U S A Mathematical Talent Search

PROBLEMS / SOLUTIONS / COMMENTS

Round 3 - Year 10 - Academic Year 1998-99
Gene A. Berg <gaberg@ieee.org>, Editor

1/3/10. Determine the leftmost three digits of the number

1 2 3 999 1000
1 + 2 + 3 + + 999 + 1000 .

Solution 1 by Jason Oh (11/MD): The leftmost three numbers of this sum are 100. To prove this
I will show that the only term that contributes to the leftmost three digits is the 10001000 term.
The way to see this is by considering the series
1 2 3 999 1000
1000 + 1000 + 1000 + + 1000 + 1000
Without much consideration, this series can be seen to add up to
1001001001001000
In this series the terms from 10001 to 1000999 have no effect on the three leftmost digits. In the
original series, the sum of the terms from 11 to 999999 will be smaller than the sum of the terms
10001 to 1000999 in the second series since each term is smaller and the original series is domi
nated by the second series. Thus in the original series, the rst 999 terms will also not contribute
to the leftmost three digits of the sum. The leftmost three terms of the sum come from the
10001000 term, and are therefore 100.

Solution 2 by Marcus Aaron (10/TX): Setting

1 2 3 999 1000
x = 1 + 2 + 3 + + 999 + 1000
it is clear that
1000 1 2 3 999 1000
1000 < x < 1000 + 1000 + 1000 + + 1000 + 1000
or that x is between 1000000... and 100100100... . Since both these numbers have exactly 3001
digits, the rst three digits of x must be 100.

Editors comment: Thanks are due to Professor Gregory Galperin of Eastern Illinois University
for this nice problem.

2/3/10. There are innitely many ordered pairs (m, n) of positive integers for which the sum
m + (m + 1 ) + (m + 2 ) + + (n 1 ) + n
is equal to the product mn . The four pairs with the smallest values of m are (1, 1), (3, 6), (15,
35), and (85, 204). Find three more (m, n) pairs.
Solution 1 by Lucy Jiang (12/MD): The equation can be rewritten as:
nm+1
(m + n) ---------------------- = mn ,
2
which simplies to
2 2
m m(1 2n) n n = 0 .
Although 0 is not a positive integer, it can be used in determining a pattern for m and n because
0 + 0 = 0 0.

Suppose (m 0, n 0 ) = (0, 0) , (m 1, n 1 ) = (1, 1) , (m 2, n 2 ) = (3, 6) , and so on. We observe that

n i + 2 = 6n i + 1 n i , as shown below:
(0,
0)
(1, 1) 1
6 0 = 6

(3, 6) 6
6 1 = 35

(15,
35 ) 35 6 6 = 204

(85, 204 )

Using this pattern, the next value for n is 204 6 35 = 1189 . Substituting n into the equation
2 2
m m(1 2n) n n = 0 , we obtain m = 493 . Thus the ordered pair is (493, 1189), and
this works.

By the same method, two other ordered pairs are: (2871, 6930) and (16731, 40391).

i(i + 1)
Solution 2 by Trevor Bass (12/NY): Using the fact that the sum 1 + 2 + 3 + + i = ----------------- ,
2
where i is an integer, we can rewrite the equation
m + (m + 1 ) + (m + 2 ) + + (n 1 ) + n = mn
as
n(n + 1) m(m + 1)
-------------------- ----------------------- = mn ,
2 2
so
2 2
m + (2n 1 )m + ( n n) = 0 .
In order for m to be an integer, the discriminant of this quadratic equation must be the square of an
integer, so
2 2 2
(2n 1 ) 4( n n) = i ,
where i is an arbitrary integer. This equation can be rewritten as
2 2
8n + 1 = i .
The rst seven values of n for which there exist is are: 1, 6, 35, 204, 1189, 6930, and 40391. For
each n, the corresponding m can be computed by solving the quadratic equation
2 2
m + (2n 1 )m + (n n) = 0 . The rst four ns are given in the problem statement, so the
next three (n, m) pairs are (493, 1189), (2871, 6930), and (16731, 40391).
Solution 3 by David Walker (11/NE): The expression m + (m + 1 ) + + (n 1 ) + n can be
nm+1
written (m + n) ---------------------- . Set this equal to mn and obtain (m + n)(n m + 1 ) = 2mn . If you
2
put 1, 3, 15, or 85 in for m in this equation, you get two solutions for each m. The solutions for 1
are 1 and 0. The solutions for 3 are 6 and -1. The solutions for 15 are 35 and -6, and for 85 are
204 and -35. I noticed the pattern that the solution to an m that is negative is the positive solution
to the previous m value. So I put -204 into the equation and got 493. Then I put -493 back in for
m and got 1189. I continued doing this and got the three (m, n) pairs (493, 1189), (2871, 6930),
and (16731, 4039).

Solution 4 by Oaz Nir (10/CA): Three pairs are: (m, n) = (493, 1189), (2871, 6930), and (16731,
40391).

Claim: If (m, n) is a solution, then (2n + m, 5n + 2m - 1) also is.

Using this claim, we easily recursively compute the three pairs above. Starting with the given
solution (85, 204), we use the transformation (m, n) (2n + m, 5n + 2m 1 ) three times in suc
cession.

To prove the claim, we rst use the arithmetic series sum formula to transform the given equation
into the following equivalent form:
(sum of the rst and last terms)(number of terms)/2 = mn
(m + n)(n m + 1 ) = 2mn .
Now we see that (2n + m, 5n + 2m 1 ) is a solution if
([2n + m] + [5n + 2m 1 ])([5n + 2m 1 ] [2n + m] + 1 ) = 2 [ 2n + m][5n + 2m 1 ]
(7n + 3m 1 )(3n + m) = 2 [ 2n + m][5n + 2m 1 ]
2 2 2 2
21n + 16mn + 3m 3n m = 20n + 18mn 4n + 4m 2m
2 2
n + n m + m = 2mn
2 2
(n + n mn ) + ( m + mn + m) = 2mn
n ( n m + 1 ) + m ( n m + 1 ) = 2mn
(n + m)(n m + 1 ) = 2mn .
That is, if (m, n) is a solution then (2n + m, 5n + 2m - 1) is also a solution.

Solution 5 by Reid Barton (10/MA): Answer: (493, 1189), (2871, 6930), and (16731, 40391).

Since m + (m + 1 ) + + (n 1 ) + n = (n m + 1 )(m + n) 2 , we want to nd m and n such

that
(n m + 1 )(m + n)
mn = ---------------------------------------------- .
2
We have
(n m + 1 )(m + n)
mn = ---------------------------------------------- .
2
 2 2
2mn = nm + n m mn + m + n
2 2
m + 2mn n m n = 0
2 2
(m + n) (m + n) 2n = 0
2 2
4(m + n) 4(m + n) + 1 8n = 1
2 2
(2m + 2n 1 ) 8n = 1
2 2
t 8n = 1
This is Pells equation with primitive solution (3,1), so its solutions are given by
k
t n 8 = (3 8) for k 1 . For k = 1, 2, 3, and 4 we obtain (t, n) = (3, 1), (17, 6), (99, 35),
(577, 204), which give (m, n) = (1, 1), (3, 6), (15, 35), (85, 204) respectively, as m = (t-2n +1)/2.
Letting k = 5, 6, and 7 gives (t, n) = (3363, 1189), (19601, 6930), and (114243, 40391), which
give (m, n) = (493, 1189), (2871, 6930), (16731, 40391), three more solutions to the original equa
tion.

Solution 6 by Mike Fliss (12/NJ): The sum on the left side of the equation (the sum of the inte
gers from m to n) can be rewritten as (n m + 1 )(n + m) 2 since this is the number of terms
times the average term. We have
(n m + 1 )(n + m) 2 = nm
2 2
n m + n + m = 2nm
2
Now, in Ax + Bx + C = 0 form, this is
2 2
n + (1 2m)n + (m m ) = 0
We use the quadratic formula to obtain solutions for n. We want positive solutions for n. For all
2
m 1 , observe that 2m 1 8m 8m + 1 , so we do not consider the , just the + in the
quadratic formula to get positive values for n. So
2
2m 1 + 8m 8m + 1
n = -----------------------------------------------------------
2
2
8m 8m + 1 1
n = m + ---------------------------------------------
2
Set
2
8m 8m + 1 1
k = ---------------------------------------------
2
Since m and n are integers, by the additive closure of the integers, k is also an integer.
2
8m 8m + 1 = 2k + 1
2 2
8m 8m + 1 = (2k + 1 )
2 2
8m 8m + 2 = (2k + 1 ) + 1
 2 2
(2k + 1 ) 2 (m 1 ) = 1
2 2
s 2r = 1
where s = 2k + 1 and r = m 1 . This is Pells equation, solvable by a continued fraction
expansion of 2 , because
2
s +1
2 = -------------
2
-.
r
Continued fraction expansion yields fractions (s/r). If (s, r) is a solution to the equation
2 2
s 2r = 1 , then the following argument shows that (s+2r, s+r) is also a solution:
2 2 2 2 2 2
(s + 2r ) 2(s + r) = s + 4sr + 4r 2s 4sr 2r
2 2
= s + 2r
= + 1
Therefore, the sequence of fractional expansions follows this form:
s s i + 2r i
---i , ----------------
- , ...
ri si + r i
Thus, the answers for the (s/r) expansion are

1 3 7 17 41 99 239 577 1393 3363 8119 19601 47321 114243 275807

1 2 5 12 29 70 169 408 985 2378 5741 13860 33461 80782 195025

BUT, these are the answers for +1 and -1; we only want solutions for -1. Therefore, we take every
other term.

1 7 41 239 1393 8119 47321 275807

1 5 29 169 985 5741 33461 195025

Since the bottom number corresponds to r, and 2m-1 = r, we get the m values (rst 7) in the table
2
8m 8m + 1 1
below. Plugging these ms back into the original n = m + ---------------------------------------------- , we get the cor
2
responding ns.

m 1 3 15 85 493 2871 16731

n 1 6 35 204 1189 6930 40391
Three new pairs (m, n) are listed in bold in the table.

Editors comment: We are indebted to Professor Suresh T. Thaker of India for this excellent
problem.
3/3/10. The integers from 1 to 9 can be arranged into a 3 3 array (as shown
on the right) so that the sum of the numbers in every row, column, and diag
A B C
onal is a multiple of 9.
D E F
(a.) Prove that the number in the center of the array must be a multiple of 3.

(b.) Give an example of such an array with 6 in the center.

G H I

Solution 1 by Rachel Johnson (9/MN): (a) There are nine possible combina
tions of three distinct numbers from 1 to 9 that have a sum of 9 or 18 (multiples of 9). They are:
{1, 2, 6} and {2, 3, 4} have a sum of 9, while {1, 8, 9}, {2, 7, 9}, {3, 6, 9}, {3, 7, 8}, {4, 5, 9}, and
{4, 6, 8} have a sum of 18. These combinations have either one or three multiples of 3 in each of
them.

Since all possible combinations have a multiple of 3 and there are only three
multiples of 3, the multiples of 3 must be shared. This can be done by placing 9 1 8
two (of 3, 6, and 9) in opposite corners. This provides a multiple of 3 for the
four outside combinations. To provide a multiple of three for the other combi 5 6 7
nations, the remaining multiple of 3 would have to be placed in the center. 4 2 3
(b) At the right is a possible arrangement of the array with 6 in the middle.

Solution 2 by Michelle Rengarajan (8/CA): (a) According to the given information:

1) A + B + C = 9k1
2) D + E + F = 9k2
3) G + H + I = 9k3
4) A + D + G = 9k4
5) B + E + H = 9k5
6) C + F + I = 9k6 1 8 9
7) A + E + I = 9k7 5 6 7
8) C + E + G = 9k8
3 4 2
Solve for E in equations 2, 7, and 8.
E = 9k2 -(D + F)
E = 9k7 -(A + I)
E = 9k8 -(C + G)
Add these three equations together to nd the value of 3E:
3E = 9k2 + 9k7 + 9k8 -(A + D + G + C + F + I).
So, according to equations 4 and 6 7 8 3
3E = 9k2 + 9k7 + 9k8 - 9k4 - 9k6.
2 6 1
Therefore, 3E is a multiple of 9, so E must be a multiple of 3.
9 4 5
(b) An array with 6 in the middle is shown on the right.

Editors comment: This clever problem was posed by Dr. Erin Schram of the National Security
Agency. He claims there are 24 possible arrays for part (b), all of which are rotations or reec
tions of one of the three arrays shown above.
4/3/10. Prove that if 0 < x < 2 , then
6 6 6 6
sec x + csc x + ( sec x)( csc x) 80 .

Solution 1 by Lucy Jin (11/MI): (Proof by contradiction.)

6 6 6 6
sec x + sec x + ( sec x)( csc x)
1 1 1
= -------------6
- + ------------
6
- + ------------------------------------
6 6

cos x sin x ( cos x)( sin x)
1 1 1
= --------------------3- + -------------------3- + ------------------------------------
6 6

( cos x)
2
( sin x)
2
( cos x)( sin x)
2 3 2 3
( sin x) + ( cos x) + 1
= --------------------------------------------------------
6 6

( sin x)( cos x)
2 2 4 2 2 4
( sin x + cos x)( sin x sin x cos x + cos x) + 1
= ----------------------------------------------------------------------------------------------------------------------
6 6
( sin x)( cos x)
4 2 2 4
sin x sin x cos x + cos x + 1
= ---------------------------------------------------------------------------
6 6

( sin x)( cos x)
2 2 2 2 2
( sin x + cos x) 3 sin x cos x + 1
= --------------------------------------------------------------------------------------
6 6

( sin x)( cos x)
2 2
2 3 sin x cos x
= --------------------------------------
6 6
( sin x)( cos x)
The last expression is equivalent to the original trigonometric expression.

Now we begin the proof by contradiction. Assume

2 2
2 3 sin x cos x
---------------------------------------
6 6
< 80 .
( sin x)( cos x)
We show that this assumption leads to a contradiction, and therefore must be false.

Let u = sin x cos x . The inequality above can be written

2
2 3u
-----------------
6
< 80
u

6 2
80 u + 3 u 2 > 0
 u + 1--- u 1--- (80u 4 + 20u 2 + 8) > 0
2 2

Observe 80u + 20u + 8 is always positive. Therefore u + --- u --- > 0 , so u > --- and either
4 2 1 1 2 1
2 2 4
1 1
u < --- or u > --- . Recall u = sin x cos x , so
2 2
1 1
sin x cos x < --- sin x cos x > --
2 2

sin2 x < 1 sin2 x > 1 .

Both these cases are contradictory, because 1 sin 1 .

Therefore, our assumption was false and we must have

2 2
2 3( sin x cos x)
6 6
- 80 .
-------------------------------------------
( sin x)( cos x)

6 6 6 6

and sec x + csc x + ( sec x)( csc x) 80 .

Solution 2 by Mark Tong (12/VA): This proof makes extensive use of the Arithmetic Mean-
Geometric Mean (AM-GM) Inequality, which states that for positive numbers a and b,
12 2
(a +
b) 2 (ab) . For the rst step of the proof, substitute into AM-GM with a = sin x and
2 2 2 12 2 2

b = cos x . So ( sin x cos x) ( sin x + cos x) 2 sin x cos x 1 2

6 6 6 6
sin x cos x 1 64 sec x csc x 64 .
6 6
Using this new equation and substituting into AM-GM again with a = sec x and b = csc x :
6 6 6 6 12 6 6 12
( sec x + csc x) 2 ( sec x csc x) sec x + csc x 2 ( 64 ) = 16 .

6 6 6 6
Since sec x csc x 64 and sec x + csc x 16 ,

6 6 6 6
sec x + csc x + sec x csc x 80 .
5/3/10. In the gure on the right, O is the center A
of the circle, OK and OA are perpendicular to

one another, M is the midpoint of OK, BN is

parallel to OK, and AMN = NMO .

Determine the measure of ABN in degrees.

B N
Solution by Chi-Bong Chan (12/NJ): For conve
nience and without loss of generality let the radius
of the circle be 2. Then OM = 1, OA = 2, and the K O
M
Pythagorean Theorem gives MA = 5 . The
angle bisector theorem gives MO/MA = NO/NA.
Rewrite as
MO NO
------------------------- = ----------------------- .
MA + MO NA + NO
Since AN + NO = OA = OB, we have
NO MO 1 51
-------- = ------------------------- = ---------------- = ---------------- .
OB MA + MO 5+1 4

OO 2NO 51
If we reect O over BN to get O, then ---------- = ------------ = ---------------- .
OB OB 2

This is the golden mean, and one interest

ing fact involving the golden mean is that A
the ratio of the side to the diagonal of a
regular pentagon is the golden mean. O

Proof of the fact: Keep in mind that B N

1) because it is a regular pentagon, the ve

arcs are equal, and K O
M

2) an angle subtending an arc has one-half

the measure of the arc.

We shall rst show that each diagonal is

parallel to the opposite side, for example
in the diagram of the regular pentagon
below, AD || BC . We have
ADC = (arcAC ) 2 = ((2 5)(360)) 2 = 72 ,
and BCD = (arcBAD) 2 = ((3 5)(360)) 2 = 108 .
Now AD || BC follows directly from the fact that ADC and BCD are complementary.
 Now, without loss of generality, let x be the length of a
A
side and 1 be the length of a diagonal, so that their ratio is
simply x. Since each diagonal is parallel to the opposite
side, AFDE is a parallelogram, and therefore DF = AB =
E B x. Moreover, since AD || BC , CBF = FDA and
1 1-x BCF = FAD . Therefore, ADF = BCF , so
F (BC ) (DA) = (BF ) (DF ) . That means
2
x 1 = (1 x) x or x = 1 x , which has the golden
x mean as the positive root, as desired. This completes the
proof of the fact.
D C

Back to our problem, we see that OBO

A
would be similar to DAC in the pentagon
because their sides are proportional
(OO/BO and DC/DA are both equal to the O
golden mean, and both triangles are isosce
les). Therefore BOO = ADC of the
B N B
pentagon, and so BOO = 72 from the
earlier calculation. 72
K O
M
Finally, to nd ABN , reect B across
AO to get B. Then
B = (arcAB) 2 = (72) 2 = 36
because an angle subtending an arc has
half the measure of the arc. Since B is
the image of ABN under the reection,
we have at last ABN = B = 36 .

Editors comment: Again we thank Professor Gregory Galperin of Eastern Illinois University
for this, his second beautiful problem included in this round.