arbitrary integers. We say that a is congruent to
b modulo m i a b is a multiple of m; in symbols, a b (mod m). If a
and m are coprime, and b is arbitrary, then there always exists an integer
x that satises the congruence ax b (mod m). A complete set of residues
modulo m is any nite set of integers that contains an integer congruent
to each i with 0 i m 1. In particular, when a and m are coprime,
{ak : 0 k m 1} is a complete set of residues.

10.4. Finite number fields. Let p be a prime number. The set Zp

is the eld whose elements are {0, 1, 2, . . . , p 1}, where addition and mul-
tiplication are dened modulo p. The set Z consisting of {1, 2, , p 1}
constitutes a group under multiplication. With these operations of addition
and multiplication, Zp is a eld.

10.5. The divisor function. Let n be a positive integer and (n) be

the number of its positive divisors. If m and n are coprime, then (mn) =
= pa where the product is taken over all prime divisors of
(m) (n). If n 
n, then (n) = (a + 1).

10.6. Fermats Little Theorem. Let p be a prime and a an integer

not divisible by p. Then ap1 1 (mod p).
 APPENDIX B

Top Ranking Students

Generally, prizes were awarded to the top three to ve students in the contest.
Because the top grades were closer to each other in some years than in others,
the way in which the students were recognized was not consistent from year
to year.
Each problem was marked out of 10. After the rst few contests, to
discourage students from frittering their eorts over too many problems, only
the ve highest marks for the problems were counted unless the total of these
marks exceeded 30, in which case the scores for all the problems were included
in the total.
Competition 1: March 18, 2001 (13 candidates)
1. Jonathan Sparling
2. Jimmy Chui
3. Pavel Gyrya
Honourable mention: Ari Brodsky, Nicholas Martin, Al Momin
Competition 2: March 9, 2002 (11 candidates)
1. Jimmy Chui
2. David Varodayan
3. Isaac Li
4. Fred Dupuis; Robert Ziman (tied)
6. Emily Redelmeier
Competition 3: March 16, 2003 (20 candidates)
1. Jonathan Sparling
2. David Varodayan
3. Garry Goldstein
4. Benjamin Moull
5. Jimmy Chui
Honourable Mention: Samuel Huang, Emily Redelmeier, Robert Ziman

Springer International Publishing Switzerland 2016 201

E.J. Barbeau, University of Toronto Mathematics Competition
(20012015), Problem Books in Mathematics,
DOI 10.1007/978-3-319-28106-3
202 B. TOP RANKING STUDENTS

Competition 4: March 14, 2004 (18 candidates)

1. Garry Goldstein
2. Ali Feiz Mohammadi
3. Robert Barrington Leigh
4. Roger Mong
Honourable mention: Samuel Huang, Emily Redelmeier, David Shiroko,
Ilya Sutskever
Competition 5: March 12, 2005 (21 candidates)
1. Robert Barrington Leigh
2. Jacob Tsimerman
3. Garry Goldstein
Honourable mention: Ali Feizmohammadi, Tianyi David Han, Samuel
Huang, Emily Redelmeier, Ilya Sutskever
Competition 6: March 12, 2006 (17 candidates)
1. Jacob Tsimerman
2. Robert Barrington Leigh
3. Janos Kramar
4. Garry Goldstein
5. Ali Feizmohammadi
Competition 7: March 11, 2007 (10 candidates)
1. Janos Kramar
2. Mu Cai
3. Georey Siu
Honourable mention: Zhiqiang Li, Samuel Huang, Elena Flat
Competition 8: March 9, 2008 (14 candidates)
1. Janos Kramar
2. Konstantin Matveev
3. Sida Wang
4. Oleg Ivrii
Honourable mention: Viktoriya Krakovna
Competition 9: March 8, 2009 (21 candidates)
1. Konstantin Matveev
2. Alexander Remorov
3. Viktoriya Krakovna
4. Junjiajia Long
5. Sergio DaSilva
Competition 10: March 7, 2010 (31 candidates)
1. Alexander Remorov
2. Sida Wang
3. Keith Ng
 B. TOP RANKING STUDENTS 203

4. Viktoriya Krakovna
5. Sergei Sagatov
Competition 11: March 6, 2011 (26 candidates)
1. Keith Ng; Alexander Remorov (tied)
3. Sergei Sagatov
4. Philip Chen
5. Fengwei Sun
6. Jungwei Huo
7. Mengdi Hua
Competition 12: March 10, 2012 (38 candidates)
1. Jonathan Zung
2. Yu Wu
3. Keith Ng
4. Jialin Song
5. Jonathan Love
Competition 13: March 9, 2013 (38 candidates)
The top three students are listed in alphabetical order as their scores
were very close:
Jonathan Love
Jialin Song
Jonathan Zung
The next three ranking students also listed in alphabetical order were:
RongXi (Bill) Guo
Shia Xiao Li
Ruize Luo
Competition 14: March 9, 2014 (31 candidates)
1. Jonathan Love
2. Brian Bi
3. Shijie Xiu
4. Yizhou Sheng
5. David Perchersky
Competition 15: March 8, 2015 (33 candidates)
1. Jonathan Love
2. Shijie Xiu
3. Michael Chow
4. Chengyuan Yao
5. Jerry Jia Wang
6. Gal Gross
Since 2007, the contest papers have been set in memory of Robert M.
Barrington Leigh (19862006), a brilliant and promising mathematics student
from Edmonton, Alberta, whose untimely death in August, 2006 occasioned
great sorrow among his friends, mentors and professors.
 Index

A diophantine equation, 177

Abels partial summation, 67 directrix, 195
Abels Partial Summation Formula, 190 discriminant, 50
adjugate matrix, 117, 130, 192 divisors, 171, 172, 177, 199
alternating group, 149, 197 dominoes, 158
area, 82, 143
arithmetic progression, 174 E
arithmetic-geometric means inequality, eigenvalue, 193
54, 55, 85, 102, 131, 189 eigenvalues, 119, 127
asymptote, 196 eigenvector, 121, 127, 193
automorphism, 150, 197 ellipse, 140, 142, 196
directrix, 145
B focus, 145
badminton, 40 equations
base of numeration, 176, 198 several unknowns, 39, 42
basis, 107, 174 equilateral triangle, 141
binary operator, 151 Eulers equation, 198
Extreme Value Theorem, 99
C
Cauchy-Schwarz Inequality, 43, 96, F
112, 189 Fermats Little Theorem, 199
characteristic polynomial, 193 Fibonacci sequence, 44
circle, 141 nite eld, 121, 178
circuit, 156 nite number eld, 199
cobweb diagram, 66 xpoint, 191
colouring, 153 focus, 195
companion matrix, 122 fractional part, 85, 178
complete set of residues, 173, 199 friends, 125
complex numbers, 51, 129 functional equation, 97
composition of functions, 37, 104
concave function, 50, 51, 190 G
concave vertex, 138 graph, 156, 198
convex function, 101, 190 greatest common divisor, 172, 198
convex set, 195 group, 149
convex vertex, 138
Cramers Rule, 115 H
cross product, 113, 116, 123 heptagon, 24, 139
cycle, 150 hermitian, 127, 194
Herons area formula, 195
D homomorphism, 197
dense, 97 hyperbola, 142, 146, 196
determinant, 111, 119, 125, 192
dierential equation, 77, 81 I
dierentiation operator, 61 inection points, 82
dimension, 131 inner product, 122, 129

Springer International Publishing Switzerland 2016 205

E.J. Barbeau, University of Toronto Mathematics Competition
(20012015), Problem Books in Mathematics,
DOI 10.1007/978-3-319-28106-3
206 INDEX

integration by parts, 64, 90 polynomial

Intermediate Value Theorem, 91, integer values, 172, 179
95, 96, 99, 100, 104 polynomials, 31, 43, 96
invertible matrix, 108, 111, 113, 117, composition, 37
125, 126, 192 greatest common divisor, 189
integer values, 182
J roots, 33, 43, 86, 88, 188
Jensen Inequality, 190 several variables, 121
Jensens inequality, 51 symmetric function of roots, 35
Jordan canonical form, 118, 193 prime, 178, 180182
primes, 172
K
probability, 139
kernel, 107, 192, 197
Ptolemys Theorem, 26, 194
L
lH
opitals Rule, 78, 82 Q
Lagranges Theorem, 149, 197 quadratic equation, 45, 75
lagranges theorem, 149 quadratic function, 96
lattice, 153 quadratic polynomial, 98
Law of Cosines, 25, 76 quadrilateral, 135
least common multiple, 172, 180, 198 R
lightbulbs, 154 rank, 110, 115, 125, 192
linearly dependent, 126 rational function, 188
linearly independent, 107, 123, 131, 192 rational functions, 35
locus, 141, 146 rational length, 172
M recursion, 46, 70, 154
maximum and minimum, 86, 87 reection property, 196
Mean Value Theorem, 86, 93, removable singularity, 78
97, 103, 106 Rolles Theorem, 92
Integral, 106 roots of unity, 26, 27, 41
means inequalities, 189 S
minimal polynomial, 128 scalar product, 194
N Sherman-Morrison Formula, 109
noninvertible matrix, 192 singular matrix, see noninvertible
nonsingular matrix, see invertible matrix
matrix span, 121
nullity, 116, 192 span of set of vectors, 192
surd equation, 33
O Sylverters Theorem, 139
optimization, 52, 54, 55, 75, 81, 146, 185 Sylvesters Theorem, 195
orthogonal matrix, 193 symmetric group, 126, 149, 151, 197
orthogonal projection, 116, 122 symmetric matrix, 125

P T
palindrome, 90 Taylors Theorem, 63, 93
parabola, 50, 76, 136, 195 tetrahedron, 133
parallelogram, 123 tight sequence, 69
partition, 161, 167 tournament, 169
path, 165 trace, 128
permanent, 130 transpose, 192
permutation, 151 transposition, 149, 151
permutations, 52, 197 tree, 198
pigeonhole principle, 108 trinomial theorem, 189
 INDEX 207

U V
uniformly convergent series, 64 vector equation, 113, 123
unitary, 127 vector product, 194
upper triangular, 127 vector space, 98