USA Mathematical Talent Search
Round 1 Solutions
Year 33 — Academic Year 2021–2022
www.usamts.org
1/1/33. 33 counters are shown in the left grid below. Choose a counter to start at and remove
it from the grid. At each subsequent step, choose a direction (up, down, left, or right), move
along the grid line from your current position to the nearest counter in that direction, and
remove that counter. You cannot choose a direction that reverses your previous one (e.g.,
left then right is not allowed). Your goal is to pick up all 33 counters in a single sequence of
steps. When you find the right sequence, write the numbers 1 to 33 on the counters so that
N is written on the N th counter you removed.
A smaller example of a solved grid is shown to the right below. (Note that the final move
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer
from 8 to 9 is possible because counters 3, 4, and 5 have been removed in earlier steps.)
There is a unique solution, but you do not need to prove that your answer is the only
one possible. You merely need to find an answer that satisfies the constraints above. (Note:
In any other USAMTS problem, you need to provide a full proof. Only in this problem is
an answer without justification acceptable.)
7 8
6 5
4
3 2
1
9
Solution
28 27
33
26 25
29 16 15
30 31 32 17 14
18 13 12 11 19
24 23
10 20
7 6
8 9 21 22 5
4
3
2
1
USA Mathematical Talent Search
Round 1 Solutions
Year 33 — Academic Year 2021–2022
www.usamts.org
2/1/33. Find, with proof, the minimum positive integer n with the following property: for
any coloring of the integers {1, 2, . . . , n} using the colors red and blue (that is, assigning the
color “red” or “blue” to each integer in the set), there exist distinct integers a, b, c between
1 and n, inclusive, all of the same color, such that 2a + b = c.
Solution
Call a triple (a, b, c) good if 2a + b = c and all of a, b, c are the same color.
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer
We can color {1, . . . , 14} as 1–3 red, 4–12 blue, and 13–14 red, and there is no good
triple:
• Choosing {a, b} ∈ {1, 2, 3} gives 4 ≤ 2a + b ≤ 8, so a and b are red but c = 2a + b is
blue.
• Choosing {a, b} ∈ {4, 5, . . . , 12} gives 2a + b ≥ 13, so a and b are blue but c = 2a + b
is either red or too big.
• Choosing either a ≥ 13 or b ≥ 13 gives 2a + b ≥ 15, so c = 2a + b is too big.
Thus we must have n > 14 to have a good triple.
We prove that any coloring of {1, . . . , 15} produces a good triple, by contradiction. Sup-
pose we have a coloring of {1, . . . , 15} for which there is no good triple. We have two cases:
Case 1: 1 and 2 are the same color. Without loss of generality, suppose 1 and 2
are both red. Using our assumption that there are no good triples, the following colors are
required:
• 4 must be blue, because otherwise (1, 2, 4) would be a red good triple.
• 5 must be blue, because otherwise (2, 1, 5) would be a red good triple.
• 13 must be red, because otherwise (4, 5, 13) would be a blue good triple.
• 15 must be blue, because otherwise (1, 13, 15) would be a red good triple.
• 7 must be red, because otherwise (4, 7, 15) would be a blue good triple.
• 3 must be blue, because otherwise (2, 3, 7) would be a red good triple.
• 9 must be blue, because otherwise (1, 7, 9) would be a red good triple.
And now we have a contradiction, because (3, 9, 15) is a blue good triple.
USA Mathematical Talent Search
Round 1 Solutions
Year 33 — Academic Year 2021–2022
www.usamts.org
Case 2: 1 and 2 are different colors. Without loss of generality, suppose 1 is red
and 2 is blue. If any 5 ≤ x ≤ 13 is colored red, then since the triple (1, x, x + 2) can-
not be a red good triple, we see that x + 2 must be colored blue. But then, since the
triple (2, x − 2, x + 2) cannot be a blue good triple, we see that x − 2 must be colored red.
But then (1, x − 2, x) is a red good triple, which is not allowed. So all of {5, . . . , 13} must
be colored blue. And now we have a contradiction: in particular (2, 5, 9) is a good blue triple.
So, by contradiction, any coloring of {1, . . . , 15} must produce a good triple, and therefore
our final answer is n = 15 .
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer
USA Mathematical Talent Search
Round 1 Solutions
Year 33 — Academic Year 2021–2022
www.usamts.org
3/1/33. Let S be a subset of {1, 2, . . . , 500} such that no two distinct elements of S have a
product that is a perfect square. Find, with proof, the maximum possible number of elements
in S.
Solution
Let f (n) denote the square-free part of n: that is, f (n) = nd , where d is the greatest
divisor of n that is a perfect square. We note that mn is a perfect square if and only if
f (m) = f (n). In particular, every element s ∈ S must have a different value of f (s). So S
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer
is in 1-1 correspondence (via f ) with a subset of T , where T is the set of square-free pos-
itive integers up to 500. Thus |S| ≤ |T |, and choosing |S| = |T | provides a maximal example.
So it remains to compute |T |. It’s easier to count non-square-free integers up to 500:
these are multiples of perfect squares greater than 1. We count multiples of p2 for p prime:
There are b 500
4
c = 125 multiples of 22 = 4.
There are b 500
9
c = 55 multiples of 32 = 9. However, b 55
4
c = 13 of these are also multiples
of 4, which we have already counted. Thus, we get 55 − 13 = 42 new numbers that are
multiples of 9.
There are b 500
25
c = 20 multiples of 52 = 25. However, b 20
4
c = 5 of these are also multiples
20
of 4, and b 9 c = 2 of these are also multiples of 9. Thus, we get 20 − 5 − 2 = 13 new numbers
that are multiples of 25.
There are b 500
49
c = 10 multiples of 72 = 49. However, b 10
4
c = 2 of these are also multiples
10
of 4, and b 9 c = 1 of these is also a multiple of 9. Thus, we get 10 − 2 − 1 = 7 new numbers
that are multiples of 49.
There are b 500
121
c = 4 multiples of 112 = 121. However, b 44 c = 1 of these is also a multiple
of 4. Thus, we get 4 − 1 = 3 new numbers that are multiples of 121.
500
There are b 169 c = 2 multiples of 132 = 169.
There is b 500
289
c = 1 multiple of 172 = 289.
There is b 500
361
c = 1 multiple of 192 = 361.
The next prime square is 233 = 529 > 500, which is too large, so these are all the multi-
ples of prime squares.
So our final answer is |T | = 500−(125+42+13+7+3+2+1+1) = 500−194 = 306 .
USA Mathematical Talent Search
Round 1 Solutions
Year 33 — Academic Year 2021–2022
www.usamts.org
4/1/33. Let m, n, k be positive integers such that k ≤ mn. Let S be the set consisting of the
(m + 1)-by-(n + 1) rectangular array of points on the Cartesian plane with coordinates (i, j)
where i, j are integers satisfying 0 ≤ i ≤ m and 0 ≤ j ≤ n. The diagram below shows the
example where m = 3 and n = 5, with the points of S indicated by black dots:
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer
Prove that there exist points A, B, C in S such that the area of 4ABC is k2 .
Solution
Let A = (0, 1) and B = (m, 0). If we let D = (k, 1), then [ABD] = k2 (using AD = k as
the base and the height from B to AD of 1), but if k > m then D 6∈ S.
Note that the slope of AB is − m1 , so if C = (k − mt, 1 + t) for any t, then CD k AB,
and hence [ABC] = [ABD] = k2 , as desired.
So if suffices to prove that we can always choose such a t so that C ∈ S, which means
that our t must satisfy
0 ≤ k − mt ≤ m, (1)
0 ≤ 1 + t ≤ n. (2)
We claim that the choice t = b k−1
m
c works. (Note that if k ≤ m, then t = 0 and C = D,
which works since because in this case D ∈ S to begin with.)
Since 1 ≤ k ≤ mn, we have 0 ≤ t ≤ n − 1, and hence 1 ≤ 1 + t ≤ n, which gives (2).
Then, by the definition of the floor function we have k−1
m
−1 < t ≤ k−1
m
. Thus we have
k − 1 − m < mt ≤ k − 1, so 1 ≤ k − mt < m + 1, which gives (1).
USA Mathematical Talent Search
Round 1 Solutions
Year 33 — Academic Year 2021–2022
www.usamts.org
The picture below shows an example of the construction for m = 3, n = 5, and k = 10.
Note that D = (10, 1) gives triangle ABD with height 1 and base AD = 10, so [ABD] = 5.
Our construction above gives t = b 10−1
3
c = 3, so C = (10 − 3 · 3, 1 + 3) = (1, 4) satisfies
CD k AB, so that [ABC] = [ABD] = 5. One can further check that applying Pick’s Theo-
rem also verifies that [ABC] = 4 + 42 − 1 = 5.
y
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer
A
D
x
B
USA Mathematical Talent Search
Round 1 Solutions
Year 33 — Academic Year 2021–2022
www.usamts.org
5/1/33. Define a sequence of positive rational numbers x0 , x1 , x2 , x3 , . . . by x0 = 2, x1 = 3, and
for all n ≥ 2,
x2 + 5
xn = n−1 .
xn−2
(a) Prove that xn is an integer for all n ≥ 0.
(b) Prove that if xn is prime, then either n = 0 or n = 2k for some integer k ≥ 0.
Solution
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer
(a) Rewrite the recurrence as xn xn−2 − x2n−1 = 5. This is true for all n ≥ 2, so
xn xn−2 − x2n−1 = xn+1 xn−1 − x2n .
We can rearrange this as
xn xn−2 + x2n = xn+1 xn−1 + x2n−1 ,
and dividing by xn xn−1 yields
xn−2 + xn xn+1 + xn−1
= .
xn−1 xn
x21 + 5 32 + 5
Thus this quantity is constant for all n ≥ 2. In particular, since x2 = = =7
x0 2
x2 + x 0 7+2
and = = 3, we have
x1 3
xn−2 + xn
=3
xn−1
for all n ≥ 2. Therefore xn = 3xn−1 − xn−2 for all n ≥ 2, and we see that all terms in
the sequence must be integers. We also note that the sequence is strictly increasing and in
particular all elements of the sequence are greater than 1.
(b) We can rewrite the recurrence as xn−2 = 3xn−1 − xn . Note that this is the same
recurrence but in the opposite direction, and this allows us to extend the sequence to xn for
n < 0. In particular, note that x−1 = 3x0 − x1 = 3(2) − 3 = 3, so since x−1 = x1 = 3, we
have x−n = xn for all n.
Lemma: Let p be prime and n, r be integers, with r > 0, such that xn and xn+r are
both multiples of p. Then xn+mr is a multiple of p for any integer m.
Proof: Let xn+1 ≡ a (mod p) and xn+r+1 ≡ b (mod p). If a ≡ 0 then there is nothing to
prove: every subsequent element is a multiple of p. Otherwise, by induction, for all k ≥ 0,
xn+k ≡ ck a (mod p) and xn+r+k ≡ ck b (mod p) for some ck . (This is because the recurrence
USA Mathematical Talent Search
Round 1 Solutions
Year 33 — Academic Year 2021–2022
www.usamts.org
relation is linear mod p.) But when k = r, we must have cr ≡ 0 (mod p), since xn+r ≡ 0
(mod p), and thus xn+2r ≡ 0 (mod p) as well. The full result then follows by straightforward
induction. 2
Applying this to x−n and r = 2n, we get that if xn is a multiple of p, then so is xmn for any
odd integer m. Considering this fact for all primes, we see that xn |xmn for any odd integer m.
In particular, if n = m2k for some odd m, then x2k |xn . So if m > 1, then xn is a nontrivial
multiple of x2k > 1, and thus must be composite.
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer
Hence, xn is prime only if n = 0 or n = 2k for some k.
Note 1: The condition is certainly not sufficient for x2k to be prime. One can check that
(x0 , x1 , x2 , x4 , x8 ) = (2, 3, 7, 47, 2207)
are all prime, but x16 = 4870847 = 1087 · 4481 is composite. Indeed, it is an open question
if x2k is prime for any k > 3: there are no known examples of x2k prime for k > 3, but it is
unproven that they must all be composite.
Note 2: It can be shown that xn = L2n , where Lk is the k th Lucas number, defined by
the recurrence L0 = 2, L1 = 1, and Ln = Ln−1 + Ln−2 for all n ≥ 2. (Note that this is the
same recurrence relation as the Fibonacci numbers, but with different initial conditions.) A
solution is acceptable if it proves this fact and then uses known facts about Lucas numbers,
cited from a reputable source, to complete the proof.
Problems by Valentin Vornicu, Carl Yerger, and USAMTS Staff.
c 2021 Art of Problem Solving Initiative, Inc.
