IMO 2003 Solution Notes

Evan Chen《陳誼廷》
15 December 2024

This is a compilation of solutions for the 2003 IMO. The ideas of the
solution are a mix of my own work, the solutions provided by the competition
organizers, and solutions found by the community. However, all the writing
is maintained by me.
These notes will tend to be a bit more advanced and terse than the “official”
solutions from the organizers. In particular, if a theorem or technique is not
known to beginners but is still considered “standard”, then I often prefer to
use this theory anyways, rather than try to work around or conceal it. For
example, in geometry problems I typically use directed angles without further
comment, rather than awkwardly work around configuration issues. Similarly,
sentences like “let R denote the set of real numbers” are typically omitted
entirely.
Corrections and comments are welcome!

Contents
0 Problems 2

1 Solutions to Day 1 3
1.1 IMO 2003/1, proposed by Carlos Gustavo Moreira (BRA) . . . . . . . . . 3
1.2 IMO 2003/2, proposed by Aleksander Ivanov (BGR) . . . . . . . . . . . . 4
1.3 IMO 2003/3, proposed by Waldemar Pompe (POL) . . . . . . . . . . . . . 5

2 Solutions to Day 2 7
2.1 IMO 2003/4, proposed by Matti Lehtinen (FIN) . . . . . . . . . . . . . . 7
2.2 IMO 2003/5, proposed by Finbarr Holland (IRL) . . . . . . . . . . . . . . 8
2.3 IMO 2003/6, proposed by Johan Yebbou (FRA) . . . . . . . . . . . . . . 10

1
IMO 2003 Solution Notes web.evanchen.cc, updated 15 December 2024

§0 Problems
1. Let A be a 101-element subset of S = {1, 2, . . . , 106 }. Prove that there exist
numbers t1 , t2 , . . . , t100 in S such that the sets

Aj = {x + tj | x ∈ A}, j = 1, 2, . . . , 100

are pairwise disjoint.

2. Determine all pairs of positive integers (a, b) such that

a2
2ab2 − b3 + 1
is a positive integer.

3. Each pair of opposite sides of√ convex hexagon has the property that the distance
between their midpoints is 23 times the sum of their lengths. Prove that the
hexagon is equiangular.

4. Let ABCD be a cyclic quadrilateral. Let P, Q and R be the feet of perpendiculars

from D to lines BC, CA and AB, respectively. Show that P Q = QR if and only if
the bisectors of angles ABC and ADC meet on segment AC.

5. Let n be a positive integer and let x1 ≤ x2 ≤ · · · ≤ xn be real numbers. Prove that

 2
n X
n n n
X 2(n2 − 1) X X
 |xi − xj | ≤ (xi − xj )2
3
i=1 j=1 i=1 j=1

with equality if and only if x1 , x2 , . . . , xn form an arithmetic sequence.

6. Let p be a prime number. Prove that there exists a prime number q such that for
every integer n, the number np − p is not divisible by q.

2
IMO 2003 Solution Notes web.evanchen.cc, updated 15 December 2024

§1 Solutions to Day 1
§1.1 IMO 2003/1, proposed by Carlos Gustavo Moreira (BRA)
Available online at https://aops.com/community/p261.

Problem statement

Let A be a 101-element subset of S = {1, 2, . . . , 106 }. Prove that there exist numbers
t1 , t2 , . . . , t100 in S such that the sets

Aj = {x + tj | x ∈ A}, j = 1, 2, . . . , 100

are pairwise disjoint.

A greedy algorithm works: suppose we have picked

T = {t1 , . . . , tn }

as large as possible,
 meaning it’s impossible to add any more elements to T . That
means, for each t ∈ 1, . . . , 106 either t ∈ T already or there exists two distinct elements
a, b ∈ A and ti ∈ T such that

t = ti + b − a (?).

There are at most |T | · |A| · (|A| − 1) = n · 101 · 100 possible values for the right-hand
side of (?). So we therefore must have

101 · 100 · n + n ≥ 106

which implies n > 99, as desired.

Remark. It is possible to improve the bound significantly with a small optimization; rather
than adding any t, we require that t1 < · · · < tn and that at each step we add the least
t ∈ S which is permitted. In that case, one finds we only need to consider b > a in
(?), and
so this will save us a factor of 2 + o(1) as the main term 101 · 100 becomes 1012 instead.
This proves it’s possible to choose 198 elements.
See, e.g., https://aops.com/community/p22959828 for such a write-up.

3
IMO 2003 Solution Notes web.evanchen.cc, updated 15 December 2024

§1.2 IMO 2003/2, proposed by Aleksander Ivanov (BGR)

Available online at https://aops.com/community/p262.

Problem statement

Determine all pairs of positive integers (a, b) such that

a2
2ab2 − b3 + 1
is a positive integer.

The answer is (a, b) = (2`, 1), (a, b) = (`, 2`) and (a, b) = (8`4 − `, 2`), for any `. Check
these work.
In the sequel, assume b > 1, and integers a, b, k obey k = 2ab2 −b a2
3 +1 . Expanding, we
have the polynomial
X 2 − 2kb2 · X + k(b3 − 1) = 0
has two integer roots, one of which is X = a. This means solutions to the original
problem come in pairs (even with k fixed):
 3 
2

k(b − 1)
(a, b) ←→ 2kb − a, b = ,b .
a

(Here, the first representation ensures 2kb2 − a ∈ Z, while the latter representation and
3
the hypothesis b > 1 ensures that k(b a−1) > 0.)
On the other hand, we claim that:

Claim — For any solution (a, b), either 2a = b or a > b.

Proof. Since the denominator is positive, a ≥ b/2. Now,

a2 ≥ 2ab2 − b3 + 1 ⇐⇒ a2 ≥ b2 (2a − b) + 1

and so if 2a − b > 0 then a2 > b2 =⇒ a > b.

Now assume we have pair (a1 , b) and (a2 , b) of solutions with b 6= 2a1 , 2a2 . Then
assume a1 > a2 > b and

a1 + a2 = 2k · b2
a1 a2 = k(b3 − 1)

That’s impossible, since then a1 > a1 +a

2
2
= kb2 and hence a1 a2 > kb2 · b = kb3 . Thus the
only solutions are the ones we claimed at the beginning.

Remark. Important to notice that the problem is positive divides, not just divides. There
is an implicit inequality built in to the problem statement and it is essentially impossible to
solve without. I would be interested in a pair (a, b) for which k < 0, k ∈ Z yet a, b > 0.

4
IMO 2003 Solution Notes web.evanchen.cc, updated 15 December 2024

§1.3 IMO 2003/3, proposed by Waldemar Pompe (POL)

Available online at https://aops.com/community/p263.

Problem statement

Each pair of opposite sides√ of convex hexagon has the property that the distance
between their midpoints is 23 times the sum of their lengths. Prove that the hexagon
is equiangular.

Unsurprisingly, this is a geometric inequality. Denote the hexagon by ABCDEF . Then

we have that

~ +E~ ~+B ~ √ ~ −A
B ~ + E
~ −D
~ √ ~ − A)
~ − (E
~ − D)
~
D A (B
− = 3· ≥ 3·
2 2 2 2

and cyclic variations. Suppose we define the right-hand sides as variables

~x = (B~ − A)
~ − (E
~ − D)
~
~ − C)
~y = (D ~ − (A
~ − F~ )
~z = (F~ − E)
~ − (C
~ − B).
~

Then we now have

√
|~y − ~z| ≥ 3 |~x|
√
|~z − ~x| ≥ 3 |~y |
√
|~x − ~y | ≥ 3 |~z| .

We square all sides (using |~v |2 = ~v · ~v ) and then sum to get

X X
(~y − ~z) · (~y − ~z) ≥ 3 ~x · ~x
cyc cyc

which rearranges to
− |~x + ~y + ~z|2 ≥ 0.
This can only happen if ~x + ~y + ~z = 0, and moreover all the inequalities above were
actually equalities. That means that our triangle inequalities above were actually sharp
(and already we have AB k DE and so on).
Working with just x and y now we have

3(~x · ~x) = (2~y − ~x) · (2~y − ~x)

= ~x · ~x − 4~y · ~x + 4~y · ~y
=⇒ −~x · ~x + 2(~y · ~y ) = 2~x · ~y
2(~x · ~x) − ~y · ~y = 2~x · ~y .

which implies ~x · ~x = ~y · ~y , that is, ~x and ~y have the same magnitude. In this way we
find ~x, ~y , ~z all have the same magnitude, and since ~x + ~y + ~z = 0 they are related by
120◦ rotations, as desired.

5
IMO 2003 Solution Notes web.evanchen.cc, updated 15 December 2024

Remark. In fact one can show further that the equiangular hexagons which work are exactly
those formed by taking an equilateral triangle and cutting off equally sized corners. This
equality case helps motivate the solution.

Remark. One can note this “must” be an inequality because the space of such hexagons is
2-dimensional, even though a priori the space of hexagons satisfying three given conditions
should have dimension 9 − 3 = 6.

6
IMO 2003 Solution Notes web.evanchen.cc, updated 15 December 2024

§2 Solutions to Day 2
§2.1 IMO 2003/4, proposed by Matti Lehtinen (FIN)
Available online at https://aops.com/community/p264.

Problem statement

Let ABCD be a cyclic quadrilateral. Let P, Q and R be the feet of perpendiculars

from D to lines BC, CA and AB, respectively. Show that P Q = QR if and only if
the bisectors of angles ABC and ADC meet on segment AC.

Let γ denote the circumcircle of ABCD. The condition on bisectors is equivalent to

(AC; BD)γ = −1. Meanwhile if ∞ denotes the point at infinity along Simson line P QR
then P Q = QR if and only if (P R; Q∞) = −1.
Let rays BQ and DQ meet the circumcircle again at F and E.

A Q C

R
D
F

Lemma (EGMO Proposition 4.1)

Then BE k P QR.

Proof. Since ]DP R = ]DAR = ]DAB = ]DEB.

Now we have
B Q
(P R; Q∞) = (CA; F E)γ = (AC; BD)γ
as desired.

7
IMO 2003 Solution Notes web.evanchen.cc, updated 15 December 2024

§2.2 IMO 2003/5, proposed by Finbarr Holland (IRL)

Available online at https://aops.com/community/p265.

Problem statement

Let n be a positive integer and let x1 ≤ x2 ≤ · · · ≤ xn be real numbers. Prove that

 2
n X
n n n
X 2(n2 − 1) X X
 |xi − xj | ≤ (xi − xj )2
3
i=1 j=1 i=1 j=1

with equality if and only if x1 , x2 , . . . , xn form an arithmetic sequence.

Let d1 = x2 − x1 , . . . , dn−1 = xn − xn−1 . The inequality in question becomes:

!2  
X 2
n −1  X X
i(n − i)di ≤ · i(n − i)d2i + 2 i(n − j)di dj  .
3
i i i<j

Clearing the square on the right-hand side we want to show

X X
3ij(n − i)(n − j) − (n2 − 1)i(n − j) · 2di dj ≤ (n2 − 1 − 3i(n − i)) · i(n − i)d2i .

i<j i

We use AM-GM directly on 2di dj ≤ d2i + d2j : this actually solves the problem. The
annoying part is to check that the coefficients actually match:

Claim (Big bash) — For an index 1 ≤ k ≤ n − 1, we have

X
3ik(n − i)(n − k) − (n2 − 1)i(n − k)

i<k
X
3kj(n − k)(n − j) − (n2 − 1)k(n − j)


+
j>k
2
=(n − 1 − 3k(n − k)) · k(n − k).

Proof. Rewrite as:

!  
X X X
3k(n − k) −k(n − k) + i(n − i) = (n2 − 1) (n − k) i+k (n − j)
i i<k j>k
2
+ (n − 1 − 3k(n − k)) · k(n − k)
 
X X X
⇐⇒ 3k(n − k) i(n − i) = (n2 − 1) (n − k) i+k (n − j)
i i<k j>k

+ (n − 1)k(n − k) − 3k (n − k)2
2 2

!  
X X X
⇐⇒ 3k(n − k) i(n − i) = (n2 − 1) (n − k) i+k i
i i≤k i<n−k
 
(n − 1)n(n + 1) k(k + 1)
⇐⇒ 3k(n − k) = (n2 − 1) (n − k)
6 2

8
IMO 2003 Solution Notes web.evanchen.cc, updated 15 December 2024
 
2 (n − k)(n − k − 1)
+ (n − 1) k
2
(n − 1)n(n + 1) n
⇐⇒ 3k(n − k) = (n2 − 1)k(n − k) ·
6 2
which is visibly true.

Equality occurs only if all di are equal because the coefficient of di dj is nonzero for any
i ≤ n/2 and j ≥ n/2.

9
IMO 2003 Solution Notes web.evanchen.cc, updated 15 December 2024

§2.3 IMO 2003/6, proposed by Johan Yebbou (FRA)

Available online at https://aops.com/community/p266.

Problem statement

Let p be a prime number. Prove that there exists a prime number q such that for
every integer n, the number np − p is not divisible by q.

By orders, we must have q = pk + 1 for this to be possible (since if q 6≡ 1 (mod p), then
np can be any residue modulo q). Since p ≡ np (mod q) =⇒ pk ≡ 1 (mod q), it suffices
to prevent the latter situation from happening.
So we need a prime q ≡ 1 (mod p) such that pk 6≡ 1 (mod q). To do this, we first
recall the following lemma.

Lemma
Let Φp (X) = 1 + X + X 2 + · · · + X p−1 . For any integer a, if q is a prime divisor of
Φp (a) other than p, then a (mod q) has order p. (In particular, q ≡ 1 (mod p).)

Proof. We have ap − 1 ≡ 0 (mod q), so either the order is 1 or p. If it is 1, then a ≡ 1

(mod q), so q | Φp (1) = p, hence q = p.

Now the idea is to extract a prime factor q from the cyclotomic polynomial
pp − 1
Φp (p) = ≡1+p (mod p2 )
p−1

such that q 6≡ 1 (mod p2 ); hence k 6≡ 0 (mod p), and as p (mod q) has order p we have
pk 6≡ 1 (mod q).

10