Solutions to the 71st William Lowell Putnam Mathematical Competition

Saturday, December 4, 2010

Kiran Kedlaya and Lenny Ng

A1 The largest such k is ⌊ n+1 n

2 ⌋ = ⌈ 2 ⌉. For n even, this Write n = 2m k with m a nonnegative integer and k a
value is achieved by the partition positive odd integer. For any nonnegative integer j,

102 j
≡ (−1)j (mod 102 + 1).
m m
{1, n}, {2, n − 1}, . . . ;

for n odd, it is achieved by the partition Since 10n ≥ n ≥ 2m ≥ m + 1, 10n isndivisible by 2n

andnhence by 2m+1 , and similarly 1010 is divisible by
{n}, {1, n − 1}, {2, n − 2}, . . . . 210 and hence by 2m+1 . It follows that

(mod 102 + 1).

m
One way to see that this is optimal is to note that the N ≡ 1 + 1 + (−1) + (−1) ≡ 0
common sum can never be less than n, since n itself
Since N ≥ 1010 > 10n + 1 ≥ 102 + 1, it follows
n m
belongs to one of the boxes. This implies that k ≤ (1 +
· · · + n)/n = (n + 1)/2. Another argument is that if that N is composite.
k > (n + 1)/2, then there would have to be two boxes
with one number each (by the pigeonhole principle), but A5 We start with three lemmas.
such boxes could not have the same sum.
Lemma 1. If x, y ∈ G are nonzero orthogonal vectors, then
Remark. A much subtler question would be to find x ∗ x is parallel to y.
the smallest k (as a function of n) for which no such
arrangement exists.
Proof. Put z = x × y 6= 0, so that x, y, and z = x ∗ y are
A2 The only such functions are those of the form f (x) = nonzero and mutually orthogonal. Then w = x × z 6= 0, so
cx + d for some real numbers c, d (for which the prop- w = x ∗ z is nonzero and orthogonal to x and z. However, if
erty is obviously satisfied). To see this, suppose that f (x∗x)×y 6= 0, then w = x∗(x∗y) = (x∗x)∗y = (x∗x)×y
has the desired property. Then for any x ∈ R, is also orthogonal to y, a contradiction.

2f ′ (x) = f (x + 2) − f (x) Lemma 2. If x ∈ G is nonzero, and there exists y ∈ G

= (f (x + 2) − f (x + 1)) + (f (x + 1) − f (x)) nonzero and orthogonal to x, then x ∗ x = 0.
= f ′ (x + 1) + f ′ (x).
Proof. Lemma 1 implies that x ∗ x is parallel to both y and
′
Consequently, f (x + 1) = f (x). ′ x × y, so it must be zero.
Define the function g : R → R by g(x) = f (x + 1) − Lemma 3. If x, y ∈ G commute, then x × y = 0.
f (x), and put c = g(0), d = f (0). For all x ∈ R,
g ′ (x) = f ′ (x + 1) − f ′ (x) = 0, so g(x) = c identically,
and f ′ (x) = f (x + 1) − f (x) = g(x) = c, so f (x) = Proof. If x × y 6= 0, then x ∗ y = x × y = −y × x = −y ∗ x,
cx + d identically as desired. so x ∗ y 6= y ∗ x.

A3 If a = b = 0, then the desired result holds trivially,

We proceed now to the proof. Assume by way of con-
so we assume that at least one of a, b is nonzero. Pick
tradiction that there exist a, b ∈ G with a × b 6= 0. Put
any point (a0 , b0 ) ∈ R2 , and let L be the line given
c = a × b = a ∗ b, so that a, b, c are nonzero and lin-
by the parametric equation L(t) = (a0 , b0 ) + (a, b)t
early independent. Let e be the identity element of G.
for t ∈ R. By the chain rule and the given equation, we
d Since e commutes with a, b, c, by Lemma 3 we have
have dt (h◦ L) = h◦ L. If we write f = h◦ L : R → R,
e × a = e × b = e × c = 0. Since a, b, c span R3 ,
then f ′ (t) = f (t) for all t. It follows that f (t) = Cet
e × x = 0 for all x ∈ R3 , so e = 0.
for some constant C. Since |f (t)| ≤ M for all t, we
must have C = 0. It follows that h(a0 , b0 ) = 0; since Since b, c, and b × c = b ∗ c are nonzero and mutually
(a0 , b0 ) was an arbitrary point, h is identically 0 over orthogonal, Lemma 2 implies
all of R2 .
b ∗ b = c ∗ c = (b ∗ c) ∗ (b ∗ c) = 0 = e.
A4 Put
10n
Hence b ∗ c = c ∗ b, contradicting Lemma 3 because
N = 1010 + 1010 + 10n − 1.
n
b × c 6= 0. The desired result follows.
 2

A6 First solution. Note that the hypotheses on f imply converges absolutely. The additional measure-theoretic
that f (x) > 0 for all x ∈ [0, +∞), so the integrand is a argument at the beginning is needed because one cannot
continuous function of f and the integral makes sense. bound − log(1 − t) by a fixed multiple of t uniformly
Rewrite the integral as for all t ∈ [0, 1).
Z ∞
f (x + 1)
 Second solution. (Communicated by Paul Allen.) Let
1− dx, b > a be nonnegative integers. Then
0 f (x)
b b−1 Z 1
f (x) − f (x + 1) f (x + k) − f (x + k + 1)
Z
and suppose by way of contradiction that it converges
X
dx = dx
to a finite limit L. For n ≥ 0, define the Lebesgue a f (x) f (x + k)
k=a 0
measurable set Z 1Xb−1
f (x + k) − f (x + k + 1)
f (x + n + 1) = dx
In = {x ∈ [0, 1] : 1 − ≤ 1/2}. 0 k=a f (x + k)
f (x + n)
Z 1Xb−1
1
f (x + k) − f (x + k + 1)
Then L ≥ ∞ ≥ dx
P
n=0 2 (1 − µ(In )), so the latter sum con- f (x + a)
verges. In particular, there exists a nonnegative integer 0 k=a

N for which ∞
P Z 1
n=N (1 − µ(In )) < 1; the intersection f (x + a) − f (x + b)
= dx.
∞ ∞ 0 f (x + a)
[ \
I= In = [0, 1] − ([0, 1] − In )
Now since f (x) → 0, given a, we can choose an in-
n=N n=N
teger l(a) > a for which f (l(a)) < f (a + 1)/2; then
f (x+a)−f (x+l(a))
then has positive Lebesgue measure. f (x+a) ≥ 1− ff(a+1)
(l(a))
> 1/2 for all x ∈ [0, 1].
By Taylor’s theorem with remainder, for t ∈ [0, 1/2], Thus if we define a sequence of integers an by a0 = 0,
  an+1 = l(an ), then
1
− log(1 − t) ≤ t + t2 sup ∞ ∞ Z an+1
f (x) − f (x + 1) f (x) − f (x + 1)
2
Z
t∈[0,1/2] (1 − t)
X
dx = dx
4 5 0 f (x) n=0 an
f (x)
= t + t2 ≤ t.
3 3 X∞ Z 1
> (1/2)dx,
For each nonnegative integer n ≥ N , we then have n=0 0
Z n 
f (x + 1) and the final sum clearly diverges.
L≥ 1− dx
N f (x) Third solution. (By Joshua Rosenberg, communicated
n−1
XZ 1 by Catalin Zara.) If the original integral converges, then

f (x + i + 1)
= 1− dx on one hand the integrand (f (x) − f (x + 1))/f (x) =
f (x + i)
i=N 0 1 − f (x + 1)/f (x) cannot tend to 1 as x → ∞. On the
n−1
XZ  
other hand, for any a ≥ 0,
f (x + i + 1)
≥ 1− dx
f (x + i)
i=N I f (a + 1)
n−1 Z
0<
3X f (x + i) f (a)
≥ log dx Z a+1
5 f (x + i + 1) 1
i=N I < f (x) dx
Z n−1 ! f (a) a
3 X f (x + i) 1
Z ∞
= log dx = (f (x) − f (x + 1)) dx
5 I f (x + i + 1) f (a) a
i=N
3 f (x + N )
Z Z ∞
f (x) − f (x + 1)
= log dx. ≤ dx,
5 I f (x + n) a f (x)
For each x ∈ I, log f (x + N )/f (x + n) is a strictly and the last expression tends to 0 as a → ∞. Hence by
increasing unbounded function of n. By R the mono- the squeeze theorem, f (a + 1)/f (a) → 0 as a → ∞, a
tone convergence theorem, the integral I log(f (x + contradiction.
N )/f (x + n)) dx grows without bound as n → +∞,
a contradiction. Thus the original integral diverges, as B1 First solution. No such sequence exists. If it did, then
desired. the Cauchy-Schwartz inequality would imply
Remark. This solution is motivated by the commonly- 8 = (a21 + a22 + · · · )(a41 + a42 + · · · )
used fact that an infinite product (1 + x1 )(1 + x2 ) · · ·
converges absolutely if and only if the sum x1 +x2 +· · · ≥ (a31 + a32 + · · · )2 = 9,
 3

contradiction. This sequence has the property that for any positive in-
Second solution. (Communicated by Catalin Zara.) teger j, the sum of the j-th powers of the terms of sn,z
Suppose that such a sequence exists. If a2k ∈ [0, 1] for equals 1/z j if j is divisible by n and 0 otherwise. More-
all k, then a4k ≤ a2k for all k, and so over, any partial sum of j-th powers is bounded in ab-
solute value by n/|z|j .
4 = a41 + a42 + · · · ≤ a21 + a22 + · · · = 2, The desired sequence will be constructed as follows.
contradiction. There thus exists a positive integer k for Suppose that we have a finite sequence which has the
which a2k ≥ 1. However, in this case, for m large, correct sum of j-th powers for j = 1, . . . , m. (For
a2m 2m 2m instance, for m = 1, we may start with the single-
k > 2m and so a1 + a2 + · · · 6= 2m.
ton sequence 1.) We may then extend it to a new se-
Third solution. We generalize the second solution to quence which has the correct sum of j-th powers for
show that for any positive integer k, it is impossible for j = 1, . . . , m + 1, by appending k copies of sm+1,z for
a sequence a1 , a2 , . . . of complex numbers to satisfy suitable choices of a positive integer k and a complex
the given conditions in case the series ak1 +ak2 +· · · con- number z with |z| < m−2 . This last restriction ensures
verges absolutely. This includes the original problem by that the resulting infinite sequence a1 , a2 , . . . is such
taking k = 2, in which case the series a21 + a22 + · · · that for each positive integer m, the series am m
1 +a2 +· · ·
consists of nonnegative real numbers and so converges is convergent (though not absolutely convergent). Its
absolutely if it converges at all. partial sums include a subsequence equal to the con-
P∞
Since the sum i=1 |ai |k converges byPhypothesis, we stant value m, so the sum of the series must equal m as
∞
can find a positive integer n such that i=n+1 |ai |k < desired.
1. For each positive integer d, we then have
  B2 The smallest distance is 3, achieved by A = (0, 0), B =
n ∞
 X 
kd 
X (3, 0), C = (0, 4). To check this, it suffices to check
kd − ai  ≤ |ai |kd < 1.

  that AB cannot equal 1 or 2. (It cannot equal 0 because
i=1 i=n+1
if two of the points were to coincide, the three points
We thus have |a1 |, . . . , |an | ≤ 1, or else the would be collinear.)
Pn cannotkd
sum i=1 ai would be bounded in absolute value The triangle inequality implies that |AC − BC| ≤ AB,
by n independently of d. But if we put r = with equality if and only if A, B, C are collinear. If
max{|a1 |, . . . , |an |} > 1, we obtain another contradic- AB = 1, we may assume without loss of generality
tion because for any ǫ > 0, that A = (0, 0), B = (1, 0). To avoid collinearity, we

X n

 must have AC = BC, but this forces C = (1/2, y) for
lim sup(r − ǫ) −kd 
 kd 
ai  > 0. some y ∈ R, a contradiction. (One can also treat this
d→∞ 
i=1
 case by scaling by a factor of 2 to reduce to the case
AB = 2, treated in the next paragraph.)
For instance, this follows from applying the root test to
the rational function If AB = 2, then we may assume without loss of gener-
ality that A = (0, 0), B = (2, 0). The triangle inequal-
n n
∞
!
X 1 X X ity implies |AC − BC| ∈ {0, 1}. Also, for C = (x, y),
kz
= akd
i zd, AC 2 = x2 +y 2 and BC 2 = (2−x)2 +y 2 have the same
i=1
1 − a i d=0 i=1 parity; it follows that AC = BC. Hence c = (1, y) for
which has a pole within the circle |z| ≤ r−1/k . (An some y ∈ R, so y 2 and y 2 + 1 = BC 2 are consecutive
elementary proof is also possible.) perfect squares. This can only happen for y = 0, but
then A, B, C are collinear, a contradiction again.
FourthPsolution. (Communicated by Noam Elkies.)
Since k a2k = 2, for each positive integer k we have Remark. Manjul Bhargava points out that more gener-
a2k ≤ 2 and so a4k ≤ 2a2k P , with equality only for ally, a Heronian triangle (a triangle with integer sides
a2k ∈ {0, 2}. Thus to have k a4k = 4, there must and rational area) cannot have a side of length 1 or 2
be a single index k for which 2 (and again it is enough to treat the case of length 2).
Pak = 2, and the other ak
The original problem follows from this because a tri-
must all equal 0. But then k a2m k = 2m 6= 2m for
any positive integer m > 2. angle whose vertices have integer coordinates has area
equal to half an integer (by Pick’s formula or the ex-
Remark. Manjul Bhargava points out it is easy to con- plicit formula for the area as a determinant).
struct sequences of complex numbers with the desired
property if we drop the condition of absolute conver- B3 It is possible if and only if n ≥ 1005. Since
gence. Here is an inductive construction (of which sev-
eral variants are possible). For n = 1, 2, . . . and z ∈ C, 2009 × 2010
1 + · · · + 2009 = = 2010 × 1004.5,
define the finite sequence 2
 
1 2πij/n for n ≤ 1004, we can start with an initial distribution
sn,z = e : j = 0, . . . , n − 1 .
z in which each box Bi starts with at most i − 1 balls (so
 4

in particular B1 is empty). From such a distribution, no so we get a solution if and only if bc − ad = 1, as

moves are possible, so we cannot reach the desired final claimed.
distribution. Second solution. (Communicated by Catalin Zara.)
Suppose now that n ≥ 1005. By the pigeonhole prin- Again, note that p and q must be nonzero. Write
ciple, at any time, there exists at least one index i for
which the box Bi contains at least i balls. We will de- p(x) = p0 + p1 x + · · · + pm xm
scribe any such index as being eligible. The following q(x) = q0 + q1 x + · · · + qn xn
sequence of operations then has the desired effect.
with pm , qn 6= 0, so that m = deg(p), n = deg(q).
(a) Find the largest eligible index i. If i = 1, proceed It is enough to derive a contradiction assuming that
to (b). Otherwise, move i balls from Bi to B1 , max{m, n} > 1, the remaining cases being treated as
then repeat (a). in the first solution.
(b) At this point, only the index i = 1 can be eligi- Put R(x) = p(x)q(x+1)−p(x+1)q(x). Since m+n ≥
ble (so it must be). Find the largest index j for 2 by assumption, the coefficient of xm+n−1 in R(x)
which Bj is nonempty. If j = 1, proceed to (c). must vanish. By easy algebra, this coefficient equals
Otherwise, move 1 ball from B1 to Bj ; in case (m − n)pm qn , so we must have m = n > 1.
this makes j eligible, move j balls from Bj to B1 . For k = 1, . . . , 2m − 2, the coefficient of xk in R(x) is
Then repeat (b).
X  j   i 
(c) At this point, all of the balls are in B1 . For i = − (pi qj − pj qi )
2, . . . , 2010, move one ball from B1 to Bi n times. k−i k−j
i+j>k,j>i

After these operations, we have the desired distribution. and must vanish. For k = 2m − 2, the only summand
is for (i, j) = (m − 1, m), so pm−1 qm = pm qm−1 .
B4 First solution. The pairs (p, q) satisfying the given
equation are those of the form p(x) = ax + b, q(x) = Suppose now that h ≥ 1 and that pi qj = pj qi is
cx + d for a, b, c, d ∈ R such that bc − ad = 1. We will known to vanish whenever j > i ≥ h. (By the pre-
see later that these indeed give solutions. vious paragraph, we initially have this for h = m − 1.)
Take k = m + h − 2 and note that the conditions
Suppose p and q satisfy the given equation; note that
i+j > h, j ≤ m force i ≥ h−1. Using the hypothesis,
neither p nor q can be identically zero. By subtracting
we see that the only possible nonzero contribution to the
the equations
coefficient of xk in R(x) is from (i, j) = (h − 1, m).
Hence ph−1 qm = pm qh−1 ; since pm , qm 6= 0, this im-
p(x)q(x + 1) − p(x + 1)q(x) = 1
plies ph−1 qj = pj qh−1 whenever j > h − 1.
p(x − 1)q(x) − p(x)q(x − 1) = 1,
By descending induction, we deduce that pi qj = pj qi
we obtain the equation whenever j > i ≥ 0. Consequently, p(x) and q(x)
are scalar multiples of each other, forcing R(x) = 0, a
p(x)(q(x + 1) + q(x − 1)) = q(x)(p(x + 1) + p(x − 1)). contradiction.
Third solution. (Communicated by David Feldman.)
The original equation implies that p(x) and q(x) have As in the second solution, we note that there are no so-
no common nonconstant factor, so p(x) divides p(x + lutions where m = deg(p), n = deg(q) are distinct
1) + p(x − 1). Since each of p(x + 1) and p(x − 1) has and m + n ≥ 2. Suppose p, q form a solution with
the same degree and leading coefficient as p, we must m = n ≥ 2. The desired identity asserts that the matrix
have  
p(x) p(x + 1)
p(x + 1) + p(x − 1) = 2p(x). q(x) q(x + 1)

If we define the polynomials r(x) = p(x + 1) − p(x), has determinant 1. This condition is preserved by re-
s(x) = q(x + 1) − q(x), we have r(x + 1) = r(x), and placing q(x) with q(x)−tp(x) for any real number t. In
similarly s(x + 1) = s(x). Put particular, we can choose t so that deg(q(x) − tp(x)) <
m; we then obtain a contradiction.
a = r(0), b = p(0), c = s(0), d = q(0).
B5 First solution. The answer is no. Suppose other-
Then r(x) = a, s(x) = c for all x ∈ Z, and hence wise. For the condition to make sense, f must be
identically; consequently, p(x) = ax+b, q(x) = cx+d differentiable. Since f is strictly increasing, we must
for all x ∈ Z, and hence identically. For p and q of this have f ′ (x) ≥ 0 for all x. Also, the function f ′ (x) is
form, strictly increasing: if y > x then f ′ (y) = f (f (y)) >
f (f (x)) = f ′ (x). In particular, f ′ (y) > 0 for all
p(x)q(x + 1) − p(x + 1)q(x) = bc − ad, y ∈ R.
 5

For any x0 , if f (x0 ) = b and f ′ (x0 ) = a > 0, then verge. One then gets a contradiction from any reason-
f ′ (x) > a for x > x0 and thus f (x) ≥ a(x − x0 ) + b able lower bound on f (y) for y large, e.g., the bound
for x ≥ x0 . Then either b < x0 or a = f ′ (x0 ) = f (x) ≥ αx2 from the second solution. (One can also
f (f (x0 )) = f (b) ≥ a(b − x0 ) + b. In the latter case, start with a linear lower bound f (x) ≥ βx, then use the
b ≤ a(x0 + 1)/(a + 1) ≤ x0 + 1. We conclude in either integral expression for g to deduce that g(x) ≤ γ log x,
case that f (x0 ) ≤ x0 + 1 for all x0 ≥ −1. which in turn forces f (x) to grow exponentially.)
It must then be the case that f (f (x)) = f ′ (x) ≤ 1 for
all x, since otherwise f (x) > x + 1 for large x. Now B6 For any polynomial p(x), let [p(x)]A denote the n × n
by the above reasoning, if f (0) = b0 and f ′ (0) = a0 > matrix obtained by replacing each entry Aij of A by
0, then f (x) > a0 x + b0 for x > 0. Thus for x > p(Aij ); thus A[k] = [xk ]A. Let P (x) = xn +
max{0, −b0/a0 }, we have f (x) > 0 and f (f (x)) > an−1 xn−1 + · · · + a0 denote the characteristic poly-
a0 x + b0 . But then f (f (x)) > 1 for sufficiently large nomial of A. By the Cayley-Hamilton theorem,
x, a contradiction.
Second solution. (Communicated by Catalin Zara.) 0 = A · P (A)
Suppose such a function exists. Since f is strictly
= An+1 + an−1 An + · · · + a0 A
increasing and differentiable, so is f ◦ f = f ′ . In
particular, f is twice differentiable; also, f ′′ (x) = = A[n+1] + an−1 A[n] + · · · + a0 A[1]
f ′ (f (x))f ′ (x) is the product of two strictly increasing = [xp(x)]A.
nonnegative functions, so it is also strictly increasing
and nonnegative. In particular, we can choose α > 0 Thus each entry of A is a root of the polynomial xp(x).
and M ∈ R such that f ′′ (x) > 4α for all x ≥ M . Then
for all x ≥ M , Now suppose m ≥ n + 1. Then

f (x) ≥ f (M ) + f ′ (M )(x − M ) + 2α(x − M )2 . 0 = [xm+1−n p(x)]A

In particular, for some M ′ > M , we have f (x) ≥ αx2 = A[m+1] + an−1 A[m] + · · · + a0 A[m+1−n]
for all x ≥ M ′ .
Pick T > 0 so that αT 2 > M ′ . Then for x ≥ T , since each entry of A is a root of xm+1−n p(x). On the
f (x) > M ′ and so f ′ (x) = f (f (x)) ≥ αf (x)2 . Now other hand,
2T 2T
1 1 f ′ (t)
Z Z
− = dt ≥ α dt; 0 = Am+1−n · P (A)
f (T ) f (2T ) f (t)2
T T = Am+1 + an−1 Am + · · · + a0 Am+1−n .
however, as T → ∞, the left side of this inequality
tends to 0 while the right side tends to +∞, a contra- Therefore if Ak = A[k] for m + 1 − n ≤ k ≤ m,
diction. then Am+1 = A[m+1] . The desired result follows by
Third solution. (Communicated by Noam Elkies.) induction on m.
Since f is strictly increasing, for some y0 , we can de- Remark. David Feldman points out that the result is
fine the inverse function g(y) of f for y ≥ y0 . Then best possible in the following sense: there exist exam-
x = g(f (x)), and we may differentiate to find that ples of n × n matrices A for which Ak = A[k] for
1 = g ′ (f (x))f ′ (x) = g ′ (f (x))f (f (x)). It follows k = 1, . . . , n but An+1 6= A[n+1] .
that g ′ (y) = 1/f (y) for y ≥ yR0 ; since g takes arbi-
∞
trarily large values, the integral y0 dy/f (y) must di-