7th Bay Area Mathematical Olympiad

February 22, 2005

Problems and Solutions

1 An integer is called formidable if it can be written as a sum of distinct powers of 4, and successful if it
can be written as a sum of distinct powers of 6. Can 2005 be written as a sum of a formidable number
and a successful number? Prove your answer.

Solution: Suppose that 2005 = a + b, where a is formidable and b is successful. Then a must be
a sum of some of the powers of 4 less than 2005, namely 1, 4, 16, 64, 256, 1024. Similarly, b must be a
sum of some of the numbers 1, 6, 36, 216, 1296. So, a + b must be a sum of some distinct entries from
the list
1, 1, 4, 6, 16, 36, 64, 216, 256, 1024, 1296.
If we use both 1024 and 1296, we get at least 1024 + 1296 = 2320 which is too big. But if we omit one
of them, the most we can get is

1 + 1 + 4 + 6 + 16 + 36 + 64 + 216 + 256 + 1296 = 1896

which is too small. So there is no way to achieve a sum of 2005.

Note: We apologize for the ambiguity of the question, because we did not carefully define what
was meant by a “power.” Some students assumed that powers had to be positive, which meant formidable
and successful numbers are always even, and consequently 2005 cannot be a sum of a formidable and a
successful number. Other students assumed that negative powers were possible (which did not alter the
solution above in any important ways). Still others assumed that fractional powers were allowable. In
this case, it is possible to express 2005 as a sum of a formidable number and a successful number. All
of these “alternative” solutions, if carefully done, got full credit.

2 Prove that if two medians in a triangle are equal in length, then the triangle is isosceles.

L N1 C
M1

N M

A B

Solution 1: Let equal medians AD and B E in triangle ABC meet at F. It is well known that

AF : F D = B F : F E = 2 : 1.
 Triangle ABC is isosceles if we can show AC = BC or equivalently, that AE = B D. But triangles
AF E and B F D are congruent with vertical angles plus sides that are 1/3 and 2/3 of the equal median
length.

Solution 2: Let medians AM = B N in
ABC. Extend each median to AM1 and B N1 so that M and
N are the midpoints of AM1 and B N1 , respectively. By the property of bisecting diagonals, AB M1 C
and ABC N1 are parallelograms. Hence C M1 and C N1 are each parallel and equal to AB. We conclude
that C lies on N1 M1 , C is the midpoint of N1 M1 , and AM1 = B N1 as they are twice the lengths of the
original medians AM and B N . Summarizing, AB M1 N1 is a trapezoid with equal diagonals.
It is easy to see that such a trapezoid is isosceles. One way to see this is to draw a line through A
parallel to diagonal B N1 , until it intersects line N1 M1 in point L. Thus, AB N1 L is a parallelogram,
so  AL N1 =  AB N1 . On the other hand,
AM1 L is isosceles since AL = B N1 = AM1 ; hence,
 AL N1 =  AM1 N1 . Finally, AB||N1 M1 implies  AM1 N1 =  B AM1 . We conclude that  B AM1 =
 AB N1 , and
AB N1 and
B AM1 are congruent by two equal sides and angles between these sides.
Therefore, B M1 = AN1 and our trapezoid is isosceles. Hence  AN1 C =  B M1 C.
Finally,
AC N1 and
BC M1 are congruent by AN1 = B M1 , C N1 = C M1 and  AN1 C =  B M1 C.
We conclude that AC = BC and our original
ABC is also isosceles.

Solution 3: As a variation of the above solution, note that N M is the midsegment of
ABC, and as
such it is parallel to AB. Thus AB M N is a trapezoid with equal diagonals, which by a similar argument
as in Solution 1 is isosceles. Therefore,  B AC =  ABC and AC = BC.

Solution 4: A well-known formula for a parallelogram AB M1 C says: 2(AB 2 + AC 2 ) = AM12 + BC 2

(it can be easily proved with vectors for example). From here one derives a formula for the median AM
of a triangle
ABC:
1
1
AM 2 = AB 2 + AC 2 ) − BC 2 .
2 4
Similarly, the other median B N in
ABC satisfies:
1
2 1
BN2 = AB + BC 2 ) − AC 2 .
2 4
Since AM = B N , easy algebraic cancellations lead to AC 2 = BC 2 , i.e. AC = BC and our triangle is
isosceles.

3 Let n be an integer greater than 12. Points P1 , P2 , . . . , Pn , Q in the plane are distinct. Prove that for
some i, at least n/6 − 1 of the distances

P1 Pi , P2 Pi , . . . , Pi−1 Pi , Pi+1 Pi , . . . , Pn Pi

are less than Pi Q.

 Solution: Cut the plane into six 60◦ “pizza slices” with vertex Q. Rotating if necessary, we may
assume that none of the P j lie on the cuts. By the pigeonhole principle, one slice contains at least n/6
of the P j . Let Pi be a point in this slice farthest from Q. It remains to show that all other points P j in
this slice satisfy P j Pi < Pi Q.
The average of the angles of
P j Pi Q is 180◦ /3 = 60◦ , so  P j Q Pi , which is less than 60◦ , is less
than one of the other angles. Smaller angles of a triangle are opposite shorter sides, so P j Pi is less than
one of P j Q and Pi Q. By choice of i, P j Q ≤ Pi Q, so in any case, P j Pi < Pi Q. (Alternatively, one
could use the Law of Cosines to show that the side of a triangle opposite an angle smaller than 60◦ is not
the longest.)

4 There are 1000 cities in the country of Euleria, and some pairs of cities are linked by dirt roads. It is
possible to get from any city to any other city by traveling along these roads. Prove that the government
of Euleria may pave some of the roads so that every city will have an odd number of paved roads leading
out of it.
Solution 1: Call a city “even” or “odd” according to whether the number of paved roads coming out of
it is even or odd. Note the following.

Lemma: No matter which roads are paved, there will be an even (possibly zero) number of
even cities.

To see why this is true, let di be the number of paved roads leading out of city i. The sum

d1 + d2 + · · · + d1000

counts each paved road exactly twice, and hence will be an even number (this is known as the “Handshake
Lemma”). If there were an odd number of even cities, then there would be an odd number of odd cities
(since there are 1000 cities and 1000 is even). But then the sum above would be odd, a contradiction.
Using this lemma, we can create an algorithm which will eventually pave the roads so that all cities
are odd.

1. Start by paving all the roads. If each city is odd, we are done.
2. Otherwise, find an even city x. By the lemma, there will be at least one other even city, y. Consider
the path joining x and y. Change the “state” of all the roads in this path (in other words, if a road
is paved, unpave it; if it is dirt, pave it). This procedure changes the parity of x and y (i.e., changes
them from even to odd), but does not alter the parity of any other city in Euleria, because if z is a
city on the path from x to y, both the road going into z (from the x-direction) and the road leaving
it (heading towards y) will have changed.
3. Step 2 thus reduces the number of even cities by 2. Repeat this step as much as needed until the
number of even cities is zero.

Solution 2: We will think of Euleria as a connected graph (a network of vertices joined by edges), where
each city is a vertex and each dirt road joining two cities is an edge. Number the vertices 1, 2, . . . , 1000.
For each i = 1, 2, . . . , 500, consider any path from vertex 2i − 1 to vertex 2i (we know such a path
exists, since the graph is connected), and place a mark on each edge used in the path. Then pave all the
 roads (edges) that have an odd number of marks. To show that this scheme meets the requirements, it
is enough to prove that, for each vertex v, the edges incident to it have an odd total number of marks,
since the number of such edges with an odd number of marks will then be odd. But consider all the
occurrences of v in any of our 500 paths. Notice that v is an endpoint of one such path, which therefore
contributes one mark on an edge incident to v; any other occurrence of v is internal to a path, which
therefore contributes two marks, one on the edge leading into v and another on the edge leading out of
it. It follows that the total number of marks on edges incident to v is odd. This completes the proof.

5 Let D be a dodecahedron which can be inscribed in a sphere with radius R. Let I be an icosahedron
which can also be inscribed in a sphere of radius R. Which has the greater volume, and why?
Note: A regular polyhedron is a geometric solid, all of whose faces are congruent regular polygons,
in which the same number of polygons meet at each vertex. A regular dodecahedron is a polyhedron
with 12 faces which are regular pentagons and a regular icosahedron is a polyhedron with 20 faces which
are equilateral triangles. A polyhedron is inscribed in a sphere if all of its vertices lie on the surface of
the sphere.
The illustration below shows a dodecahdron and an icosahedron, not necessarily to scale.

Solution 1: D has the greater volume.

The key idea that we will use is a beautiful relationship between I and D; they are dual polyhedra. To
understand this, note that D has 12 faces, 30 edges, and 20 vertices. (The 12 pentagonal faces produce 60
edges and 60 vertices, but these are, respectively, double- and triple-counted. Hence there are 60/2 = 30
edges and 60/3 = 20 vertices). Likewise, I has 20 faces, 30 edges, and 12 vertices. Notice that I has
as many vertices as D has faces, and vice-versa.
Consequently, if we join each center of the faces of an icosahedron to the centers of adjacent faces,
we will get a dodecahedron, and vice-versa. (For the other platonic solids, the cube and octahedron are
duals of each other, and the tetrahedron is its own dual).
We will use duality to prove an important lemma.

Lemma: If a dodecahedron and an icosahedron can be inscribed in the same sphere, then
they circumscribe the same sphere as well.
 To see why this is true, imagine I and D both sitting inside a sphere of radius R with center O, which
circumscribes both polyhedra. Thus all the vertices of both polyhedra lie on the surface of this sphere.
Because the polyhedra are dual, we can place the polyhedra so that for every vertex A of I , the line AO
is perpendicular to and passes through the center of a face of D. Likewise, for every vertex B of D, the
line B O is perpendicular to and passes through the center of a face of I .
The illustration below is an imperfect depiction of a portion of this situation. The left picture is a
“top view,” as seen by an observer looking directly down at the center of a face of D. Point A is a vertex
of I , and B is a vertex of D.

B
A r
E C

B
h
A R

Top View Side View

The picture on the right is a “side view” (not-to-scale), showing the projections of A and B down to
the center O of the sphere. Let C O be the perpendicular projection of AO onto B O. Note that point
C is the center of a face of the icosahedron I (in the Top View, you cannot see C because it is directly
“underneath” B), and h = C O is the radius of the inscribed sphere of I . Also r = AC is the radius of
the circumscribed circle about a face of I .
Let E O be the perpendicular projection of B O on to AO. Clearly E O = C O = h (since AO =
B O = R), but by duality, E must be a center of a face of the dodecahedron D. Consequently, the radius
of the inscribed sphere of D must also equal h.
Now it is a simple matter to compute and compare the two volumes. Notice that D consists of 12
pyramids with regular pentagonal bases (each with circumradius r ) and height h. Likewise, I consists
of 20 pyramids with equilateral triangular bases (each with circumradius r ) and height h. Recall that
the volume of a pyramid with base area B and height h is Bh/3. Now we can easily compare the two
volumes.1
h
vol(D) × 12 × area of pentagon
= 3
vol(I ) h
× 20 × area of triangle
3
1
We are using the fact that a regular n-gon that can be inscribed in a circle of radius r can be dissected into n “pie
slices” that are isosceles triangles with sides of length r and vertex angle (360/n)◦ . Thus the area of the regular n-gon is
n × (r 2 /2) × sin(360/n)◦ .
 3 × area of pentagon
=
5 × area of triangle
r2
3 × 5 × sin 72◦
= 2
r2
5 × 3 × sin 120◦
2
sin 72◦
=
sin 60◦
> 1.

Solution 2: (sketch) Instead of showing that the inscribed spheres have equal radii, we can use duality
and the “top view” used in the previous solution to show that the radius of the circumscribed circle of
each face of D is equal to the radius of the circumscribed circle of each face of I . In other words, the two
circles depicted below have the same radii. This fact is attributed to the ancient Greek mathematician
Apollonius.

Once this is known, it is easy to deduce that the two polyhedra have identical inscribed sphere radii,
and the rest of the solution proceeds as before.