USAMO 2008 Solution Notes

Evan Chen《陳誼廷》
15 December 2024

This is a compilation of solutions for the 2008 USAMO. 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 USAMO 2008/1, proposed by Titu Andreescu . . . . . . . . . . . . . . . . 3
1.2 USAMO 2008/2, proposed by Zuming Feng . . . . . . . . . . . . . . . . . 4
1.3 USAMO 2008/3, proposed by Gabriel Carroll . . . . . . . . . . . . . . . . 6

2 Solutions to Day 2 9
2.1 USAMO 2008/4, proposed by Gregory Galperin . . . . . . . . . . . . . . . 9
2.2 USAMO 2008/5, proposed by Kiran Kedlaya . . . . . . . . . . . . . . . . 10
2.3 USAMO 2008/6, proposed by Sam Vandervelde . . . . . . . . . . . . . . . 11

1
USAMO 2008 Solution Notes web.evanchen.cc, updated 15 December 2024

§0 Problems
1. Prove that for each positive integer n, there are pairwise relatively prime integers
k0 , . . . , kn , all strictly greater than 1, such that k0 k1 . . . kn − 1 is the product of
two consecutive integers.

2. Let ABC be an acute, scalene triangle, and let M , N , and P be the midpoints
of BC, CA, and AB, respectively. Let the perpendicular bisectors of AB and
AC intersect ray AM in points D and E respectively, and let lines BD and CE
intersect in point F , inside triangle ABC. Prove that points A, N , F , and P all
lie on one circle.

3. Let n be a positive integer. Denote by Sn the set of points (x, y) with integer
coordinates such that
1
|x| + y + < n.
2
A path is a sequence of distinct points (x1 , y1 ), (x2 , y2 ), . . . , (x` , y` ) in Sn such that,
for i = 2, . . . , `, the distance between (xi , yi ) and (xi−1 , yi−1 ) is 1.
Prove that the points in Sn cannot be partitioned into fewer than n paths.

4. For which integers n ≥ 3 can one find a triangulation of regular n-gon consisting
only of isosceles triangles?

5. Three nonnegative real numbers r1 , r2 , r3 are written on a blackboard. These

numbers have the property that there exist integers a1 , a2 , a3 , not all zero, satisfying
a1 r1 + a2 r2 + a3 r3 = 0. We are permitted to perform the following operation: find
two numbers x, y on the blackboard with x ≤ y, then erase y and write y − x in its
place. Prove that after a finite number of such operations, we can end up with at
least one 0 on the blackboard.

6. At a certain mathematical conference, every pair of mathematicians are either

friends or strangers. At mealtime, every participant eats in one of two large
dining rooms. Each mathematician insists upon eating in a room which contains
an even number of his or her friends. Prove that the number of ways that the
mathematicians may be split between the two rooms is a power of two (i.e. is of
the form 2k for some positive integer k).

2
USAMO 2008 Solution Notes web.evanchen.cc, updated 15 December 2024

§1 Solutions to Day 1
§1.1 USAMO 2008/1, proposed by Titu Andreescu
Available online at https://aops.com/community/p1116186.

Problem statement

Prove that for each positive integer n, there are pairwise relatively prime integers
k0 , . . . , kn , all strictly greater than 1, such that k0 k1 . . . kn − 1 is the product of two
consecutive integers.

In other words, if we let

P (x) = x(x + 1) + 1
then we would like there to be infinitely many primes dividing some P (t) for some integer
t.
In fact, this result is true in much greater generality. We first state:

Theorem 1.1 (Schur’s theorem)

If P (x) ∈ Z[x] is nonconstant and P (0) = 1, then there are infinitely many primes
which divide P (t) for some integer t.

Proof. If P (0) = 0, this is clear. So assume P (0) = c 6= 0.

Let S be any finite set of prime numbers. Consider then the value
 
Y
P k p
p∈S

for some integer k. It is 1 (mod p) for each prime p, and if k is large enough it should
not be equal to 1 (because P is not constant). Therefore, it has a prime divisor not in
S.

Remark. In fact the result holds without the assumption P (0) 6= 1. The proof requires
only small modifications, and a good exercise would be to write down a similar proof that
works first for P (0) = 20, and then for any P (0) 6= 0. (The P (0) = 0 case is vacuous, since
then P (x) is divisible by x.)

To finish the proof, let p1 , . . . , pn be primes and xi be integers such that

P (x1 ) ≡ 0 (mod p1 )
P (x2 ) ≡ 0 (mod p2 )
..
.
P (xn ) ≡ 0 (mod pn )

as promised by Schur’s theorem. Then, by Chinese remainder theorem, we can find x

such that x ≡ xi (mod pi ) for each i, whence P (x) has at least n prime factor.

3
USAMO 2008 Solution Notes web.evanchen.cc, updated 15 December 2024

§1.2 USAMO 2008/2, proposed by Zuming Feng

Available online at https://aops.com/community/p1116181.

Problem statement

Let ABC be an acute, scalene triangle, and let M , N , and P be the midpoints of
BC, CA, and AB, respectively. Let the perpendicular bisectors of AB and AC
intersect ray AM in points D and E respectively, and let lines BD and CE intersect
in point F , inside triangle ABC. Prove that points A, N , F , and P all lie on one
circle.

We present four solutions.

¶ Barycentric solution. First, we find the coordinates of D. As D lies on AM , we

know D = (t : 1 : 1) for some t. Now by perpendicular bisector formula, we find
c2 + b2 − a2
0 = b2 (t − 1) + (a2 − c2 ) =⇒ t = .
b2
Thus we obtain
D = 2SA : c2 : c2 .

Analogously E = (2SA : b2 : b2 ), and it follows that

F = 2SA : b2 : c2 .

The sum of the coordinates of F is

(b2 + c2 − a2 ) + b2 + c2 = 2b2 + 2c2 − a2 .

Hence the reflection of A over F is simply

2F − A = 2(b2 + c2 − a2 ) − (2b2 + 2c2 − a2 ) : 2b2 : 2c2 = −a2 : 2b2 : 2c2 .

It is evident that F 0 lies on (ABC) : −a2 yz − b2 zx − c2 xy = 0, and we are done.

¶ Synthetic solution (harmonic). Here is a synthetic solution. Let X be the point so

that AP XN is a cyclic harmonic quadrilateral. We contend that X = F . To see this it
suffices to prove B, X, D collinear (and hence C, X, E collinear by symmetry).

T N
P D
O
X
E

B M C

4
USAMO 2008 Solution Notes web.evanchen.cc, updated 15 December 2024

Let T be the midpoint of P N , so 4AP X ∼ 4AT N . So 4ABX ∼ 4AM N , ergo

]XBA = ]N M A = ]BAM = ]BAD = ]DBA

as desired.

¶ Angle chasing solution (Mason Fang). Obviously AN OP is concyclic.

Claim — Quadrilateral BF OC is cyclic.

Proof. Write

]BF C = ]F BC + ]BCF = ]F BA + ]ABC + ]BCA + ]ACF

= ]DBA + ]ABC + ]BCA + ]ACE
= ]BAD + ]ABC + ]BCA + ]EAC
= 2∠BAC = ∠BOC.

Define Q = AA ∩ BC.

Claim — Point Q lies on F O.

Proof. Write
◦
]BOQ = ]BOA + ]AOQ = 2]BCA + 90 + ]AQO
◦
= 2]BCA + 90 + ]AM O
◦ ◦
= 2]BCA + 90 + ]AM C + 90
= ]BCA + ]M AC = ]BCA + ]ACE
= ]BCE = ]BOF.

As Q is the radical center of (AN OP ), (ABC) and (BF OC), this implies the result.
√
¶ Inversive solution (Kelin Zhu). Invert about A with radius bc followed by a
reflection over the angle bisector of ∠A, and denote the image of a point X by X 0 . The
inverted problem now states the following:

In triangle AP ∗ N ∗ , let B ∗ , C ∗ be the midpoints of AP ∗ , AN ∗ and D∗ , E ∗

be the intersection of the A symmedian with (AP ∗ ), (AN ∗ ), respectively.
(AB ∗ D∗ ), (AC ∗ E ∗ ) intersect at a point F ∗ ; prove that it lies on P ∗ N ∗ .

I claim that, in fact, the midpoint of P ∗ N ∗ is the desired intersection. Redefine that
point as F ∗ and I will prove that (AB ∗ D∗ ), (AC ∗ E ∗ ) pass through it.
Note that
∠AD∗ B ∗ = ∠D∗ AB ∗ = ∠F ∗ AN ∗ = ∠AF ∗ B ∗ ,
where the first equality is due to B ∗ being the circumcircle of AD∗ P ∗ , the second
equality is due to the definition of the symmedian, and the third equality is due to the
parallelogram AB ∗ F ∗ C ∗ . A symmetric argument for C finishes.

5
USAMO 2008 Solution Notes web.evanchen.cc, updated 15 December 2024

§1.3 USAMO 2008/3, proposed by Gabriel Carroll

Available online at https://aops.com/community/p1116367.

Problem statement

Let n be a positive integer. Denote by Sn the set of points (x, y) with integer
coordinates such that
1
|x| + y + < n.
2
A path is a sequence of distinct points (x1 , y1 ), (x2 , y2 ), . . . , (x` , y` ) in Sn such that,
for i = 2, . . . , `, the distance between (xi , yi ) and (xi−1 , yi−1 ) is 1.
Prove that the points in Sn cannot be partitioned into fewer than n paths.

¶ First solution (local). We proceed by induction on n. The base case n = 1 is clear,

so suppose n > 1. Let S denote the set of points
 
1
S = (x, y) : x + y + ≥n−2 .
2
An example when n = 4 is displayed below.

For any minimal partition P of Sn , let P denote the path passing through the point
a = (n − 1, 0). Then the intersection of P with S consists of several disconnected paths;
let N be the number of nodes in the component containing a, and pick P such that N is
maximal. We claim that in this case P = S.
Assume not. First, note a = (n − 1, 0) must be connected to b = (n − 1, −1) (otherwise
join them to decrease the number of paths).
Now, starting from a = (n − 1, 0) walk along P away from b until one of the following
three conditions is met:
• We reach a point v not in S. Let w be the point before v, and x the point in S
adjacent to w. Then delete vw and add wx. This increases N while leaving the
number of edges unchanged: so this case can’t happen.

6
USAMO 2008 Solution Notes web.evanchen.cc, updated 15 December 2024

• We reach an endpoint v of P (which may be a), lying inside the set S, which is
not the topmost point (0, n − 1). Let w be the next point of S. Delete any edge
touching w and add edge vw. This increases N while leaving the number of edges
unchanged: so this case can’t happen.

• We reach the topmost point (0, n − 1).

Thus we see that P must follow S until reaching the topmost point (0, n − 1). Similarly
it must reach the bottom-most point (0, −n). Hence P = S.
The remainder of Sn is just Sn−1 , and hence this requires at least n − 1 paths to cover
by the inductive hypothesis. So Sn requires at least n paths, as desired.

Remark (Motivational comments from Evan). Basically the idea is that I wanted to peel
away the right path S highlighted in red in the figure, so that one could induct. But the
problem is that the red path might not actually exist, e.g. the set of paths might contain
the mirror of S instead.
Nonetheless, in those equality cases I found I could perturb some edges (e.g. change from
(−1, n − 2)–(0, n − 2) to (0, n − 2)–(1, n − 2)). So the idea then was to do little changes and
try to convert the given partition into one where the red path S exists, (and then peel it
away for induction) without decreasing the total number of paths.
To make this work, you actually want the incisions to begin ear the points a and b,
because that’s the point of S that is most constrained (e.g. you get a–b right away for free),
and assemble the path from there. (If you try to do it from the top, it’s much less clear
what’s happening.) That’s why the algorithm starts the mutations from around a.

¶ Second solution (global). Here is a much shorter official solution, which is much
trickier to find, and “global” in nature.
Color the upper half of the diagram with a blue/red checkerboard pattern such that
the uppermost point (n − 1, 0) is blue. Reflect it over to the bottom, as shown.

Assume there are m paths. Cut in two any paths with two adjacent blue points; this
occurs only along the horizontal symmetry axis. Thus:

• After cutting there are at most m + n paths, since at most n cuts occur.

7
USAMO 2008 Solution Notes web.evanchen.cc, updated 15 December 2024

• On the other hand, there are 2n more blue points than red points. Hence after
cutting there must be at least 2n paths (since each path alternates colors, except
possibly for double-red pairs).

So m + n ≥ 2n, hence m ≥ n.

Remark. This problem turned out to be known already. It appears in this reference:
Nikolai Beluhov, Nyakolko Zadachi po Shahmatna Kombinatorika, Matematika
Plyus, 2006, issue 4, pages 61–64.
Section 1 of 2 was reprinted with revisions as Nikolai Beluhov, Dolgii Put Korolya, Kvant,
2010, issue 4, pages 39–41. The reprint is available at http://kvant.ras.ru/pdf/2010/
2010-04.pdf.

Remark (Nikolai Beluhov). As pointed out in the reference above, this problem arises
naturally when we try to estimate the greatest possible length of a closed king tour on the
chessboard of size n × n with n even, a question posed by Igor Akulich in Progulki Korolya,
Kvant, 2000, issue 3, pages
√ 47–48. Each one of the two references above contains a proof
that the answer is n + 2(n2 − n).

8
USAMO 2008 Solution Notes web.evanchen.cc, updated 15 December 2024

§2 Solutions to Day 2
§2.1 USAMO 2008/4, proposed by Gregory Galperin
Available online at https://aops.com/community/p1116177.

Problem statement

For which integers n ≥ 3 can one find a triangulation of regular n-gon consisting
only of isosceles triangles?

The answer is n of the form 2a (2b + 1) where a and b are nonnegative integers not both
zero.
Call the polygon A1 . . . An with indices taken modulo n. We refer to segments A1 A2 ,
A2 A3 , . . . , An A1 as short sides. Each of these must be in the triangulation. Note that

• when n is even, the isosceles triangles triangle using a short side A1 A2 are 4An A1 A2
and 4A1 A2 A3 only, which we call small.

• when n is odd, in addition to the small triangles, we have 4A 1 (n+3) A1 A2 , which

2
we call big.

This leads to the following two claims.

Claim — If n > 4 is even, then n works iff n/2 does.

Proof. All short sides must be part of a small triangle; after drawing these in, we obtain
an n/2-gon.

Thus the sides of P must pair off, and when we finish drawing we have an n/2-gon.

Since n = 4 works, this implies all powers of 2 work and it remains to study the case
when n is odd.

Claim — If n > 1 is odd, then n works if and only if n = 2b + 1 for some positive
integer b.

Proof. We cannot have all short sides part of small triangles for parity reasons, so some
side, must be part of a big triangle. Since big triangles contain the center O, there can
be at most one big triangle too.
Then we get 12 (n − 1) small triangles, pairing up the remaining sides. Now repeating
the argument with the triangles on each half shows that the number n − 1 must be a
power of 2, as needed.

9
USAMO 2008 Solution Notes web.evanchen.cc, updated 15 December 2024

§2.2 USAMO 2008/5, proposed by Kiran Kedlaya

Available online at https://aops.com/community/p1116189.

Problem statement

Three nonnegative real numbers r1 , r2 , r3 are written on a blackboard. These

numbers have the property that there exist integers a1 , a2 , a3 , not all zero, satisfying
a1 r1 + a2 r2 + a3 r3 = 0. We are permitted to perform the following operation: find
two numbers x, y on the blackboard with x ≤ y, then erase y and write y − x in its
place. Prove that after a finite number of such operations, we can end up with at
least one 0 on the blackboard.

We first show we can decrease the quantity |a1 | + |a2 | + |a3 | as long as 0 ∈
/ {a1 , a2 , a3 }.
Assume a1 > 0 and r1 > r2 > r3 without loss of generality and consider two cases.

• Suppose a2 > 0 or a3 > 0; these cases are identical. (One cannot have both a2 > 0
and a3 > 0.) If a2 > 0 then a3 < 0 and get

0 = a1 r1 + a2 r2 + a3 r3 > a1 r3 + a3 r3 =⇒ a1 + a3 < 0

so |a1 + a3 | < |a3 |, and hence we perform (r1 , r2 , r3 ) 7→ (r1 − r3 , r2 , r3 ).

• Both a2 < 0 and a3 < 0. Assume for contradiction that |a1 + a2 | ≥ −a2 and
|a1 + a3 | ≥ −a3 both hold (if either fails then we use (r1 , r2 , r3 ) 7→ (r1 − r2 , r2 , r3 )
and (r1 , r2 , r3 ) 7→ (r1 − r3 , r2 , r3 ), respectively). Clearly a1 + a2 and a1 + a3 are
both positive in this case, so we get a1 + 2a2 and a1 + 2a3 ≥ 0; adding gives
a1 + a2 + a3 ≥ 0. But

0 = a1 r1 + a2 r2 + a3 r3
> a1 r2 + a2 r2 + a3 r2
= r2 (a1 + a2 + a3 )
=⇒ 0 < a1 + a2 + a3 .

Since this covers all cases, we see that we can always decrease |a1 | + |a2 | + |a3 | whenever
0∈/ {a1 , a2 , a3 }. Because the ai are integers this cannot occur indefinitely, so eventually
one of the ai ’s is zero. At this point we can just apply the Euclidean Algorithm, so we’re
done.

10
USAMO 2008 Solution Notes web.evanchen.cc, updated 15 December 2024

§2.3 USAMO 2008/6, proposed by Sam Vandervelde

Available online at https://aops.com/community/p1116182.

Problem statement

At a certain mathematical conference, every pair of mathematicians are either friends

or strangers. At mealtime, every participant eats in one of two large dining rooms.
Each mathematician insists upon eating in a room which contains an even number
of his or her friends. Prove that the number of ways that the mathematicians may
be split between the two rooms is a power of two (i.e. is of the form 2k for some
positive integer k).

Take the obvious graph interpretation where we are trying to 2-color a graph. Let A be
the adjacency matrix of the graph over F2 , except the diagonal of A has deg v (mod 2)
instead of zero. Then let d~ be the main diagonal. Splittings then correspond to A~v = d.
~
It’s then immediate that the number of ways is either zero or a power of two, since if it
is nonempty it is a coset of ker A.
Thus we only need to show that:

Claim — At least one coloring exists.

Proof. If not, consider a minimal counterexample G. Clearly there is at least one odd
vertex v. Consider the graph with vertex set G − v, where all pairs of neighbors of v have
their edges complemented. By minimality, we have a good coloring here. One can check
that this extends to a good coloring on G by simply coloring v with the color matching
an even number of its neighbors. This breaks minimality of G, and hence all graphs G
have a coloring.

It’s also possible to use linear algebra. We prove the following lemma:

Lemma (grobber)
Let V be a finite dimensional vector space, T : V → V and w ∈ V . Then w is in the
image of T if and only if there are no ξ ∈ V ∨ for which ξ(w) 6= 0 and yet ξ ◦ T = 0.

Proof. Clearly if T (v) = w, then no ξ exists. Conversely, assume w is not in the image
of T . Then the image of T is linearly independent from w. Take a basis e1 , . . . , em for
the image of T , add w, and then extend it to a basis for all of V . Then have ξ kill all ei
but not w.

Corollary
In a symmetric matrix A mod 2, there exists a vector v such that Av is a copy of
the diagonal of A.

Proof. Let ξ be such that ξ ◦ T = 0. Look at ξ as a column vector w> , and let d be the
diagonal. Then
0 = w> · T · w = ξ(d)

11
USAMO 2008 Solution Notes web.evanchen.cc, updated 15 December 2024

because this extracts the sum of coefficients submatrix of T , and all the symmetric entries
cancel off. Thus no ξ as in the previous lemma exists.

This corollary gives the desired proof.

12