2013 UI MOCK PUTNAM EXAM

September 25, 2013, 5 pm 7 pm

Solutions

1. Let f (n) denote the n-th term in the sequence 1, 2, 2, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 5, . . . , obtained by writ-
ing one 1, two 2s, three 3s, four 4s, etc.
(a) Find, with proof, f (2013).
(b) Find, with proof, a simple general formula for f (n). (The formula can involve the floor or ceiling
function.)
p
Solution. Answer: (a) f (2013) = 63 ; (b) f (n) = d 2n + 1/4 1/2e .
(a) The first k blocks occupy 1 + 2 + + k = k(k + 1)/2 positions. Thus, we have f (n) = k if and
only if n occurs among the last k of these positions, i.e., if and only if
k(k 1) k(k + 1)
(1) <n .
2 2
For the case n = 2013, a direct calculation gives the appropriate k-value: 62 63/2 = 63 31 = 1953
and 63 64/2 = 63 32 = 2016, so the k-value satisfying (1) is k = 63, and hence f (n) = 63.
(b) To get a general formula, we multiply (1) by 2 and complete the square on both sides:
 2  2
1 1 1
k < 2n + k + ,
2 4 2
r
1 1
k 1 < 2n + k.
4 2
p
This shows that f (n) = k = d 2n + 1/4 1/2e.

2. (Problem B1, Putnam 1971) Let S be a set and a binary operation on S satisfying x x = x for all
x S and (x y) z = (y z) x for all x, y, z S. Prove that is commutative, i.e., that x y = y x
holds for all x, y S.
Solution. Applying the two given rules repeatedly, we have, for any x, y S,

x y = (x y) (x y) = ((x y) x) y = ((y x) x) y = ((x x) y) y

= (x y) y = (y y) x = y x,
which proves the commutativity of .

3. (Problem B1, Putnam 1980) For which real numbers c is

1 x 2
e + ex ecx


2
for all real numbers x? Prove your answer.
Solution. Answer: c 1/2
To show that any c 1/2 works, expand both sides into a Taylor series and compare coefficients

1
 of corresponding powers of x:

1 x x

1X xn + (x)n X x2m
(1) e +e = = ,
2 2 n=0 n! m=0
(2m)!

cx2
X cm x2m
(2) e = .
m=0
m!

The constant terms on both sides are equal, and for m 1, the coefficients of x2m in (1) and
(2) are 1/(2m)! and cm /m!, respectively. Since (2m)! = m!(m + 1) . . . (2m) 2m m!, we have
1/(2m)! cm /m!, for m 1, when c 1/2, so all terms in the series (1) are less than or equal
to the corresponding terms in the series (2). This proves that the inequality holds when c 1/2.
To show that c < 1/2 does not work, we consider the behavior of both sides when x 0: From (1)
and (2) we see that
cx2 x2
   
cx2 1 x x 4 4 2c 1 2
e (e + e ) = 1 + + O(x ) 1 + + O(x ) = x + O(x4 ),
2 1! 1! 2
and hence  
1 cx2 1 x x 2c 1
lim 2 e (e + e ) = < 0,
x0 x 2 2
when c < 1/2. Thus, if c < 1/2, the inequality fails when x is sufficiently close to 0.
Alternative solution (Krishna Srinivasan): First note that, since both sides of the inequality
are even functions of x, we may restrict to values x 0.
Taking logarithms on both sides, we can rewrite the inequality as cx2 ln cosh(x), or
(1) f (x) = cx2 ln cosh(x) 0.
Now compute
sinh x
(2) f 0 (x) = 2cx ,
cosh x
1
(3) f 00 (x) = 2c
(cosh x)2
At x = 0, f (0) = 0 ln 1 = 0 by (1), and f 0 (0) = 0 by (2).
If c 1/2, then (3) gives f 00 (x) 1 1/(cosh x)2 1 1 0 for x 0, so f 0 (x) is nondecreasing for
x 0, and hence f 0 (x) f 0 (0) = 0. Therefore f (x) f (0) = 0 for x 0, so (1) holds for all x 0
in the case c 1/2.
Conversely, if c < 1/2, then (3) shows f 00 (0) = 2c 1 < 0, so f 0 (x) is decreasing for sufficiently small
x, and hence f 0 (x) < f 0 (0) = 0 and f (x) < 0 for such x. Thus, (1) fails for small enough x in the case
c < 1/2.
4. Find, with proof, a simple formula for the sum
2n  
X k 2nk
2 .
k=n
n

Solution. Answer: 4n
Let S(n) denote the given sum. We will show by induction that
() S(n) = 22n .

2
 For n = 1, we have S(1) = 10 + 21 21 = 2, so () holds in this case. Now let n 1 and suppose ()

holds for this n. Then, using the recurrence formula nk = k1 k1

n
+ n1 we get
2(n+1) 
X k
S(n + 1) = 22n+2k
k=n+1
n + 1
2n+2
X k 1 2n+2
X k 1
2n+2k
= 2 + 22n+2k
k=n+2
n+1 k=n+1
n
2n+1
X  h  2n+1
X h
2n+1h
= 2 + 22n+1h (set h = k 1)
h=n+1
n + 1 h=n
n
     
1 2n + 2 1 2n + 1
= S(n + 1) + 2 S(n) +
2 n+1 2 n
1
= S(n + 1) + 2S(n),
2
since      
2n + 2 (2n + 2)! 2n + 2 2n + 1 2n + 1
= = =2 .
n+1 (n + 1)!(n + 1)! n+1 n n
Hence S(n + 1) = 4S(n) = 4 4n = 4n+1 . This proves the formula for n + 1 and completes the
induction.
5. (B3, Putnam 1982) Let f (n) be the number of ordered pairs (a, b) of integers from {1, 2, . . . , n}
such that a + b is a perfect square (i.e., of the form k 2 , for some integer k). Prove that the limit
limn f (n)n3/2 exists and express this limit in the form r( s t), where s and t are integers and
r is a rational number.
4
Solution. Answer: ( 2 1) .
3
By definition, f (n) is the number of pairs (a, b) of integers from {1, 2, . . . , n} such that a + b = k 2 for
some positive integer k. We first count the number of such pair for fixed k.
If k 2 n, then there are exactly k 2 1 such pairs, namely (1, k 2 1), (2, k 2 2), . . . , (k 2 1, 1).
If n < k 2 2n, then the possible pairs are (k 2 n, n), (k 2 n + 1, n 1), . . . , (n, k 2 n), so there
are n (k 2 n) + 1 = 2n + 1 k 2 such pairs in this case.
If k 2 > 2n, there are no such pairs.
Adding up these counts, we get

b nc b 2nc
X X
(1) f (n) = (k 2 1) + (2n + 1 k 2 )

k=1 k=b nc+1
 
= S(b nc) b nc + (2n + 1) b 2nc b nc S(b 2nc) + S(b nc),

where S(t) = tk=1 k 2 . Now,

P

S(t) t(t + 1)(2t + 1) 1

lim
3
= lim 3
= .
t t t 6t 3
Hence, dividing (1) by n3/2 and letting n , we get
3
f (n) 1 2 1 4
lim 3/2 = 0 + 2( 2 1) + = ( 2 1).
n n 3 3 3 3

3
6. Let x1 , x2 , x3 , . . . and y1 , y2 , y3 , . . . be sequences of positive real numbers satisfying
(1) y1 y2 y3 ,

and
(2) x1 x2 . . . xk y1 y2 . . . yk
for k = 1, 2, 3 . . . . Prove that

x1 + x2 + + xk y1 + y2 + + yk
for k = 1, 2, 3 . . . .
Solution. From (2) and the AGM inequality, we get
k  1/k
X xi x1 xk
(3) k ... k
i=1
yi y1 yk

Set di = (xi yi )/yi . From (3) we get

k k
X X xi
(4) Sk := di = k 0.
i=1 i=1
yi

On the other hand, we have (setting S0 = 0)

k
X k
X
(xi yi ) = di y i
i=1 i=1
Xk
= (Si Si1 )yi
i=1
k1
X
= Si (yi yi+1 ) + Sk yk 0,
i=1

since Si 0 by (4) and yi yi+1 0 by (1). Thus,

k
X k
X
xi yi (k = 1, . . . , n),
i=1 i=1

as claimed.