PMO 2021 Qualifying Stage

Carl Joshua Quines

February 20, 2021

Due to the pandemic, the PMO was held virtually. In lieu of a qualifying stage and an area stage,
only a single qualifying stage was held, explaining the different format. This test continues last
year’s numbering scheme, with the numbering continuing throughout the test. Are any explanations
unclear? If so, contact me at cj@cjquines.com External-Link-Alt. More material is available on my website:
https://cjquines.com.

PART I. Choose the best answer. Figures are not drawn to scale. Each correct answer is worth
two points.

1. In a convex polygon,the number of diagonals is 23 times the number of its sides. How many sides
does it have?

(a) 46 (b) 49 (c) 66 (d) 69

Answer. (b) 49 .

Solution. If a polygon has n vertices, the number of diagonals it has is n2 − n. This is because

a diagonal
joins two vertices, but we overcounted the n sides of the polygon. From the problem,
we get n2 − n = 23n, and solving yields n = 49.

Remark. Another way to get the same formula. A diagonal is formed by joining each of the n vertices to one
of n − 3 vertices: any other vertex except itself and the vertices it’s next to. This counts each diagonal twice,
though, so we divide by two to get 12 n(n − 3), which is the same formula.

2. What is the smallest real number a for which the function f (x) = 4x2 − 12x − 5 + 2a will always
be nonnegative for all real numbers x?
3 5
(a) 0 (b) (c) (d) 7
2 2

Answer. (d) 7 .

Solution 1. For a quadratic to be always nonnegative, its discriminant has to be nonpositive:

(−12)2 − 4(4)(−5 + 2a) ≤ 0

144 − (−80 + 32a) ≤ 0
224 ≤ 32a.

Hence a ≥ 7, and the minimum value for a is 7.

Solution 2. Note that 2a−5 only affects the constant term of f , so we are considering translating
the graph of 4x2 − 12x upward or downward. It opens upward, and to make it tangent to the
x-axis, it needs to be 4x2 − 12x + 9 = (2x − 3)2 . Hence we want 2a − 5 ≥ 9, or a ≥ 7.
2 Carl Joshua Quines

3. In how many ways can the letters of the word P AN ACEA be arranged so that the three As are
not all together?

(a) 540 (b) 576 (c) 600 (d) 720

Answer. (d) 720 .

Solution. We do complementary counting: count the total number of arrangements, and then
subtract the ones where the As are all together. The total number of arrangements is 7!
3! . There
are 7 letters, but we overcounted by a factor of 3!, because the As can be arranged in 3! ways
without changing the arrangement.

The number of arrangements where the As are all together is 5!. You can imagine combining the
As as a single, big letter AAA. Then there would be 5 letters to arrange. The final answer is
 
7! 7 · 6 · 5! 7·6
− 5! = − 5! = 5! − 1 = 720.
3! 3! 3!

4. How many ordered pairs of positive integers (x, y) satisfy 20x + 21y = 2021?

(a) 4 (b) 5 (c) 6 (d) infinitely many

Answer. (b) 5 .

Solution. From 2021 = 2000 + 21, we get the solution (100, 1). To produce another solution, note
that 20(−21) + 21(20) = 0. We can add this to both sides to get:
20(100) + 21(1) = 2021
20(−21) + 21(20) = 0
20(100 − 21) + 21(1 + 20) = 2021,
giving us the solution (79, 21). If we keep doing this, we get the solutions (58, 41), (37, 61), and
(16, 81). This gives 5 solutions.

To prove there are no other solutions, let’s say that (x, y) was another solution. We can subtract
the equation 20x + 21y = 2021 from 20(100) + 21(1) = 2021 to get 20(100 − x) + 21(1 − y) = 0.
Note that 21 is a factor of 0, and it is also a factor of 21(1 − y). This means 21 must be a factor
of 20(100 − x). Since it doesn’t share factors with 20, it has to be a factor of 100 − x. This limits
the possible x, and similarly y; from here we can show that there are only 5 solutions.

Remark. There is a general theory for solving linear Diophantine equations External-Link-Alt. Here’s another way to visualize
this. Consider the graph of 20x +21y = 2021 in the plane. It’s a line, and (100, 1) is one of the points on it. The
slope of the line is − 20
21 . Interpreting this as rise over run, it means that if we go up 20 units in the y -coordinate,
we go back 21 units in the x -coordinate. So another point on the line would be (100−21, 1+20) = (79, 21), and
if we think about drawing the line on graph paper, it wouldn’t cross any other points with integer coordinates.

5. Find the sum of all k for which x5 + kx4 − 6x3 − 15x2 − 8k 3 x − 12k + 21 leaves a remainder of
23 when divided by x + k.
3 5 3
(a) −1 (b) − (c) (d)
4 8 4
3 Carl Joshua Quines

3
Answer. (b) − .
4

Solution. From the remainder theorem, we know that the remainder when divided by x + k is
the result of substituting −k for x. Setting this to 23, we get
(−k)5 + k(−k)4 − 6(−k)3 − 15(−k)2 − 8k 3 (−k) − 12k + 21 = 23
−k 5 + k 5 + 6k 3 − 15k 2 + 8k 4 − 12k − 21 = 0
8k 4 + 6k 3 − 15k 2 − 12k − 21 = 0.
From Vieta’s formulas, we know that the sum of possible values of k is − 68 = − 34 .

6. In rolling three fair twelve-sided dice simultaneously, what is the probability that the resulting
numbers can be arranged to form a geometric sequence?
1 5 1 7
(a) (b) (c) (d)
72 288 48 288

7
Answer. (d) .
288

Solution. There are 123 possible ordered triplets of the results. We’ll count the number of triplets
that can be arraged to form a geometric sequence. We’ll work up from the possible common
ratios, and within each one, work from the smallest term:

• When the ratio is 1, the possibilities are (1, 1, 1), (2, 2, 2), . . . , (12, 12, 12). These are 12
possibilities in all.
• When the ratio is 32 , the only possibility that works is (4, 6, 9), giving 3! = 6 permutations.
• When the ratio is 2, the possibilites are (1, 2, 4), (2, 4, 8), (3, 6, 12). This gives 3 · 6 = 18 in
total.
• When the ratio is 3, the only possibility is (1, 3, 9), which gives 6 possible triplets.

Note that the ratio can’t be any other value; the ratio of the last to first term needs to be a
perfect square, and the only squares are 1, 4, 9. Thus the total number of triplets is 42 and the
answer is 12
42
3 = 288 .
7

n
7. How many positive integers n are there such that is a positive integer?
120 − 2n

(a) 2 (b) 3 (c) 4 (d) 5

Answer. (b) 3 .

Solution 1. Let’s say this integer is m. Then 120−2n

n
= m rearranges to 2mn − 120m + n = 0. We
now complete the rectangle by using Simon’s Favorite Factoring Trick External-Link-Alt. The −120m suggests
that it’s from (2m)(−60), so we want something like (2m + __)(n − 60), and we can fill it in
with 1. This adds an extra term 60 to both sides:
2mn − 120m + n + 60 = 60
(2m + 1)(n − 60) = 60.
4 Carl Joshua Quines

Now, 2m + 1 is an odd factor of 60. The odd factors of 60 are 1, 3, 5, 15. Of these, 2m + 1 can’t
be 1, because then m wouldn’t be positive, but the other ones work. This means there are 3
solutions for m, and for each, we can find the corresponding value of n.

Solution 2. It’s easier to work with a complicated numerator than a complicated denominator.
Since the fraction is positive and n is positive, we know 120 − 2n must also be positive. Let’s let
60 − n = m, and that way n = 60 − m. This means
n n 60 − m
= = .
120 − 2n 2(60 − n) 2m
Because the denominator is even, and this is an integer, the numerator must also be even. Hence
60 − m is even, and thus m is even, so m = 2` for some positive integer `. That makes
60 − m 60 − 2` 30 − `
= = .
2m 4` 2`
Through similar reasoning, ` must also be even, so letting ` = 2k,
30 − ` 30 − 2k 15 − k
= = .
2` 4k 2k
At this point, we can see that k is odd, but it’s now small enough to just check the remaining
choices. The choices k = 1, 3, and 5 work, which means there are 3 solutions.

Remark. Compare PMO 2018 Areas I.7 External-Link-Alt “Determine the area of the polygon formed by the ordered pairs
(x , y ) where x and y are positive integers that satisfy the equation x1 + y1 = 13 1
.” and PMO 2020 Areas I.9 External-Link-Alt
“A brick with dimensions 3 by a by b units is painted blue and then cut into 3ab unit cubes. Exactly 1/8 of
the cubes have all faces unpainted. Given a and b are positive integers, what is the volume of the brick?”

8. Three real numbers a1 , a2 , a3 form an arithmetic sequence. After a1 is increased by 1, the three
numbers now form a geometric sequence. If a1 is a positive integer, what is the smallest positive
value of the common difference?
√ √
(a) 1 (b) 2 + 1 (c) 3 (d) 5 + 2
√
Answer. (b) 2+1 .

Solution. Let’s say the common difference is d, and let’s write a for a1 . Then the arithmetic
sequence is a, a + d, a + 2d, and the geometric sequence is a + 1, a + d, a + 2d. Because this is a
geometric sequence, the ratios of consecutive terms have to be the same:
a+d a + 2d
=
a+1 a+d
2
(a + d) = (a + 2d)(a + 1)
a2 + 2ad + d2 = a2 + 2ad + a + 2d
d2 − 2d − a = 0.
We can now work up from values of a, starting from 1, and solve for d. Alternatively, we can use
the quadratic formula, or complete the square:
d2 − 2d + 1 − 1 − a = 0
(d − 1)2 = a + 1
√
d = 1 ± a + 1.
5 Carl Joshua Quines
√
Given that a is a positive integer, we can see the minimum positive value for d is 2 + 1.

9. Point G lies on side AB of square ABCD and square AEF G is drawn outwards ABCD, as
shown in the figure below. Suppose that the area of triangle EGC is 1/16 of the area of pentagon
DEF BC. What is the ratio of the areas of AEF G and ABCD?

E F

G B
A

D C

(a) 4 : 25 (b) 9 : 49 (c) 16 : 81 (d) 25 : 121

Answer. (a) 4 : 25 .

Solution 1. The tricky part is finding the area of EGC. It’s a good idea to consider the possible
bases to compute the area from. Here, EG is a promising base. We now need the height from C
to EG. In fact, this is equal to the height from A to EG.

E F

G B
A

D C

To see this, draw AC. Note that AC and EG are parallel lines. The height from C to EG is
thus the distance between these two parallel lines, which is equal to the height from A to EG.
So triangles EGC and EGA have the same area! Now we can compute. Let the smaller square
have side length x and the larger square have side length y. Then
x2
[EGC] [EAG] 2 x2 1
= = x(y−x)
= 2 2
= .
[DEF BC] [AEF G] + [GF B] + [ABCD] x2 + + y2 x + xy + 2y 16
2

Cross-multiplying this last equation and factoring, we get

15x2 − xy − 2y 2 = 0
(3x + y)(5x − 2y) = 0.
6 Carl Joshua Quines

The case 3x + y = 0 isn’t possible, because then one of x and y would have to be negative. So
5x − 2y = 0, which means xy = 25 . Squaring this gives us the ratio of the areas, 4 : 25.

Solution 2. If the smaller square has side length x and the larger square have side length y,

x2 y(y − x) y(x + y) x2
[EGC] = [EAG] + [ABCD] − [GBC] − [EDC] = + y2 − − = ,
2 2 2 2
and the rest of the solution proceeds as in Solution 1.

Solution 3. We use Cartesian coordinates. Since only the ratio of the area matters, we can say
that the larger square has side length 1. Let a be the side length of the smaller square. Taking
D to be the origin, we get these coordinates:

E = (0, a + 1) F = (a, a + 1)

B = (1, 1)
A = (0, 1)
G = (a, 1)

D = (0, 0) C = (1, 0)

We can now use the shoelace formula External-Link-Alt to find the areas of EGC and DEF BC:
1
[EGC] =|0 · 1 + a · 0 + 1 · (a + 1) − a · (a + 1) − 1 · 1 − 0 · 0|
2
1 1
= −a2 = (a2 ).
2 2
1
[DEF BC] = |a · 1 − a · (a + 1) − 1 · (a + 1) − 1|
2
1 1
= −a2 − a − 2 = (a2 + a + 2).
2 2
Here, in the shoelace formula for [DEF BC], we only write the terms that don’t have a zero
factor. Each of the absolute values follow from a > 0. Now we can solve for a:
a2 1
2
= ⇐⇒ 15a2 − a − 2 = (3a + 1)(5a − 2) = 0,
a +a+2 16

so a = 25 , and the ratio of the areas is 4 : 25.

Remark. Compare with PMO 2020 Qualifying I.12 External-Link-Alt “In parallelogram ABC D , C D = 18. Point F lies inside
ABC D and AB and DF meet at E . If AE = 12 and the areas of F E B and F C D are 30 and 162, find the
area of triangle BF C .” and PMO 2017 Qualifying III.1 External-Link-Alt “A paper cut-out in the shape of an isosceles right
triangle is folded in such a way that one vertex meets the edge of the opposite side, and that the constructed
edges m1 and m2 are parallel to each other. If the length of the triangle’s leg is 2 units, what is the area of the
shaded region?”
7 Carl Joshua Quines

10. In how many ways can 2021 be written as a sum of two or more consecutive integers?

(a) 3 (b) 5 (c) 7 (d) 9

Answer. (c) 7 .

Solution. Let’s say that the consecutive integers begin with a, and there are n of them. From
the formula for the sum of an arithmetic series,
n
2021 = a + (a + 1) + · · · + (a + (n − 1)) = (2a + n − 1) ⇐⇒ 4042 = n(2a + n − 1).
2
Note that if n is odd, then n + (2a − 1) is an odd number plus an odd number, and is even.
Similarly, if n is even, the other factor is odd. Thus we need to write 4042 as a product of an
even and an odd number: one of them will be n, and the other will be 2a + n − 1. As long as
this is true, we can always find an integer a that works.

Note 4042 = 2 · 43 · 47. Because there’s only one factor of 2, if we pick any factor as n, the other
will be the opposite parity. For example, if n = 2 · 43, which is even, then 2a + n − 1 = 47, which
is odd. So any factor of 4042 corresponds to a way to write it as a sum of consecutive integers.
From a well-known formula, we know 4042 has 8 factors. Subtracting the case n = 1, because
the problem asks for “two or more consecutive integers”, we get the final answer, 7.

Remark. From here, it’s possible to solve “Which integers can be written as the sum of two or more consecutive
integers?” Because n and 2a + n − 1 need to have opposite parity, the ones that can’t are the powers of 2.

11. In quadrilateral
√ ABCD, ∠CBA
√ = 90◦ , ∠BAD = 45◦ , and ∠ADC = 105◦ . Suppose that
BC = 1 + 2 and AD = 2 + 6. What is the length of AB?
√ √ √ √
(a) 2 3 (b) 2 + 3 (c) 3 + 2 (d) 3 + 3
√
Answer. (c) 3 + 2.

Solution. Let E and F be the feet of the perpendiculars from D to AB, and C to DE, respectively.
Because ∠EAD = 45◦ , that means 4AED is a 45–45–90 triangle. Then ∠EDA = 45◦ , so
∠CDF = ∠ADC − ∠EDA = 105◦ − 45◦ = 60◦ , so 4DF C is a 30–60–90 triangle.

C 60◦
F

B E A
√ √
From the fact that 4AED is 45–45–90, we know that AE = ED = AD √ =
2
2 + 3. From
√ √
√
√
rectangle BEF C, BC = F E, so we can find DF = ED − F E = 2 + 3 − 1 + 2 = 3 − 1.
√
Then we use the fact that 4DF C is a 30–60–90 triangle
√ to
√ get

CF =√3−
3, which
√ from rectangle
BEF C is also BE. Finally, AB = AE + BE = 2 + 3 + 3 − 3 = 3 + 2.
8 Carl Joshua Quines

12. Alice tosses two biased coins, each of which has a probability p of obtaining a head, simultaneously
and repeatedly until she gets two heads. Suppose that this happens on the rth toss for some
integer r ≥ 1. Given that there is a 36% chance that r is even, what is the value of p?
√ √
7 2 2 3
(a) (b) (c) (d)
4 3 2 4
√
7
Answer. (a) .
4

Solution 1. The probability that Alice gets two heads on the rth toss is p2 , times the probability
she didn’t get two heads on any of the previous r − 1 tosses, which is 1 − p2 . Hence the

r−1

probability r is even is the total probability that r = 2, 4, 6, . . ., which is

1
3
5 (1 − p2 )p2
1 − p2 p2 + 1 − p2 p2 + 1 − p2 p2 + · · · = ,
1 − (1 − p2 )2

where we used the formula for an infinite geometric series. Setting it equal to 100 ,
36
we get

(1 − p2 )p2 36
2 2
=
1 − (1 − p ) 100
100 − 100p = 72 − 36p2 .
2

√
Hence p2 = 28
64 = 16 ,
7
so p = 4 .
7

Solution 2. Either r is even, which happens with probability 36%, or r is odd, which must
happen with probability 100% − 36% = 64%. These are related—r is even is just like r being
odd, if you started counting after the first flip. That is, the probability r is even is the probability
that Alice doesn’t get two heads in the first flip, times the probability that r is odd. This means
36 64
= (1 − p2 ) ,
100 100
√
and we can solve for p = 4 .
7

Remark. Solution 2 uses the fact that r , being a geometric random variable, is memoryless External-Link-Alt.

13. For a real number t, btc is the greatest integer less than or equal to t and {t} = t − btc is the
√
fractional part of t. How many real numbers between 1 and 23 satisfy bxc {x} = 2 x?

(a) 18 (b) 19 (c) 20 (d) 21

Answer. (a) 18 .

Solution 1. It helps to think about what the graph of bxc {x} looks like. Consider a given
interval, say, [2, 3). Here, bxc is always 2, while {x} goes [0, 1). So the graph is a line from 0 to 2.
√ √ √
In this interval, what does 2 x look like? It goes from 2 2 ≈ 2.82 to 2 3 ≈ 3.46. It’s also
√
increasing. So in the interval [2, 3), the value of 2 x is always at least 2.82. But the maximum
value of bxc {x} is 2. This means that there are no solutions in the interval [2, 3).
9 Carl Joshua Quines

Let’s look at a different example,

√ like [5, 6). √Again, the bxc {x} part would go from 0 to 5.
√
The 2 x part would go from 2 5 ≈ 4.47 to 2 6 = 4.90. That means that their graphs would
intersect at some point in the interval. Since both graphs are increasing, that means they also
intersect at only one point.

We can do similar reasoning for the rest of the intervals. Each of [1, 2), . . . , [4, 5) have no solutions,
while each of [5, 6), . . . , [22, 23) have one solution, giving 18 such real numbers.

Solution 2. Let n = bxc and d = {x}. Then

√
bxc {x} = 2 x
√
nd = 2 n + d
n2 d2 − 4n − 4d = 0
p
16 − 4(n2 )(−4n)
4±
d= 2
√ 2n
2 + 2 1 + n3
d= .
n2
Here, we use the quadratic formula to solve for d, and take the positive solution because d ≥ 0.
The value of n determines the value of d, and thus the value of x = n + d. Thus, we only need to
count the number of n that make d < 1:
√
2 + 2 1 + n3
<1
pn2
2 1 + n3 < n 2 − 2
4 + 4n3 < n4 − 4n2 + 4
n2 n2 − 4n − 4 > 0.

This becomes (n − 2)2 > 8, which is satisfied by each of n = 5, 6, . . . , 22, giving 18 solutions.
2021
14. Find the remainder when nn is divided by 5.
X

n=2

(a) 1 (b) 2 (c) 3 (d) 4

Answer. (d) 4 .

Solution. To find nn mod 5, we’ll simplify both the base and the exponent. The base can just be
taken mod 5. For the exponent, we know by Fermat’s little theorem External-Link-Alt that n4 ≡ 1 (mod 5), as
long as n isn’t 0. This means we only need to take the exponent mod 4, because if the exponent
is, say, 4k + 2, then n4k+2 ≡ n4 n2 ≡ 1 · n2 ≡ n2 (mod 5).

k

Because we’re taking the base mod 5 and the exponent mod 4, this means that n mod 20
completely determines the value of nn . So we only need to find the value of the first 20 numbers,
and then multiply by the number of times they appear in the sum.
11 22 33 44 55 11 22 33 44 0 1 4 2 1 0
6 6 77 88 9 9 1010 12 23 30 41 0 1 3 1 4 0
11 12 13 14 15 ≡ 3 ≡ (mod 5)
11 12 13 14 15 1 20 31 42 0 1 1 3 1 0
16 16 17 17 18 18 19 19 2020 10 21 32 43 0 1 2 4 4 0
10 Carl Joshua Quines

The sum of nn from 1 to 20 is thus 4. By our previous argument, this is also the sum of nn from
21 to 40, and from 41 to 60, and so on. Going from 1 to 2020, this repeats 101 times, so the sum
would be 404 ≡ 4 (mod 5). Finally, note that the sum starts from 2 and ends at 2021, so we
have to subtract 11 and add 20212021 . This means the final answer is 4 − 1 + 1 ≡ 4 (mod 5).

Remark. An interesting observation is the columns of the previous table sum to either 4 or 0 mod 5. This
follows from the fact that, for a prime p , a0 + a1 + · · · + ap −1 ≡ 0 (mod p ) when a 6≡ 1, which can be proven
using the geometric series formula.

15. In the figure below, BC is the diameter of a semicircle centered at O, which intersects AB and
AC at D and E respectively. Suppose that AD = 9, DB = 4, and ∠ACD = ∠DOB. Find the
length of AE.

C O B

117 39 √ √
(a) (b) (c) 2 13 (d) 3 13
16 5

39
Answer. (b) .
5

Solution 1. Our strategy is to use power of a point on A, and to do that, we want to find AC.
The key observation is that, because ∠DCB is an inscribed angle, its measure is half of ∠DOB,
and thus, half of ∠ACD. This encourages us to draw the angle bisector of ∠ACD, so let F be
on segment AD such that CF bisects ∠ACD.

F
E

C O B

Now if we let ∠ACD = ∠DOB = 2θ for some θ, then ∠DCB = 12 ∠DOB = θ, and ∠ACF =
∠F CD = 12 ∠ACD = θ. Because BC is a diameter of a semicircle, it follows ∠CDB = ∠CDF =
90◦ , and hence 4F CD ∼
= 4BCD by ASA. Thus DF = DB = 4 and AF = AD − DF = 5.
11 Carl Joshua Quines

We now apply the angle bisector theorem on 4ACD with angle bisector CF . This tells us that
CD = F D = 4 . Hence, let AC = 5x, CD = 4x for some x. Using the Pythagorean theorem
AC AF 5

on right 4ADC, we get CD2 + AD2 = AC 2 , or (4x)2 + 92 = (5x)2 . It follows that x = 3

and AC = 15. Finally, applying power of a point on A, we get that AD · AB = AE · AC, or
9 · 13 = AE · 15, and hence AE = 39
15 .

Solution 2. We pick up from Solution 1, after deducing ∠ACD = 2θ, ∠DCB = θ and
∠CDB = ∠CDA = 90◦ . Then using right 4ADC and 4BDC, we get that CD = tan9 2θ = tan4 θ .
From the tangent double angle formula,
9 4
2 tan θ
=
1−tan2 θ
tan θ
4 · 2 tan θ
1 − tan2 θ = ,
9 tan θ
and hence tan θ = 13 . (We discard tan θ = 0 and tan θ = − 13 because θ is acute.) Hence CD = 12,
and from the Pythagorean theorem, AC = 15, and the rest proceeds as in Solution 1.

Solution 3. There’s a solution that involves no geometric insight, although it is a lot of

algebra. We pick up from Solution 1, after noticing ∠CDB = ∠CDA = 90◦ . Somehow we
have to use the fact that ∠ACD = ∠DOB. But 4ADC is right, so we can find sin ∠ACD
using AC. Then we can use the cosine law on ∠DOB to get cos ∠DOB, and then try to use
sin2 ∠ACD + cos2 ∠DOB = 1.

Let AC = x. Let AC = x. From right 4ADC, sin ∠ACD = AD AC = x . Now let OB = OC =

9

OD = r, the radius of the semicircle. If we use the cosine law on 4DOB, we can find

DB 2 = DO2 + OB 2 − 2 · DO · OB · cos ∠DOB

r2 + r2 − 42
cos ∠DOB =
2·r·r
8
cos ∠DOB = 1 − 2 .
r
So sin2 ∠ACD + cos2 ∠DOB = 1 relates r and x, but to solve for them, we need one more way
to relate r and x. Well, we can use the Pythagorean theorem on right 4ADC and 4BDC. Note
that CD2 = BC 2 − BD2 = AC 2 − AD2 , so (2r)2 − 42 = x2 − 92 , giving us r2 = x −65
4 . Finally,
2

sin2 ∠ACD + cos2 ∠DOB = 1

 2 
8 2

9
+ 1− 2 =1
x r
 2
81 32
+ 1− 2 =1
x2 x − 65
81 64 1024
2
+1− 2 + 4 =1
x x − 65 x − 130x2 + 4225
17x4 − 5346x2 + 342225
=0
x6 − 130x4 + 4225x2
17x2 − 392 x2 − 152 = 0.



12 Carl Joshua Quines

The positive possibilities are x = √3917 and x = 15. Of these, the former can be ruled out as being
too small, or because it doesn’t lead to an answer in the choices. The latter gives us AC = 15,
and the rest proceeds as in Solution 1.

Remark. In Solution 1, from AC : AD = 5 : 4, we can deduce that 4AC D is 3–4–5, and then get AC = 15.

PART II. All answers are positive integers. Do not use commas if there are more than 3 digits, e.g.
type 1234 instead of 1, 234. A positive fraction a/b is in lowest terms if a and b are both positive
integers whose greatest common factor is 1. Each correct answer is worth five points.

16. Consider all real numbers c such that |x − 8| + 4 − x2 = c has exactly three real solutions.
The sum of all such c can be expressed as a fraction a/b in lowest terms. What is a + b?

Answer. 93 .

Solution. It helps to think about the graph of |x − 8| + 4 − x2 = y to get a sense of what the y
would be. When does it increase and decrease? To analyze it, we can split it up based on the
value of x. We’ll split on x = −2, 2, 8, because these are where the absolute values would change.

• When x ≤ −2, it’s (8 − x) + x2 − 4 = y, or y = x2 − x + 4. This is a parabola, pointing

up, whose vertex is at x = 12 , so it just decreases in this interval.

• When −2 ≤ x ≤ 2, it’s (8 − x) + 4 − x2 = y, or y = −x2 − x + 12. This is a parabola,

pointing down, whose vertex is at x = − 12 . In this interval, it increases, then decreases.

• When 2 ≤ x ≤ 8, it’s (8 − x) + x2 − 4 = y, or y = x2 − x + 4. This is the parabola we

saw earlier, which means that it just increases in this interval.

• Finally, when x ≥ 8, it’s (x − 8) + x2 − 4 = y, or y = x2 + x − 12. Again, this is a parabola

pointing up with vertex at − 12 , so it continues increasing in this interval.

The graph changes direction thrice, at x = −2, − 12 , and 2. We can compute the y values at these
points as 10, 49
4 , 6. Using this information, we can sketch what the graph would look like, and
determine that the y that produce three solutions are 10 and 49 4 . Their sum is 4 , so the answer
89

is 89 + 4 = 93.

17. Find the smallest positive integer n for which there are exactly 2323 positive integers less than
or equal to n that are divisible by 2 or 23, but not both.

Answer. 4644 .

Solution. The number of integers at most that are divisible by is 2 , and similarly, the
n
n 2
number divisible by 23 is 23 . But this double-counts the numbers divisible by both. To not
n

count those numbers, we can subtract 2 46n

. So we’re looking for the smallest n such that


jnk j n k jnk
+ −2 = 2323.
2 23 46
Let’s over-estimate n and then go down bit-by-bit until we find the right one. bxc is always at
most x, so we want n2 + 23
n
46 ≤ 2323, which solves to n ≤ 4646. Plugging into the original
− 2n
equation, we see that 4646 works, but is it the smallest? We can check that 4645 and 4644 both
work, but 4643 gives 2322, which is too small. So the answer must be 4644.
13 Carl Joshua Quines

18. Let P (x) be a polynomial with integer coefficients such that P (−4) = 5 and P (5) = −4. What
is the maximum possible remainder when P (0) is divided by 60?

Answer. 41 .

Solution 1. By the remainder theorem, we know that P (x) = (x + 4)Q(x) + 5, where Q(x) is
some other polynomial with integer coefficients. (To see this, consider substituting x = −4.) We
want P (5) = −4, so substituting x = 5 gives

P (5) = (5 + 4)Q(5) + 5
Q(5) = −1
Q(x) = (x − 5)R(x) − 1,

where we again use the remainder theorem. Plugging it back into the first equation and
substituting x = 0,

P (x) = (x + 4)Q(x) + 5
P (x) = (x + 4) ((x − 5)R(x) − 1) + 5
P (0) = 4 (−5R(0) − 1) + 5
P (0) = −20R(0) + 1.

Now R(0) is some constant. Modulo 60, the value of −20R(0) + 1 is either 1, −20 + 1, or −40 + 1.
These are 1, 41, and 21 modulo 60, so the largest possible remainder is 41.

Solution 2. We use the fact that a−b | P (a)−P (b) to get 4 | P (0)−P (−4) and −5 | P (0)−P (5).
This means P (0) ≡ P (−4) ≡ 1 (mod 4) and P (0) ≡ P (5) ≡ 1 (mod 5). This means P (0) ≡ 1
(mod 20), and the possible values of P (0) mod 60 would be 1, 21, 41, the largest of which is 41.

19. Let 4ABC be an equilateral triangle with side length 16. Points D, E, F are on CA, AB, and
BC, respectively, such that √
DE ⊥√ AE, DF√ ⊥ CF , and BD = 14. The perimeter of 4BEF can
be written in the form a + b 2 + c 3 + d 6, where a, b, c, and d are integers. Find a + b + c + d.

Answer. 31 .

Solution 1. Let AE = x and√CF = y. Then note that√4DAE and 4DCF are 30–60–90
triangles, so AD = 2x, DE = x 3, DC = 2y, and DF = y 3.

A
E
D

B F C

From AD + DC = AC we get 2x + 2y = 16, or x + y = 8. We also have BE = 16 − x and

BF = 16 − y. So BE + BF = 32 − (x + y) = 24, and the only thing we need to find is EF .
14 Carl Joshua Quines

Applying Stewart’s theorem on cevian BD, we get

AC · AD · DC + BD2 · AC = BC 2 · AD + AB 2 · DC
16 · 2x · 2y + 142 · 16 = 162 · 2x + 162 · 2y
4xy + 196 = 32(x + y)
xy = 15.

Because ∠BED = ∠BF D = 90◦ , it follows ∠BED + ∠BF D = 180◦ and BEDF is cyclic. We
can now apply Ptolemy’s theorem to find EF :

BD · EF = BF · DE + BE · DF
 √   √ 
14 · EF = (16 − y) x 3 + (16 − x) y 3
√ √
14 · EF = 16 3(x + y) − 2xy 3
√
EF = 7 3.
√
Hence the perimeter is BE + BF + EF = 24 + 7 3 and the answer is 24 + 0 + 7 + 0 = 31.

Solution 2. We proceed from the first paragraph of Solution 1. Using the Pythagorean theorem
on right 4AED, we get that ED2 + EB 2 = BD2 , or
 √ 2
x 3 + (16 − x)2 = 142
3x2 + x2 − 32x + 256 = 196.

This is (x − 5)(x − 3) = 0. Hence x = 3, 5, and as x + y = 8, we get (x, y) = (3, 5) or (5, 3).

These are symmetric about swapping A and C, so we know that both choices will give the same
perimeter. From here, we can proceed using Ptolemy’s, as in Solution 1, to find EF .

Solution 3. Alternatively, we could also use the cosine law on, say, 4DEF . From quadrilateral
BEDF we get ∠EDF = 120◦ , hence

DE 2 + DF 2 − 2 · DE · DF · cos ∠EDF = EF 2
 √ 2  √ 2  √  √ 
x 3 + y 3 − 2 x 3 y 3 cos 120◦ = EF 2
 
2 2 1
3x + 3y − 6xy − = EF 2
2
3 (x + y)2 − xy = EF 2 .

If we knew x + y and xy from Solution 1, or if we knew the values of x and y from Solution 2,
we can now find EF . It’s also possible to use the cosine law on 4BEF itself.
√
Solution 4. Let G be the foot of the perpendicular from B to AC. Then BG = 8 3, because
it’s the height of an isosceles triangle with side length 16. Using the Pythagorean theorem on
right 4BGD, we get DG2 = BD2 − BG2 = 196 − 192 = 4, so DG = 2. From here we can get
AD = 6 and DC = 10, and we can proceed as in Solution 1.

20. How many subsets of the set {1, 2, 3, . . . , 9} do not contain consecutive odd integers?

Answer. 208 .
15 Carl Joshua Quines

Solution. Such a set can contain any subset of {2, 4, 6, 8}, and then a subset of {1, 3, 5, 7, 9} with
no two consecutive odd integers. There are 24 ways to pick a subset of the even numbers. Then
we want to pick a subset of 5 things, no two of which are consecutive. It’s well-known that this
is F7 , the seventh Fibonacci number, which is 13. So the answer is 24 · 13 = 208.

Why is Fn+2 the number of ways to choose a subset of {1, 2, . . . , n} containing no consecutive
integers? Say there are an such subsets. We count based on whether they contain n or not:

• If it doesn’t contain n, then it can be a subset of {1, 2, . . . , n − 1} with no consecutive

integers. So there are an−1 subsets that don’t contain n.
• If it does contain n, then it can’t contain n − 1. So it’s a subset of {1, 2, . . . , n − 2}, with
no consecutive integers, with n added in. There are an−2 of these, so there are an−2 subsets
that do contain n.

This means an = an−1 + an−2 . Now, a0 = 1, because there’s only one way, the empty set. And
a1 = 2, because it’s either the whole set or the empty set. This means a0 = F2 and a1 = F3 , and
from the recursion, we can prove an = Fn+2 .

Remark. A common interpretation of the Fibonacci numbers is the number of ways to tile a 2 × n rectangle
with 2 × 1 dominoes. It’s possible to construct a bijection from this to the number of subsets of {1, 2, . . . , n − 1}
with no consecutive integers. See A000045 External-Link-Alt for more interpretations.

Remark. It’s also possible to construct the recursion directly. Let an be the number of subsets of {1, 2, . . . , n}
with no consecutive odd integers. Then we can show, by considering whether or not 1 is part of the subset,
that an = 2an−2 + 4an−4 .

21. For a positive integer n, define s(n) as the smallest positive integer t such that n is a factor of
t!. Compute the number of positive integers n for which s(n) = 13.

Answer. 792 .

Solution. An important fact about the factorials is that t! = t · (t − 1)!. So any factor of (t − 1)!
is also a factor of t!. By induction, if s < t, then any factor of s! is also a factor of t!. This means
that if s(n) = 13, then n is a factor of 13!, and it isn’t a factor of 12!. If τ (n) is the number of
factors of n, the answer must be τ (13!) − τ (12!). We can compute that 12! = 210 · 35 · 52 · 71 · 111 ,
and from a well-known formula, we get that τ (12!) = (1 + 10)(1 + 5)(1 + 2)(1 + 1)(1 + 1) = 792.
Similarly τ (13!) = 1584, so the answer is 1584 − 792 = 792.

Remark. From multiplicativity External-Link-Alt, τ (13!) − τ (12!) = τ (13)τ (12!) − τ (12!) = τ (12!). So we don’t need to find
the factors of 13!, although it’s not hard to do so if we already have the factors of 12!.

22. Alice and Bob are playing a game with dice. They each roll a die six times, and they take the
sums of the outcomes of their own rolls. The player with the higher sum wins. If both players
have the same sum, then nobody wins. Alice’s first three rolls are 6, 5, and 6, while Bob’s first
three rolls are 2, 1, and 3. The probability that Bob wins can be written as a fraction a/b in
lowest terms. What is a + b?

Answer. 3895 .
16 Carl Joshua Quines

Solution. Let’s say that Alice’s next three rolls are a, b, c and Bob’s next three rolls are d, e, f . We
count the number of possibilities for a, b, c, d, e, f such that Bob wins. We want 6+5+6+a+b+c <
2 + 1 + 3 + d + e + f , or

a + b + c + (7 − d) + (7 − e) + (7 − f ) < 10.

Here, we write 7 − d so that it becomes a positive integer; because d is between 1 and 6, 7 − d

would also be between 1 and 6. We can check that a, b, c are also forced to be between 1 and 6,
otherwise the sum would be more than 10.

So we have six positive integers that sum to some integer less than 10. We can count the
number
of possibilities with balls and urns External-Link-Alt. When the sum is n, the number of solutions is n−1
5 . Since
the sum can be either 6, 7, 8, or 9, the total number of solutions is
         
6−1 7−1 8−1 9−1 9
+ + + = = 84,
5 5 5 5 6

where we use the hockeystick identity External-Link-Alt. The number of possible rolls is 66 , so the probability is
84
66
7
= 3888 and the answer is 7 + 3888 = 3895.

23. Let ABC be an isosceles triangle with a right angle at A, and suppose that the diameter of its
circumcircle Ω is 40. Let D and E be points on the arc BC not containing A such that D lies
between B and E, and AD and AE trisect ∠BAC. Let I1 and I2 be the incenters √ of 4ABE
√ √
and 4ACD respectively. The length of I1 I2 can be expressed in the form a + b 2 + c 3 + d 6,
where a, b, c, and d are integers. Find a + b + c + d.

Answer. 20 .

Solution 1. We have an incenter and a circumcircle, so the key idea is to use the incenter–
excenter lemma External-Link-Alt. Because AD and AE trisect ∠BAC, it follows AD is the angle bisector of
∠BAE. Hence, by the lemma, DI1 = DB = DE. Similarly, EI2 = EC = DE.

I1 I2
B C

D J1 J2 E

Now we claim that DI1 = BD = DE = EC = EI2 = 20, half the radius of the circumcircle.
One way to see this is to imagine rotating quadrilateral BDEC 180◦ about the center of the
circle, which produces a regular hexagon, which is well-known to have side length equal to the
radius of the circle. Another way is to note that ∠DAE is one-third of ∠BAC, so it’s 30◦ , and
by the extended law of sines DE = 2 · 20 · sin 30◦ = 20.
17 Carl Joshua Quines

Let J1 and J2 lie on DE such that I1 J1 and I2 J2 are both perpendicular to DE. Finally,

I1 I2 = J1 J2 = DE − DJ1 − EJ2 ,

so using right triangles DJ1 I1 and EJ2 I2 ,

√ √ ! √ √ !
6− 2 6− 2 √ √
DE − DI1 cos 75◦ − EI2 cos 75◦ = 20 − 20 − 20 = 20 − 10 6 + 10 2,
4 4

so the final answer is 20 + 10 + 0 − 10 = 20.

Solution 2. We use Cartesian coordinates. Center Ω on the origin O and choose A = (0, 20),
B = (−20, 0), and C = (20, 0). By symmetry, we know that I1 I2 is parallel to BC and is bisected
by OA, so it’s enough to just find the distance of, say, I2 to OA, and then double it. But this is
just the x-coordinate of I2 !

To find the x-coordinate of I2 , we can use the formula for the incenter of a triangle in Cartesian
coordinates: it’s the weighted average of the triangle’s vertices, where each vertex is weighted
by the length of the opposite side. This means we need to coordinates of D. Because ∠BAD =
1 ◦ , we know ∠BOD = 2∠BAD = 60◦ , and using some trigonometry we get
3 ∠BAC = 30 √
D = −10, −10 3 .

√
Now we √ need the side lengths
√ of √
4ACD. Using the distance formula, we get AC = 20 2,
CD = 20 3 and DA = 10 2 + 10 6. Then the x-coordinate of I2 would be
√
√ √
√
√ √ √ √
0 20 3 + 20 10 2 + 10 6 − 10 20 2 200 2 + 200 6 − 200 2 20 6
√ √ √ √ = √ √ √ = √ √ √ .
20 3 + 10 2 + 10 6 + 20 2 30 2 + 20 3 + 10 6 3 2+2 3+ 6
√ √ √
We want to write
√ this√ in the√form a + b 2 + c 3 + d 6. The simplest way is to do this is equate
it with a + b 2 + c 3 + d 6, and cross-multiply to get
√ √ √ √
20 6 = (6b + 6c + 6d) + (3a + 3c + 6d) 2 + (2a + 2b + 6d) 3 + (a + 2b + 3c) 6.

Solving the
√ system
√ of equations gives (a, b, c, d) = (10, 5, 0, −5). Hence the
√ x-coordinate
√ of I2
is 10 + 5 2 − 5 6, which means the length of I1 I2 is double that, 20 + 10 2 − 10 6, and the
answer is 20 + 10 + 0 − 10 = 20.
√ √ √ √
Remark. In Solution 2, the fact that we could rewrite 3√2+2 20√6 √
in the form a + b 2 + c 3 + d 6 is a
√ √
3+ 6 √
fancy consequence of the fact that Q 2, 3 is a field extension External-Link-Alt of Q. This is why the 6 is necessary,
√ √
for example, 1+√42+√3 = 2 + 2 − 6. Sorry, it’s been a while since I’ve done number theory, and I need to
cite little facts like these to prove to myself that I can still do it.

24. Find the number of functions f from the set S = {0, 1, 2, . . . , 2020} to itself such that, for all
a, b, c ∈ S, all three of the following conditions are satisfied:

(i) If f (a) = a, then a = 0;

(ii) If f (a) = f (b), then a = b; and
(iii) If c ≡ a + b (mod 2021), then f (c) ≡ f (a) + f (b) (mod 2021).
18 Carl Joshua Quines

Answer. 1845 .

Solution. The important condition is (iii). We can show through induction that, because
f (a) ≡ f (a − 1) + f (1) mod 2021, then f (a) ≡ af (1) (mod 2021). If we determine the value of
f (1), we know what the entire function is. Now we consider the other two conditions and what
restrictions they give for f (1).

Condition (i) means that for each a 6= 0, we want af (1) 6= a. Because f (a) ≡ af (1) (mod 2021),
this means af (1) 6≡ a (mod 2021), or a (f (1) − 1) 6≡ 0 (mod 2021). Now, note that this has
to be true for every a 6= 0, so we can choose the a we want. Picking a = 43, we get that
43 (f (1) − 1) 6≡ 0 (mod 2021). Now we can divide by 43 to get f (1) − 1 6≡ 0 (mod 47). This
means f (1) 6≡ 1 (mod 47). Similarly, by picking a = 47, we can prove that f (1) 6≡ 1 (mod 43).

Let’s look at condition (ii) with b = 0. Note that f (0) = 0f (1) = 0, regardless of the choice of
f (1). This means that, for each a 6= 0, we want f (a) 6= 0, or af (1) 6≡ 0 (mod 2021). Again, we
choose a = 43 to show f (1) 6≡ 0 (mod 47), and a = 47 to show f (1) 6≡ 0 (mod 43).

So, we have that f (1) 6≡ 0, 1 (mod 43) and f (1) 6≡ 0, 1 (mod 47). The values of f (1) mod 43 and
47 completely determine it mod 2021. Since f (1) ≡ 2, . . . , 42 (mod 43), it has 41 possibilities
mod 43, and similarly 45 possibilities mod 47, it has 41 · 45 = 1845 possibilities mod 2021.

Remark. To give a full proof, we have to show that if f (1) 6≡ 0, 1 (mod 43, 47), then the function satisfies all
three conditions. This isn’t necessary for the contest and follows a similar idea to the solution above, but it
does need to be shown when the proof is required.

25. A sequence {an } of positive real numbers is defined by a1 = 1 and for all integers n ≥ 1,
√
an n2 + n
an+1 = p .
n2 + n + 2a2n

Compute the sum of all positive integers n < 1000 for which an is a rational number.

Answer. 131 .
√ √ √ √
Solution. The first few terms are 1, 22 , 721 , 510 , 1365 , . . .. There’s a lot of square roots here, so
let’s try squaring all the terms. This gives us the numbers 1, 12 , 37 , 25 , 13
5
, . . .. Now it looks like the
numerators are 1, 2, 3, . . .. In fact, we can rewrite the squares of the terms as 11 , 24 , 37 , 10 4 5
, 13 , . . . ,
so we can guess that an = 3n−2 . In fact, we can prove this with induction. The base case is
q
n

clear, and for the inductive step,

√ √ p s
an n2 + n n2 + n n2 (n + 1)
r
n n+1
p = ·q =p = ,
n2 + n + 2a2n 3n − 2 n2 + n + 2 · n 2
n (3n + 1) 3 (n + 1) − 2
3n−2

as desired. Now we need to find all positive integers n < 1000 for which is a rational
q
n
3n−2
number. For this to happen, the fraction 3n−2 ,
when put in simplest terms, must have a perfect
n

square in the numerator and a perfect square in the denominator. To put it in simplest forms,
we divide n and 3n − 2 by their GCD. By a property of GCD, we get that (n, 3n − 2) =
(n, 3n − 2 − 3n) = (n, −2), so the GCD is either 2 or 1.
19 Carl Joshua Quines

If the GCD is 2, that means n = 2n0 for some n0 , and 3n−2 = 6n2n0 −2 = 3nn0 −1 , which is now in
n 0 0

simplest terms. Now the numerator must be a perfect square, so let’s say n0 = x2 for some integer
x. That means 3n0 − 1, which is 3x2 − 1, must also be a perfect square. But modulo 3, this is
−1, and −1 is not a perfect square modulo 3. So 3n0 − 1 can’t be a perfect square, and there are
no solutions in this case.

The remaining case is when the GCD is 1. Then both n and 3n − 2 are perfect squares. Letting
n = x2 for some integer x, we get that 3x2 − 2 is also a perfect square. At this point we can just
try some values of x until we get one that works: x = 1 works, so does x = 3, and x = 11. Since
n = x2 , we only have to check up to 31, because 322 > 1000. We find that the only ones that
work are 1, 3, 11, which means n = 1, 9, 121 are the solutions, and they have sum 131.

Remark. The equation 3x 2 − 2 = y 2 is a generalized Pell equation External-Link-Alt; it can be written in the more familiar
form x 2 − 3y 2 = −2. The morally correct method is to use the solution (x , y ) = (1, 1), and then generate
solutions through multiplying by the solutions of the regular Pell equation u 2 − 3v 2 = 1. For example, the Pell
equation has solution (u , v ) = (2, 1), and note that
 √  √  √ √ √
x + y 3 u + v 3 = (1 + 3)(2 + 3) = 5 + 3 3,

and indeed, (x , y ) = (5, 3) is a solution to x 2 − 3y 2 = −2. We can get the solutions of the regular Pell equation
u 2 − 3v 2 = 1 through computing the powers of the fundamental solution External-Link-Alt (u , v ) = (2, 1). For example,
√ √
(2 + 3)3 = 26 + 15 3, and (u , v ) = (26, 15) is a solution to u 2 − 3v 2 = 1.

With thanks to Nathanael Joshua Balete and Richard Eden for comments.