Putnam 1986 Solutions

Evan O’Dorney
MATH 191, Prof. Alexander Givental
September 12, 2009

A1. The inequality x4 + 36 ≤ 13x2 factors as (x + 3)(x + 2)(x − 2)(x − 3) ≤ 0, i.e. −3 ≤ x ≤ −2 or 2 ≤ x ≤ 3.

We compute f 0 (x) = 3(x + 1)(x − 1), so f is increasing on both these intervals. Since f (−2) = −2 and f (3) = 18,
18 is the maximum.
A2.
∞ 199
1020000 1019900 X
i 19900−100i
X
X = 100 = = (−3) 10 ≈ (−3)i 1019900−100i = Y.
10 + 3 1 + 103100 i=0 i=0

The finite sum Y is a positive integer, and all of its terms but the last are divisible by 10: Y ≡ (−3)199 ≡ 3
(mod 10). Since X is an alternating series, the difference X − Y has the same sign and smaller magnitude than
the first omitted term (−3)200 10−100 = (9/10)100 , which is positive and less than 1. Consequently bXc = Y , and
the rightmost digit of bXc is 3.
A3. The identity Arccot(n2 + n + 1) = Arccot(n) − Arccot(n + 1) is provable from the addition formula for
PN
tangent when n > 0 and becomes the identity π/4 = π/2 − π/4 when n = 0. Thus n=0 Arccot(n2 + n + 1) =
Arccot(0) − Arccot(N + 1), which as N → ∞ tends to π/2 − 0 = π/2.
A4. It is not hard to see, by completing the pairs (α1,1 , αi,j ) and (αi,1 , α1,j ) to a transversal in the same way
when i, j 6= 1, that all transversals of a matrix A = [αi,j ] have the same sum if and only if αi,j = αi,1 +α1,j −α1,1 for
all i and j. For any matrix A of the given type, let X = max1≤i,j≤n (α1,i −α1,j ) and Y = max1≤k,m≤n (αk,1 −αm,1 ).
Note that X + Y ≤ 2, since (α1,i − α1,j ) + (αk,1 − αm,1 ) = αk,i − αm,j ≤ 1 − (−1) = 2. Using inclusion-exclusion
to compute the number of matrices having each of the six possible pairs (X, Y ), we get:
(X, Y ) Matrices
(0, 0) 1
(0, 1) 2 · 2n − 4
(1, 0) 2 · 2n − 4
(0, 2) 3n − 2 · 2n + 1
(2, 0) 3n − 2 · 2n + 1
(1, 1) 4n − 4 · 2n + 4
Total 4n + 2 · 3n − 4 · 2n − 1.

1
 P
A5. Let hi (x) = j<i cij xj − fi (x). Plugging into the given condition easily yields ∂hi /∂xj = ∂h Pj /∂xi for
all i and j, so the hi are the partial derivatives of some function g. Note that fi + ∂g/∂xi = fi + hi = j<i cij xj
is linear.
A6. We differentiate the given equality k times, k = 0, 1, . . . , n, divide through by k!, and then plug in x = 1
to get
X
ai = −1
i
X
ai bi = 0
i
 
X bi
ai =0
i
2
···
 
X bi
ai =0
i
n−1
X bi 
ai = (−1)n f (1).
i
n
Pn x


Let P (x) =Q(x − b1 ) · · · (x − bn ). We can uniquely express P (x) = k=0 ck k ; plugging in x = 0 yields
c0 = (−1)n bi , while considering P the leading term gives
Q cn = n!. Linearly combining Q the above equations with
coefficients c0 , . . . , cn gives 0 = ai P (bi ) = (−1)n+1 bi + (−1)n n!f (1), so f (1) = bi /n!.
B1. Since they have the same area and the same base, the triangle is twice as high as the rectangle. The
diameter of the circle through the apex of the triangle is thus cut by the sides of the rectangle in the ratio 2 : 1 : 2,
and so h = 2/5.
B2. Answer: (0, 0, 0), (1, 0, −1), (0, −1, 1), (−1, 1, 0). Let x − y = u, y − z = v, and z − x = w. Moving all
terms of the first equation to the left side and factoring gives (x − y)(x + y + 1 − 2z) = 0, i.e. u = 0 or w − v = 1.
Similarly, the second equation reads v = 0 or u − w = 1. Finally, the triples (u, v, w) corresponding to triples
(x, y, z) are exactly those satisfying u + v + w = 0. Solving the systems of three linear equations corresponding
to each of the four cases yields the four solutions above.
B3. We induct on n with the base case n = 1 already given. Let Fn and Gn be the polynomials in the
induction hypothesis and let u = h − Fn Gn ≡ 0 (mod pn ). Let Fn+1 = Fn + us and Gn+1 = Gn + ur, noting
that Fn+1 ≡ Fn ≡ f (mod p) and similarly for Gn+1 . We get

Fn+1 Gn+1 = Fn Gn + u(rFn + sGn ) + u2 = Fn Gn + u + u(rFn + sGn − 1) + u2 .

The last term clearly drops out (mod pn+1 ) since it is divisible by p2n . Also, the next-to-last term is the product
of a multiple of pn and a multiple of p, so it is also 0 (mod pn+1 ). We are left with Fn Gn + u = h, as desired.√
B4. The assertion is true. Let M be a fixed positive integerpand M ≤ r < M +1; √ we prove 0 ≤ G(r)
√ ≤ 3/ M ,
2 2 2 2
√ → ∞. Specifically, we prove that M + 2(n +√1) − M +√2n < 3/ M whenever
which clearly goes to 0 as M
n is an integer with M ≤ M 2 2 2 2 To do this,
√ + 2n < M + 1, so that every r is within 3/pM of some M + 2n . √
we first solve the condition M 2 + 2n2 < M + 1 for n, yielding 0 ≤ n < M + 1/2, i.e. 0 ≤ n ≤ M . Then
rationalizing the numerator yields
p p 4n + 2 2n + 1 2 1 3
M 2 + 2(n + 1)2 − M 2 + 2n2 = p √ < ≤√ + ≤√ .
M2 + 2(n + 1)2 + M2 + 2n2 M M M M

B5. The assertion

 is false,
 as the example p(x, y, z) = x, q(x, y, z) = y, r(x, y, z) = xy − z shows.
 
A B T O I
B6. Let W = . The conditions of the problem reduce to W SW S = −I, where S = . This
C D −I O
says that W S and −W T S are inverses, which is equivalent to W T SW S = −I. Expanding this in A, B, C, D
and examining the upper left n × n submatrix yields AT D − B T C = I.