37th English Language Master’s Open

1. Let ABCD be a convex quadralateral. Let E, F, G, H be points on segments AB, BC,

CD, DA, respectively, and let P be intersection of EG and F H. Given that quadrilaterals
HAEP, EBF P, F CGP, GDHP all have inscribed circles, prove that ABCD also has an
inscribed circle.

Solution: Let us label the points of tangency of the four given incircles as shown in the
diagram.
Then, to prove that ABCD has an inscribed circle, it suffices to show that AB + CD =
AD + BC. Since common tangents from a point to a circle share the same length, we get

AB + CD = AD + BC

⇔ (AA1 +A1 B2 +B2 B)+(CC1 +C1 D2 +D2 D) = (AA2 +A2 D1 +D1 D)+(BB1 +B1 C2 +C2 C)

⇔ A1 B2 + C1 D2 = A2 D1 + B1 C2 .

We first want to show that A2 D1 = A4 D3 . If AD||EG, then this is true because A2 , D1 , D3 , A4

form the corners of a rectangle. Otherwise, consider the intersection of EG and AD. Note

1
 that, of the incircles of AEP H and HP GD, one is an incircle and the other an excircle
of the triangle with the intersection point as a vertex.
Consequently, A4 is the reflection of D3 over the midpoint of HP and we have

A2 D1 = A2 H + HD1 = HA4 + HD3 = P D3 + A4 P = A3 P + P D4 = A3 D4 .

Similarly, B1 C2 = B3 C4 , A1 B2 = A3 B4 , and D2 C1 = D4 C3 .
Combining, we get

A2 D1 + B1 C2 =A4 D3 + B3 C4
=P A4 + P D3 + P B + 3 + P C4
=P A3 + P D4 + P B4 + P C3
=A3 B4 + D4 C3
=A1 B2 + C1 D2 ,

so we are done.
This problem was proposed by Evan O’Dorney.

2. Wanda the Worm likes to eat Pascal’s triangle. One day, she starts at the top of the

triangle and eats 00 = 1. Each move, she travels to an adjacent positive integer and eats
it, but she can never return to a spot that she has previously eaten. If Wanda can never
eat numbers a, b, c such that a + b = c, proof that it is possible for her to eat 100, 000
numbers in the first 2011 rows given that she is not restricted to traveling only in the first
2011 rows.
 
(Here, the n+ 1st row of Pascal’s triangle consists of entries of the form nk for integers 0 ≤
    n−1  n   n 
k ≤ n. Thus, the entry nk is considered adjacent to the entries n−1 k−1
, k , k−1 , k+1 ,
n+1 n+1
k
, k+1 .)

Solution: We will prove by induction on n that it is possible for Wanda to eat 3n numbers
in the first 2n rows of Pascal’s triangle. Our inductive hypothesis includes the following
conditions on the first 2n rows of Pascal’s triangle when all the entries are taken modulo
2:

• Row 2n contains only odd numbers.

• The 2n rows contain a total of 3n odd numbers.

2
 • The triangle of rows has 120 degree rotational symmetry.
• There is a path for Wanda to munch that starts at any corner of these rows, contains
all the odd numbers, and ends at any other corner.

Our base case is n = 1; it is not difficult to check that all of these conditions hold. Wanda’s
  
path in these two rows is 00 → 01 → 11 .
Now, assume that these hold for the first 2m rows of Pascal’s triangle. We will show that
they also hold for the first 2m+1 rows. Note that a single 1 surrounded by 2m − 1 0’s to
either side generated the first 2m rows since each element is equal to the sum of the two
numbers directly above it. However, by our inductive hypothesis, all of the entries in the
2m row were 1’s. Hence, the first and last entires of the 2m+1 row are also both 1, and the
remainder of the entires are 0. Consequently, we note that these 1’s and 0’s generate two
other copies of the first 2m rows of Pascal’s triangle, along with an inverted triangle of all
0’s in the middle.
Now it suffices to check that our conditions hold:

• As row 2m+1 simply contains two side-by-side copies of the 2m th row modulo 2, it
also consists all of 1’s.
• The first 2m+1 rows contain three copies of the first 2m rows along with a triangle of
0’s, so they contain 3(3m ) = 3m+1 odd numbers.
• As each of the three 2m row triangles had rotational symmetry, so does the larger
one.
 m 
• By our inductive hypothesis, Wanda can travel from 00 to 2 0−1 and eat all the
 m
odd numbers in those rows. She can then travel to 20 , eat all the numbers in the
 m+1 −1  m+1 
lower-left triangle and end at 22m −1 , travel to 2 2m−1 , eat all the odd numbers
 m+1 
in the lower-right triangle, and finally end at 22m+1 −2
−1
. Due to rotational symmetry,
she can also start and end at any corner.

We have now proved our induction.

Note that if Wanda only eats odd numbers, then she will never eat three numbers a, b, c
such that a + b = c. We have 210 < 2011 < 2048 = 211 .
It suffices to check that there are sufficient odd numbers in the first 2011 rows. We have
showed that there are 311 odd numbers in the first 2048 rows. Also, row n has n elements

3
 and thus contains at most n odd numbers. Hence, there are at least
1
311 − 2048 − 2047 − . . . − 2012 = 311 − (2048 + 2012)(2048 − 2011) > 1000000
2
odd numbers in the first 2011 rows.
This problem was proposed by Linus Hamilton.

3. Determine whether there exists a sequence {an }∞

n=0 of real numbers such that the following
holds:

• For all n ≥ 0, an = 0.
• There exist real numbers x and y such that an+2 = xan+1 + yan for all n ≥ 0.
• For all positive real numbers r, there exists positive integers i and j such that
|ai | < r < |aj |.

Solution: The answer is yes.

···2

22 . Then, let θ =
Let xn =  π
2
1
x1
+ 1
x2
+ . . . , and let r = 2 and an = r n cos(nθ).
2n 2’s
We will prove that this sequences satisfies the three given conditions.
First, note that
an+2 = 2r cos(θ)an+1 − r 2 an
for all n by the addition formula for cosine, so the recursion condition is satisfied by setting
x = 2r cos(θ) and y = −r 2 .
Second, we note that if there exists any integer n such that an = 0, then we would have
nθ = π(k + 12 ) for some k ∈ {0, 1, 2, . . .}, implying that πθ is rational. However, we have


θ 1 1 1
= + + ... ,
π 2 x1 x2
which has a non-periodic binary expansion and is therefore irrational. Hence, we know
hat the second condition is satisfied.
Third, consider the subsequence

bn = axn
= r xn cos(xn θ)
 ∞ 
π xn
= r xn cos .
2 k=1 xk

4
 n xn
Where a = k=1 xk is an odd integer, we note that
  ∞ 
 
 π xn 
|bn | < |r|xn cos 
 2 k=1 xk 
   
 ∞ 
 π xn 
= |r|xn cos a+ 
 2 xk 
k=n+1
  

∞ 
xn  π xn 
= |r| sin 
 2 xk 
k=n+1

π
∞
xn
≤ 2xn
2 k=n+1 xk

∞
xn π xn
≤2
2 k=n+! xn+1 · 2k−n−1
xn
= 2xn · π · ,
xn+1
which becomes arbitrarily small as n approaches infinity.
Consequently, {an } has a subsequence with arbitrarily small magnitude. By Kronecker’s
 
Theorem, there is also a sequence n1 , n2 , . . . with { n2π1 θ } ∈ − π6 , π6 for i = 1, 2, . . .. Then,
the sequence an1 , an2 , an3 , . . . tends to infinity. Thus, {an } has both a subsequence with
magnitude tending to 0 and a subsequence with magnitude tending to infinity, so the third
property also holds.
This problem was proposed by Alex Zhu.

4. Find all functions f : R+ → R+ , where R+ denotes the positive reals, such that whenever
a > b > c > d > 0 are reel numbers with ad = bc,

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

Solution: Since f (a+ d) −f (a−d) only depends on ad, we can have a function g mapping
positive reals to reals such that whenever a > d,

g(ad) = f (a + d) − f (a − d).

5
Also,

g(kad + k(k + 1)d2 ) = g((a + (k + 1)d)(kd))

= f (a + (2k + 1)d) − f (a + d)
and
g((k + 1)ad + k(k + 1)d2 ) = g((a + kd)((k + 1)d))
= f (a + (2k + 1)d) − f (a − d)
= g(ad) + g(kad + k(k + 1)d2)

for any constant k > 0.

Let a = 2d, and let x = d2 . Then we have the following:

g((k 2 + 3k + 2)x) = g(2x) + g((k 2 + 3k)x)

However, (k 2 +3k)x is surjective over the positive reals as k > 0, so if we let y = (k 2 +3k)x,
we obtain
g(x + y) = g(x) + g(y)
for all positive real numbers x and y. Consequently, for any positive real number x, we
can always find a unique λ > 0 such that λ(λ + 1) = x. Thus,

g(x) = g(λ(λ + 1)) = f (2λ + 1) − f (1) ≥ −f (1)

Because g is bounded below and satisfies Cauchy’s Functional Equation, there exists a real
number a such that g(x) = ax for all x > 0. That gives, for u > 1,
a −a + 4f (1)
f (u) = f (1) + g((u2 − 1)/4) = u2 +
4 4
and for u < 1
a −a + 4f (1)
f (u) = f (1) − g((1 − u2 )/4) = u2 +
4 4
Thus there exist constants c, d such that for u = 1, f (u) = cu2 + d. Finally,

f (1) = f (4 − 3) = f (4 + 3) + f (6 − 2) − f (6 + 2) = 49c + d + 16c + d − 64c − d = c + d

Thus equations of the form f (u) = cu2 + d for all u > 0 are the only possible solutions. It
is not hard to see that this is a solution to the functional equation if and only if c and d
are nonnegative real numbers which are not both zero.
This problem was proposed by Calvin Deng.

6
5. Let p > 13 be a prime of the the form 2q + 1, where q is prime. Find the number of ordered
pairs of integers (m, n) such that 0 ≤ m < n < p − 1 and

3m + (−12)m ≡ 3n + (−12)n (mod p).

Solution:
Lemma 1: −4 is a primitive root modulo p.
Proof of Lemma 1: Note that ordp (−4)|p−1 = 2q, so ordp (−4) is one of 1, 2, q, 2q. Because
−4 ≡ 1 (mod p) we have ordp (−4) = q). As 16 = 4n ≡ 1 (mod p), we have ordp (−4) = 2.
Also, we have



−4 −1 4
= · = −1 · 1 · −1,
p p p

−1 p−1
since p
= 1 follows from p > 13 ⇒ 2
= q being odd.
Thus, following from the fact that −4 is not a quadratic residue modulo p, we have that
p−1 2q
2| =
ordp (−4) ordp (−4)

⇒ 2 · ordp (−4)  | 2q

⇒ ordp (−4)  | q

Consequently, ordp (−4) = 2q, as desired.

Lemma 2: The order of 3 modulo p is exactly q.
Proof of Lemma 2: Note that ordp (3)|p − 1 = 2q, so ordp (3) is one of 1, 2, q, 2q. Because
3 ≡ 1 (mod p), we have ordp (3) = 1. As 32 = 9 ≡ 1 (mod p), we have ordp (3) = 2. Then,
we have

p = 2q + 1
≡ 2 · 1 + 1 or 2 · 2 + 1 (mod 3)
≡ 0 or 2 (mod 3)

Because p > 13, we know that q = 3 and p = 3, giving

p
2
= = 1.
3 3

7
Also, by quadratic reciprocity, we have

 
3 p (3−1)(p−1)
· = (−1) 4
p 3


3
(−1) = (−1)q
p
= −1


3
⇒ = 1.
p

We now know that 3 is a quadratic residue modulo p, so ordp (3) = 2q, giving us ordp (3) = q,
as desired.
Lemma 3: −12 is a primitive root modulo p.
Proof of Lemma 3: Note that ordp (−12)|p − 1 = 2q, so ordp (−12) is one of 1, 2, q, 2q.
Because −12 ≡ 1 (mod p), we have ordp (−12) = 1. As (−12)2 = 144 ≡ 1 (mod p), we
have ordp (−12) = 2. Then, after substituting for the values found in Lemma 1 and Lemma
2, we obtain





−12 −1 4 3
= · ·
p p p p
= (−1) · (1) · (1)
= −1.

Thus, −12 is not a quadratic residue modulo p, giving us

p−1 2q
2| =
ordp (−12) ordp (−12)
⇒ ordp (−12)  | q

It follows that ordp (−12) = 2q, as desired.

Main Proof: We now simplify the given equation:

3m + (−12)m ≡ 3n + (−12)n (mod p)

≡ 3n−m · 3m + 3n−m · 3m · (−4)n (mod p)
1 + (−4)m ≡ 3n−m + 3n−m · (−1)n (mod p)
1 − 3n−m ≡ 3n−m · (−4)n − (−4)m (mod p)
≡ (−4)m · ((−12)n−m − 1) (mod p).

8
 We ignore for the moment the condition that m < n and count all pairs m, n ∈ Zp−1 = Z2q .
So, if n ≡ m (mod 2q), then (−12)n−m − 1 ≡ 0 (mod p), giving us

(−4)m ≡ (1 − 3n−m )((−12)n−m − 1)−1 (mod p).

Because −4 is primitive modulo p, we have that any non-zero residue of (−4)m (mod p)
uniquely determines the residue of m (mod 2q). So each non-zero residue of n−m (mod 2q)
uniquely determines m (mod 2q), so long as

3n−m − 1 ≡ 0 (mod p) ⇔ q  |n − m.

Consequently, for each (n − m) ∈ {1, 2, . . . , q − 1, q + 1, q + 2, . . . , 2q − 1} (mod 2q), we

unitely determine the ordered pair (m, n) ∈ Z22q . However, on taking remainders on
divison by 2q of m, n, we must have m < n. Thus, for each x ≡ q, 0, the solutions for
n − m ≡ x (mod 2q) and n − m ≡ −1 (mod 2q) give exactly 1 solution (m, n) with m < n.
2q−2
Thus, we have a total of 2
= q − 1 solutions.
This problem was proposed by Alex Zhu.

6. Consider the infinite grid of lattice points in Z3 . Little D and Big Z play a game, where
Little D first loses a shoe on an unmunched point in the grid. Then, Big Z munches a
shoe-free plane perpendicular to one of the coordinate axes. They continue to alternature
turns in this fashion, with Little D’s goal to loose a shoe on each of n consecutive lattice
points on a line parrallel to one of the coordinate axes. Determine all n for which Little
D can accomplish his goal.

Solution: We claim that Little D can accomplish this for all n.

We will start by separating out the three coordinate axes: thus, if Little D loses a shoe
at the point (i, j, k) for integers i, j, and k, he plays on i on the x-axis, j on the y-axis,
and k on the z-axis in the same move. Meanwhile, when Big Z munches a plane, he plays
on only one point on one of the coordinate axes. Hence, since Big Z can only munch a
shoe-free plane, he cannot munch point l on a particular axis if Little D has already placed
a shoe there.
We will call a string of points marked (by shoes) on one of these coordinate axes unbounded
if Big Z has not munched any point on that axis within 2n + 1 of at least one endpoint of
the string.

9
Lemma: For any integers m and l, Little D can create l unbounded strings of m consecutive
points on a single coordinate axis.
Proof of lemma: We will prove this by induction on m.
Our base case is m = 1. Then, we note that if Little D makes 1.5l triplets of moves
over the three axes, making sure that he distributes any marked points in the same axis
at least 5n apart, then Big Z can bound at most 1.5l strings because he can only bound
at most one string on each move. However, this leaves 3x unbounded strings of length 1;
by the pigeonhole principle, at least l of these must be in the same cooordinate axis.
Now, suppose that this is true for some m. We will show that it is also true for
m + 1. Without loss of generality, we note from our induction hypothesis that Little D
can construct  x(m+1)
2
 unbounded strings of length m on the x-axis. Consequently, he
can create m + 3 strings of length m + 1 in m + 1 moves: in each move, he lengthens
one unbounded string of the x-axis, while on each of the y and z-axes he builds up a new
string of m + 1 marked points. However, Big Z can, in these m + 1 moves, bound at most
m + 1 of these strings. Hence, Little D can construct at least 2 strings of length m + 1
for every m + 1 strings of length m used up. It follows that he can achieve x unbounded
strings of length m + 1. We have now proved our desired induction.
Main Proof: Now, without loss of generality, by our lemma Little D can mark n consecutive
points on the x-axis. Then, he has established n consecutive yz-planes that Big Z can never
much. Suppose that one of the points he has played on is (i, j, k) for some i, j, k ∈ Z. Then,
Big Z can never munch any part of the line y = j, z = k in those n consecutive xy planes.
Hence Little D can lose shoes in the remainder of those n points over his next few moves,
at which point he has achieved his goal.
This problem was proposed by David Yang.

10