OLYMON

VOLUME 1

April - December, 2000

Problems 1-54

This is the Mathematical Olympiads Correspondence Program sponsored by the Canadian Mathematical
Society and the University of Toronto Department of Mathematics. The organizer and editor is Edward J.
Barbeau of the University of Toronto Department of Mathematics, and the problems and solutions for this
volume of Olymon were prepared by Edward J. Barbeau of the University of Toronto, Dragos Hrimiuk
of the University of Alberta and Valeria Pandeleva in Ottawa.

Notes: The inradius of a triangle is the radius of the incircle, the circle that touches each side of the
polygon. The circumradius of a triangle is the radius of the circumcircle, the circle that passes through its
three vertices.
A set of lines of concurrent if and only if they have a common point of intersection.
The word unique means exactly one. A regular octahedron is a solid figure with eight faces, each of which
is an equilateral triangle. You can think of gluing two square pyramids together along the square bases. The
symbol buc denotes the greatest integer that does not exceed u.
An acute triangle has all of its angles less than 90◦ . The orthocentre of a triangle is the intersection
point of its altitudes. Points are collinear iff they lie on a straight line.
For any real number x, bxc (the floor of x) is equal to the greatest integer that is less than or equal to
x.

1. Let M be a set of eleven points consisting of the four vertices along with seven interior points of a square
of unit area.

(a) Prove that there are three of these points that are vertices of a triangle whose area is at most 1/16.

(b) Give an example of a set M for which no four of the interior points are collinear and each nonde-
generate triangle formed by three of them has area at least 1/16.

2. Let a, b, c be the lengths of the sides of a triangle. Suppose that u = a2 + b2 + c2 and v = (a + b + c)2 .
Prove that
1 u 1
≤ <
3 v 2
and that the fraction 1/2 on the right cannot be replaced by a smaller number.

3. Suppose that f (x) is a function satisfying

n
|f (m + n) − f (m)| ≤
m
for all rational numbers n and m. Show that, for all natural numbers k,
k
X k(k − 1)
|f (2k ) − f (2i )| ≤ .
i=1
2

4. Is it true that any pair of triangles sharing a common angle, inradius and circumradius must be con-
gruent?

1
 5. Each point of the plane is coloured with one of 2000 different colours. Prove that there exists a rectangle
all of whose vertices have the same colour.

6. Let n be a positive integer, P be a set of n primes and M a set of at least n + 1 natural numbers, each
of which is divisible by no primes other than those belonging to P . Prove that there is a nonvoid subset
of M , the product of whose elements is a square integer.

7. Let
12 22 32 5002
S= + + + ··· + .
1·3 3·5 5·7 999 · 1001
Find the value of S.

8. The sequences {an } and {bn } are such that, for every positive integer n,

1 1
an > 0 , bn > 0 , an+1 = an + , bn+1 = bn + .
bn an

Prove that a50 + b50 > 20.

9. There are six points in the plane. Any three of them are vertices of a triangle whose sides are of different
length. Prove that there exists a triangle whose smallest side is the largest side of another triangle.

10. In a rectangle, whose sides are 20 and 25 units of length, are placed 120 squares of side 1 unit of length.
Prove that a circle of diameter 1 unit can be placed in the rectangle, so that it has no common points
with the squares.

11. Each of nine lines divides a square into two quadrilaterals, such that the ratio of their area is 2:3. Prove
that at least three of these lines are concurrent.

12. Each vertex of a regular 100-sided polygon is marked with a number chosen from among the natural
numbers 1, 2, 3, · · · , 49. Prove that there are four vertices (which we can denote as A, B, C, D with
respective numbers a, b, c, d) such that ABCD is a rectangle, the points A and B are two adjacent
vertices of the rectangle and a + b = c + d.

13. Suppose that x1 , x2 , · · · , xn are nonnegative real numbers for which x1 + x2 + · · · + xn < 12 . Prove that

1
(1 − x1 )(1 − x2 ) · · · (1 − xn ) > ,
2

14. Given a convex quadrilateral, is it always possible to determine a point in its interior such that the four
line segments joining the point to the midpoints of the sides divide the quadrilateral into four regions
of equal area? If such a point exists, is it unique?

15. Determine all triples (x, y, z) of real numbers for which

x(y + 1) = y(z + 1) = z(x + 1) .

16. Suppose that ABCDEZ is a regular octahedron whose pairs of opposite vertices are (A, Z), (B, D)
and (C, E). The points F, G, H are chosen on the segments AB, AC, AD respectively such that
AF = AG = AH.
(a) Show that EF and DG must intersect in a point K, and that BG and EH must intersect in a point
L.
(b) Let EG meet the plane of AKL in M . Show that AKM L is a square.

2
17. Suppose that r is a real number. Define the sequence xn recursively by x0 = 0, x1 = 1, xn+2 = rxn+1 −xn
for n ≥ 0. For which values of r is it true that

x1 + x3 + x5 + · · · + x2m−1 = x2m

for m = 1, 2, 3, 4, · · ·.

18. Let a and b be integers. How many solutions in real pairs (x, y) does the system

bxc + 2y = a

byc + 2x = b
have?

19. Is it possible to divide the natural numbers 1, 2, · · · , n into two groups, such that the squares of the
members in each group have the same sum, if (a) n = 40000; (b) n = 40002? Explain your answer.

20. Given any six irrational numbers, prove that there are always three of them, say a, b, c, for which a + b,
b + c and c + a are irrational.

21. The natural numbers x1 , x2 , · · ·, x100 are such that

1 1 1
√ + √ + ··· + √ = 20 .
x1 x2 x100

Prove that at least two of the numbers are equal.

22. Let R be a rectangle with dimensions 11 × 12. Find the least natural number n for which it is possible
to cover R with n rectangles, each of size 1 × 6 or 1 × 7, with no two of these having a common interior
point.

23. Given 21 points on the circumference of a circle, prove that at least 100 of the arcs determined by pairs
of these points subtend an angle not exceeding 120◦ at the centre.

24. ABC is an acute triangle with orthocentre H. Denote by M and N the midpoints of the respective
segments AB and CH, and by P the intersection point of the bisectors of angles CAH and CBH. Prove
that the points M , N and P are collinear.

25. Let a, b, c be non-negative numbers such that a + b + c = 1. Prove that

ab bc ca 1
+ + ≤ .
c+1 a+1 b+1 4
When does equality hold?

26. Each of m cards is labelled by one of the numbers 1, 2, · · · , m. Prove that, if the sum of labels of any
subset of cards is not a multiple of m + 1, then each card is labelled by the same number.

27. Find the least number of the form |36m − 5n | where m and n are positive integers.

28. Let A be a finite set of real numbers which contains at least two elements and let f : A −→ A be a
function such that |f (x) − f (y)| < |x − y| for every x, y ∈ A, x 6= y. Prove that there is a ∈ A for which
f (a) = a. Does the result remain valid if A is not a finite set?
√
29. Let A be a nonempty set of positive integers such that if a ∈ A, then 4a and b ac both belong to A.
Prove that A is the set of all positive integers.

30. Find a point M within a regular pentagon for which the sum of its distances to the vertices is minimum.

3
31. Let x, y, z be positive real numbers for which x2 + y 2 + z 2 = 1. Find the minimum value of
xy yz zx
S= + + .
z x y

32. The segments BE and CF are altitudes of the acute triangle ABC, where E and F are points on
the segments AC and AB, respectively. ABC is inscribed in the circle Q with centre O. Denote
the orthocentre of ABC by H, and the midpoints of BC and AH be M and K, respectively. Let
∠CAB = 45◦ .
(a) Prove, that the quadrilateral M EKF is a square.
(b) Prove that the midpoint of both diagonals of M EKF is also the midpoint of the segment OH.
(c) Find the length of EF , if the radius of Q has length 1 unit.

33. Prove the inequality a2 + b2 + c2 + 2abc < 2, if the numbers a, b, c are the lengths of the sides of a
triangle with perimeter 2.

34. Each of the edges of a cube is 1 unit in length, and is divided by two points into three equal parts.
Denote by K the solid with vertices at these points.
(a) Find the volume of K.
(b) Every pair of vertices of K is connected by a segment. Some of the segments are coloured. Prove that
it is always possible to find two vertices which are endpoints of the same number of coloured segments.

35. There are n points on a circle whose√radius is 1 unit. What is the greatest number of segments between
two of them, whose length exceeds 3?

36. Prove that there are not three rational numbers x, y, z such that

x2 + y 2 + z 2 + 3(x + y + z) + 5 = 0 .

37. Let ABC be a triangle with sides a, b, c, inradius r and circumradius R (using the conventional notation).
Prove that
r abc
≤p .
2R 2(a + b )(b2 + c2 )(c2 + a2 )
2 2

When does equality hold?

38. Let us say that a set S of nonnegative real numbers if hunky-dory if and only if, for all x and y in
S, either x + y or |x − y| is in S. For instance, if r is positive and n is a natural number, then
S(n, r) = {0, r, 2r, · · · , nr} is hunky-dory. Show that every hunky-dory set with finitely many elements
is {0}, is of the form S(n, r) or has exactly four elements.

39. (a) ABCDEF is a convex hexagon, each of whose diagonals AD, BE and CF pass through a common
point. Must each of these diagonals bisect the area?
(b) ABCDEF is a convex hexagon, each of whose diagonals AD, BE and CF bisects the area (so that
half the area of the hexagon lies on either side of the diagonal). Must the three diagonals pass through
a common point?

40. Determine all solutions in integer pairs (x, y) to the diophantine equation x2 = 1 + 4y 3 (y + 2).

41. Determine the least positive number p for which there exists a positive number q such that

√ √ xp
1+x+ 1−x≤2−
q

4
 for 0 ≤ x ≤ 1. For this least value of p, what is the smallest value of q for which the inequality is
satisfied for 0 ≤ x ≤ 1?

42. G is a connected graph; that is, it consists of a number of vertices, some pairs of which are joined by
edges, and, for any two vertices, one can travel from one to another along a chain of edges. We call two
vertices adjacent if and only if they are endpoints of the same edge. Suppose there is associated with
each vertex v a nonnegative integer f (v) such that all of the following hold:
(1) If v and w are adjacent, then |f (v) − f (w)| ≤ 1.
(2) If f (v) > 0, then v is adjacent to at least one vertex w such that f (w) < f (v).
(3) There is exactly one vertex u such that f (u) = 0.
Prove that f (v) is the number of edges in the chain with the fewest edges connecting u and v.

43. Two players pay a game: the first player thinks of n integers x1 , x2 , · · ·, xn , each with one digit,
and the second player selects some numbers a1 , a2 , · · ·, an and asks what is the value of the sum
a1 x1 + a2 x2 + · · · + an xn . What is the minimum number of questions used by the second player to find
the integers x1 , x2 , · · ·, xn ?

44. Find the permutation {a1 , a2 , · · · , an } of the set {1, 2, · · · , n} for which the sum

S = |a2 − a1 | + |a3 − a2 | + · · · + |an − an−1 |

has maximum value.

45. Prove that there is no polynomial p(x) = an xn + an−1 xn−1 + · · · + a0 with integer coefficients ai for
which p(m) is a prime number for every integer m.
an +2
46. Let a1 = 2, an+1 = 1−2an for n = 1, 2, · · ·. Prove that
(a) an 6= 0 for each positive integer n;
(b) there is no integer p ≥ 1 for which an+p = an for every integer n ≥ 1 (i.e., the sequence is not
periodic).

47. Let a1 , a2 , · · · , an be positive real numbers such that a1 a2 · · · an = 1. Prove that

n
X 1
≤1
s − ak
k=1

where s = 1 + a1 + a2 + · · · + an .

48. Let A1 A2 · · · An be a regular n−gon and d an arbitrary line. The parallels through Ai to d intersect its
circumcircle respectively at Bi (i = 1, 2, · · · , n. Prove that the sum

S = |A1 B1 |2 + · · · + |An Bn |2

is independent of d.

49. Find all ordered pairs (x, y) that are solutions of the following system of two equations (where a is a
parameter):
x−y =2
  
2 2
x− y− = a2 − 1 .
a a
Find all values of the parameter a for which the solutions of the system are two pairs of nonnegative
numbers. Find the minimum value of x + y for these values of a.

50. Let n be a natural number exceeding 1, and let An be the set of all natural numbers that are not
relatively prime with n (i.e., An = {x ∈ N : gcd (x, n) 6= 1}. Let us call the number n magic if for each
two numbers x, y ∈ An , their sum x + y is also an element of An (i.e., x + y ∈ An for x, y ∈ An ).

5
 (a) Prove that 67 is a magic number.
(b) Prove that 2001 is not a magic number.
(c) Find all magic numbers.

51. In the triangle ABC, AB = 15, BC = 13 and AC = 12. Prove that, for this triangle, the angle bisector
from A, the median from B and the altitude from C are concurrent (i.e., meet in a common point).
√
52. One solution of the equation 2x3 + ax2 + bx + 8 = 0 is 1 + 3. Given that a and b are rational numbers,
determine its other two solutions.

53. Prove that among any 17 natural numbers chosen from the sets {1, 2, 3, · · · , 24, 25}, it is always possible
to find two whose product is a perfect square.

54. A circle has exactly one common point with each of the sides of a (2n + 1)−sided polygon. None of the
vertices of the polygon is a point of the circle. Prove that at least one of the sides is a tangent of the
circle.

Solutions

1. Let M be a set of eleven points consisting of the four vertices along with seven interior points of a square
of unit area.

(a) Prove that there are three of these points that are vertices of a triangle whose area is at most 1/16.

(b) Give an example of a set M for which no four of the interior points are collinear and each nonde-
generate triangle formed by three of them has area at least 1/16.

Solution. (a) We begin by covering the square with non-overlapping triangles whose vertices are found
among the eleven points. Begin by drawing one of the diagonals of the square. We then select the remaining
seven points in turn. Suppose, as an induction hypothesis, that we have selected k ≥ 0 points and covered
the square with 2(k + 1) triangles whose vertices are among the four vertices of the square and the k points
already selected. Consider the next point. If it is in the interior of an existing triangle, join it to each of
the three vertices of the triangle. If it is in the interior of an edge common to two triangles, join it to the
remaining vertex of each of the triangles. In each case, we have two more triangles than before, for a total of
2(k + 3) triangles. When all seven interior points have been selected, we have a total of 2 × 8 = 16 triangles.
The total area of these sixteen nonoverlapping triangles is 1, so at least one of them must have area not
exceeding 1/16.

(b) Let the square have vertices at (0, 0), (1, 0), (0, 1), (1, 1) and let the interior points be at ( 18 , 58 ),
( 28 , 28 ),
( 83 , 78 ), ( 84 , 48 ), ( 85 , 81 ), ( 68 . 68 ) and ( 78 , 38 ). There is a triangulation for which each of the triangles has
area exactly equal to 1/16. (Exercise: produce the diagram.) Any other triangle determined by three of the
points will have area at least as large as some triangle in the triangulation.

Comment. (a) Most students realized that one had only to look at the triangles involved in some
triangulation of the square. There are two issues that need to be addressed: that such a triangulation exists,
and the number of triangles obtained. This was neglected by some solvers. The induction argument above
looks after both of these issues. Given that there is a triangulation, the number of triangles can be counted
in another way. Suppose that there are N triangles. Then the total of all of the angles of the triangles is 2N
right angles. The angles of the triangles at each corner points of the square total to one right angle, while
at the interior seven points of the square total to four right angles, for a total of 4 + 4 × 7 = 32 right angles.
Since 32 = 2N , there are 16 triangles.

(b) Most people did not produce an example. The ones who did either transgressed the condition about
not having too many collinear points, or else did not ensure that the area of every triangle, even those not

6
involved in the triangulation, was at least 1/16.

2. Let a, b, c be the lengths of the sides of a triangle. Suppose that u = a2 + b2 + c2 and v = (a + b + c)2 .
Prove that
1 u 1
≤ <
3 v 2
and that the fraction 1/2 on the right cannot be replaced by a smaller number.

1 u
Solution. The numerator of the difference 2 − v is equal to

v − 2u = 2(ab + bc + ca) − (a2 + b2 + c2 )

= a(b + c − a) + b(c + a − b) + c(a + b − c) .

By the triangle inequality, a < b + c, b < c + a and c < a + b, so that the right side is always positive. Since
all variables are positive, the right inequality follows.
u 1
The numerator of v − 3 is equal to

3u − v = 2(a2 + b2 + c2 − ab − bc − ca)
= (a − b)2 + (b − c)2 + (c − a)2 ,

The right side, being a sum of squares, is nonnegative and it vanishes if and only if a = b = c. The left
inequality follows.

To see that u/v can be arbitrarily close to 1/2, let (a, b, c) = (, 1, 1) where 0 <  < 4. Then
1 u (4 − )
− = .
2 v (2 + )2
This can be made as close to 0 as desired by taking  sufficiently close to 0.

Coment. The inequality v ≤ 3u can be obtained by applying the Cauchy-Schwarz Inequality (try it!).
Robert Barrington Leigh found it convenient to look at the reciprocal v/u. In showing that this could be as
close as desired to 1/2, note that, when (a, b, c) = (, 1, 1),
v 2(ab + bc + ca) 4 + 2 4 + 2
2< =1+ =1+ <1+ = 2 + 2 .
u a2 + b2 + c2 +2 2

Some students tried to argue the inequality by looking at extreme situations, without attempting to
draw a relationship between the extreme and general situations. Others tried a kind of variation argument,
seeming to feel that since we can change the variables to a situation where the inequality holds, it will hold
on the way there. While it is often useful to think in terms of extreme cases and variation of the unknowns in
an inequality problem in order to get a handle on the situation, this approach generally becomes completely
unmanageable in the write-up. It is best to think of the variables as being fixed at certain values and perhaps
try to compare the value of a function to a purported extreme value. In many inequalities, including this
one, it is often conceptually cleanest to look at the difference between two sides; this difference can usually
be manipulated into a form which is clearly positive or negative, in a way in which the reader can easily
follow the steps. If you ever depart from this format, you should have a good reason for doing so. Avoid in
the write-up starting with what you have to prove and working backwards; this makes the logic harder to
follow and could lead to trouble with steps that are not reversible. Begin with what you know and proceed
by logical steps to what has to be obtained.

3. Suppose that f (x) is a function satisfying

n
|f (m + n) − f (m)| ≤
m

7
 for all positive rational numbers n and m. Show that, for all natural numbers k,
k
X k(k − 1)
|f (2k ) − f (2i )| ≤ .
i=1
2

Solution. Taking the case m = n = 2i , we find that

|f (2i+1 ) − f (2i )| ≤ 1

for each nonnegative integer i. Hence

|f (2k ) − f (2i )| ≤ |f (2k ) − f (2k−1 )| + |f (2k−1 ) + f (2k−2 )| + · · · + |f (2i+1 − f (2i )| ≤ k − i .

so that
k k k−1
X X X k(k − 1)
|f (2k ) − f (2i )| ≤ (k − i) = j= .
i=1 i=1 j=0
2

Comment. One student picked up that the condition does not make sense when m/n is negative, so
I have corrected the statement of the problem. Others (including myself) either rolled with the context
or corrected the statement to something reasonable. It does happen that a problem can be misstated on
a competition; if you feel that this has happened, you should draw attention to the mistake and make a
reasonable nontrivial interpretation of the problem and solve that.

This problem provided a nice illustration of the maxim that a little knowledge is a dangerous thing.
Students who had seen some calculus thought that somehow a derivative was involved, and then were lost
forever beneath the waves. The context does not suggest a place for calculus. Calculus deals with functions
defined on the real numbers (not just the rationals) and possessing derivatives. The most important thing
for you to learn about calculus is when not to use it. More seriously, turning to calculus right off the bat
inhibits the ability to take the problem on its own terms and come up with a more elementary solution.
Resist the urge to categorize problems too quickly. Once the key idea of looking at consecutive powers of 2
is found, the solution then falls right out.

4. Is it true that any pair of triangles sharing a common angle, inradius and circumradius must be con-
gruent?

Solution. Yes. Let ABC be the triangle, with sides a, b, c, circumradius R, inradius r, semiperimeter
s, area ∆ and u, v, w the respective lengths of the tangents to the incircle from vertices A, B, C. We are
given that angle A, and radii r and R are fixed. Then also fixed are a = 2R sin A, b + c − a = u = r cot A2 ,
b + c = u + a, s = (u + 2a)/2 and ∆ = rs. But ∆ = 21 bc sin A, so we find that bc and b + c are both fixed.
Now b and c are the uniquely determined roots of a quadratic equation (whose coefficients are fixed by the
sum and product of the roots), and the result follows.

Comment. There is a natural dynamical way of looking at the situation which is hard to capture in
a rigorous argument. We can imagine a fixed circumcircle with A fixed as one point on its circumference.
Let the angle at A be fixed, but let its arms vary within the circumcircle. Since we know the radius of the
incircle enveloped by the arms, this incircle is a fixed distance from A (why?); as we move the arms around,
we can convince ourselves that there are two positions (giving congruent triangles) where the incircle will be
tangent to the chord of the circumcircle determined by the arms of the angle at A. Nailing all this down is
not so easy, and seems to need at the least a continuity argument, which takes us out of the realm of pure
geometry into that of analysis.

5. Each point of the plane is coloured with one of 2000 different colours. Prove that there exists a rectangle
all of whose vertices have the same colour.

8
 Solution. Let N = 1000 and let S consist of points (x, y) with integer coordinates for which 0 ≤ x ≤
N, 0 ≤ y. Each row Ry := {(x, y) ∈ S : 0 ≤ x ≤ N } has N
+ 1 points, so that by the Pigeonhole Principle,
there is at least one colour used twice. There are N N2+1 possible ways in which a colour can be used in
two positions on Ry . Since there are more than this many rows Ry , the same colour must occur in the same
two positions in two of these rows. The four points in these two positions on the two rows determine the
desired rectangle.

Comment. Many students essentially had this pigeonhole argument, which was set up in a variety of
ways.

6. Let n be a positive integer, P be a set of n primes and M a set of at least n + 1 natural numbers, each
of which is divisible by no primes other than those belonging to P . Prove that there is a nonvoid subset
of M , the product of whose elements is a square integer.

Solution. Let P consist of the distinct primes p1 , p2 , · · · , pn . Let S be a nonvoid subset of M , and write
the product of the numbers in S in the form x = y 2 z, where y 2 is the largest square divisor of x. Then z
must have the form
pa1 1 pa2 2 · · · pann
where the vector (a1 , a2 , · · · , an ) has all of its entries equal to either 0 or 1. There are 2n possibilities for
this vector. However, there are at least 2n+1 − 1 > 2n nonvoid subsets of M and so that many possible
products x. Hence there are two such products, x = y 2 z and u = v 2 w which give rise to the same vector. It
may transpire that both x and u have factors from M in common; divide through by the product of these
factors to obtain products of disjoint subsets P and Q of M . These two products will have the same vector,
and so must have the form r2 t and s2 t, where t is a product of distinct primes from P . The product of the
numbers in P ∪ Q is r2 s2 t2 , and this is the desired square.

7. Let
12 22 32 5002
S= + + + ··· + .
1·3 3·5 5·7 999 · 1001
Find the value of S.

Solution 1. [Miranda Holmes] Each term of S is of the form

i2
 
1 i i
= + .
(2i − 1)(2i + 1) 4 2i − 1 2i + 1

Thus
500  
1X i i
S= + .
4 i=1 2i − 1 2i + 1
500  
X i i 1 1 2 2 3 3 500 500
4S = + = + + + + + + ··· + + .
i=1
2i − 1 2i + 1 1 3 3 5 5 7 999 1001

Since
i i+1
+ =1,
2i + 1 2i + 1
we can combine all such terms to get

500 501000
4S = 1 + 499 + = .
1001 1001
Thus,
125250
S= .
1001

9
 Comment. If you can guess at the formula for the sum of the series to n terms, it takes a straightforward
induction argument to establish that
n
X i2 n(n + 1)
= .
i=1
(2i + 1)(2i − 1) 2(2n + 1)

Solution 2. [Samer Seraj] We have that

n n 
i2 i2 i2

X 1X
= −
i=1
(2i − 1)(2i + 1) 2 i=1 2i − 1 2i + 1
n−1
i2 (i + 1)2 n2
 X  
1
= 1+ − + −
2 i=1
2i + 1 2(i + 1) − 1 2n + 1
n−1
n2 n2
   
1 X 1 n(n + 1)
= 1+ 1− = n− = ,
2 i=1
2n + 1 2 2n + 1 2(2n +1

When n = 500, we get the answer 125250/1001.

8. The sequences {an } and {bn } are such that, for every positive integer n,

1 1
an > 0 , bn > 0 , an+1 = an + , bn+1 = bn + .
bn an
Prove that a50 + b50 > 20.

Solution 1. Note that x + (1/x) ≥ 2 for every positive real x. Consider the sequence cn = (an + bn )2 .
Since  2    2
1 1 1 1
c2 = (a2 + b2 )2 = a1 + + b1 + = a1 + + b1 + ,
b1 a1 a1 b1
we find that c2 ≥ (2 + 2)2 = 16.

For each positive integer n,

 2
1 1
cn+1 = (an+1 + bn+1 )2 = an + + bn +
bn an
 
2 2 1 1 1 1
= [(an + bn + 2an bn ) + + 2 +2· ·
a2n bn an bn
 
1 1 an bn
+ 2 · an · + 2 · bn · +2 +
an bn bn an
 2  
1 1 an bn
= cn + + +2+2+2· + > cn + 8 ,
bn an bn an

Thus, cn+1 > cn + 8 . It follows that

c50 > c49 + 8 > c48 + 2 · 8 > · · · > c2 + 48 · 8 ≥ 16 + 48 · 8 = 400 .

Since a50 + b50 > 0, it follows that a50 + b50 > 20.

Solution 2. [F. Tian] For n ≥ 1,

1
an+1 bn+1 = an bn + 2 + .
an bn

10
Also, a2 b2 = a1 b1 + 2 + 1/(a1 b1 ) ≥ 4. By induction, it can be shown that an bn > 2n for all n ≥ 3. Hence,
by the Airthmetic-Geometric Means Inequality,
p √
a50 + b50 ≥ 2 a50 b50 > 2 100 = 20 .

9. There are six points in the plane, no three of them collinear. Any three of them are vertices of a triangle
whose sides are of different length. Prove that there exists a triangle whose smallest side is the largest
side of another triangle.

Comment. Before giving the solution to the problem, we present a result that you should be aware of;
pay also close attention to the proof.

Proposition. There are six points in the plane or in space, no three of them collinear. Each of the
segments between two of them is coloured in one of two colours. There exists a triangle whose vertices are
three of the given points and whose sides are of the same colour.

Proof. Let A be one of the points. There are five segments joining A to the other points. Since they
have one of two colours, by the Pigeonhole Principle, at least three of them must have the same colour.
Wolog, suppose that AB, AC, AD are coloured the same. If any of the segments BD, BC, CD has this
colour, then we will have a triangle in this colour. Otherwise, BCD must be a triangle all of whose edges
have the other colour.

Note 1: The minimum number of points with such a property is 6. If there are five points, it is possible
to colour the segments between any two of them so that a triangle with edges of a single colour does not
exist. For example, for a regular pentagon, we can colour all the sides with one colour and all the diagonals
with the other.

Note 2: A triangle of one-colour always exists when we have 17 points in the plane (no three collinear)
and three colours are used for the segments. This can be given a similar proof. From any point, at least
six of the segments emanating from it have the same colour. Now look at the six points terminating these
segments.

Now we can solve the problem.

Solution. Consider the six points and all triangles whose vertices are any three of them. Colour the
(uniquely determined) largest side of each triangle black, and colour the remaining edges red. There must
be a triangle all of whose edges are the same colour. This colour cannot be red. (Why?) So there must be
a triangle all of whose edges are black; its smallest edge must be the largest edge of some other triangle.

10. In a rectangle, whose sides are 20 and 25 units of length, are placed 120 squares of side 1 unit of length.
Prove that a circle of diameter 1 unit can be placed in the rectangle, so that it has no common points
with the squares.

Solution. [M. Holmes] If a circle of diameter 1 can be placed, it means that there must be a point in
the rectangle such that every point of every square is more than 1/2 units away from it to the centre of the
circle. The maximum area A around each square in which the centre of the circle cannot be located is the
area of the figure F formed by
(a) the square;
(b) four rectangles of dimensions 1 × 12 external to the sides of the square;
(c) four quarters of circles with radius 1/2 units external to the square with and centres at the vertices
of the square.
Hence, A = 1 + 12 · 1 · 4 + ( 12 )2 π = 3 + π4 . As there are 120 squares, the sum of all such areas within the
rectangle does not exceed 120 · (3 + π4 ) < 455.

As the circle should be placed inside of the rectangle, its centre cannot be less than 1/2 units away
from the rectangle’s sides, i.e., it can be only in the rectangle with sides 19 and 24 units of length, whose

11
sides are parallel to the rectangle’s sides on the distance 1/2 units from them. The area of this rectangle is
19 × 24 = 456. But 456 − 455 > 0, so at least one point is not covered by any of the 120 figures F described
above.. This point can be the centre of a circle of diameter 1 lying within the rectangle and having no point
in common with any of the squares.

11. Each of nine lines divides a square into two quadrilaterals, such that the ratio of their area is 2:3. Prove
that at least three of these lines are concurrent.

Solution. [M. Holmes] Since the lines divide the square into two quadrilaterals, they cut opposite sides
of the square. Let the vertices of the square be A, B, C, D (counterclockwise), and let one of the lines
intersect AB at M and CD at N . We can represent these points in an appropriate coordinate plane as
A(0, 0), B(1, 0), C(1, 1), D(0, 1), M (m, 0), N (n, 1).

Let [AM N D] : [M BCN ] = 2 : 3. Then [AM N D] = (m + n)/2 = 2/5, because the area of the whole
square is 1. The midpoint of M N is the point S( 21 (m + n), 12 ) = S( 52 , 12 ), which does not depend on the
points of intersection of M and N and, hence, is the same for all such lines. So, each line which divides the
square into two quadrilaterals in this way, must go through the point S. Because of the symmetry, there
are three other possible points in the square ( 53 , 12 ), ( 12 , 25 ), ( 12 , 53 ), and each of the given 9 points must pass
through one of them. Applying the Pigeonhole Principle for 9 lines and 4 points, we find that at least three
of the lines must pass through the same point, and because of that, they are concurrent.

12. Each vertex of a regular 100-sided polygon is marked with a number chosen from among the natural
numbers 1, 2, 3, · · · , 49. Prove that there are four vertices (which we can denote as A, B, C, D with
respective numbers a, b, c, d) such that ABCD is a rectangle, the points A and B are two adjacent
vertices of the rectangle and a + b = c + d.

Solution. Since the given polygon is regular, it can be inscribed in a circle. There are exactly 50
diagonals of the polygon which pass through the centre of the circle. As they are diameters, they are of
equal length. Consider the positive differences of two vertices which are endpoints of the same diagonal.
Since the mark numbers are from among 1, 2, · · · , 49, the range of the differences is between 0 and 48. So, we
have 49 possible values for 50 differences. Hence, there are at least two diagonals with the same difference.
Without loss of generality, denote these diagonals as as AC and BD and suppose that a ≥ c and d ≥ b.
Then a − c = d − b, so that a + b = c + d. The quadrilateral ABCD has two diagonals of equal length and
with the same midpoint, so it is a rectangle, which satisfies all of the required conditions.

13. Suppose that x1 , x2 , · · · , xn are nonnegative real numbers for which x1 + x2 + · · · + xn < 12 . Prove that

1
(1 − x1 )(1 − x2 ) · · · (1 − xn ) > ,
2

Comments. Some of you tried a “moving variables” argument, or an “extreme case” argument, i.e., if we
make xi as large as possible, we reduce the product, so that if it works for (x1 , x2 , · · · , xn ) = ( 12 , 0, 0, · · · , 0),
then we are in business. Some felt that if it works for the extreme case with each xi = 1/2n, then we are
ok. Unless you back this up with solid argument and detailed analysis that relates the general case to the
extreme case, then it is worthless. “Moving variable” arguments are always risky and best avoided. Both
approaches often muddy the situation rather than clarify it.

Solution 1. If 0 < u, v < 1, then

(1 − u)(1 − v) = 1 − (u + v) + uv > 1 − (u + v) .

It can be shown by induction that

(1 − x1 )(1 − x2 ) · · · (1 − xk ) > 1 − (x1 + x2 + · · · + xk )

12
for 2 ≤ k ≤ n, whence
1
(1 − x1 ) · · · (1 − xn ) > 1 − (x1 + · · · + xn ) > .
2

Comments. You should carry out the necessary induction argument. Note that the problem asks for >
rather than ≥.

Solution 2. [R. Barrington Leigh] Let

X
tk = {xi1 x12 · · · xik : 1 ≤ i1 < i2 < · · · < ik ≤ n}

for 1 ≤ k ≤ n. Then

(1 − x1 )(1 − x2 ) · · · (1 − xn ) = 1 − t1 + t2 − t3 + t4 − · · · + (−1)n tn .

For 1 ≤ k ≤ n − 1, we have that

X
tk = {xi1 xi2 · · · xik : 1 ≤ i1 < i2 < · · · < ik ≤ n}
X
> {xi1 xi2 · · · xik (xik +1 + xik +2 + · · · + xn ) : 1 ≤ i1 < i2 < · · · < ik ≤ n} = tk+1

so that
1
(1 − x1 )(1 − x2 ) · · · (1 − xn ) = (1 − t1 ) + (t2 − t3 ) + · · · > (1 − t1 ) = 1 − (x1 + · · · + xn ) > .
2

14. Given a convex quadrilateral, is it always possible to determine a point in its interior such that the four
line segments joining the point to the midpoints of the sides divide the quadrilateral into four regions
of equal area? If such a point exists, is it unique?

Note. Let ABCD be the quadrilateral and let P, Q, R, S be the respective midpoints of AB, BC, CD,
DA. Recall that P SkQR and P QkSR (prove it!).
1 1
Solution 1. Yes, and yes. Observe that [AP S] = 4 [ABD] and [CRQ] = 4 [CDB], so that [AP S] +
[CRQ] = 14 [ABCD] (why?). Since

1
[ABCD] = BD · (dist (A, BD) + dist (C, BD))
2
1
[AP S] = BD · dist (A, P S)
4
1
[CRQ] = BD · dist (C, QR)
4
we have that
1
dist (A, P S) + dist (C, QR) = [ dist (A, BD) + dist (C, BD)
2
= dist(P S, QR) .
Hence it is possible to draw a line m parallel to and between P S and QR for which dist (m, QR) = dist
(A, P S) and dist (m, P S) = dist (C, QR). For any point M on m, we have that [M QR] = [AP S] and
[M SP ] = [CRQ], so that [M QCR] = [AP M S] = 41 [ABCD].

In a similar way, we can draw a line n parallel to and between P Q and SR for which dist (n, P Q) = dist
(D, SR) and dist (n, SR) = dist (B, P Q), so that, for any point N on n, [N P Q] = [DSR], [N RS] = [BRP ]
and [N P BQ] = [DSN R] = 14 [ABCD].

13
 The lines m and n are not parallel, and, so, intersect in a unique point O for which

1
[OQCR] = [AP OS] = [OP BQ] = [DSOR] = [ABCD] .
4
Note that m and n intersect inside P QRS and so inside ABCD.

We show that O is the only point with this property. Let L be another point inside ABCD. Then L
lies in one of the four partitioning quadrilaterals, say [OQCR], so that [LQCR] < [OQCR] = 14 [ABCD].
Hence, L will satisfy the conditions of the problem.

Comments. A careful solution will contain the observation that m and n are not parallel, and will
intersect within the quadrilateral. The fact that m and n have a unique point of intersection does not, in
and of itself, establish that the requisite point cannot be found in some other way. Therefore, uniqueness
needs to be explicitly handled, either by showing that the required point must lie on m and n or by some
other argument.

Solution 2. Analysis. Suppose such a point O exists. Then [OP BQ] = [OCQR] = 41 [ABCD], so that

[AOB] = 2[P OB] = 2([OP BQ] − [OBQ]) = 2([OQCR] − [OQC]) = 2[ROC] = [DOC] .

Since [AOB] = 21 AB · dist (O, AB) and [DOC] = 21 CD · dist (O, CD), we must have that

dist (O, AB) CD

= .
dist (O, CD) AB

The locus of points O with this property is a line h through the intersection of AB and CD (or parallel to
them if they do not intersect) which lies between them (prove this!). Similarly, O must lie on a line k defined
by
dist (O, BC) AD
=
dist (O, AD) BC
through the intersection of BC and AD (if they intersect) that lies between them. These lines are not
parallel and intersect within ABCD (why?). If O exists, it must lie on the intersection of h and k.

Synthesis. Construct lines h and k to satisfy the foregoing conditions. They must intersect in a unique
point O within ABCD. Now

1 1
[AOB] = AB · dist (O, AB) = CD · dist (O, CD) = [DOC] .
2 2
Hence [AOP ] = [P OB] = [COR] = [DOR]. Let α be the common value. Similarly, [BOQ] = [COQ] =
[AOS] = [DOS] = β, say. Then each of [AP OS], [BQOP ], [CROQ] and [DSOR] has the value α + β and
O is the desired point.

15. Determine all triples (x, y, z) of real numbers for which

x(y + 1) = y(z + 1) = z(x + 1) .

Comments. In solving a system of equations, one begins by assuming a solution and determining what
properties it must have. Since this may involve one-way reasoning, such properties may not necessarily yield
a solution and extraneous solutions may arise. Thus, when you have solved the equations, for a complete
solution, you should check that the solution is valid.

If, in your manipulations, you divide by a certain quantity or find a quantity in the denominator of
a fraction, you should explore the possibility that the quantity could vanish, Remember that you cannot

14
divide by zero. When you write up your final solution, it is often a good idea to deal with this possibility
ahead of time and get it out of the way.

Solution 1. Suppose one of the variable, say x, is 0. Then z(x + 1) = 0, so z = 0, and y(z + 1) = 0 so
y = 0. Suppose one of the variables, say x is −1. The x(y + 1) = 0, so y = −1 and y(z + 1) = 0, so z = −1.
Hence, if any variable assumes either of the values −1 and 0, then all are equal. Henceforth, we assume that
none of them have either value.

From the equations, we find that

y(z + 1) yz + y − z
=x=
y+1 z

whence
yz(z + 1) = (y + 1)yz + (y + 1)(y − z)
or
yz(z − y) = (y + 1)(y − z) .
Hence, either y = z or yz + y + 1 = 0.

If y = z, then z + 1 = x + 1. so that z = x and x = y = z. Conversely, the system is satisfied when

x = y = z.

Suppose that yz +y +1 = 0. Then x(y +1) = z(x+1) = y(z +1) = −1, and so xy +x+1 = xz +z +1 = 0.
Suppose z = t, Then x = −(t + 1)/t and y = −1/(t + 1). Conversely, it is straightforward to check that the
system is satisfied by  
t+1 1
(x, y, z) = − ,− ,t
t 1+t
where t 6= 0, −1. Thus, we have obtained exactly the complete set of solutions.

Comment. The solution with x, y, z unequal may look non-symmetrical. Verify that, if x = s =
−(t + 1)/t, then y = −1/(1 + t) = −(s + 1)/s and z = t = −1/(1 + s). Check that x and z can similarly be
expressed in terms of y = r.

Solution 2. From the equations, we find that

x(y − z) = z − x , y(z − x) = x − y , and z(x − y) = y − z .

Multiplying these equations yields

xyz(y − z)(z − x)(x − y) = (z − x)(x − y)(y − z) .

Hence, either two variables (and so all) are equal or xyz = 1. The system is satisfied when all variables are
equal.

Suppose that xyz = 1. Then

xy + x + 1 = xy + x + xyz = x(yz + y + 1)

and
yz + y + 1 = yz + y + xyz = y(zx + z + 1) .
Since xy + x = yz + y = zx + z, either x = y = z = 1 or xy + x + 1 = yz + y + 1 = zx + z + 1 = 0. We need
to check that there are solutions of this type. Select arbitrary x 6= 0, −1, then y to satisfy xy + x + 1 = 0
and z to satisfy xyz = 1. Then

zx + z + 1 = zx + z + xyz = z(xy + x + 1) = 0

15
and
yz + y + 1 = yz + y + xyz = y(zx + z + 1) = 0
so that x(y + 1) = y(x + 1) = z(x + 1) = −1 as desired.

Solution 3. As before, we can check that (x, y, z) = (0, 0, 0) is the only solution in which any variable
vanishes. Henceforth, suppose that xyz 6= 0. Let z = vx. Then y + 1 = v(x + 1), whence y = vx + v − 1.
Therefore
x(vx + v) = x(y + 1) = y(z + 1) = (vx + v − 1)(vx + 1)
so that
v(v − 1)x2 + v(v − 1)x + (v − 1) = 0 .
Either v = 1 and we are led to x = y = z, which works, or

vx2 + vx + 1 = 0 .

Hence √ √ √
v 2 − 4v (v − 2) ± v 2 − 4v −v ± v 2 − 4v
 
−v ±
(x, y, z) = , , .
2v 2 2
where v < 0 or v ≥ 4. It can be checked that these values satisfy the equations. (Exercise: Check that
this colution is consistent with the other solutions.)

Solution 4. As in the foregoing solutions, we can check that (x, y, z) = (0, 0, 0), (−1, −1, −1) are the
only solutions in which any variable assumes either of the values 0 or −1. Henceforth, suppose x, y, z all
differ from 0 and −1.
Let x(y + 1) = y(z + 1) = z(x + 1) = k. Then z = k/(x + 1) and y = (k/x) − 1 = (k − x)/x. Therefore,
  
k−x k+x+1
k = y(z + 1) =
x x+1
=⇒ k(x2 + x) = k 2 − x2 + k − x = k(k + 1) − (x2 + x)
=⇒ (k + 1)(x2 + x − k) = 0
=⇒ k = −1 or x2 + x = k .

Suppose that x2 + x = k. Then x(x + 1) = z(x + 1). Since x 6= −1, x = z, and so y + 1 = x + 1. Thus,
x = y = z. Suppose that k = −1. Then z = −1/(x + 1) and y = −(x + 1)/x. It can be checked that this
works.

Comment. It is not hard to check independently that the only nonzero solution in which x, y, z have
the same sign is in fact given by x = y = z. For in this case, the ratio of any pair is positive, and the system
can be rewritten
1 1 z
− =1−
x y x
1 1 y
− =1−
z x z
1 1 x
− =1−
y z y
whence 3 = (x/y) + (y/z) + (z/x). By the Arithmetic-Geometric Means Inequality (applicable since the
three summands are positive), (x/y) + (y/z) + (z/x) ≥ 3 with equality if and only if x/y = y/z = z/x or
x = y = z.

16. Suppose that ABCDEZ is a regular octahedron whose pairs of opposite vertices are (A, Z), (B, D)
and (C, E). The points F, G, H are chosen on the segments AB, AC, AD respectively such that
AF = AG = AH.

16
 (a) Show that EF and DG must intersect in a point K, and that BG and EH must intersect in a point
L.
(b) Let EG meet the plane of AKL in M . Show that AKM L is a square.

Comment. Many students had complicated arguments involving similar triangles. You should try to
envisage the situation in terms of transformations, as this gives you a better sense of what is going on. Of
course, if a synthetic argument cannot be found, you always have recourse to the “refuge of the destitute”,
coordinate geometry and the hope that the computations will not be too horrendous. In this case, they are
not bad at all.

Solution 1. (a) Since AF : AB = AG : GC, it follows that F GkBCkED, while F G < BC = ED.
Hence, F GDE is a coplanar isosceles trapezoid and so EF and DG must intersect in a point K. A 90◦
rotation about the axis AZ takes B → C, F → G, C → D, G → H, D → E, E → B. Hence EF → BG and
DG → EH, so that BG and EH must intersect in a point L, which is the image of K under the rotation.
(b) KE and AB intersect in F , so that the two lines are coplanar. Also KF : KE = F G : ED =
F G : BC = AF : F B so that ∆KAF ∼ ∆EF B and AKkEB. Hence K lies in a plane through A parallel
to BCDE. Because the 90◦ rotation about the axis AZ (which is perpendicular to the planes BCDE and
AKL takes K → L, AK = AL and ∠KAL = 90◦ .
Consider a dilation with centre E and factor |AB|/|F B|. Let I be on AE with AI = AF . The the
dilation takes F → K, H → L, I → A and the plane of F GHI to the parallel plane AKL. The image of
G under this dilation is the intersection of EG and the plane of AKL, namely M . Thus the square F GHI
goes to KM LA which must also be a square.

Solution 2. The dilation-reflection perpendicular to a plane through F G perpendicular to BE and CD

with factor |AF |/|F B| takes B → E, C → D, and fixes F and G. The lines BF and CG with intersection
A gets carried to lines EF and DG which intersect in a point K for which AK is perpendicular to the plane
and so parallel to BE and CD, and the distance from K to the plane is |AF |/|F B| times the distance from
A to the plane.
Similarly, considering a dilation-reflection to a plane through GH perpendicular to BC and DE with
the same factor produces the point L with AL perpendicular to this plane and so parallel to BC and ED.
Thus ∠KAL = 90◦ .
The reflection in the plane AECZ fixes A, C, E, G and interchanges the points in each of the pairs
(B, D) and (F, H). Hence the line pairs (EF, DG) and (EH, BG) are interchanged as is the pair K and L.
Thus AK = AL and KL is perpendicular to AECZ. The triangle AKL is in a plane through A parallel to
BCDE. The proof that AKM L is a square can be completed as in Solution 1.

Solution 3. Assign solid coordinates: A ∼ (0, 0, 1), B ∼ (0, −1, 0), C ∼ (1, 0, 0), D ∼ (0, 1, 0), E ∼
(−1, 0, 0). For some t with 0 < t < 1, F ∼ (0, −(1 − t), t), G ∼ (1 − t, 0, t), H ∼ (0, 1 − t, t). The line EF
consists of points (−u, −(1 − u)(1 − t), (1 − u)t), with real u and the line DG consists of points ((1 − v)(1 −
t), v, (1 − v)t) with real v. They meet in K ∼ ((1 − t)/t, (t − 1)/t, 1). Similarly, L ∼ ((1 − t)/t, (1 − t)/t, 1),
so A, K, L lie on the plane z = 1. The line EG consists of points (−w + (1 − t)(1 − w), 0, t(1 − w)) with real
w, and this intersects the plane z = 1 in the point (2(1 − t)/t, 0, 1). The desired result can now be verified.

17. Suppose that r is a real number. Define the sequence xn recursively by x0 = 0, x1 = 1, xn+2 = rxn+1 −xn
for n ≥ 0. For which values of r is it true that

x1 + x3 + x5 + · · · + x2m−1 = x2m

for m = 1, 2, 3, 4, · · ·.

Answer. The property holds for all values of r.

17
 Solution 1. Define x−1 = −1; the recurrence still holds with this extension of the index. Note that, for
n ≥ 0,
(x2n+1 − x2n ) − (xn xn+2 − xn−1 xn+1 ) = xn+1 (xn+1 + xn−1 ) − xn (xn + xn+2 )
= xn+1 (rxn ) − xn (rxn+1 ) = 0
so that x2n+1 − x2n = xn xn+2 − xn−1 xn+1 .

We prove by induction that for each nonnegative integer m,

x2m = xm (xm+1 − xm−1 )

x2m+1 = x2m+1 − x2m = xm xm+2 − xm−1 xm+1 .

These equations check out for m = 0. Suppose they hold for m = k. Then

x2k+2 = rx2k+1 − x2k = r(x2k+1 − x2k ) − xk (xk+1 − xk−1 )

= xk+1 (rxk+1 − xk ) − xk (rxk − xk−1 )
= xk+1 xk+2 − xk xk+1 = xk+1 (xk+2 − xk )

x2k+3 = rx2k+2 − x2k+1 = r(xk+1 xk+2 − xk+1 xk ) − (x2k+1 − x2k )

= xk+1 (rxk+2 − xk+1 ) − xk (rxk+1 − xk )
= xk+1 xk+3 − xk xk+2 = x2k+2 − x2k+1
so the result holds for m = k + 1.

Hence
x1 + x3 + · · · + x2m−1 = (x21 − x20 ) + (x22 − x21 ) + · · · + (x2m − x2m−1 ) = x2m
as desired.

Solution 2. [R. Mong] We prove by induction that, for m ≥ 1,

x1 + x3 + · · · + x2m−1 = x2m

x2 + x4 + · · · + x2m = xm xm+1 .
These hold for m = 1. Assume they hold for 1 ≤ k ≤ m. Then

x1 + x3 + · · · + x2k+1 = 1 + (rx2 − x1 ) + (rx4 − x3 ) + · · · + (rx2k − x2k−1 )

= 1 + r[x2 + x4 + · · · + x2k ] − [1 + r(x2 + x4 + · · · + x2k−2 ) − (x1 + x3 + · · · + x2k−3 )]
= r(xk xk+1 − xk−1 xk ) + x2k−1 = rxk xk+1 − xk−1 (rxk − xk−1 )
= rxk xk+1 − xk−1 xk+1 = xk+1 (rxk − xk−1 ) = x2k+1

x2 + x4 + · · · + x2k+2 = (rx1 − x0 ) + (rx3 − x2 ) + · · · + (rx2k+1 − x2k )

= rx2k+1 − xk xk+1 = xk+1 (rxk+1 − xk ) = xk+1 xk+2 .
Hence the result holds for all m and the result follows.

Solution 3. The recursion xn+2 = rxn+1 − xn has characteristic polynomial t2 − rt + 1 with discriminant
2
r − 4. Its roots are distinct as long as r 6= ±2, so we deal with r = ±2 separately.
The solution for r = 2 is xn = n and for r = −2 is xn = (−1)n−1 n, and it is straightforward to establish
the desired relation in these cases. When r 6= ±2, the characteristic polynomial has distinct roots σ and
1/σ, where σ + 1/σ = r, i.e. σ 2 = rσ − 1. Note that σ 6= ±1 since r 6= ±2. The solution of the recursion is
σ
xn = (σ n − σ −n ) .
σ2 − 1

18
Then
x1 + x3 + · · · + x2m−1
σ σ
= 2 (σ + σ 3 + · · · + σ 2m−1 ) − 2 (σ −1 + σ −3 + · · · + σ −(2m−1) )
σ −1 σ −1
 −2m
σ2
 2m  
σ −1 1 σ −1
= 2 − 2
σ − 1 σ2 − 1 σ − 1 σ −2 − 1
2
σ −2m σ2
= 2 [σ 2m
− 1 − 1 + σ ] = (σ m − σ −m )2 = x2m .
(σ − 1)2 (σ 2 − 1)2

Comment. Solvers who used the approach of Solution 3 failed to consider the case in which the charac-
teristic polynomial had a double root.

18. Let a and b be integers. How many solutions in real pairs (x, y) does the system

bxc + 2y = a

byc + 2x = b
have?

Comments. Several of the solvers got all mixed up with the status of the variables in the problem,
and, for example, found infinitely many solutions to the equations. a and b are fixed in advance; they
are parameters, and your final answer will be conditioned by the the characteristics of various pairs (a, b).
Rather than rush blindly into the problem, it is a good strategy to gain some understanding of the situation
by looking at particular cases. For example, if one takes (a, b) = (0, 0), it is not too hard to see that
(x, y) = (0, 0) is the only solution. Similarly, if (a, b) = (1, 1), one arrives at the sole solution (x, y) = ( 21 , 12 ).
However, if (a, b) = (1, 0), we find two distinct solutions, (x, y) = (− 21 , 1), (0, 12 ). This not only clarifies the
situation, but gives you a couple of examples against which you can check your final answer. Get in the habit
of using examples to help understanding. It is an excellent exercise to plot, on the same axes, the graphs of
the two equations for various values of a and b.

Solution 1. Since 2x and 2y are the difference of two integers, x and y must have the form m + u2 and
n + v2 respectively, where m and n are integers and u and v each take one of the values 0 and 1. Plugging
these in yields
m + 2n + v = a
n + 2m + u = b .
Solving for m and n gives
3m = 2b − a + v − 2u
3n = 2a − b + u − 2v .
For a viable solution, u and v must be such that the right side of each equation is a multiple of 3 (and this
is where the characteristics of the parameters a and b enter in). We consider three cases:

(i) Let a + b ≡ 0 (mod 3). Then, for a solution, we must have

0 ≡ 2b − a + v − 2u = 3b − (a + b) + v − 2u ≡ v − 2u

and
0 ≡ 2a − b + u − 2v ≡ u − 2v
modulo 3. Since {u, v} ⊆ {0, 1}, we must have u = v = 0, and we obtain the solution
 
2b − a 2a − b
(x, y) = , .
3 3

19
This checks out.

(ii) Let a + b ≡ 1 (mod 3). Then, for a solution, we must have

0 ≡ 2b − a + v − 2u ≡ −1 + v − 2u

and
0 ≡ 2a − b + u − 2v ≡ −1 + u − 2v
modulo 3. The only solutions of u − 2v ≡ v − 2u ≡ 0 (mod 3) with u and v equal to 0 or 1 are given by
(u, v) = (1, 0), (0, 1), so, either

1
(x, y) = (m + , n) with 3m = 2b − a − 2, 3n = 2a − b + 1
2
or
1
(x, y) = (m, n + ) with 3m = 2b − a + 1, 3n = 2a − b − 2 .
2
Thus, we get two solutions
   
4b − 2a − 1 2a − b + 1 2b − a + 1 4a − 2b − 1
(x, y) = , , , .
6 3 3 6

These check out.

(iii) Let a + b ≡ 2 (mod 3). Then

0 ≡ 2b − a + v − 2u ≡ −2 + v − 2u

and
0 ≡ 2a − b + u − 2v ≡ −2 + u − 2v
modulo 3. For a solution, we must have v − 2u ≡ u − 2v ≡ 2 (mod 3), so that (u, v) = (1, 1) and
(x, y) = (m + 21 , n + 12 ) with 3m = 2b − a − 1 and 3n = 2a − b − 1. Hence there is a unique solution
 
4a − 2a + 1 4a − 2b + 1
(x, y) = , .
6 6

This checks out.

Hence, when a + b ≡ 0 or a + b ≡ 2 (mod 3), there is one solution to the system, while when a + b ≡ 1
(mod 3), there are two solutions.

Solution 2. Observe that x and y must be half integers. Since

bxc + byc + 2x + 2y = a + b
1 1
and x − 2 ≤ bxc ≤ x, y − 2 ≤ byc ≤ y, we must have that

a + b ≤ 3(x + y) ≤ a + b + 1 .

Hence, x + y, being a half integer, must have exactly one of the values

a + b a + b + 12 a + b + 1
, ,
3 3 3
depending on the divisibility of the numerator of the fractions by 3. Consider cases:

20
 (i) Let a + b ≡ 0 (mod 3). Then x + y is the integer (a + b)/3. If x and y are not themselves integers,
then
1 1
bxc + byc = (x − ) + (y − ) = x + y − 1
2 2
so that
3(x + y) − 1 ≡ a + b ≡ 0
modulo 3, a contradiction. Hence, x and y are both integers and
 
2b − a 2a − b
(x, y) = , .
3 3

This checks out.

(ii) Let a + b ≡ 1 (mod 3). Then 2(a + b) + 1 ≡ 0 (mod 3), and so

1
a+b+ 2 2(a + b) + 1
x+y = = ,
3 6

a half-integer. Hence there are integers m and n for which

1 1
(x, y) = (m + , n) or (m + 1, n − )
2 2

where m + n = (a + b − 1)/3. We find that

   
4b − 2a − 1 2a − b + 1 2b − a + 1 4a − 2b − 1
(x, y) = , , , .
6 3 3 6

These check out.

(iii) a + b ≡ 2 (mod 3). Then x + y = (a + b + 1)/3, an integer. In this case, we can check that x and y
cannot be integers, so that x = bxc + 1/2 and y = byc + 1/2, and so
 
4a − 2a + 1 4a − 2b + 1
(x, y) = , .
6 6

This checks out.

19. Is it possible to divide the natural numbers 1, 2, · · · , n into two groups, such that the squares of the
members in each group have the same sum, if (a) n = 40000; (b) n = 40002? Explain your answer.

Solution. [M. Holmes] (a) Yes, it is possible. Partition the numbers into two sets A and B such that
- if i ≡ 1, 4, 6, 7 (mod 8), put i in set A;
- if i ≡ 2, 3, 5, 8 (mod 8), put i in set B.
Since 40000 is a multiple of 8, there are 5000 strings of eight consecutive natural numbers. For each of them,
it is straightforward to see that

(8k + 1)2 + (8k + 4)2 + (8k + 6)2 + (8k + 7)2 = 4 × (8k)2 + 16 × (1 + 4 + 6 + 7) + (1 + 16 + 36 + 49)
= 4 × (8k)2 + 16 × (2 + 3 + 5 + 8) + (4 + 9 + 25 + 64)
= (8k + 2)2 + (8k + 3)2 + (8k + 5)2 + (8k + 8)2

for 0 ≤ k ≤ 4999. So, if the numbers are put into the sets as suggested, the squares of the numbers in each
group will have the same sum.

21
 (b) No, it is impossible. Suppose it were possible to partition the numbers from 1 to 40002 inclusive
into two sets A and B as required. There is a well-known formula for the sum of the squares of the first n
natural numbers,
n(n + 1)(2n + 1)
12 + 22 + · · · + n2 =
6
which we recommend that you prove by induction. When n = 40002, this sum is odd, and so we cannot
express it as the sum of two equal numbers, the sums of the squares in A and in B. Hence, the desired
partition is not possible.

Comment. One does not need the formula for the sum of squares to establish that the sum is odd; just
note that the sum has 20001 odd summands.

20. Given any six irrational numbers, prove that there are always three of them, say a, b, c, for which a + b,
b + c and c + a are irrational.

Solution. [O. Bormashenko] Recall the result given in the solution to Problem 9 in Olymon 1:5 (August,
2000): For any six points in space, let the full graph of all fifteen edges between two of them be coloured
with two colours. There exists a triangle of three of its vertices, each edge of which has the same colour.
Let each of the six irrational numbers be assigned a point in space, and colour an edge joining two points
representing a pair (u, v) red if u + v is rational and green if u + v is irrational. Then there must be a red
triangle or a green triangle. Suppose, if possible, there is a red triangle. Then three of the numbers, say a,
b, c, have a + b, b + c, c + a all rational. But then 2a = (a + b) + (c + a) − (b + c) would be rational, contrary
to hypothesis. So there is no red triangle, and so there must be a green triangle. The triple corresponding
to the vertices of this triangle must satisfy the requirement of the problem.

21. The natural numbers x1 , x2 , · · ·, x100 are such that

1 1 1
√ + √ + ··· + √ = 20 .
x1 x2 x100

Prove that at least two of the numbers are equal.

Solution. [R. Barrington Leigh] We construct a proof by contradiction. Assume that the natural numbers
are distinct, and, wolog, in increasing order. Thus, 1 ≤ x1 , 2 ≤ x2 , 3 ≤ x3 , · · ·, 100 ≤ x100 , so that

1 2 1 1 1 1
√ + √ + ··· + √ ≤ √ + √ + ··· + √ .
x1 x2 x100 1 2 100

On the other hand, for each natural number n,

√ √ √ 1
2 n> n+ n−1= √ √
n− n−1

whence
1 √ √
√ < 2( n − n − 1) .
n
It follows that
1 1 1
20 = √ + √ + · · · + √
x1 x2 x100
1 1 1
≤ √ + √ + ··· + √
1 2 100
√ √ √ √ √ √
< 2[( 1 − 0) + ( 2 − 1) + · · · + ( 100 − 99)] = 20
or 20 < 20, which is palpably false. Therefore, the assumption that the numbers x1 , x2 , · · ·, x100 is false,
and the desired result holds.

22
22. Let R be a rectangle with dimensions 11 × 12. Find the least natural number n for which it is possible
to cover R with n rectangles, each of size 1 × 6 or 1 × 7, with no two of these having a common interior
point.

Comment. Clearly, we can use 22 rectangles of size 1 × 6, so the optimum number is no greater than
this. We show in fact that the optimum value of n is 20. First, we establish that at least 20 rectangles are
needed, and then display a case in which 20 suffice.

23
 Solution. Coordinatize the squares (i, j) with 1 ≤ i ≤ 11, 1 ≤ j ≤ 12. Mark the twenty squares
(1 + t, 1 + t) (0 ≤ t ≤ 10), (1 + s, 8 + s) (0 ≤ s ≤ 4), (8 + r, 1 + r) (0 ≤ r ≤ 3). It is impossible to cover
more than one of the marked squares with a small rectangle, whether 1 × 6 or 1 × 7, so we need at least as
many rectangles as marked squares, i.e., at least 20. On the other hand, we can achieve the covering with
12 rectangles of size 1 × 7 and 8 of size 1 × 6. Cover the top seven rows with 12 rectangles of size 7 × 1 and
the bottom four rows with rectangles of size 1 × 6.

23. Given 21 points on the circumference of a circle, prove that at least 100 of the arcs determined by pairs
of these points subtend an angle not exceeding 120◦ at the centre.

Comment. Before providing the solution of this problem, it is necessary to recall some basic concepts
from graph theory and to prove one theorem. A graph is a topological combination of two sets, points and
line segments between some pairs of the points. The points are referred to as the vertices or nodes of the
graph and the line segments as edges or arcs. Let G be a graph. If A is a vertex of G, then the degree, d(A),
of A is the number of edges emanating from A. The number of edges is denoted by l(G). If there are three
vertices, A, B, C, of G such that any two of them are connected by an edge, the set of vertices A, B, C
and edges AB, BC, CA is a triangle in G, and the number of triangles in G is denoted by t(G). If there
are four vertices with any pair connected by an edge, then the set of the four vertices and their six edges is
a tetrahedron in G, and the number of all tetrahedra is denoted by T (G). The angle of an arc is the angle
subtended by the arc at the centre of the circle.

Turan’s Theorem. Let G be a graph with n vertices. If t(G) = 0, then l(G) ≤ bn2 /4c.
(This theorem is named after the Eastern European mathematician Pal Turan and is very useful in a
variety of problems, for which it is possible to model the given objects and their relationships with a graph.
In other words, it says that in a graph with n vertices with no triangles, there are no more than bn2 /4c
edges. It is easy to check that the result holds for n = 3 and n = 4, and you should do this.)
Proof of Turan’s theorem. In the graph G, let A be the vertex of greatest degree (i.e., the number
of edges from any other vertex does not exceed the number from A). Suppose that d(A) = k (a natural
number), that B1 , B2 , · · ·, Bk are the vertices connected to A by an edge, and that G0 is the graph that can
be obtained from G by removing A and all edges emanating from A. Then, l(G) = d(A) + l(G0 ). Obviously,
there is no edge Bi Bj because, otherwise, the vertices A, Bi , Bj will form a triangle. So, in G0 are counted
only all edges emanating from the n − k − 1 vertices of G other than A, B1 , B2 , · · ·, Bk . Since d(A) = k is
the maximum number of edges emanating from any vertex, l(G0 ) ≤ (n − k − 1)k. Therefore

l(G) ≤ k + (n − k − 1)k = k(n − k) .

Applying the arithmetic-geometric means inequality, we get k(n − k) ≤ [k + (n − k)]2 /4 = n2 /4, so that
l(G) ≤ n2 /4. Since l(G) is a natural number, the desired result follows. ♠
Two examples show that this value can be reached, so that it is the largest possible value of l(G).
Example 1. Let n = 2p be an even number. Partition the vertices into two sets with p vertices in each.
Draw all edges connecting a vertex from one set to the other. There are example p2 = n2 /4 edges, with no
triangles formed. Any additional edge will form a triangle with some other two.
Example 2. Let n = 2p + 1 be an odd number. Partition the vertices into two sets with p and p + 1
vertices. As in Example 1, connect ay vertex from one set with one in the other to obtain a total of
p(p + 1) = bn2 /4c edges, with no triangle in the graph.
(There is a similar theorem about tetrahedra in the graph, to wit: for a graph with n vertices with
T (G) = 0, then l(G) ≤ bn2 /3c. This can be proved using similar ideas to those for the triangle case. Here
is an example of a graph with T (G) = 0 and l(G) = bn2 /3c. Divide the vertices of G into three “almost
equal” sets (the difference between the numbers of vertices in any two of the sets is at most 1). Connect two
vertices with an edge if and only if they are from two different sets.)
Now we apply Turan’s Theorem to solve Problem 23.

24
 ◦
Solution

1. Count all arcs not exceeding 180 and ending in any two of the given 21 points. ◦There are
21
210 = 2 arcs in all. Connect with a segment any two points determining an arc exceeding 120 ; consider
all such segments as edges in a graph G whose vertices are the 21 given points. There are no triangles in G
(otherwise, each of its angles would exceed 60◦ giving an angle sum in excess of 180◦ . According to Turan’s
theorem, there are no more than b212 /4c = 110 such segments, so there must be at least 210 − 110 = 100
arcs that do not exceed 120◦ .

Solution 2. [M. Tancer, O. Bormashenko] There are 210 arcs not exceeding 180◦ for the 21 points on
the circumference of the circle, one for each pair. Given three points on the circle, at least one of the arcs
between two of them must not exceed 120◦ . If all of the 210 arcs do not exceed 120◦ , then the problem is
solved.
Suppose that there is an arc AB exceeding 120◦ . For any third point C, among the three arcs AB, AC,
BC, at least one must not exceed 120◦ ; since it is not AB, it must be one of the other two. Similarly, for
each of the nineteen points other than A and B, there must be at least one arc determined by that point
and one of A and B not exceeding 120◦ . Thus, there are at least 19 arcs one of whose endpoints is either A
or B exceeding 120◦ . Since there are 2 × 19 + 1 = 39 arcs with at least one endpoint A or B, the maximum
number of arcs among them exceeeding 120◦ is 20. If there are no further arcs exceeding 120◦ , the problem
is solved.
Otherwise, let DE be a second arc exceeding 120◦ , with D and E distinct from A and B. There are at
least 19 arcs with at least one endpoint D or E not exceeding 120◦ . Since all arcs emanating form A and
B have been counted, there are at least 17 new such arcs, and at most 18 arcs exceeding 120◦ . Continuing
this procedure, we find that at most 20 + 18 + 16 + · · · + 2 = 110 arcs exceeding 120◦ . So there must be at
least 100 arcs that do not exceed 120◦ .
Comment. Here is an example in which the number of arcs not exceeding 120◦ is exactly 100. Let the
centre of the circle be O, and let AB and CD be two diameters with ∠AOC = 60◦ . On the circumference of
the circle, put 10 points on the smaller arc AC and 11 on the smaller arc BD, all distinct from A, B, C, D.
Consider all arcs between a point from the first set and a point from the second set; there are 10 × 11 = 110
such arcs, all exceeding 120◦ . The remaining arcs do not exceed 120◦ .

24. ABC is an acute triangle with orthocentre H. Denote by M and N the midpoints of the respective
segments AB and CH, and by P the intersection point of the bisectors of angles CAH and CBH. Prove
that the points M , N and P are collinear.

Solution. Let ha and hb be the altitudes from A and B respectively, and let them BC ∩ ha = {E}
and AC ∩ hb = {D}. Let l1 and l2 be the respective angle bisectors of angles CAH and CBH. Since
∠CDH = ∠CEH = 90◦ , the quadrilaterial CDHE is inscribed in a circle k1 with centre N and diameter
CH. Hence N D = N E, as radii of the circle k1 . Similarly, since ∠ADB = ∠AEB = 90◦ , the quadrilateral
ADEB is inscribed in a circle k2 with centre M and diameter AB. Hence M D = M E as radii of the circle
k2 . It follows that M N is the right bisector of the segment DE; denote it by SDE . So, to prove M , N and
P collinear, it suffices to prove that P ∈ SDE , or P D = P E.
Consider the circle ADEB. Let G be the centre of the arc DE. Since ∠DAG = ∠GAE, l1 must pass
through G; since ∠EBG = ∠GED, l2 must pass through G. But then the point of intersection of l1 and l2
is P , so that P must coincide with G and therefore be the midpoint of the arc DE. Now it is easy to see
that the triangles DP M and EP M are congruent (DM = P M = EM as radii and ∠DM P = ∠P M E).
Hence, P D = P E and so M , N and P are collinear.

25. Let a, b, c be non-negative numbers such that a + b + c = 1. Prove that

ab bc ca 1
+ + ≤ .
c+1 a+1 b+1 4

When does equality hold?

25
 Solution. It is straightforward to show that the inequality holds when one of the numbers is equal to
zero. Equality holds if and only if the other two numbers are each equal to 1/2. Henceforth, assume that
all values are positive.
Since a + b + c = 1, at least one of the numbers is less than 4/9. Assume that c < 4/9. Let E denote
the left side of the inequality. Then
 
1 1 1 1 1 1
E = abc + + − − − .
a b c a+1 b+1 c+1

Since  
1 1 1 1 1 1 1 9
+ + = [(a + 1) + (b + 1) + (c + 1)] + + ≥ ,
a+1 b+1 c+1 4 a+1 b+1 c+1 4
(by the Cauchy-Schwarz Inequality for example), we have that
 
1 1 1 9 9
E ≤ abc + + − = ab + bc + ca − abc .
a b c 4 4

On the other hand, (1 − c)2 = (a + b)2 ≥ 4ab, whence ab ≤ 41 (1 − c)2 . Therefore

1 9 1
E− ≤ ab + c(a + b) − abc −
4   4 4
9 1
= ab 1 − c + c(1 − c) −
4 4
 
1 9 1
≤ (1 − c)2 1 − c + c(1 − c) −
4 4 4
1
= − c(3c − 1)2 ≤ 0 .
16
Equality occurs everywhere if and only if a = b = c = 1/3.

26. Each of m cards is labelled by one of the numbers 1, 2, · · · , m. Prove that, if the sum of labels of any
subset of cards is not a multiple of m + 1, then each card is labelled by the same number.
Pn
Solution. Let ak be the label of the kth card, and let sn = k=1 ak for n = 1, 2, · · · , m. Since the sum
of the labels of any subset of cards is not a multiple of m + 1, we get different remainders when we divide the
sn by m + 1. These remainders must be 1, 2, · · · , m in some order. Hence there is an index i ∈ {1, 2, · · · , m}
for which a2 ≡ si (mod m + 1). If i were to exceed 1, then we would have a contradiction, since then si − a2
would be a multiple of m + 1. Therefore, a2 ≡ s1 = a1 , so that a2 ≡ a1 (mod m + 1), whence a2 = a1 . By
cyclic rotation of the ak , we can argue that all of the ak are equal.

27. Find the least number of the form |36m − 5n | where m and n are positive integers.

Solution. Since the last digit of 36m is 6 and the last digit of 5n is 5, then the last digit of 36m − 5n is
1 when 36m > 5n and the last digit of 5n − 36m is 9 when 5n > 36m . If 36m − 5n = 1, then

5n = 36m − 1 = (6m + 1)(6m − 1)

whence 6m + 1 must be a power of 5, an impossibility. 36m − 5n can be neither −1 nor 9. If 5n − 36m = 9,

then 5n = 9(4 · 36m−1 + 1), which is impossible. For m = 1 and n = 2, we have that 36m − 5n = 36 − 25 = 11,
and this is the least number of the given form.

28. Let A be a finite set of real numbers which contains at least two elements and let f : A −→ A be a
function such that |f (x) − f (y)| < |x − y| for every x, y ∈ A, x 6= y. Prove that there is a ∈ A for which
f (a) = a. Does the result remain valid if A is not a finite set?

26
 Solution 1. Let a ∈ A, a1 = f (a), and, for n ≥ 2, an = f (an−1 ). Consider the sequence {xn } with

xn = |an+1 − an |

where n = 1, 2, · · ·. Since A is a finite set and each an belongs to A, there are only a finite number of distinct
xn . Let xk = minn≥1 {xn }; we prove by contradiction that xk = 0.
Suppose if possible that xk > 0. Then

xk = |ak+1 − ak | > |f (ak+1 ) − f (ak )| = |ak+2 − ak+1 | = xk+1 .

But this does not agree with the selection of xk . Hence, xk = 0, and this is equivalent to ak+1 = ak or
f (ak ) = ak . The desired result follows.

Solution 2. We first prove that f (A) 6= A. Suppose, if possible, that f (A) = A. Let M be the largest
and m be the smallest number in A. Since f (A) = A, there are elements a1 and a2 in A for which M = f (a1 )
and m = f (a2 ). Hence

M − m = |f (a1 ) − f (a2 )| < |a1 − a2 | ≤ |M − m| = M − m

which is a contradiction. Therefore, f (A) ⊂ A.

Note that A ⊇ f (A) ⊇ f 2 (A) ⊇ · · · ⊇ f n (A) ⊇ · · ·. In fact, we can extend the foregoing argument to
show strict inclusion as long as the sets in question have more than one element:

A ⊃ f (A) ⊃ f 2 (A) ⊃ · · · ⊃ f n (A) ⊃ · · · .

(The superscripts indicate multiple composites of f .) Since A is a finite set, there must be a positive integer
m for which f m (A) = {a}, so that f m+1 (A) = f m (A). Thus, f (a) = a.
Example. If A is not finite, the result may fail. Indeed, we can take A = (0, 21 ) (the open interval of
real numbers strictly between 0 and 21 ) or A = {2−2n : n = 1, 2, · · ·} and f (x) = x2 .
√
29. Let A be a nonempty set of positive integers such that if a ∈ A, then 4a and b ac both belong to A.
Prove that A is the set of all positive integers.

Solution. (i) Let us first prove that 1 ∈ A. Let a ∈ A. Then we have

ba1/2 c ∈ A , bba1/2 c1/2 c ∈ A , ··· , b· · · ba1/2 c1/2 c1/2 · · ·c ∈ A , ··· .

Also, the following inequalities are true

2 n
1 ≤ ba1/2 c ≤ a1/2 , 1 ≤ bba1/2 c1/2 c ≤ a1/2 , · · · , 1 ≤ b· · · ba1/2 c1/2 c1/2 · · ·c ≤ a1/2 ,

where there are n brackets in the general inequality. There is a sufficiently large positive integer k for which
k
a1/2 ≤ 1.5, and for this k, we have, with k brackets,
k
1 ≤ b· · · ba1/2 c1/2 c1/2 · · ·c ≤ a1/2 ≤ 1.5 ,

and thus
b· · · bba1/2 c1/2 c1/2 · · ·c = 1 ,

(ii) We next prove that 2n ∈√A for n = 1, 2, · · ·. Indeed, since 1 ∈ A, we obtain that, for each positive
integer n, 22n ∈ A so that 2n = b 22n c ∈ A.

27
 (iii) We finally prove that an arbitrary positive integer m is in A. It suffices to show that there is a
k
positive integer k for which m2 ∈ A. For each positive integer k, there is a positive integer pk such that
k k
2pk ≤ m2 < 2pk +1 (we can take pk = blog2 m2 c). For k sufficiently large, we have the inequality

 2 k
1 1
1+ ≥1+ · 2k > 4 .
m m

This combined with the foregoing inequality produces

k k
2pk ≤ m2 < 2pk +1 < 2pk +2 < (m + 1)2 . (∗)

Since 22(pk +1)+1 ∈ A, we have that

p √
b 22(pk +1)+1 c = b2pk +1 2c ∈ A .

Hence, with k + 1 brackets,

√
b· · · bb2(pk +1) 2c1/2 c1/2 · · ·c ∈ A .

On the other hand, using (∗), we get

k √ k
m2 < 2pk +1 ≤ b2(pk +1) 2c < (m + 1)2

and, then, with k + 1 brackets,

√
m ≤ b· · · bb2(pk +1) 2c1/2 c1/2 · · ·c < m + 1 .

Thus,
√
m = b· · · bb2(pk +1) 2c1/2 c1/2 · · ·c ∈ A .

30. Find a point M within a regular pentagon for which the sum of its distances to the vertices is minimum.

Solution. We solve this problem for the regular n−gon A1 A2 · · · An . Choose a system of coordinates
centred at O (the circumcentre) such that
 
2kπ 2kπ −−→
Ak ∼ r cos , r sin , r = kOAk k
n n

for k = 1, 2, · · · , n. Then

n n    Xn n 
X −−→ X 2kπ 2kπ 2kπ X 2kπ
OAk = r cos , r sin = r cos ,r sin = (0, 0) .
n n n n
k=1 k=1 k=1 k=1

Indeed, letting ζ = cos 2π 2π n

n + i sin n and using DeMoivre”s Theorem, we have that ζ = 1 and

n n n  k
X 2kπ X 2kπ X 2π 2π
cos +i sin = cos
+ i sin
n n n n
k=1 k=1 k=1
n
1 − ζn
X  
= ζk = ζ =0,
1−ζ
k=1

28
 Pn Pn
whence k=1 cos(2πk/n) = k=1 sin(2πk/n) = 0. On the other hand,
n n
X −−−→ 1 X −−→ −−→ −−→
kM Ak k = kOAk − OM kkOAk k
r
k=1 k=1
n
1 X −−→ −−→ −−→
≥ (OAk − OM ) · OAk
r
k=1
n
X n 
1 −−→ 2 −−→ X −−→
= kOAk k − kOM k OAk
r
k=1 k=1
n n
X X −−→
= r= kOAk k .
k=1 k=1

Equality occurs if and only if M = O, so that O is the desired point.

31. Let x, y, z be positive real numbers for which x2 + y 2 + z 2 = 1. Find the minimum value of
xy yz zx
S= + + .
z x y

Solution 1. [S. Niu] Let a = yz/x, b = zx/y and c = xy/z. Then a, b, c are positive, and the problem
becomes to minimize S = a + b + c subject to ab + bc + ca = 1. Since

2(a2 + b2 + c2 − ab − bc − ca) = (a − b)2 + (b − c)2 + (c − a)2 ≥ 0,

we have that a2 + b2 + c2 ≥ ab + bc + ca. Thus,

1 = ab + bc + ca ≤ a2 + b2 + c2
= (a + b + c)2 − 2(ab + bc + ca) = (a + b + c)2 − 2 = S 2 − 2
√ √
so S ≥ 3 with equality if and only if a = b = c, or x = y = z = 1/ 3. The desired result follows.

Solution 2. We have that

x2 y 2 y2 z2 z 2 x2
S2 = + + + 2(x2 + y 2 + z 2 )
z2 x2 y2
 2 2
z 2 x2
  2 2
y2 z2
  2 2
z 2 x2

1 x y x y y z
= + 2 + + 2 + + 2 +2
2 z2 y z2 x x2 y
≥ x2 + y 2 + z 2 + 2 = 3

by the Arithmetic-Geometric Means √

Inequality. Equality holds if and√only if xy/z = yz/x = zx/y, which is
equivalent to x = y = z. Hence S ≥ 3 if and only if x = y = z = 1/ 3.

32. The segments BE and CF are altitudes of the acute triangle ABC, where E and F are points on
the segments AC and AB, respectively. ABC is inscribed in the circle Q with centre O. Denote
the orthocentre of ABC by H, and the midpoints of BC and AH be M and K, respectively. Let
∠CAB = 45◦ .

(a) Prove, that the quadrilateral M EKF is a square.

(b) Prove that the midpoint of both diagonals of M EKF is also the midpoint of the segment OH.

(c) Find the length of EF , if the radius of Q has length 1 unit.

29
 Solution 1. (a) Since AH is the hypotenuse of right triangles AF H and AHE, KF = KH = KA =
KE. Since BC is the hypotenuse of each of the right triangles BCF and BCE, we have that M B =
M F = M E = M C. Since ∠BAC = 45◦ , triangles ABE, HF B and ACF are isosceles right triangles, so
∠ACF = ∠ABE = ∠F BH = ∠F HB = 45◦ and F A = F C, F H = F B.

Consider a 90◦ rotation with centre F that takes H → B. Then F A → F C, F H → F B, so ∆F HA →

∆F BC and K → M . Hence F K = F M and ∠KF M = 90◦ .

But F K = KE and F M = M E, so M EKF is an equilateral quadrilateral with one right angle, and
hence is a square.

(b) Consider a 180◦ rotation (half-turn) about the centre of the square. It takes K ↔ M , F ↔ E and
H ↔ H 0 . By part (a), ∆F HA ≡ ∆F BC and AH ⊥ BC. Since KHkM H 0 (by the half-turn), M H 0 ⊥ BC.
Since AH = BC, BM = 21 BC = 21 AH = KH = M H 0 , so that BM H 0 is a right isosceles triangle and
∠CH 0 M = ∠BH 0 M = 45◦ . Thus, ∠BH 0 C = 90◦ . Since ∠BAC = 45◦ , H 0 must be the centre of the circle
through ABC. Hence H 0 = O. Since O is the image of H by a half-turn about the centre of the square, this
centre is the midpoint of OH as well as of the diagonals.

√ √
(c) |EF | = 2|F M | = 2|BM | = |OB| = 1.

Solution 2. [M. Holmes] (a) Consider a Cartesian plane with origin F (0, 0) and x−axis along the line
AB. Let the vertices of the triangle be A(−1, 0), C(0, 1), B(b, 0). Since the triangle is acute, 0 < b < 1.
The point E is at the intersection of the line AC (y = x + 1) and a line through B with slope −1, so that
E = ( 12 (b − 1), 12 (b + 1). H is the intersection point of the lines BE and CF , so H is at (0, b); K is the
midpoint of AH, so K is at (− 21 , 2b ); M is the midpoint of BC, so M is at ( 2b , 12 ). It can be checked that
the midpoints of EF and KM are both at ( 14 (b − 1), 14 (b + 1)). The slope of EF is (b + 1)/(b − 1) and that
of KM is the negative reciprocal of this, so that EF ⊥ KM . It is straightforward to check that the lengths
of EF and KM are equal, and we deduce that EKF M is a square.

(b) O is the intersection point of the right bisectors of AB, AC and BC. The line x + y = 0 is the
right bisector of AC and the abscissae of points on the right bisector of BC are all 12 (b − 1). Hence O is at
( 12 (b − 1), 12 (1 − b)). It can be checked that the midpoint of OH agress with the joint midpoint of EF and
KM .

(c) This can be checked by using the coordinates of points already identified.

Comment. One of the most interesting theorems in triangle geometry states that for each triangle there
exists a circle that passes through the following nine special points: the three midpoints of the sides; the
three intersections of sides and altitudes (pedal points); and the three midpoints of the segments connecting
the vertices to the orthocentre. This circle is called the nine-point circle. If H is the orthocentre and O is
the circumcentre, then the centre of the nine-point circle is the midpoint of OH. Note that in this problem,
the points M, E, F, K belong to the nine-point circle.

33. Prove the inequality a2 + b2 + c2 + 2abc < 2, if the numbers a, b, c are the lengths of the sides of a
triangle with perimeter 2.

Solution 1. Let u = b + c − a, v = c + a − b and w = a + b − c, so that 2a = v + w, 2b = u + w

and 2c = u + v. Then u, v, w are all positive and u + v + w = 2. The difference of the right and left sides

30
multiplied by 4 is equal to

4[2 − (a2 + b2 + c2 + 2abc)]

= 8 − (v + w)2 − (u + w)2 − (u + v)2 − (v + w)(u + w)(u + v)]
= 8 − (u + v + w)2 − (u2 + v 2 + w2 ) − (2 − u)(2 − v)(2 − w)
= 8 − 4 − (u2 + v 2 + w2 ) − 8 + 4(u + v + w) − 2(vw + uw + uv) + uvw
= 4 − (u + v + w)2 − 8 + 4 × 2 + uvw
= 4 − 4 − 8 + 8 + uvw = uvw > 0

as desired.

Solution 2. [L. Hong] The perimeter of the triangle is a + b + c = 2. We have that

a2 + b2 + c2 + 2abc = (a + b + c)2 + 2(a − 1)(b − 1)(c − 1) + 2 − 2(a + b + c)

= 4 + 2(a − 1)(b − 1)(c − 1) + 2 − 4 = 2(a − 1)(b − 1)(c − 1) + 2 .

Since a < b + c, b < c + a and c < a + b, it follows that a < 1, b < 1, c < 1, from which the result follows.

34. Each of the edges of a cube is 1 unit in length, and is divided by two points into three equal parts.
Denote by K the solid with vertices at these points.

(a) Find the volume of K.

(b) Every pair of vertices of K is connected by a segment. Some of the segments are coloured. Prove that
it is always possible to find two vertices which are endpoints of the same number of coloured segments.

Solution. (a) The solid figure is obtained by slicing off from each corner a small tetrahedron, three of
whose faces are pairwise mutually perpendicular at one vertex; the edges emanating from that vertex all
have length 1/3, and so the volume of each tetrahedron removed is 1/3(1/2 · 1/3 · 1/3)(1/3) = 1/162. Since
there are eight such tetrahedra removed, the volume of the resulting solid is 1 − 4/81 = 77/81. (The numbers
of vertices, edges and faces of the solid are respectively 24, 36 and 14.)

(b) The polyhedron has 3 × 8 = 24 vertices. Each edge from a given vertex is joined to 23 vertices. The
possible number of coloured segments emanating from a vertex is one of the twenty-four numbers, 0, 1, 2,
· · ·, 23. But it is not possible for one vertex to be joined to all 23 others and another vertex to be joined to
no other vertex. So there are in effect only 23 options for the number of coloured segments emanating from
each of the 24 vertices. By the Pigeonhole Principle, there must be two vertices with the same number of
coloured segments emanating from it.

35. There are n points on a circle whose√radius is 1 unit. What is the greatest number of segments between
two of them, whose length exceeds 3?
√
Solution. [O.Bormashenko]
√ The side of the equilateral triangle inscribed in a circle of unit radius is 3.
So the segment with length 3 is a chord subtending an angle of 120◦ at the√centre. Therefore, there is no
triangle with three vertices on the circle each of whose sides are longer than 3.√ Consider the graph whose
vertices are all n given points and whose arcs all have segments longer than 3. This graph contains no
triangles.
Recall Turan’s theorem (see the solution of problem 23 in Olymon 1:4: Let G be a graph with n vertices.
Denote by l(G) the number of its edges and t(G) the number of triangles contained in the graph. If t(G) =√0,
then l(G) ≤ bn2 /4c. From this theorem, it follows that the number of segments with chords exceeding 3
is at most bn2 /4c.
To show that this maximum number can be obtained, first construct points A, B, C, D on the circle, so
that the disjoint arcs AB and CD subtend angles of 120◦ at the centre. If n = 2k + 1 is odd, place k points

31
on the arc BC and k + 1 points on the arc DA. Any segment √ containing a point in BC to a point in DA
must subtend an angle exceeding 120◦ , so its length exceeds 3. There are exactly k(k + 1) = b(2k + 1)2 /4c
such segments. If n = 2k is even, place k points in each of the arcs BC and CA, so that there are exactly
√
k 2 = b(2k)2 /4c such segments. In either case, the maximum number of segments whose length exceeds 3
is bn2 /4c.

36. Prove that there are not three rational numbers x, y, z such that

x2 + y 2 + z 2 + 3(x + y + z) + 5 = 0 .

Solution. Suppose the x = u/m, y = v/m and z = w/m, where m is the least common multiple of the
denominators of x, y and z. Then, multiplying the given equation by m2 yields

u2 + v 2 + w2 + 3(um + vm + wm) + 5m2 = 0 .

A further multiplication by 4 and a rearrangement of terms yields

(2u + 3m)2 + (2v + 3m)2 + (2w + 3m)2 = 7m2 .

This is of the form

p2 + q 2 + r2 = 7n2 (∗)
for integers p, q, r, n. Suppose that n is even. Then, considering the equation modulo 4, we deduce that
p, q, r must also be even. We can divide the equation by 4 to obtain another equation of the form (*) with
smaller numbers. We can continue to do this as long as the resulting n turns out to be even. Eventually, we
arrive at an equation for which n is odd, so that the right side is congruent to 7 modulo 8. But there is not
combination of integers p, q and r for which p2 + q 2 + r2 ≡ 7 (mod 8), so that the equation is impossible.

37. Let ABC be a triangle with sides a, b, c, inradius r and circumradius R (using the conventional notation).
Prove that
r abc
≤p .
2R 2(a + b )(b2 + c2 )(c2 + a2 )
2 2

When does equality hold?

Solution. Suppose, first, that all angle of the triangle are acute.

a2 − (b2 + c2 )(1 − cos A) = b2 + c2 − 2bc cos A − (b2 + c2 ) + (b2 + c2 ) cos A

≥ (b − c)2 cos A ≥ 0

=⇒ a2 ≥ (b2 + c2 )(1 − cos A) = 2(b2 + c2 ) sin2 (A/2) .

With similar inequalities for b and c, we find that

a2 b2 c2 ≥ 8(a2 + b2 )(b2 + c2 )(c2 + a2 ) sin2 (A/2) sin2 (B/2) sin2 (C/2) .

Since r = 4R sin(A/2) sin(B/2) sin(C/2), the desired result follows. Equality holds if and only if the triangle
is equilateral.
Suppose that the angle A is obtuse. Then cos A < 0,

a2 = b2 + c2 − 2bc cos A > b2 + c2

and
2[ABC] = bc sin A < bc .

32
Since b + c − a, a2 − b2 − c2 and b2 − bc + c2 are all positive, we have that

a2 (a + b + c)2 − 2(a2 + b2 )(c2 + a2 )

= 2(b + c − a)a3 + (a2 − b2 − c2 )(a2 + 2bc) + 2bc(b2 − bc + c2 ) > 0 ,

whence
2(a2 + b2 )(c2 + a2 ) < a2 (a + b + c)2 .
Hence
(b2 + c2 ) · 16[ABC]4 · 2(a2 + b2 )(c2 + a2 ) < a2 · b4 c4 · a2 (a + b + c)2 .
Since [ABC] = abc/4R = (a + b + c)r/2, this becomes

(abc)2 (a + b + c)2 r2
 
2 2 2 2 2 2
(b + c )(c + a )(a + b ) · < (abc)4 · (a + b + c)2 ,
4R2

from which the result follows.

Comment. The identity in the solution can be obtained as follows. Let 2s = a + b + c. Then

r A sin A2
= tan =
s−a 2 cos A2

while
A A
a = 2R sin A = 4R sin cos .
2 2
Hence
ar A
= 4R sin2 .
s−a 2
Using similar identities for the other sides, we find that

abcr3 A B C
= 64R3 sin2 sin2 sin2 . (∗)
(s − a)(s − b)(s − c) 2 2 2

Note that the area ∆ of the triangle is given by

abc p
∆ = rs = = s(s − a)(s − b)(s − c) ,
4R

so that the left side of (∗) becomes 4R∆r2 (rs)∆−2 = 4Rr2 . Substituting this in, dividing by 4R and taking
the square root yields
A B C
r = 4R sin sin sin .
2 2 2

38. Let us say that a set S of nonnegative real numbers if hunky-dory if and only if, for all x and y in
S, either x + y or |x − y| is in S. For instance, if r is positive and n is a natural number, then
S(n, r) = {0, r, 2r, · · · , nr} is hunky-dory. Show that every hunky-dory set is {0}, is of the form S(n, r)
or has exactly four elements.

Solution 1. {0} and sets of the form {0, r} are clearly hunky-dory. Let S be a nontrivial hunky-dory
set with largest postive element z. Then 2z 6∈ S, so 0 = z − z ∈ S. Thus, every hunky-dory set contains 0.
Suppose that S has at least three elements, with least positive element a.
Suppose, if possible, that S contains an element that is not a positive integer multiple of a. Let b be
the least nonmultiple of a. Then 0 < b − a < b. Since b − a cannot be a multiple of a (why?), we must
have b − a 6∈ S and b + a ∈ S. Since z is the largest element of S, z − a and z − b belong to S. However,

33
(z −a)−(z −b) = b−a does not belong to S, so 2z −(a+b) = (z −a)+(z −b) ∈ S. Therefore, 2z −(a+b) ≤ z,
whence z ≤ a + b, so that z = a + b. Thus, S contains {0, a, b, a + b}, with a + b the largest element. This
subset is already hunky-dory. But suppose, if possible, S contains more elements. Let c be the smallest such
element. Then 0 < (a + b) − c ∈ S ⇒ a ≤ (a + b) − c ⇒ c < b ⇒ c = ma for some positive integer m ≥ 2.
Since b + ma > b + a, b − ma must belong to S, and so be a multiple of a. This yields a contradiction. Hence,
S must be equal to {0, a, b, a + b}.
The only remaining case is that S consists solely of nonnegative multiples of some element a. Let na be
the largest such multiple. If n = 2, then S = S(2, a). Suppose that n > 2. Then (n − 1)a ∈ S, so S contains
{0, a, (n − 1)a, na}, which is hunky-dory.
Suppose S contains a further multiple ma with 2 ≤ m ≤ n − 2. Since a ∈ S and na + a > na,
(n − 1)a ∈ S, so that n − (m + 1)a = (n − 1)a − ma ∈ S ⇒ (m + 1)a = n − [n − (m + 1)a] ∈ S. By induction,
it can be shown that ka ∈ S for m ≤ k ≤ n. In particular, (n − 2)a ∈ S so that 2a = na − (n − 2)a ∈ S.
But then 3a, 4a, · · · , na are in S and so S = S(n, a). The desired result follows,

Solution 2. [S. Niu] Let S = {a0 , a1 , · · · , an }, with a0 < a1 < · · · < an . The elements an − an , an − an−1 ,
· · ·, an − a0 are n + 1 distinct elements of S listed in increasing order, and so a0 = 0, and for each i with
0 ≤ i ≤ n, we must have that an − ai = an−i . Let i ≤ n2 . Then i ≤ n − i and so ai ≤ an−i ; thus,
ai ≤ (an )/2 ≤ an−i . Thus, if j > k ≥ n/2, aj − ak ∈ S.
Since 0 < an−1 − an−2 < an − an−2 = a2 , it follows that an−1 − an−2 = a1 . Also, 0 < an−1 − an−2 =
a1 < an−1 − an−3 < an − an−3 = a3 , so that an−1 − an−3 = a2 . Continuing on in this way, we find that, for
i ≥ n/2,
0 < an−1 − an−2 < an−1 − an−3 < · · · < an−1 − an−i < an − an−i = ai ,

whence an−1 − an−j = aj−1 for 1 ≤ j ≤ (n/2).

Now 0 < an−2 − an−3 < an−1 − an−3 = a2 so an−2 − an−3 = a1 . We can proceed in this fashion to
obtain that, for j ≥ n/2, aj+1 − aj = a1 . Hence, for i ≤ (n/2) − 1, ai+1 − ai = (an − an−i−1 ) − (an − an−i ) =
an−i − an−i−1 = a1 .
Let n = 2m. Then ai = ia1 and an−i = an − ia1 for 1 ≤ i ≤ m, so that am = ma1 and an = am + ma1 =
na1 . It follows that ak = ka1 for 1 ≤ k ≤ n and S = S(n, a1 ).
Let n = 2m + 1. If m = 0, then S = {0, a1 } = S(1, a1 ). If m = 1, then S = {0, a1 , a3 − a1 , a3 } =
{0, a1 , a2 , a1 + a2 } is a 4-element hunky-dory set. Let m ≥ 2. Then, for 1 ≤ i ≤ m, ai = ia1 and
an−i = an − iai . Now am+1 = an − am > an−1 − am > · · · ≥ am+2 − am > am+1 − am ≥ a1 . Since
{an − am = am+(m+1) − am , an−1 − am , · · · , am+2 − am , am+1 − am } contains m + 1 elements, we must have
am+j − am = aj for 1 ≤ j ≤ n − m = m + 1. Therefore, ai = iai for 1 ≤ i ≤ n. (Why does this last statement
fail to follow when m = 1?)

39. (a) ABCDEF is a convex hexagon, each of whose diagonals AD, BE and CF pass through a common
point. Must each of these diagonals bisect the area?
(b) ABCDEF is a convex hexagon, each of whose diagonals AD, BE and CF bisects the area (so that
half the area of the hexagon lies on either side of the diagonal). Must the three diagonals pass through a
common point?

Solution 1. (a) No, they need not bisect the area. Let the vertices of the hexagon have coordinates
(−1, 0), (−1, −1), (1, −1), (1, 0), (−t, t), (t, −t) with t > 0 but t 6= 1. The diagonals with equations y = 0,
y = x and y = −x intersect in the origin but do not bisect the area of the hexagon.

(b) Let the hexagon be ABCDEF and suppose that the intersection of the diagonals AD and BE is
on the same side of CF as the side AB. Thus, AB, CD and EF border on triangles whose third vertices
form a triangle at the centre of the hexagon (we will show this triangle to be degenerate). Let a, b, c, d,
e, f be the lengths of the rays from the respective vertices A, B, C, D, E, F to the vertices of the central

34
triangle, whose sides are x, y, z so that the lengths of AD, BE and CF are respectively a + x + d, b + y + e,
c + z + f . All lower-case variables represent nonnegative real numbers.
Let the areas of the bordering on F A, AB, BC, CD, DE, EF be respectively α, β, γ, δ, , φ, and let
the area of the central triangle be λ. Then, since each diagonal bisects the area of the hexagon, we have that

α+β+γ+λ=δ++φ

+φ+α+λ=β+γ+δ
γ+δ++λ=φ+α+β .
From the first two equations, we find that δ = α + λ. Similarly, φ = γ + λ and β =  + λ.
Using the fact that the area of a trangle is half the product of adjacent sides and the sine of the angle
between them, and the equality of opposite angles, we find that

α+λ (a + x)(f + z)
1= =
δ cd
γ+λ (b + y)(c + z)
1= =
φ ef
+λ (d + x)(e + y)
1= = .
β ab
Multiplying these three equations together yields that

abcdef = (a + x)(b + y)(c + z)(d + x)(e + y)c + z) ,

whence x = y = z = 0. Thus, the central triangle degenerates and the three diagonals intersect in a common
point.

Solution 2. (a) No. Let ABCDEF be a regular hexagon. The diagonals AD, BE, CF intersect and each
diagonal does bisect the area. Let X be any point other than F on the diagonal CF for which ABCDEX
is still a convex hexagon. The diagonals of this hexagon are the same as those of the regular hexagon, and
so have a common point of intersection. However, the diagonals AD and BE no longer intersect the area of
the hexagon.

(b) [X. Li] Let ABCDEF be a given convex hexagon, each of whose diagonals bisect its area. Suppose
that the diagonals AD and CF intersect at G. As in Solution 1, we can determine that the areas of triangles
AGF and DGC are equal, whence AG·GF = CG·GD, or AG/GD = CG/GF . Therefore, ∆AGC ∼ ∆DGF
(SAS). It follows that AC/DF = AG/GD = CG/GF , ∠CAG = ∠F DG, and so ACkDF . In a similar way,
we find that BF kCE and AEkBD, so that ∆ACE ∼ ∆DF B and AC/DF = CE/F B = EA/BD.
Suppose diagonals AD and BE intersect at H. Then, as above, we find that AG/GD = AC/DF =
EA/BD = EH/HB = AH/HD, so that H = G. Hence, the three diagonals have the point G in common.

40. Determine all solutions in integer pairs (x, y) to the diophantine equation x2 = 1 + 4y 3 (y + 2).

Solution 1. Clearly, (x, y) = (±1, 0), (±1, −2) are solutions. When y = −1, the right side is negative
and there is no solution. Suppose that y ≥ 1; then

(2y 2 + 2y)2 = 4y 4 + 8y 3 + 4y 2 > 4y 4 + 8y 3 + 1

and
(2y 2 + 2y − 1)2 = 4y 4 + 8y 3 − 4y + 1 < 4y 4 + 8y 3 + 1
so that the right side is between two consecutive squares, and hence itself cannot be square.

35
 Suppose that y ≤ −3. We first observe that for a given product p of two positive integers, the sum of
√
these positive integers has a minium value of 2 p (why?) and a maximum value of 1 + p. This follows from
the fact that, for integers u with 1 ≤ u ≤ p,

(1 + p) − (u + p/u) = (u − 1)[(p/u) − 1] ≥ 0 .

We have that
[(2y 2 + 2y − 1) + x][(2y 2 + 2y − 1) − x] = (2y 2 + 2y − 1)2 − x2
= (4y 4 + 8y 3 − 4y + 1) − (4y 4 + 8y 3 + 1)
= −4y .
Since 2y 2 + 2y − 1 = y 2 + (y + 1)2 − 2 is positive, at least one of the factors on the left is positive. Since the
product is positive, both factors are positive. By our observation on the sum of the factors, we find that

4y 2 + 4y − 2 ≤ 1 − 4y ,

which is equivalent to
4(y − 1)2 ≤ 7 .
However, this does not hold when y ≤ −3. Therefore, the only solutions are the four that we identified at
the outset.

Solution 2. Since x must be odd, we can let x = 2z + 1 for some integer z, so that the equation
becomes z(z + 1) = y 4 + 2y 3 . We can deal with the cases that y = 0, −1, −2 directly to obtain the solutions
(x, y, z) = (1, 0, 0), (−1, 0, −1), (1, −2, 0), (−1, −2, −1). Henceforth, suppose that y ≥ 1 or y ≤ −3, so that
y 4 + 2y 3 is positive. Let φ(t) = t(t + 1). Then φ(t) is increasing for t ≥ 0 and φ(−t) = φ(t − 1) for every
integer t; thus, we need check only that y 4 + 2y 3 does not coincide with a value taken by φ(t) for nonnegative
values of t.
Now
φ(y 2 + y) = y 4 + 2y 3 + 2y 2 + y > y 4 + 2y 3 ;
φ(y 2 + y − 1) = y 4 + 2y 3 − y 6= y 4 + 2y 3 ;
φ(y 2 + y − 2) = y 4 + 2y 3 − 2y 2 − 3y + 2
= y 4 + 2y 3 − (2y + 1)(y + 1) + 3 < y 4 + 2y 3 .
It follows that φ(t) can never assume the value y 4 + 2y 3 for any positive t, and hence for any t. Thus, the
solutions already listed comprise the complete solution set.

41. Determine the least positive number p for which there exists a positive number q such that
√ √ xp
1+x+ 1−x≤2−
q

for 0 ≤ x ≤ 1. For this least value of p, what is the smallest value of q for which the inequality is
satisfied for 0 ≤ x ≤ 1?

Comments. Recall the binomial expansion

n(n − 1) 2 n(n − 1)(n − 2) 3 n(n − 1) · · · (n − r + 1) r

(1 + x)n = 1 + nx + x + x + ··· + x + ··· .
2! 3! r!
When n is not a nonnegative integer, this is an infinite series that converges when 0 ≤ |x| < 1 to (1 + x)n .
The partial sums constitute a close approximation. When n = 12 , we have that

1 1 1 1 5 4
(1 ± x) 2 = 1 ± x − x2 ± x3 − x ± ···
2 8 16 128

36
so that
1 1 x2 x4 x4
(1 + x) 2 + (1 − x) 2 ∼ 2 − − ≤2− .
4 8 4
This suggests that we are looking for (p, q) = (2, 4). However, the approximation approach is not sufficiently
rigorous, and we need to find an argument in finite terms that will work.

Solution 1. Observe that, for 0 ≤ x ≤ 1,

√ 1 1 1
1 ± x ≤ 1 ± x − x2 ± x3
2 8 16
5 4 1 1 6
⇔1±x≤1±x+ x ∓ x5 + x
64 64 256
⇔ 0 ≤ 5 ∓ x + 4x2 .
The last inequality clearly holds, so the first must as well. Hence
√ √ 1 1
1+x+ 1 − x ≤ 2(1 − x2 ) = 2 − x2
8 4
so the pair (p, q) = (2, 4) works for all x ∈ [0, 1].
Suppose, for some constants p and c with 0 < p < 2 and c > 0,
√ √
1 + x + 1 − x ≤ 2 − 2cxp

for 0 ≤ x ≤ 1. For this range of x, this is equivalent to

p
2 + 2 1 − x2 ≤ 4 − 8cxp + 4c2 x2p
p
⇔ 1 − x2 ≤ 1 − 4cxp + 2c2 x2p
⇔ 1 − x2 ≤ 1 − 8cxp + 20c2 x2p − 16c3 x3p + 4c4 x4p
8c ≤ x2−p + 4c2 xp (5 − 4cx2p + c3 x3p ) .
However, for x sufficiently small, the right side can be made less than 8c, yielding a contradiction. Hence,
when 0 < p < 2, there is no value that yields the desired inequality.
Now we look at the situation when p = 2 and q > 0. For 0 ≤ x ≤ 1,

√ √ x2
1+x+ 1−x≤2−
q
p 4x2 x4
⇔ 2(1 + 1 − x2 ) ≤ 4 − + 2
q q
p 2x2 x4
⇔ 1 − x2 ≤ 1 − + 2
q 2q
4x2 5x4 2x6 x8
⇔ 1 − x2 ≤ 1 − + 2 − 3 + 4
q q q 4q
2 2
x4
   
2 4 x 2x
⇔0≤x 1− + 2 5− + 2 .
q q q 4q
If q < 4, then the quantity in square brackets is negative for small values of x. Hence, for the inequality to
hold for all x in the interval [0, 1], we must have q ≥ 4. Hence, p must be at least 2, and for p = 2, q must
be at least 4.

37
 Solution 2. [R. Furmaniak] The given inequality is equivalent to
xp
q≥ √ √
2− 1+x− 1−x
√ √
xp (2 + 1 + x + 1 − x)
= √
4 − (2 + 2 1 − x2 )
√ √
xp (2 + 1 + x + 1 − x)
= √
2(1 − 1 − x2 )
√ √ √
xp (2 + 1 + x + 1 − x)(1 + 1 − x2 )
= .
2x2
If p < 2, then the right side becomes arbitrarily large as x gets close to zero, so the inequality becomes
unsustainable for any real q. Hence, for the inequality to be viable, we require p ≥ 2. When p = 2, we can
cancel x2 and see by taking x = 0 that q ≥ 4. It remains to verify the√inequality when (p, q) = (2, 4). We
have the following chain of logically equivalent statements, where y = 1 − x2 (note that 0 ≤ x ≤ 1):
√ √ x2
1+x+ 1−x≤2−
4
p x4
⇔ 2 + 2 1 − x2 ≤ 4 − x2 +
16
p
2 4 2
⇔ 32 1 − x ≤ x − 16x + 32
⇔ 32y ≤ 1 − 2y 2 + y 4 − 16 + 16y 2 + 32
⇔
0 ≤ y + 14y − 32y + 17 = (y − 1) (y 2 + 2y + 17) = (y − 1)2 [(y + 1)2 + 16] .
4 2 2

Since the last inequality is clearly true, the first holds and the result follows.

42. G is a connected graph; that is, it consists of a number of vertices, some pairs of which are joined by
edges, and, for any two vertices, one can travel from one to another along a chain of edges. We call two
vertices adjacent if and only if they are endpoints of the same edge. Suppose there is associated with
each vertex v a nonnegative integer f (v) such that all of the following hold:
(1) If v and w are adjacent, then |f (v) − f (w)| ≤ 1.
(2) If f (v) > 0, then v is adjacent to at least one vertex w such that f (w) < f (v).
(3) There is exactly one vertex u such that f (u) = 0.
Prove that f (v) is the number of edges in the chain with the fewest edges connecting u and v.

Solution. We prove by induction that f (x) = n if and only if the shortest chain from u to x has n
members. This is true for n = 0 (and for n = 1). Suppose that this holds for 0 ≤ n ≤ k.
Let f (x) = k + 1. There exists a vertex y adjacent to x for which h = f (y) < k + 1. By the induction
hupothesis, y can be connected to u by a chain of h edges, so x can be connected to u by a chain of h + 1
edges. Hence, h + 1 ≥ k + 1. From these two inequalities, we must have h = k, so x can be connected to
u by a chain of k + 1 edges. There cannot be a shorter chain, as, by the induction hypothesis, this would
mean that f (x) would have to be less than k + 1.
Let the shortest chain connecting x to u have k + 1 edges. Following along this chain, we can find an
element z adjacent to x connected to u by k edges. This must be one of the shortest chains between u and
z, so that f (z) = k. By hypothesis (1), f (x) must take one of the values k − 1 and k + 1. The first is not
admissible, since there is no chain with k − 1 edges connecting u and x. Hence f (x) = k + 1.

43. Two players play a game: the first player thinks of n integers x1 , x2 , · · ·, xn , each with one digit,
and the second player selects some numbers a1 , a2 , · · ·, an and asks what is the value of the sum

38
 a1 x1 + a2 x2 + · · · + an xn . What is the minimum number of questions used by the second player to find
the integers x1 , x2 , · · ·, xn ?

Solution. We are going to prove that the second player needs only one question to find the integers x1 ,
x2 , · · ·, xn . Indeed, let him choose a1 = 100, a2 = 1002 , · · ·, an = 100n and ask for the value of the sum

Sn = 100x1 + 1002 x2 + · · · + 100n xn .

Note that
100x1 + 1002 x2 + · · · + 100n−1 xn−1
100n
100|x1 | + 1002 |x2 | + · · · + 100n−1 |xn−1 |
≤
100n
9(100 + 100 + · · · + 100n−1 )
2
≤
100n
102n−1 1
< = .
102n 10
Hence
Sn 100x1 + 1002 x2 + · · · + 100n−1 xn−1 1
n
− xn = < ,
100 100n 10
and xn can be obtained. Now, we can find the sum

Sn−1 = Sn − 100n xn = 100x1 + 1002 x2 + · · · + 100n−1 xn−1

and similarly obtain xn−1 . The procedure continues until all the numbers are found.

44. Find the permutation {a1 , a2 , · · · , an } of the set {1, 2, · · · , n} for which the sum

S = |a2 − a1 | + |a3 − a2 | + · · · + |an − an−1 |

has maximum value.

Solution. Let a = (a1 , a2 , · · · , an ) be a permutation of {1, 2, · · · , n} and define

n−1
X
f (a) = |ak+1 − ak | + |an − a1 | .
k=1

With an+1 = a1 and 0 = n , we find that

n
X n
X
f (a) = k (ak+1 − ak ) = (k−1 − k )ak
k=1 k=1

where 1 , 2 , · · ·, n are all equal to 1 or −1. Thus

n
X
f (a) = βk k .
k=1
Pn
where each βk is one of −2, 0, 2, and k=1 βk = 0 (there are the same number of positive and negative
numbers among the βk ).

Therefore
f (a) = 2(x1 + x2 + · · · + xm ) − 2(y1 + y2 + · · · + ym )

39
where x1 , · · · , xm , y1 , · · · , ym ∈ {1, 2, · · · , n} and are distinct from each other. Hence f (a) is maximum when
{x1 , x2 , · · · , xm } = {n, n − 1, · · · , n − m + 1}, {y1 , y2 , · · · , ym } = {1, 2, · · · , m} with m = bn/2c, i.e., m is as
large as possible. The maximum value is
   
n n
2 n− .
2 2

This value is attained by taking

a = (s + 1, 1, s + 2, 2, · · · , 2s, s) when n = 2s

and
a = (s + 2, 1, s + 3, 2, · · · , 2s + 1, s, s + 1) when n = 2s + 1 .
Since |an − a1 | = 1 for these permutations, the maximum value of the given expression is
   
n n
2 n− −1 .
2 2

This is equal to 2s2 − 1 when n = 2s, and to 2s(s + 1) − 1 when n = 2s + 1.

45. Prove that there is no nonconstant polynomial p(x) = an xn +an−1 xn−1 +· · ·+a0 with integer coefficients
ai for which p(m) is a prime number for every integer m.

Solution. Let a be an integer, for which p(a) 6= −1, 0, 1. (If there is no such a, then p cannot take all
prime values.) Suppose that b is a prime divisor of p(a). Now, for any integer k,

p(a + kb) − p(a) = an [(a + kb)n − an ] + an−1 [(a + kb)n−1 − an−1 ] + · · · + a1 [(a + kb) − a] .

It can be seen that b is a divisor of p(a + kb) − p(a) and hence of p(a + kb) for every integer k. Both of the
equations p(x) = b and p(x) = −b have at most finitely many roots. So some of the values of p(a + kb) must
be composite, and the result follows.
Comment. It should have been stated in the problem that the polynomial was nonconstant, or had
positive degree.

an +2
46. Let a1 = 2, an+1 = 1−2an for n = 1, 2, · · ·. Prove that
(a) an 6= 0 for each positive integer n;
(b) there is no integer p ≥ 1 for which an+p = an for every integer n ≥ 1 (i.e., the sequence is not
periodic).

Solution. (a) We prove that an = tan nα where α = arctan 2 by mathematical induction. This is true
for n = 1. Assume that it holds for n = k. Then
2 + an tan α + tan nα
ak+1 = = = tan(n + 1)α ,
1 − 2an 1 − tan α tan nα
as desired.
Suppose that an = 0 with n = 2m + 1. Then a2m = −2. However,
2 tan mα 2am
a2m = tan 2mα = = ,
1 − tan2 mα 1 − a2m

whence √
2am 1± 5
= −2 ⇔ am = ,
1 − a2m 2

40
which is not possible, since am has to be rational.
Suppose that an = 0 with n = 2k (2m + 1) for some positive integer k. Then

2 tan 2k−1 (2m + 1) 2an/2

0 = an = tan 2 · 2k−1 (2m + 1)α = = ,
1 − tan2 2k−1 (2m + 1) 1 − a2n/2

so that an/2 = 0. Continuing step by step backward, we finally come to a2m+1 = 0, which has already been
shown as impossible.

(b) Assume, if possible, that the sequence is periodic, i.e., there is a positive integer p such that
an+p = an for every positive integer n. Thus

sin pα
tan(n + p)α − tan nα = =0.
cos(n + p)α cos nα

Therefore sin pα = 0 and so ap = tan pα = 0, which, as we have shown, is impossible. The desired result
follows.

47. Let a1 , a2 , · · · , an be positive real numbers such that a1 a2 · · · an = 1. Prove that

n
X 1
≤1
s − ak
k=1

where s = 1 + a1 + a2 + · · · + an .

Solution. First, we recall that Chebyshev’s Inequalities:

(1) if the vectors (a1 , a2 , · · · , an ) and (b1 , b2 , · · · , bn ) are similarly sorted (that is, both rising or both falling),
then
a1 b1 + · · · + an bn a1 + · · · + an b1 + · · · + bn
≥ · ;
n n n
(2) if the vectors (a1 , a2 , · · · , an ) and (b1 , b2 , · · · , bn ) are oppositely sorted (that is, one rising and the other
falling), then
a1 b1 + · · · + an bn a1 + · · · + an b1 + · · · + bn
≤ · .
n n n

If x1 , x2 , · · · , xn are positive real numbers with x1 ≤ x1 ≤ · · · ≤ xn , then xn1 ≤ xn2 ≤ · · · ≤ xnn . From
Chebyshev’s Inequality (1), we have, for each k = 1, 2, · · · , n, that
n n n
 X n
 X 
X X 1
xni = xin−1 xi ≥ xi xn−1
i .
n−1
i=1,i6=k i=1,i6=k i=1,i6=k i=1,i6=k

The Arithmetic-Geometric Means Inequality yields

n
X
xn−1
i ≥ (n − 1)x1 · · · xk−1 xk+1 · · · xn ,
i=1,i6=k

for k = 1, · · · , n. Therefore,
n
X n
X
xni ≥ (x1 · · · xk−1 xk+1 · · · xn ) xi .
i=1,i6=k i=1,i6=k

for each k. This inequality can also be written

xn1 + · · · + xnk−1 + xnk+1 + · · · xnn + x1 x2 · · · xn

41
 ≥ x1 · · · xk−1 xk+1 · · · xn (x1 + x2 + · · · + xn ) ,
or
1 1 1
≤ · .
x1 x2 · · · xn + xn1 + · · · + xnk−1 + xnk+1 + · · · + xnn x1 + x2 + · · · + xn x1 · · · xk−1 xk+1 · · · xn

Adding up these inequalities, for 1 ≤ k ≤ n, we get

n
X 1 1
≤ .
x1 x2 · · · xn + xn1 n n n
+ · · · + xk−1 + xk+1 + . . . xn x1 x2 · · · xn
k=1

Now, let the xnk be equal to the ak in increasing order to obtain the desired result.

48. Let A1 A2 · · · An be a regular n−gon and d an arbitrary line. The parallels through Ai to d intersect its
circumcircle respectively at Bi (i = 1, 2, · · · , n. Prove that the sum

S = |A1 B1 |2 + · · · + |An Bn |2

is independent of d.

Solution. Select a system of coordinates so that O is the centre of the circumcircle and the x−axis (or
real axis) is orthogonal to d. Without loss of generality, we may assume that the radius of the circumcircle
is of length 1. Let ak the the affix (complex number representative) of Ak (1 ≤ k ≤ n). Then the ak
are solutions of the equation z n = λ, where λ is a complex number with |λ| = 1. Since Ak and Bk are
symmetrical with respect to the real axis, the affix of Bk is ak , the complex conjugate of ak , for 1 ≤ k ≤ n,
Thus
Ak Bk2 = |ak − ak |2 = (ak − ak )(ak − ak ) = 2ak ak − a2k − ak 2 = 2 − ak − ak 2 .
Summing these inequalities yields that
n
X n
X n
X
Ak Bk2 = 2n − a2k − ak 2 .
k=1 k=1 k=1

Since {ak : 1 ≤ k ≤ n} is a complete set of solutions of the equation z n = λ, their sum and the sum of their
pairwise products vanishes. Hence
Xn Xn
0= a2k = ak 2 .
k=1 k=1
Pn
Hence k=1 Ak Bk2 = 2n .

49. Find all ordered pairs (x, y) that are solutions of the following system of two equations (where a is a
parameter):
x−y =2
  
2 2
x− y− = a2 − 1 .
a a
Find all values of the parameter a for which the solutions of the system are two pairs of nonnegative
numbers. Find the minimum value of x + y for these values of a.

Solution. We need to assume that a 6= 0. Substitute y = x − 2 into the second equation to obtain

4 4
x2 − 2x − (x − 1) + 2 = a2 − 1 .
a a

42
This can be manipulated to
  2
2
x− 1+ = a2 ,
a
from which we find that

(x, y) = (1 + (2/a) + a, −1 + (2/a) + a), (1 + (2/a) − a, −1 + (2/a) − a) .

For x and y to be nonnegative for both solutions, we require all of the four inequalities:

a(a2 + a + 2) ≥ 0 , a(a2 − a + 2) ≥ 0 ,

a(a − 2)r(a + 1) = a(a2 − a − 2) ≤ 0 , a(a + 2)(a − 1) = a(a2 + a − 2) ≤ 0 .

The last two equations entail that a ≤ −2 or 0 < a ≤ 1. When a ≤ −2, the first two inequalities fail to hold,
while if 0 < a ≤ 1, all four equations hold. Thus, the solutions are two pairs of nonnegative numbers when
0 < a ≤ 1.
When both x and y are nonnegative, then x + y ≥ x − y = 2. When a = 1, we obtain the solutions
(x, y) = (4, 2) and (2, 0), and in the latter case, x + y = 2. Hence the minimum value of x + y is 2.

Comment. The minimum of x + y can be also obtained with more effort directly from the solutions in
terms of a. Now x + y = (4/a) ± 2a. Since
 
4 2(2 − a)(1 − a)
+ 2a − 6 = ≥0
a a

for 0 < a ≤ 1, (4/a) + 2a assumes its minimum value of 6 when a = 1. Since

 
4 2(2 + a)(1 − a)
− 2a − 2 = ≥0
a a

for 0 < a ≤ 1, (4/a) − 2a assumes its minimum value of 2 when a = 1. Hence the minimum possible value
of x + y is 2.

50. Let n be a natural number exceeding 1, and let An be the set of all natural numbers that are not
relatively prime with n (i.e., An = {x ∈ N : gcd (x, n) 6= 1}. Let us call the number n magic if for each
two numbers x, y ∈ An , their sum x + y is also an element of An (i.e., x + y ∈ An for x, y ∈ An ).
(a) Prove that 67 is a magic number.
(b) Prove that 2001 is not a magic number.
(c) Find all magic numbers.

Solution. [O. Bormashenko] (a) 67 is prime, so that all numbers that are not relatively prime with 67
and are elements of A67 are the multiples of 67. The sum of two such numbers is also a multiple of 67, and
hence belongs to A67 . Therefore 67 is a magic number.
(b) 2001 = 3 × 23 × 29. Now 3 and 23 belong to A2001 , but 26 = 3 + 23 does not, because gcd (26,
2001) is equal to 1. Thus, 2001 is not a magic number.
(c) First, let us prove that all prime powers are magic numbers. Suppose that n = pk for some prime
p and positive integer k. Then the numbers not relatively prime to n are precisely those that are divisible
by p (as this is the only prime that can divide any divisor of n). The sum of any two such numbers is also
divisible by p, so that An is closed under addition and n is magic.
Now suppose that n is not a power of a prime. Then n = ab, where a and b are two relatively prime
numbers exceeding 1. Clearly, a and b belong to An . However, gcd(a, a + b) = gcd(a, b) = 1, and gcd(b, a + b)

43
= gcd(b, a) = 1, so that a + b is relatively prime to both a and b, and hence to n = ab. Thus, a + b 6∈ An , so
that n is not magic.
We conclude that the set of prime powers equals the set of magic numbers.

51. In the triangle ABC, AB = 15, BC = 13 and AC = 12. Prove that, for this triangle, the angle bisector
from A, the median from B and the altitude from C are concurrent (i.e., meet in a common point).

Solution. [K. Ho] Denote by E the intersection of AC with the median from B (so that E is the midpoint
of AC), by F the intersection of AB with the altitude from C (so that CF ⊥ AB), by D the intersection of
BC and the bisector from A, and by x the length of AF . Then BF = AB−AF = 15−x. By the angle-bisector
theorem, CD : DB = AC : AB = 12 : 15 = 4 : 5. By the pythagorean theorem for the triangles AF C and
BF C, AC 2 − AF 2 = CF 2 = BC 2 − BF 2 ⇔ 122 − x2 = 132 − (15 − x)2 = −56 + 30x − x2 . Thus, x = 20/3
and 15 − x = 25/3, so that AF : F B = 4 : 5. Since (AF/F B)(BD/DC)/(CE/EA) = (4/5)(5/4)1 = 1,
Ceva’s theorem tells us that AD, BE and CF are concurrent.
√
52. One solution of the equation 2x3 + ax2 + bx + 8 = 0 is 1 + 3. Given that a and b are rational numbers,
determine its other two solutions.
√
Solution 1. [R. Barrington Leigh] Since 1 + 3 satisfies the given equation,
√ √ √ √
0 = 2(1 + 3)3 + a(1 + 3)2 + b(1 + 3) + 8 = (28 + 4a + b) + (12 + 2a + b) 3 .

Since a and b are rational, this is possible only when 28 + 4a + b = 12 + 2a + b = 0, or (a, b) = (−8, 4). The
0 = 2x3 − 8x√
equation is thus √ 2
+ 4x + 8. By inspection, we find that 2 is a root, and so the three roots turn
out to be 2, 1 + 3 and 1 − 3.
√
Comment. Once 2 and 1 + 3 are known to be roots, we can get the third root by noting that the sum
of the roots is −(−8)/2 = 4.

Solution
√ 2. A more structural way of getting the result is to √ note that the mapping that takes a
surd u + v 3 (with u and v rational) to its surd conjugate u − v 3 preserves addition, subtraction and
multiplication (i.e., the surd conjugate of the sum (resp. product) or two surds is equal to the sum (resp.
product) of the surd conjugates. The surd conjugate of a rational is the rational itself. Tranforming all
the elements of the equation to their surd√ conjugates,
√ gives the same equation with x replaced by its surd
conjugate. Thus, the surd conjugate 1 − 3 of 1 + 3 also satisfies the equation. The quadratic with these
as roots is x2 − 2x − 2, and this quadratic must be a factor of the given cubic. Since 2x3 + ax2 + bx + 8 =
(x2 − 2x − 2)(2 + [a + 4]x) + (2a + b + 12)x + (16 + 2a), we must have 2a + b + 12 = 16 + 2a = 0, whence
(a, b) = (−8, 4), and
2x3 + ax2 + bx + 8 = 2(x2 − 2x − 2)(x − 2)

and the third root is 2.

53. Prove that among any 17 natural numbers chosen from the sets {1, 2, 3, · · · , 24, 25}, it is always possible
to find two whose product is a perfect square.

Solution. [K. Ho] Consider the following 16 sets: {1, 4, 9, 16, 25}, {2, 8, 18}, {3, 12}, {5, 20}, {6, 24}, {7},
{10}, {11}, {13}, {14}, {15}, {17}, {19}, {21}, {22}, {23}. In any of the sets with more than one element,
the product of any two elements is a perfect square. Any choice of 17 numbers from among these sets must,
by the Pigeonhole Principle, yield at least two from the same set. The product of these two must be a square.
The desired result follows.

54. A circle has exactly one common point with each of the sides of a (2n + 1)−sided polygon. None of the
vertices of the polygon is a point of the circle. Prove that at least one of the sides is a tangent of the
circle.

44
 Solution. [J.Y. Jin] Assume that none of the sides is tangent to the circle. Let A1 , A2 , · · · , A2n−1 be the
consecutive vertices of a (2n + 1)−sided polygon. Each side of the polygon has exactly one common point
with the circle, none of the vertices lies on the circle and, by assumption, none of the sides is tangent to the
circle. Therefore, for each side, one of the endpoints lies inside the circle and the second endpoint lies outside
the circle. Colour the vertices inside the circle blue and the ones outside the circle red. Apparently, the
colours of the consecutive vertices alternate (i.e., blue, red, blue, red, etc.). Since there are 2n + 1 vertices,
A1 and A2n+1 must have the same colour. However, these two vertices are the endpoints of a side of the
polygon and so they cannot have the same colour. Thus, the assumption leads to a contradiction, and so at
least one of the sides is tangent to the circle.

GEOMETRY ON THE 2000 USAMO

There were two geometry problems on the 2000 USAMO, both of which were reasonably well handled
by the Canadian candidates. However, each of them illustrate the importance of not jumping to conclusions
and ending up with a score that does not reflect your understanding of the situation.

2. Let S be the set of all triangles ABC for which

 
1 1 1 3 6
5 + + − = ,
AP BQ CR min {AP, BQ, CR} r

where r is the inradius and P, Q, R are the points of tangency of the incircle with sides AB, BC, CA,
respectively. Prove that all triangles in S are isosceles and similar to one another.

Comments. Since the statement either holds or does not hold for both of a pair of similar triangles (i.e.,
it does not depend on scale), we can eliminate a possible variable by assuming that r = 1. If the angles of
the triangle at A, B and C are respectively, 2α, 2β, 2γ, we find that the lengths of AP , BQ and CR are,
respectively, cot α, cot β and cot γ. Assuming that cot α is the smallest of these, we find that the condition
is
0 = 2 tan α + 5(tan β + tan γ) − 6 .

In our quest to reduce the number of variables we have to deal with, we note that α + β + γ = 90◦ , so
we try to combine the expression involving β and γ into something depending on their sum. We need the
observation that tan x is convex for 0 ≤ x < 90◦ , so that tan((α + β)/2) ≤ (1/2)(tan α + tan β).

So let us dispose of this first. From the tangent of a difference, we have the identity

tan ρ − tan σ = tan(ρ − σ)[1 + tan ρ tan σ] .

Applying this, when 0 ≤ α, β < 90◦ yields

      
α+β α+β
tan α − tan − tan − tan β
2 2
   
α−β α+β α+β
= tan 1 + tan α tan − 1 + tan β tan
2 2 2
α−β α+β
= tan · tan · (tan α − tan β) ≥ 0 ,
2 2

whence
α+β
2 tan ≤ tan α + tan β .
2

45
 Using the convexity of tan x and making the substitution t = tan(α/2), we find that
 
β+γ
0 ≥ 2 tan α + 10 tan −6
2
 
α
= 2 tan α + 10 tan 45◦ − −6
2
4t 10(1 − t)
= + −6
1 − t2 1+t
4(2t − 1)2
= .
1 − t2
Since α/2 < 45◦ , t < 1 so that we must have equality, β = γ and t = 1/2. This yields tan 21 α = 1/2 and so
tan α = 4/3.

The triangle is isosceles, and the ratio of the semi-base to height is tan α : 1 = 4 : 3, so that ratio of the
sides of the triangle must be 8 : 5 : 5.

Several students avoided the appeal to convexity. Noting that tan α = cot(β + γ) and multiplying the
equation by tan β + tan γ, we get

0 = 2(1 − tan β tan γ) + 5(tan β + tan γ)2 − 6(tan β + tan γ)

= 5 tan2 β + 5 tan2 γ + 8 tan β tan γ − 6 tan β − 6 tan γ + 2 .

Considering this as a quadratic equation in tan2 β, we find that its discriminant is

(8 tan γ − 6)2 − 20(5 tan2 γ − 6 tan γ + 2) = −4(3 tan γ − 1)2 .

We are presuming that there is a suitable triangle, so that the quadratic for tan β has a real solution; this
happens if and only if tan γ = 1/3. By a parallel argument, we must also have tan β = 1/3. This gives the
same triangle as before.

In hindsight, we can verify that the condition holds with these values of tan β and tan γ by rewriting it
as
0 = 5(tan β − tan γ)2 + 2(3 tan β − 1)(3 tan γ − 1) .

Several students identified the circumradius in terms of the lengths x, y and z of the respective segments
AP , BQ, CR. This can be done by comnparing two expressions for the area of the triangle, the first involving
the product of the inradius and the semiperimeter x + y + z, and the second using Heron’s formula:
p
r(x + y + z) = (x + y + z)xyz .

Using this and assuming that x is the minimum of x, y, z, transforms the given condition to
r
2 5 5 x+y+z
0= + + =6 .
x y z xyz
We can clear the denominators of this equation and get a fairly complicated expression in x, y, z to analyze.
But Yin Ren of Windsor decided to deal with the reciprocals of these variables: u = 1/x, v = 1/y and
w = 1/z. This makes the condition
√
2u + 5v + 5w = 6 uv + vw + wu .

Squaring and collectin terms gives

0 = 4u2 + 25v 2 + 25w2 − 16uv − 16uw + 14vw = 4u2 + 25(v 2 + w2 ) − 16u(v + w) + 14vw ,

46
 Yin the made the assumption that v = w and found the isosceles triangle that delivers the condition.
But he still had the non-isosceles possibility to consider, and left his proof incomplete. Actually, he made
more progress than he thought, and could follow up on his inspiration by looking at a perturbation from the
isoceles case. Let v = r + e and w = r − e where r is the average of v and w and e is the perturbation from
the average. Then v 2 + w2 = 2(r2 + e2 ), v + w = 2r and vw = r2 − e2 (notice how products of re drop out).
Then the condition becomes
0 = 4(u − 4r)2 + 36e2 ,
which evidently holds if and only if u = 4r and e = 0. Thus, y = 4x and the ratio of the sides of the triangle
must be x + y : x + z : y + z = 5 : 5 : 8.

To nail the thing down completely, one should make the observation that the condition does in fact
hold when the sides are in this ratio. Logically, without this, all one has demontrated is that presence of the
condition means that the sides cannot have any other ratio.

5. Let A1 A2 A3 be a triangle and let ω1 be a circle in its plane passing through A1 and A2 . Suppose
there exist circles ω2 , ω3 , · · ·, ω7 such that, for k = 2, 3, · · · , 7, ωk is externally tangent to ωk−1 and passes
through Ak and Ak+1 , where An+3 = An for all n ≥ 1. Prove that ω7 = ω1 .

Comments. The presence of the circles¡F4¿ makes the problem conceptually quite complex, and the
students who succeeded (with two notable exceptions) got the circles out of the picture early on by making
the key observation that the line joining the centres of two tangent circles is perpendicular to their common
tangent line. This enabled a succession of angles to be related and so the problem reduced to a rather simple
angle-chasing exercise.

Note that ωk is externally tangent to ωk−1 at the point Ak . Let the centre of circle ωk be Ok and let
βk = ∠Ok Ak Ak+1 = ∠O + kAk+1 Ak , αk = ∠Ak−1 Ak Ak+1 . Then β2 = 180◦ − β1 − α2 , β3 = β1 + α2 − α3 ,
β5 = β3 + α1 − α2 = β1 + α1 − α3 and β7 = β5 + α3 − α1 . Thus ∠O7 A2 A1 = ∠O1 A2 A1 and it follows that
ω1 and ω7 have the same centre and radius.

Some students were not careful with the argument and concluded that ω1 and ω4 coincided. You may
wish to verify that this is not necessarily so. Can it ever occur?

The alternative approach, adopted by David Arthur and David Pritchard, was to use inversion. Invert
the whole configuration in a circle with centre A1 and use primes to indicate the images of entities under
inversion. Note that A1 gets transported “to infinity”, and each circle through A1 to lines passing through
the intersection points (if any) of the circle and the circle of inversion. Inversion preserves tangency.

ω10 is a straight line through A02 ; ω20 is a circle through A02 and A03 that is tangent to ω10 at A02 : ω30 is a
straight line tangent to ω20 at A03 ; ω40 is a straight line through A02 parallel to ω30 ; ω50 is a circle through A02
and A03 tangent to ω40 at A02 ; ω60 is a line through A03 tangent to ω50 ; ω70 is a line through A02 parallel to ω60 .
So it remains to see that ω50 and ω10 are tangent. One way to do this is to observe that the half-turn (180◦
rotation) about the midpoint of A02 A03 interchanges A02 and A03 and interchanges ω30 and ω40 . Thus, ω20 , a
circle through A02 and A03 tangent to ω30 at A02 , goes to a circle through the same points, but tangent to ω40
at A04 . Thus, ω20 and ω50 get switched, as do ω10 and ω60 . Hence, ω10 , ω60 and ω7 are parallel, and so ω10 and ω70
must coincide. The desired result follows.

47