1

PROBLEMS, U.S.A.

Ramona Cioran

C.N. ”Elena Ghiba Birta” Arad

1. For the real positive number x, let Bn (x) = 1x + 2x + 3x + .. + nx .

∞
X Bn (logn 2)
Prove or disprove the convergence of
n=2
(n log2 n)2

William Lowell Putnam Competition, 1982

Solution:

Since x = logn 2 > 0, Bn (x) = 1x + 2x + 3x + .. + nx 6 n · nx and

Bn (logn 2) n · nlogn 2 2
06 2
6 2
= .
(n log2 n) (n log2 n) n · (log2 n)2
∞
X 2
As converges by the integral test, the given series converges by
n=2
n · (log2 n)2
the comparison test.
Z ∞
arctan (πx) − arctan x
2. Evaluate dx.
0 x

William Lowell Putnam Competition, 1982

Solution:
Z ∞ Z ∞ Z ∞Z π
arctan (πx) − arctan x 1 π 1
dx = ·arctan ux dx = dudx =

x x 1 1 + (xu) 2
Z0 π Z ∞ Z π 0
π
0 1
1 1π π  π
dxdu = du = ln u = ln π.

1 + (xu)2 u 2 2 1 2
1 0 1

3. Let f (n) = gcd{k n −k|k = 2, 3, 4, ..}, for n > 2. Evaluate f (n). In particular,

show that f (2n) = 2.

AMM, 1981

Solution:

1) f (n) cannot contain a factor p2 for any prime p(p > 2) because p2 - k(k n−1 −1)
2

for k = p.

2) For every n, 2|f (n).

3) If p is an odd prime and if a is a primitive root (mod p), then p|a(an−1 − 1)

only if (p − 1)|(n − 1). On the other hand, if (p − 1)|(n − 1) then p|(k n − k)

for every k. Thus, if P is the product of the distinct odd primes p such that

(p − 1)|(n − 1), then f (n) = 2P .

4) (p − 1) is not a divisor of (2n − 1) for any odd prime p. Thus f (2n) = 2.

Z k
1
4. Prove that the function A(h) = [1 − (h − cos x)2 ] 2 · [4(h − cos x)2 − 1]dx,
0

k = arccos (h − 1) has at least two zeros in the interval (0,2).

AMM, 1981

Solution:
Z π Z π
We have first A(0) = sin x · (4 cos2 x − 1)dx = (1 − 4 cos2 x) · (cos x)0 dx =
 π 0 0
4 cos3 x 

2
cos x −  = 3 > 0.
3 0
3
We also have A(h) > 0 for < h < 2, since the integrand in this case is positive
2
throughout the range of integration.
Z 1
Now let w = 1−2 cos x. Then, by straightforward calculation, A(1) = F (w)dw,
s −1
1 (3 + w)(1 − w)
where F (w) = w(2 + w) .
2 (3 − w)(1 + w)
w2 (1 + w2 )
Finally, since F (w) + F (−w) = − p , we see that A(1) < 0.
(9 − w2 )(1 − w2 )
But A(h) is continuous and must therefore have at least one zero in each of the
3
intervals (0,1) and (1, ).
2
5. Let pn be the probability that c + d is a perfect square when the integers c

and d are selected independently at random from the set (1,2,3,..,n). Show that
 3

√ √
lim (pn n) exists and express this limit in the form r( s − t), where s and t
n→∞

are integers and r is a rational number.

William Lowell Putnam Competition, 1982

Solution:
√ √
Let a(n) = [ n + 1] and b(n) = [ 2n]. For t in {1, 2, .., a(n)} there are

t2 − 1 ordered pairs (c, d) with c and d in X = {1, 2, .., n} and c + d = t2 .

For t in {1 + a(n), 2 + a(n), .., b(n)} there are 2n + 1 − t2 ordered pairs (c, d)

with c and d in X and c + d = t2 . Hence the total number F (n) of favor-


a(n) b(n) a(n)
X X X
able (c, d) is F (n) = (t2 − 1) + (2n + 1 − t2 ) = 2  t2  −
t=1 t=1+a(n) t=1
b(n)
X 2 · a(n)[1 + a(n)] · [1 + 2 · a(n)]
t2 − a(n) + [b(n) − a(n)] · (2n + 1) = −
t=1
6
b(n) · [1 + b(n)] · [1 + 2 · b(n)]
− 2(n + 1) · a(n) + (2n + 1) · b(n).
6 3 3
√
 
F (n) F (n) 2·2 a(n) 2 b(n)
Since pn = , lim (pn n) = lim 3 = lim √ − · lim √ −
n2 n→∞ n→∞ n 2 6 n→∞ n √ 6 n→∞ n
a(n) b(n) 2 1 √ √ 4 2 2 √
2 · lim √ + 2 lim √ = − ( 2)3 − 2 + 2 2 = − − +2 2 =
n→∞ n n→∞ n 3 3 3 3
4 √
( 2 − 1).
3
6. For each x > ee define a sequence Sx = u0 , u1 , u2 , ... recursively as follows:

u0 = e, for n > 0, un+1 is the logarithm of x to the base un .

Prove that Sx converges to a number g(x) and that the function g defined this

way is continuous for x > ee .

William Lowell Putnam Competition, 1982

Solution:
1
Since the derivative of x x is negative for x > e, ab > ba when e 6 a < b. (1)
4

u0 = e and x = (u0 )u1 = (u1 )u2 = (u2 )u3 = ... (2)

un · ln un−1
Hence un+1 = (3)
ln un
As x > ee , u1 = ln x > e = u0 . Now u1 > u0 implies ln u1 > ln u0 and then (3)

with n u2 < u1 (??)

Also, (2) and (1) imply (u1 )u2 − (u0 )u1 > (u1 )u0 , which gives us u2 > u0 . Now

u2 < u1 and (3) with n = 2 imply u3 > u2 .

Also, (2) and (1) imply (u2 )u3 = (u1 )u2 < (u2 )u1 and hence u3 < u1 .

Similarly, u2 < u4 < u3 . Then an easy induction shows that e < u2n < u2n+2 <

u2n+1 < u2n−1 , for n = 1, 2, ..

Thus the monotonic bounded sequences u0 , u2 , u4 , .. and u1 , u3 , u5 have the lim-

its a and b, respectively, with e < a 6 b.

Also lim (u2n )u2n+1 = x = lim (u2n−1 )u2n = ba .

n→∞ n→∞

Then ab = ba , e 6 a 6 b and (1) imply a = b.

Hence lim un exists and is the unique real number g = g(x) with g > e and
n→∞

g g = x.

Since f (y) = y y is continuous and strictly increasing for y > e, its inverse func-

tion g(x) is also continuous.

Y
7. Show that det[aij ]n1 = (xq −xp )np ·nq , n1 , n2 , .., nr are positive integers
16p<q6r
with sum
 n and
 i − 1  i−k
aij =   · xq ; aij = 0, if i < k
 
k−1
with q = min{p : j 6 n1 + n2 + .. + np }, k = j − (n1 + n2 + .. + nq−1 )

AMM, 1982
 5

Note: If every nq = 1, this reduces to the standard Vandermonde formula.

Solution:

We view the generalized Vandermonde determinant as an appropriate limit in-

volving the standard Vandermonde determinant, V (h), on the nodes xj + kh

(j = 0, 1, ..m; 0 6 k < nj ) with h 6= 0.

Let R(x) denote the column vector R(x) = (1, x, x2 , .., xn−1 )T .

Set Rjk = R(xj +kh) and let ∆ be the forward difference operator on the second

index; i.e., ∆Rjk = Rj,k+1 − Rjk .

Then V (h) = det[R00 , R01 , .., R0,n0 −1 , R10 , .., Rm0 , Rm1 , .., Rm,nm −1 ] =

= det[R00 , ∆R00 , .., ∆n0 −1 R00 , R10 , .., Rm0 , ∆Rm0 , .., ∆nm −1 Rm0 ].

If we denote Dk Rj = Dk R(xj ) (and Rj = R(xj )), then the generalized Vander-

monde determinant, V , can be written as

Dn0 −1 R0 Dnm −1 Rm
 
DR0 DRm
V = det R0 , , .., , R1 , .., Rm , , ..,
1! (n0 − 1)! 1! (nm − 1)!
k
∆ R j0
But Dk Rj = lim .
h→0 hk
Consequently (after factoring out the limits and scalars)
!
i V (h)
V = lim Qm Qnj −1 .
h→0
j=1 k=0 k! Y· nk
The final formula V = (xj − xi )ni ·nj results from directly evaluating
06i<j6m
the last limit, using the standard formula for the Vandermonde V (h).