U S A Mathematical Talent Search

PROBLEMS / SOLUTIONS / COMMENTS

Round 1 - Year 11 - Academic Year 1999-2000
Gene A. Berg, Editor

1/1/11. The digits of the three-digit integers a, b, and c are the nine nonzero digits 1, 2, 3, ..., 9,
each of them appearing exactly once. Given that the ratio a:b:c is 1:3:5, determine a, b, and c.

Solution 1 by Antoinette Realica (10/GU):

a: b: c
1: 3: 5
--- --- ---

a, b, c have 9 nonzero digits with each of them appearing exactly once.

The smallest possible number: 123

The last digit of c should be a multiple of 5 ending in 5, so the first digit of a should end in an odd
digit (3, 7, or 9). The first digit of a should be 1 because if it were larger, then c would be four
digits.

Process of trial and error:

a: b: c
1: 3: 5
123: 369 615 3 and 6 repeated
127: 381: 635 1 and 3 repeated
129: 387 645 This works!

a=129, b = 387, c = 645.

Solution 2 by Vladimir Novakovski (10/VA): Let a = a2a1a0, b = b2b1b0, c = c2c1c0.

Since c = 5a < 1000, a < 200, so a2 = 1. Let us examine a0. If

a0 is even, then c0 = 0, which is impossible. If a0 = 5, then b0 = c0 = 5, which is also impossible.
If a0 = 7, then b0 = 1, which is impossible as well (since a2 = 1). So a 0 ∈ {3, 9} .

Case 1: a0 = 3
Consider a1, b1, and c1. From our hypothesis, it follows that b 1 ≡ 3a 1 (mod10) , and that
c 1 ≡ 5a 1 + 1 ( mod10 ) ( the latter since 5 ⋅ 3 = 10 + 5 , and the 10 contributes to the tenths dig­
its). Therefore a1 cannot be even, since then c1 = 1. Also, a 1 ≠ 5 since otherwise a1 = b1 = 5;
and a 1 ≠ 7 since otherwise b1 = 1. Since a0 = 3, a 1 ≠ 3 , and since b 0 ≡ 3 ⋅ a 0 mod 10 = 9,
a 1 ≠ 9 . So there are no solutions.

Case 2: a0 = 9
Again, consider a1, b1, and c1. Now b 1 ≡ 3a 1 + 2 ( mod10 ) , and c 1 ≡ 5a 1 + 4 ( mod10 ) . Now
a 1 ≠ 3 since then b1 = 1; a 1 ≠ 4 since then b1 = 4; a 1 ≠ 5 since then c1 = 9; a 1 ≠ 6 since then
b1 = 0; a 1 ≠ 7 since then c1 = 9; a 1 ≠ 1 and a 1 ≠ 9 . So the only possibilities are 2 and 8.
Now 189 ⋅ 5 > 900 , but c 0 ≠ 9 . It can be checked by arithmetic that 129 does work, so the num­
bers a, b, and c are 129, 387, and 645.

Editor’s Comments: This problem was proposed by Professor Sándor Róka, the founding editor
of Hungary’s Abacus, a problem solving journal for students in grades 4 through 8.

2/1/11. Let N = 111...1222...2, where there are 1999 digits of 1 followed by 1999 digits of 2.
Express N as the product of four integers, each of them greater than 1.

Solution 1 by Kirsten Rutschman (12/WA):

To solve, look for patterns:
1122 ÷ 11 = 102
111222 ÷ 111 = 1002
11112222 ÷ 1111 = 10002
N ÷ A = B where A is 1999 digits of 1 and B = 1000...02 has 1998 digits of 0.

102 ÷ 2 = 51
1002 ÷ 2 = 501
10002 ÷ 2 = 5001
B÷2 = C where C = 5000...01 with 1997 digits of 0.

51 ÷ 3 = 17
501 ÷ 3 = 167
5001 ÷ 3 = 1667
C÷3 = D where D = 1666...67 with 1997 digits of 6.

N = A ⋅ B ⋅ C ⋅ D = 111…1 ⋅ 2 ⋅ 3 ⋅ 1666…67 .

Solution 2 by Luke Gustafson (11/MN):

n n
10 – 1 is a string of n 9’s, hence (10 – 1 ) ⁄ 9 is a string of n ones. Then we have
1999 1999
1999 ⎛ 10
------------------------⎞ + 2 ⎛ ------------------------⎞
–1 10 –1
N = 10
⎝ 9 ⎠ ⎝ 9 ⎠
 1999
N = ⎛ ------------------------⎞ (10
10 –1 1999
+ 2)
⎝ 9 ⎠
1999
N = 2 ⋅ ⎛ ------------------------⎞ ( 5 ⋅ 10
10 –1 1998
+ 1)
⎝ 9 ⎠
1999 1998
– 1 5 ⋅ 10
N = 2 ⋅ 3 ⋅ ⎛ ------------------------⎞ ⎛ -------------------------------­⎞
10 +1
⎝ 9 ⎠ ⎝ 3 ⎠
1998
The last factor is an integer since the sum of the digits of 5 ⋅ 10 + 1 (a five followed by a
string of zeros and then a one) is 6, implying it is divisible by three. Hence

N = 2 ⋅ 3 ⋅ (111…1) ⋅ (1666…67)

1998
It might be interesting to note that N is also divisible by 23. I will show that 5 ⋅ 10 + 1 is
divisible by 23. This proof makes use of Fermat’s Theorem: if x ≠ 0 ( mod23 ) , then
22
x ≡ 1 ( mod23 ) .
1998 22 90 18
5 ⋅ 10 + 1 ≡ 5 ⋅ (10 ) (10 ) + 1 ( mod23 )
18
≡ 5 ⋅ 10 +1
9
≡ 5 ⋅ (100 ) + 1
9
≡5⋅8 +1
22 5
≡ 5 ⋅ (2 ) ⋅ (2 ) + 1
5
≡5⋅2 +1
≡ 161
≡ 0 ( mod23 )
This yields a more complete factorization of N:
1999 1998
5 ⋅ 10
N = 2 ⋅ 3 ⋅ 23 ⋅ ⎛ ------------------------⎞ ⋅ ⎛ --------------------------------⎞
10 –1 +1
⎝ 9 ⎠ ⎝ 69 ⎠

Solution 3 by Alex Utter (11/CA): N is even so it is a multiple of 2, but the last two digits are not
divisible by 4, so N is not a multiple of 4. The sum of its digits is 1999 ⋅ 1 + 1999 ⋅ 2 = 1999 ⋅ 3 ,
so it is a multiple of 3 but not 9. 5 is not a factor because N does not end with 0 or 5.

Potential third factors can be tested by a technique that I devised based on remainder repetition
intervals. I will use the division of 7 into a short series of ones and twos as an example:
 1 5 8 7 3 0 3 1 7 4 6
7 ) 11 14 16 15 12 10 22 21 25 23 24 20

As you can see, it divides evenly into 6 ones, and divides evenly into 6 twos. Because the remain­
der at the end of each section is zero, it can loop around as many times as desired. In addition, at
equal distances from the change from 1 to 2 are two remainders of 2, so the cycle can be inter­
rupted at that point. Therefore, 7 is a factor of any number like N with 6x ones and 6x twos, or a
number with 6x+5 ones and 6x+5 twos, with x as an integer greater than or equal to 0. I write this
cycle structure as 6x and 6x+5. Since 1999 does not equal 6x or 6x+5 for any integer x, 1999 does
not fit seven’s cycle, and 7 is not a factor.

Now, this method of testing primes until a factor is found can be used. The results starting after 7
are:
Prime Cycle structure
11 2x
13 6x
17 16x, 16x+2
19 18x, 18x+8
23 22x, 22x+19

1999 = 22 ⋅ 90 + 19 = 22 ⋅ x + 19 , so the third factor of N is 23. The results of division cycle

with the remainder cycles, so N/23 can be expressed as follows:

(Notation: X indicates that the previous series of numbers should be repeated the indicated num­
ber of times, ... indicates that the next number should be concatenated to the previous one)

N/23 = 0048309178743961352657X90...
004830917874396135266183574879227...
00966178357487922705314X90.

Each section alone is not a multiple of 3, but the three sections together are. In addition, the mid­
dle section (put together) is also a multiple of 3. So

N
------------- =
3 ⋅ 23
001610305958132045088566827697262479871175523349436392914653784219X30...
0016130595813204508872785829307568438...
003220611916264090177133655394524959742351046698872785829307568438X30

Only the first section is odd, but the remainder can be contained in the first digit of the next repe­
tition. Therefore, the final solution is
N = 2 * 3 * 23 *
280008051529790660225544283413848631239935587761672718196457326892109...
500805152979066022544283413848631239935587761674718196457326892109X29...
50080515297906602254436392914653784219...
001610305958132045488566827697262479871175523349436392914653784219X30

Editor’s Comments: We thank Dr. George Berzsenyi, the creator of the USAMTS and now
retired from Rose-Hulman Institute of Technology to the mountains of Colorado, for this beautiful
problem.

The solutions to these problems give us an opportunity to briefly introduce congruences and mod­
ular arithmetic, and state an important theorem from Number Theory. Our goal is to introduce
some of the notation and terminology, and briefly introduce this subject to young mathematicians
seeing this for the first time, and possibly help them understand the proofs.

Two integers a and b are said to be congruent modulo m if m divides a – b . We write

a ≡ b(modm) , and say that a is congruent to b modulo m. The positive integer m is called the
modulus, and the relation is called a congruence. For example,

86 ≡ 6 ( mod10) 1 ≡ 16 ( mod5) – 3 ≡ 4 ( mod7) 88 ≡ 0 ( mod11)

Observe that if you divide a by m and get remainder r, then a ≡ r (modm) . If a ≡ b(modm) and
c ≡ d (modm) , then a + b ≡ c + d (modm) and a ⋅ b ≡ c ⋅ d (modm) .

This gives us an interesting new arithmetic system where we can add, subtract, and multiply. We
refer to this arithmetic system as “arithmetic modulo m”. [Do you see a connection between
“clock arithmetic” and arithmetic modulo 12?]

If m is a prime integer, then we can also divide by nonzero elements. For example, observe that
1
2 ⋅ 4 ≡ 1 ( mod 7) , so we think of 2 as the multiplicative inverse of 4; that is, 2 ≡ --- in this arith­
4
metic system. Thus, dividing by 4 is the same as multiplying by 2 when you are working modulo
7. Whenever m is a prime number, and a is not congruent to 0 modulo m, then we can find a sec­
ond number b such that a ⋅ b ≡ 1 ( modm) ; dividing by a is the same as multiplying by b, so we
now can divide by any “nonzero” element a. Now we have an arithmetic system where we can
add, subtract, multiply, and divide by nonzero numbers.

This sounds good, you say, but given a, how do you find the corresponding b? The process is
straightforward and efficient, and uses the extended Euclidean algorithm for finding Greatest
Common Divisors, an algorithm already familiar to you, perhaps. If p is prime and p does not
divide a, then GCD(p,a) = 1. Therefore there are integers r and s such that 1 = ar + ps. So
1 ≡ ar + ps(mod p) ≡ ar (mod p) , and r is the multiplicative inverse of a, as desired.
 –1
As an example, we will find 19 modulo 61. That is, we find an integer r such that
19 ⋅ r ≡ 1 ( mod61 ) . First we use the division algorithm, divide 61 by 19, and get remainder 4:
4 = 61 – (3 ⋅ 19 )
Next, divide 19 by 4 and obtain remainder 3:
3 = 19 – (4 ⋅ 4)
Finally
1=4-3
So 1 is the Greatest Common Divisor of 61 and 19. Substituting back into the above equations we
obtain:
1=4-3
= 4 – (19 – 4 ⋅ 4) = 5 ⋅ 4 – 1 ⋅ 19
= 5 ⋅ ( 61 – 3 ⋅ 19 ) – 1 ⋅ 19
= –16 ⋅ 19 + 5 ⋅ 61
≡ 45 ⋅ 19 + 5 ⋅ 61 ( mod61 )
≡ 45 ⋅ 19 ( mod61 ) .
–1
Thus, 19 modulo 61 is 45.

Now we state a beautiful theorem. If n is a positive integer, then an corresponds to

a ⋅ a ⋅ a ⋅ … ⋅ a with n factors of a. We define a0 = 1. So we also know how to take positive
exponents.

p–1
Fermat’s Little Theorem: If p is a prime integer, and p does not divide a, then a ≡ 1 ( mod p) .

For example,
2 6 22 10
5 ≡ 1 ( mod 3 ) 2 ≡ 1 ( mod7 ) 123456 ≡ 1 ( mod23 ) ( –3 ) ≡ 1 ( mod11 )

This theorem is a key part of Luke Gustafson’s proof above.

3/1/11. Triangle ABC has angle A measuring 30° , angle B measuring 60° , and angle C measur­
ing 90° . Show four different ways to divide triangle ABC into four triangles, each similar to
triangle ABC, but with one quarter of the area. Prove that the angles and sizes of the smaller
triangles are correct.

Solution 1 by Neeraj Kumar (12/NC): Let us start with triangle ABC having sides AB, BC, and
3
AC of lengths 2, 1, and 3 , respectively. The area of this triangle is ------- . Therefore, each of the
2
3 3
smaller triangles must have area ------- and sides of lengths 1, 1/2, and ------- .
8 2
Here is the first way to divide ABC:
B
We choose E to be the midpoint of BC, and connect
it to AB (parallel to AC). Since angle B is common
to both triangles ABC and DBE, and E is also a right D
E
angle, the two triangles are similar. Also, since BE is
1/2, DE must equal ( 3) ⁄ 2 , and the area of DBE is
A F C
3
indeed ------- . If we now choose F to be the midpoint
8
of AC, then again we create a similar situation with
shared angle A of the two triangles ADF and ABC, and so ADF is also a valid smaller triangle.
Now DC = DB = DA, by Side-Angle-Side congruence in the triangles BED and CED (BE = EC,
angle BED = angle DEC, DE = DE), and in triangles ADF and CDF. Therefore, each of DEC and
DFC is also a valid smaller triangle.

For the second way to divide the triangle, we can simply choose the other diagonal of rectangle
DECF, i.e. FE rather than DC.

For the final two ways, we can see that triangle DBC is equilateral (because each side has length
1). Therefore, the altitude DE can be moved to any one of the three sides. Thus the four different
ways are given here:

Editor’s Comments: This problem, along with many earlier USAMTS problems, was proposed
by Dr. Erin Schram of the National Security Agency. We are also thankful to him for his thought­
ful evaluation of the problems considered for the program.

4/1/11. There are 8436 steel balls, each with radius 1 centimeter, stacked in a tetrahedral pile,
with one ball on top, 3 balls in the second layer, 6 in the third layer, 10 in the fourth, and so on.
Determine the height of the pile in centimeters.

Solution 1 by Shaili Jain (12/MI): The number of balls in each level of the pyramid are: 1, 3, 6,
10, 15, 21, ..., n(n+1)/2.
n
i(i + 1)
∑ -----------------
2
= 8436
i=1
n = number of layers in the pyramid. We want to solve for n.
n 2 n n
1 1 n ( n + 1 )(2n + 1 ) n ( n + 1 )
∑ ∑i ∑i
i +i 2
8436 = ------------ = --- + = --- ----------------------------------------- + -------------------­
2 2 2 6 2
i=1 i=1 i=1

8436 = --- n ( n + 1 )(n + 2 ) = ⎛ n + 2⎞ , a binomial coefficient.

1
6 ⎝ 3 ⎠
n = 36.
The sides of the tetrahedron are
35 ⋅ 2 = 70 cm each (they go through the 30°
centers of the outer balls). 120°

sin 120° sin30°

------------------- = ---------------- by Law of Sines
70 x
70 3 Base of
x = ------------­
3 tetrahedron

70 3 2
70 – ⎛ -------------⎞ = ------------­
2 70 6
Tetrahedron height =
⎝ 3 ⎠ 3

70 6
The height of the pile of balls is ------------- + 2 cm. because it is +1 cm. from the center of the top
3
ball to the top, and +1 cm. from the center of a bottom ball to the bottom.

Solution 2 by Wendy Pang (11/CA): 8436 is the sum of 36 trian­

gle numbers, meaning that there are 36 layers of balls. The centers
of four adjacent balls form a tetrahedron (equilateral) with slant
2 6 3 x 2
height of 2, which means that the height of the tetrahedron is ---------- .
3
Height of pile = (layers - 1)(height of tetrahedron) + 2 ⋅ radius .

= (36 – 1 ) ⎛ ----------⎞ + 2 ( 1 )
2 6
⎝ 3 ⎠ 3
70 6
= ------------- + 2 cm.
3

Editor’s Comments: This problem was suggested 20 years ago by Professor Endre Hòdi, long­
time leader of Hungary’s successful teams to the IMO’s.
5/1/11. In a convex pentagon ABCDE the sides have lengths 1, 2, 3, 4, and 5, though not neces­
sarily in that order. Let F, G, H, and I be the midpoints of the sides AB, BC, CD, and DE,
respectively. Let X be the midpoint of segment FH, and Y be the midpoint of segment GI.
The length of segment XY is an integer. Find all possible values for the length of side AE.

Solution 1 by Melody Chan (12/NY): Use the notation A = (4a1, 4a2) to mean “the coordinates
of the point A are (4a1, 4a2) .”

Let A = (4a1, 4a2), B = (4b1, 4b2), C = (4c1, 4c2), D = (4d1, 4d2), and E = (4e1, 4e2) .
Then F = (2a1+2b1, 2a2+2b2)
G = (2b1+2c1, 2b2+2c2)
H = (2c1+2d1, 2c2+2d2)
I = (2d1+2e1, 2d2+2e2)

and X = (a1+b1+c1+d1, a2+b2+c2+d2)

Y = (b1+c1+d1+e1, b2+c2+d2+e2)
Thus
2 2
XY = [(b 1 + c 1 + d 1 + e 1 ) – (a 1 + b 1 + c 1 + d 1 )] + [(b 2 + c 2 + d 2 + e 2 ) – (a 2 + b 2 + c 2 + d 2 )]
2 2
= (e 1 – a 1 ) + (e 2 – a 2 ) which is given to be an integer.
Also
2 2
AE = (4e 1 – 4a 1 ) + (4e 2 – 4a 2 )
2 2
= 4 ( e 1 – a 1 ) + ( e 2 – a 2 ) = 4 ⋅ XY

So AE is a multiple of 4. Thus AE = 4 is the only possible solution.

Solution 2 by Kartik Lamba (12/IL): The

solution becomes simple when one uses vec­
tors. Consider the pentagon shown at right with
any of the given following dimensions:
B G C
A+B B+C
F = -------------- G = --------------
2 2
F X
H
C+D D+E Y
H = -------------- I = --------------
2 2 0
A
D
A+B C+D B+C D+E
-------------- + -------------­ -------------- + -------------- I
2 2 2 2
X = ----------------------------------- Y = ----------------------------------­ E
2 2
 A+B+C+D B+C+D+E A–E
X – Y = ----------------------------------- – ----------------------------------- = ------------­
4 4 4

4 ⋅ XY = AE
In order for both lengths XY and AE to have integer values between 1 and 5, inclusive, XY must
be 1 and AE must be 4. Thus
AE = 4 .

Solution 3 by Dmitry Portnyagin (12/NY): We draw

line segment BE, and mark its midpoint Z. Now, G and B G C
H are midpoints of the sides of ΔBCD , Z and I are
midpoints of the sides of ΔBED . Thus X
F H
ΔGC H ∼ ΔBCD , ΔZEI ∼ ΔBED , and in both cases Y
the sides of the smaller triangle are half the size of the Z
A
sides of the larger triangles. Thus, ZI = GH = BD/2, D
and ZI || GH || BD . Thus, GHIZ is a parallelogram.
I
In a parallelogram the diagonals bisect each other.
Thus Y is the midpoint of HZ. E

F and Z are midpoints of AB and BE, respectively.

Thus, once again, FZ = AE/2. X and Y are midpoints
of FH and HZ, and once again XY = FZ/2. Thus, XY =
AE/4. But XY has to be an integer, and out of the five possible lengths, only 4 is divisible by 4.
AE = 4

Editor’s Comments: This clever problem was created for USAMTS by Professor Gregory Galp­
erin, who is the author of a soon-to-be published compendium of problems for the famous Mos­
cow Mathematical Olympiads.