IMO 2007 Solution Notes

Evan Chen《陳誼廷》
15 December 2024

This is a compilation of solutions for the 2007 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 2007/1, proposed by Michael Albert (NZL) . . . . . . . . . . . . . . 3
1.2 IMO 2007/2, proposed by Charles Leytem (LUX) . . . . . . . . . . . . . . 4
1.3 IMO 2007/3, proposed by Vasily Astakhov (RUS) . . . . . . . . . . . . . . 5

2 Solutions to Day 2 7
2.1 IMO 2007/4, proposed by Marek Pechal (CZE) . . . . . . . . . . . . . . . 7
2.2 IMO 2007/5, proposed by Kevin Buzzard, Edward Crane (UNK) . . . . . 8
2.3 IMO 2007/6, proposed by Gerhard Woeginger (NLD) . . . . . . . . . . . 9

1
IMO 2007 Solution Notes [Link], updated 15 December 2024

§0 Problems
1. Real numbers a1 , a2 , . . . , an are fixed. For each 1 ≤ i ≤ n we let di = max{aj :
1 ≤ j ≤ i} − min{aj : i ≤ j ≤ n} and let d = max{di : 1 ≤ i ≤ n}.
(a) Prove that for any real numbers x1 ≤ · · · ≤ xn we have
1
max {|xi − ai | : 1 ≤ i ≤ n} ≥ d.
2

(b) Moreover, show that there exists some choice of x1 ≤ · · · ≤ xn which achieves
equality.

2. Consider five points A, B, C, D and E such that ABCD is a parallelogram and

BCED is a cyclic quadrilateral. Let ` be a line passing through A. Suppose that
` intersects the interior of the segment DC at F and intersects line BC at G.
Suppose also that EF = EG = EC. Prove that ` is the bisector of angle DAB.

3. In a mathematical competition some competitors are (mutual) friends. Call a group

of competitors a clique if each two of them are friends. Given that the largest size
of a clique is even, prove that the competitors can be arranged into two rooms such
that the largest size of a clique contained in one room is the same as the largest
size of a clique contained in the other room.

4. In triangle ABC the bisector of ∠BCA meets the circumcircle again at R, the
perpendicular bisector of BC at P , and the perpendicular bisector of AC at Q.
The midpoint of BC is K and the midpoint of AC is L. Prove that the triangles
RP K and RQL have the same area.

5. Let a and b be positive integers. Show that if 4ab − 1 divides (4a2 − 1)2 , then a = b.

6. Let n be a positive integer. Consider

S = {(x, y, z) | x, y, z ∈ {0, 1, . . . , n}, x + y + z > 0}

as a set of (n+1)3 −1 points in the three-dimensional space. Determine the smallest

possible number of planes, the union of which contains S but does not include
(0, 0, 0).

2
IMO 2007 Solution Notes [Link], updated 15 December 2024

§1 Solutions to Day 1
§1.1 IMO 2007/1, proposed by Michael Albert (NZL)
Available online at [Link]

Problem statement

Real numbers a1 , a2 , . . . , an are fixed. For each 1 ≤ i ≤ n we let di = max{aj : 1 ≤

j ≤ i} − min{aj : i ≤ j ≤ n} and let d = max{di : 1 ≤ i ≤ n}.

(a) Prove that for any real numbers x1 ≤ · · · ≤ xn we have

1
max {|xi − ai | : 1 ≤ i ≤ n} ≥ d.
2

(b) Moreover, show that there exists some choice of x1 ≤ · · · ≤ xn which achieves
equality.

Note that we can dispense of di immediately by realizing that the definition of d just says

d = max (ai − aj ) .
1≤i≤j≤n

If a1 ≤ · · · ≤ an are already nondecreasing then d = 0 and there is nothing to prove

(for the equality case, just let xi = ai ), so we will no longer consider this case.
Otherwise, consider any indices i < j with ai > aj . We first prove (a) by applying the
following claim with p = ai and q = aj :

Claim — For any p ≤ q, we have either |p − ai | ≥ 12 (ai − aj ) or |q − aj | ≥ 12 (ai − aj ).

Proof. Assume for contradiction both are false. Then p > ai − 12 (ai −aj ) = aj + 12 (ai −aj ) >
q, contradiction.

As for (b), we let i < j be any indices for which ai − aj = d > 0 achieves the maximal
difference. We then define x• in three steps:

• We set xk = for k = i, . . . , j.
ai +aj
2

• We recursively set xk = max(xk−1 , ak ) for k = j + 1, j + 2, . . . .

• We recursively set xk = min(xk+1 , ak ) for k = i − 1, i − 2, . . . .

By definition, these x• are weakly increasing. To prove this satisfies (b) we only need to
check that
ai − aj
|xk − ak | ≤ (?)
2
for any index k (as equality holds for k = i or k = j).
We note (?) holds for i < k < j by construction. For k > j, note that xk ∈
{aj , aj+1 , . . . , ak } by construction, so (?) follows from our choice of i and j giving the
largest possible difference; the case k < i is similar.

3
IMO 2007 Solution Notes [Link], updated 15 December 2024

§1.2 IMO 2007/2, proposed by Charles Leytem (LUX)

Available online at [Link]

Problem statement

Consider five points A, B, C, D and E such that ABCD is a parallelogram and

BCED is a cyclic quadrilateral. Let ` be a line passing through A. Suppose that `
intersects the interior of the segment DC at F and intersects line BC at G. Suppose
also that EF = EG = EC. Prove that ` is the bisector of angle DAB.

Let M , N , P denote the midpoints of CF , CG, AC (noting P is also the midpoint of

BD).
By a homothety at C with ratio 12 , we find M N P is the image of line ` ≡ AGF .

B A

M
C D
F

N
E

However, since we also have EM ⊥ CF and EN ⊥ CG (from EF = EG = EC) we

conclude P M N is the Simson line of E with respect to 4BCD, which implies EP ⊥ BD.
In other words, EP is the perpendicular bisector of BD, so E is the midpoint of arc
BCD.
’
Finally,

](AB, `) = ](CD, M N P ) = ]CM N = ]CEN

= 90◦ − ]N CE = 90◦ + ]ECB

which means that ` is parallel to a bisector of ∠BCD, and hence to one of ∠BAD.
(Moreover since F lies on the interior of CD, it is actually the internal bisector.)

4
IMO 2007 Solution Notes [Link], updated 15 December 2024

§1.3 IMO 2007/3, proposed by Vasily Astakhov (RUS)

Available online at [Link]

Problem statement

In a mathematical competition some competitors are (mutual) friends. Call a group

of competitors a clique if each two of them are friends. Given that the largest size of
a clique is even, prove that the competitors can be arranged into two rooms such
that the largest size of a clique contained in one room is the same as the largest size
of a clique contained in the other room.

Take the obvious graph interpretation G. We paint red any vertices in one of the maximal
cliques K, which we assume has 2r vertices, and paint the remaining vertices green. We
let α(•) denote the clique number.
Initially, let the two rooms A = K, B = G − K.

Claim — We can move at most r vertices of A into B to arrive at α(A) ≤ α(B) ≤

α(A) + 1.

Proof. This is actually obvious by discrete continuity. We move one vertex at a time,
noting α(A) decreases by one at each step, while α(B) increases by either zero or one at
each step.
We stop once α(B) ≥ α(A), which happens before we have moved r vertices (since
then we have α(B) ≥ r = α(A)). The conclusion follows.

So let’s consider the situation

α(A) = k ≥ r and α(B) = k + 1.

At this point A is a set of k red vertices, while B has the remaining 2r − k red vertices
(and all the green ones). An example is shown below with k = 4 and 2r = 6.

A B

α(A) = k α(B) = k + 1
k red vertices 2r − k red vertices

Now, if we can move any red vertex from B back to A without changing the clique
number of B, we do so, and win.
Otherwise, it must be the case that every (k + 1)-clique in B uses every red vertex in
B. For each (k + 1)-clique in B (in arbitrary order), we do the following procedure.

5
IMO 2007 Solution Notes [Link], updated 15 December 2024

• If all k + 1 vertices are still green, pick one and re-color it blue. This is possible
since k + 1 > 2r − k.

• Otherwise, do nothing.

Then we move all the blue vertices from B to A, one at a time, in the same order we
re-colored them. This forcibly decreases the clique number of B to k, since the clique
number is k + 1 just before the last blue vertex is moved, and strictly less than k + 1
(hence equal to k) immediately after that.

Claim — After this, α(A) = k still holds.

Proof. Assume not, and we have a (k + 1)-clique which uses b blue vertices and (k + 1) − b
red vertices in A. Together with the 2r − k red vertices already in B we then get a clique
of size
b + ((k + 1 − b)) + (2r − k) = 2r + 1
which is a contradiction.

Remark. Dragomir Grozev posted the following motivation on his blog:

I think, it’s a natural idea to place all students in one room and begin moving
them one by one into the other one. Then the max size of the cliques in the
first and second room increase (resp. decrease) at most with one. So, there
would be a moment both sizes are almost the same. At that moment we may
adjust something.
Trying the idea, I had some difficulties keeping track of the maximal cliques
in the both rooms. It seemed easier all the students in one of the rooms to
comprise a clique. It could be achieved by moving only the members of the
maximal clique. Following this path the remaining obstacles can be overcome
naturally.

6
IMO 2007 Solution Notes [Link], updated 15 December 2024

§2 Solutions to Day 2
§2.1 IMO 2007/4, proposed by Marek Pechal (CZE)
Available online at [Link]

Problem statement

In triangle ABC the bisector of ∠BCA meets the circumcircle again at R, the
perpendicular bisector of BC at P , and the perpendicular bisector of AC at Q. The
midpoint of BC is K and the midpoint of AC is L. Prove that the triangles RP K
and RQL have the same area.

We first begin by proving the following claim.

Claim — We have CQ = P R (equivalently, CP = QR).

Proof. Let O = LQ ∩ KP be the circumcenter. Then

]OP Q = ]KP C = 90◦ − ]P CK = 90◦ − ]LCQ = ]]CQL = ]P QO.

Thus OP = OQ. Since OC = OR as well, we get the conclusion.

Denote by X and Y the feet from R to CA and CB, so 4CXR ∼

= 4CY R. Then, let
t = CR = 1 − CR .
CQ CP

Q
K

O
A P
X Y

Then it follows that

[RQL] = [XQL] = t(1 − t) · [XRC] = t(1 − t) · [Y CR] = [Y KP ] = [RKP ]

as needed.
Remark. Trigonometric approaches are very possible (and easier to find) as well: both
areas work out to be 18 ab tan 12 C.

7
IMO 2007 Solution Notes [Link], updated 15 December 2024

§2.2 IMO 2007/5, proposed by Kevin Buzzard, Edward Crane (UNK)

Available online at [Link]

Problem statement

Let a and b be positive integers. Show that if 4ab − 1 divides (4a2 − 1)2 , then a = b.

As usual,

4ab − 1 | (4a2 − 1)2 ⇐⇒ 4ab − 1 | (4ab · a − b)2 ⇐⇒ 4ab − 1 | (a − b)2 .

Then we use a typical Vieta jumping argument. Define

(a − b)2
k= .
4ab − 1
Note that k = 0 ⇐⇒ a = b. So we will prove that k > 0 leads to a contradiction.
Indeed, suppose (a, b) is a minimal solution with a > b (we have a 6= b since k 6= 0).
By Vieta jumping, (b, b a+k ) is also such a solution. But now
2

b2 + k
≥ a =⇒ k ≥ a2 − b2
a
(a − b)2
=⇒ ≥ a2 − b2
4ab − 1
=⇒ a − b ≥ (4ab − 1)(a + b)

which is absurd for a, b ∈ Z>0 . (In the last step we divided by a − b > 0.)

8
IMO 2007 Solution Notes [Link], updated 15 December 2024

§2.3 IMO 2007/6, proposed by Gerhard Woeginger (NLD)

Available online at [Link]

Problem statement

Let n be a positive integer. Consider

S = {(x, y, z) | x, y, z ∈ {0, 1, . . . , n}, x + y + z > 0}

as a set of (n + 1)3 − 1 points in the three-dimensional space. Determine the smallest

possible number of planes, the union of which contains S but does not include
(0, 0, 0).

The answer is 3n. Here are two examples of constructions with 3n planes:

• x + y + z = i for i = 1, . . . , 3n.

• x = i, y = i, z = i for i = 1, . . . , n.

Suppose for contradiction we have N < 3n planes. Let them be ai x + bi y + ci z + 1 = 0,

for i = 1, . . . , N . Define the polynomials
n
Y n
Y n
Y
A(x, y, z) = (x − i) (y − i) (z − i)
i=1 i=1 i=1
YN
B(x, y, z) = (ai x + bi y + ci z + 1) .
i=1

Note that A(0, 0, 0) = (−1)n (n!)3 6= 0 and B(0, 0, 0) = 1 6= 0, but A(x, y, z) =

B(x, y, z) = 0 for any (x, y, z) ∈ S. Also, the coefficient of xn y n z n in A is 1, while
the coefficient of xn y n z n in B is 0.
Now, define
P (x, y, z) := A(x, y, z) − λB(x, y, z).
where λ = A(0,0,0)
B(0,0,0) = (−1)n (n!)3 . We now have that

• P (x, y, z) = 0 for any x, y, z ∈ {0, 1, . . . , n}3 .

• But the coefficient of xn y n z n is 1.

This is a contradiction to Alon’s combinatorial nullstellensatz.