PROBLEMS AND SOLUTIONS

Source: The American Mathematical Monthly, Vol. 117, No. 9 (November 2010), pp. 834-843
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/10.4169/000298910X521724 .
Accessed: 21/03/2013 00:35

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access to
The American Mathematical Monthly.

http://www.jstor.org

This content downloaded from 128.197.27.9 on Thu, 21 Mar 2013 00:35:27 AM

All use subject to JSTOR Terms and Conditions
 PROBLEMS AND SOLUTIONS
Edited by Gerald A. Edgar, Doug Hensley, Douglas B. West
with the collaboration of Itshak Borosh, Paul Bracken, Ezra A. Brown, Randall
Dougherty, Tamás Erdélyi, Zachary Franco, Christian Friesen, Ira M. Gessel, László
Lipták, Frederick W. Luttmann, Vania Mascioni, Frank B. Miles, Bogdan Petrenko,
Richard Pfiefer, Cecil C. Rousseau, Leonard Smiley, Kenneth Stolarsky, Richard
Stong, Walter Stromquist, Daniel Ullman, Charles Vanden Eynden, Sam Vandervelde,
and Fuzhen Zhang.

Proposed problems and solutions should be sent in duplicate to the MONTHLY

problems address on the back of the title page. Submitted solutions should arrive
at that address before March 31, 2011. Additional information, such as general-
izations and references, is welcome. The problem number and the solver’s name
and address should appear on each solution. An asterisk (*) after the number of
a problem or a part of a problem indicates that no solution is currently available.

PROBLEMS
11501. Proposed by Finbarr Holland, University College Cork, Cork, Ireland. (Cor-
rection) Let
3
g(z) = 1 − .
1
1−az
+ 1
1−i z
+ 1
1+i z

Show that the coefficients

√ in the Taylor series expansion of g about 0 are all nonnega-
tive if and only if a ≥ 3.
11530. Proposed by Pál Peter Dályay, Szeged, Hungary. Let A be an m × m matrix
with nonnegative entries  ai, j and with the property that there exists a permutation σ
m
of {1, . . . , m} for which i=1 ai,σ (i) ≥ 1. Show that the union over n ≥ 1 of the set of
n
entries of A is bounded if and only if some positive power of A is the identity matrix.
11531. Proposed by Nicuşor Minculete, “Dimitrie Cantemir” University, Brasov, Ro-
mania. Let M be a point in the interior of triangle ABC and let λ1 , λ2 , λ3 be positive
real numbers. Let Ra , Rb , and Rc be the circumradii of triangles M BC, MC A, and
M AB, respectively. Show that
 
|M A| |M B| |MC|
λ1 R a + λ2 R b + λ3 R c ≥ λ1 λ2 λ3
2 2 2
+ + .
λ1 λ2 λ3
(Here, for V = A, B, C, |M V | denotes the length of the line segment M V . )
11532. Proposed by Cezar Lupu (student), University of Bucharest, Bucharest, Ro-
mania, and Vicenţiu Rădulescu, Institute of Mathematics “Simon Stoilow” of the
Romanian Academy, Bucharest, Romania. Find all prime numbers p such that there
exists a 2 × 2 matrix A with integer entries, other than the identity matrix I , for which
A p + A p−1 + · · · + A = p I .
doi:10.4169/000298910X521724

834 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 117

This content downloaded from 128.197.27.9 on Thu, 21 Mar 2013 00:35:27 AM

All use subject to JSTOR Terms and Conditions
11533. Proposed by Erwin Just (emeritus), Bronx Community College of the City Col-
lege of New York, Bronx, NY. Let t be a positive integer and let R be a ring, not neces-
sarily having an identity element, such that x + x 2t+1 = x 2t + x 10t+1 for each x in R.
Prove that R is a Boolean ring, that is, x = x 2 for all x in R.
11534. Proposed by Christopher Hillar, Mathematical Sciences Research Institute,
Berkeley, CA. Let k and n be positive integers with k < n. Characterize the n × n
real matrices M with the property that for all v ∈ Rn with at most k nonzero entries,
Mv also has at most k nonzero entries.
11535. Proposed by Marian Tetiva, Bı̂rlad, Romania. Let f be a continuously differ-
1
entiable function on [0, 1]. Let A = f (1) and let B = 0 x −1/2 f (x) d x. Evaluate
 
n  2  
1
k (k − 1)2 (k − 1)2
lim n f (x) d x − − f
n→∞ 0 k=1
n2 n2 n2
in terms of A and B.
11536. Proposed by Mihaly Bencze, Brasov, Romania. Let K , L, and M denote the
respective midpoints of sides AB, BC, and C A in triangle ABC, and let P be a point
in the plane of ABC other than K , L, or M. Show that
|AB| |BC| |C A| |AB| · |BC| · |C A|
+ + ≥ ,
|P K | |P L| |P M| 4|P K | · |P L| · |P M|
where |U V | denotes the length of segment U V .

SOLUTIONS

The Number of k-cycles in a Random Permutation

11378 [2008, 664]. Proposed by Daniel Troy (Emeritus), Purdue University–Calumet,
Hammond, IN. Let n be a positive integer, and let U1 , . . . , Un be random variables
defined by one of the following two processes:
A: Select a permutation of {1, . . . , n} at random, with each permutation of equal
probability. Then take Uk to be the number of k-cycles in the chosen permutation.
B: Repeatedly select an integer at random from {1, . . . , M} with uniform distribu-
tion, where M starts at n and at each stage in the process decreases by the value
of the last number selected, until the sum of the selected numbers is n. Then take
Uk to be the number of times the randomly chosen integer took the value k.
Show that the probability distribution of (U1 , . . . , Un ) is the same for both processes.
Solution by O.P. Lossers, Eindhoven University of Technology, Netherlands. First we
introduce a standard notation for the permutations: in each cycle put the lowest num-
ber in front, and list the cycles with the first elements in decreasing order. Next we
count the permutations of n objects where the last cycle has length k. The last cycle
starts with 1, and the other k − 1 elements are arbitrary, in any order. Hence there are
(n − 1)!/(n − k)! ways to fill the last cycle, and then the permutation can be completed
in (n − k)! ways. Hence the number of permutations in which the last cycle has length
k is (n − 1)!, independent of k. It follows that the length of the last cycle is uniformly
distributed, and the remaining cycles are produced by the same process on the remain-
ing n − k elements. Hence the production of cycle lengths from back to front under
process A emulates process B.

November 2010] PROBLEMS AND SOLUTIONS 835

This content downloaded from 128.197.27.9 on Thu, 21 Mar 2013 00:35:27 AM

All use subject to JSTOR Terms and Conditions
Editorial comment. Erich Bach noted that the use of process B to generate the cycle
lengths of random permutations has appeared before, such as in E. Bach, Exact Anal-
ysis of a Priority Queue Algorithm for Random Variate Generation, Proc. ACM-SIAM
Symposium on Discrete Algorithms (SODA), 1994, 48–56.
Also solved by E. Bach, D. Beckwith, R. Chapman (U. K.), S. J. Herschkorn, J. H. Lindsey II, R. Martin
(Germany), J. H. Nieto (Venezuela), M. A. Prasad (India), K. Schilling, J. H. Smith, P. Spanoudakis (U. K.),
R. Stong, BSI Problems Group (Germany), GCHQ Problem Solving Group (U. K.), and the proposer.

When HI = IO
11398 [2008, 948]. Proposed by Stanley Huang, Jiangzhen Middle School, Huaining,
China. Assume acute triangle ABC has its middle-sized angle at A. Suppose further
that the incenter I is equidistant from the circumcenter O and the orthocenter H . Show
that angle A has measure 60 degrees and that the circumradius of I BC is the same as
that of ABC.
Composite solution by the Editors. The restriction to acute triangles appears to be
unnecessary.
V. V. Garcia (Huelva, Spain) pointed to Problem E2282, this M ONTHLY, April
1972, pp. 397–8, where it is shown that (excepting only equilateral triangles, for
which I O = 0, and not excluding right or obtuse triangles) H I /I O is (1) less than 1,
(2) equal to 1, or (3) greater than 1, according as the middle-sized angle of the triangle
is (1) greater than, (2) equal to, or (3) less than 60◦ . Geometrically, this means that
with respect to the perpendicular bisector λ of the Euler segment, I is (1) on the H
side of λ, (2) on λ, or (3) on the O side of λ. Thus when I is equidistant from O and
H , i.e., on λ, the middle-sized angle must be 60◦ .
The second claim of this problem is too humble. Actually, when angle A has mea-
sure 60◦ , the reflection C of the circumcircle C of ABC across BC, which of course
has the same radius, contains not only I (making it the circumcircle of B I C) but also
O and H . A proof of this expanded claim was submitted to this M ONTHLY in 1998 by
W. W. Meyer as part of a solution to Problem 10547. Here, we will give a proof based
on the solution by Jerry Minkus (San Francisco, CA): Let the angles at A, B, and C
be α, β, and γ , respectively. We have shown that α = 60◦ .
Claim. I lies on C . Proof. Designate the midpoint of BC as M. Let P be the
point on the opposite side of BC from A at which the perpendicular bisector of BC
meets C . Triangles B P M and C P M are congruent, so arcs B P and C P are congruent.
Therefore angles B A P and C A P are congruent. Thus A P is the angle bisector of
B AC, and therefore A P contains I .
It is known that R 2 − I O 2 = 2Rr , which may also be observed by constructing the
diameter of C through I . Thus I A · I P = (R + O I ) · (R − O I ) = R 2 − O I 2 = 2Rr .
Since I A = r/ sin(α/2), we have I P = 2R sin(α/2). Similarly, B P and C P are also
equal to 2R sin(α/2). Hence B, C, and I all lie on a circle about P. When Eα = 60◦ ,
the radius of that circle is R, because sin(60◦ /2) = 1/2. Hence P is the reflection in
BC of O, and the circle just referenced containing B, C, and I is the circle C .
Claim. O lies on C . Proof. O and P are reflections of each other in BC.
Claim. H lies on C . Proof. Note that AH = 2R cos α. This may be seen by ex-
tending ray C O to meet C , say at Q. Then since C Q is a diameter, its length is 2R,
angle C B Q is right, and
B QC =
B AC = α, so B Q = 2R cos α. Now B Q is par-
allel to AH , and similarly, AQ is parallel to B H . Thus AH B Q is a parallelogram and
AH = B Q = 2R cos α. Here we have α = 60◦ and cos 60◦ = 1/2, so AH = R. We
may conclude that AO P H is a parallelogram, since AH is parallel to O P and of the

836 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 117

This content downloaded from 128.197.27.9 on Thu, 21 Mar 2013 00:35:27 AM

All use subject to JSTOR Terms and Conditions
same length. (It is in fact a rhombus.) It follows that P H = AO = R. Thus as claimed
H lies on C .
Editorial comment. The Blundon result from E2282 may be strengthened in an inter-
esting way due to Francisco Bellot Rosado (Spain), who submitted it to this M ONTHLY
in 1998 as part of a solution to Problem 10547: Let G denote the centroid of the tri-
angle. The incenter I always lies inside the circle whose diameter is G H , because the
angle G I H is always obtuse. Since the perpendicular bisector λ of the Euler segment
O H divides the circle of Bellot Rosado into a larger and a smaller piece, I is (1) in the
larger piece, (2) on line λ, or (3) in the smaller piece, according as the middle-sized
angle of ABC is (1) greater than, (2) equal to, or (3) less than 60◦ .
Also solved by M. Bataille (France), R. Chapman (U. K.), C. Curtis, Y. Dumont (France), D. Fleischman,
V. V. Garcia (Spain), D. Grinberg, J.-P. Grivaux (France), E. Hysnelaj (Australia) & E. Bojaxhiu (Albania), O.
Kouba (Syria), J. H. Lindsey II, J. Minkus, R. Stong, M. Tetiva (Romania), D. Vacaru (Romania), Z. Vörös
(Hungary), M. Vowe (Switzerland), J. B. Zacharias & K. Greeson, GCHQ Problem Solving Group (U. K.),
Microsoft Research Problems Group, and the proposer.

An Alternating Series
11409 [2009, 83]. Proposed by Paolo Perfetti, Mathematics Department, University
“Tor Vergata,” Rome, Italy. For positive real α and β, let

N

n
α + k log k
S(α, β, N ) = n log(n)(−1)n .
n=2 k=2
β + (k + 1) log(k + 1)
Show that if β > α, then lim N →∞ S(α, β, N ) exists.
Solution by Hongwei Chen, Christopher Newport University, Newport News, VA. Let
ωk = k log k. Write


n  
n
α + ωk β −α (α + ω2 )ωn
an = ωn = bn 1− , where bn = , (1)
k=2
β + ωk+1 k=3
β + ωk β + ωn+1
and suppose β > α. We will prove that

∞
(−1)n an converges,
n=2

so lim N →∞ S(α, β, N ) exists. By the alternating series test of Leibniz, and noting an >
0, it suffices to prove
(i) an+1 /an < 1 for all sufficiently large n, and
(ii) an → 0 as n → ∞.
(i) From the definition of an in (1),
an+1 ωn+1 (α + ωn+1 )
= ,
an ωn (β + ωn+2 )
so an+1 /an < 1 is equivalent to ωn+1 α + (ωn+1
2
− ωn ωn+2 ) < ωn β. Calculation shows
ωn+1 − ωn ωn+2 = (log n) + log n + 1 + o(1). Because β > α and ωn+1 ∼ ωn =
2 2

n log n, the required result follows.

(ii) Because limn→∞ bn exists, to show limn→∞ an = 0 it suffices to show that the
infinite product

∞  
β −α
1− (2)
k=3
β + ωk

November 2010] PROBLEMS AND SOLUTIONS 837

This content downloaded from 128.197.27.9 on Thu, 21 Mar 2013 00:35:27 AM

All use subject to JSTOR Terms and Conditions
 ∞
diverges to zero. Recall that if 0 < ck < 1 for all k and
 k=1 ck diverges, then
∞
k=1 (1 − c k ) diverges to 0. In the present case, the divergence of
∞
1 ∞
1
=
k=3
ωk k=3
k log k

shows that the infinite product in (2) diverges to 0. (That the sum diverges is well
known, as it follows from the integral test or Cauchy condensation test.)
Also solved by S. Amghibech (Canada), P. Bracken, R. Chapman (U. K.), P. P. Dályay (Hungary), D. Grinberg,
J. Grivaux (France), E. A. Herman, O. Kouba (Syria), J. H. Lindsey II, A. Stadler (Switzerland), R. Stong, M.
Tetiva (Romania), BSI Problems Group (Germany), GCHQ Problem Solving Group (U. K.), and the proposer.

A Fix for a Triangle Inequality

11413 [2009, 179]. Proposed
 3Let θi for 1 ≤ i ≤ 5
by Mihály Bencze, Brasov, Romania.
be nonnegative, with 31 θi = π, θ4 = θ1 , and θ5 = θ2 . Let S = i=1 sin θi . Show that
√       
3 3 1 1 √ 1
S≤ − 4 max sin2 (θi − θi+1 ) cos θi+2 + 3 sin2 (π − 3θi+2 ) .
2 1≤i≤3 (4) 2 12

Solution by Richard Stong, San Diego, CA. (The originally published statement had
a misprint, with “2” where “(4)” now stands.) If A, B, C ≥ 0 with A + B + C = π,
then
S = sin A + sin B + sin C = 4 cos(A/2) cos(B/2) cos(C/2).
Hence
S + 4 sin2 ((A − B)/4) cos(C/2) = 4 cos2 ((A + B)/4) cos(C/2)
= 4 cos2 (π − C)/4) cos(C/2).
Applying the identity
4 cos(x + 2y) cos2 (x − y) + 8 sin2 y cos x = 4 cos3 x − 4 sin2 y cos(x − 2y)
with x = π/6 and y = (π − 3C)/12, we have
√
C 2 π −C
√ 2 π − 3C 3 3 π − 3C 2π − 3C
4 cos cos + 4 3 sin = − 4 sin2 cos
2 4 12 2 12 6
or, combined with the above,
√
A−B C √ 2 π − 3C 3 3 π − 3C 2π − 3C
S + 4 sin 2
cos + 4 3 sin = − 4 sin2 cos .
4 2 12 2 12 6
Since 0 ≤ C ≤ π, the last cosine is nonnegative, and hence
√
A−B C √ 2 π − 3C 3 3
S + 4 sin 2
cos + 4 3 sin ≤ .
4 2 12 2
Apply this result three times, taking (A, B, C) to be (θ1 , θ2 , θ3 ), then (θ2 , θ3 , θ1 ), and
finally (θ3 , θ1 , θ2 ), to obtain the desired result.
Editorial comment. Some solvers corrected the problem by showing that it holds as
originally printed but with the inequality reversed.

838 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 117

This content downloaded from 128.197.27.9 on Thu, 21 Mar 2013 00:35:27 AM

All use subject to JSTOR Terms and Conditions
Also solved by R. Bagby, P. P. Dályay (Hungary), J. H. Lindsey II, GCHQ Problem Solving Group (U. K.),
and Microsoft Research Problems Group.

Blundon’s Inequality Improved

11414 [2009, 179]. Proposed by Marian Tetiva, National College “Gheorghe Roşca
Codreanu,” Bı̂rlad, Romania. Let ABC be a triangle with largest angle at A, let A also
denote the measure of that angle, let c = cot(A/2), and let s, r , and R be the semi-
perimeter, inradius, and circumradius
√ of the triangle, respectively. Show that Blundon’s
Inequality s ≤ 2R + r (3 3 − 4) can be strengthened to
 √
√ ( 3 − c)3
s ≤ 2R + r 3 3 − 4 − .
4c

Solution by Oliver Geupel, Brühl, NRW, Germany.

Lemma. If a, b, c are positive real numbers a + b + c = abc and c =

√ such that √
min{a, b, c}, then (a − 1)(b − 1)(c − 1) ≤ 6 3 − 10 − ( 3 − c)3 /(2c).

Proof. Note that ba = (abc)/c > (abc)/(a + b + c) = 1, and similarly bc > 1 and
ca > 1. Thus at most one of the numbers a, b, c can be less than 1. Hence a ≥ 1 and
b ≥ 1. The equality a + b + c = abc yields c = (a + b)/(ab − 1). We must show that
if a, b ≥ 1 and ab > 1, then f (a, b) ≤ 0, where
  √ 3
a+b √ 3(ab − 1) − (a + b)
f (a, b) = (a − 1)(b − 1) − 1 − (6 3 − 10) − .
ab − 1 2(ab − 1)2 (a + b)
Put a = 1 + x and b = 1 + y with x, y ≥ 0, and rewrite the function as

f (1 + x, 1 + y) = −2(x + y + 2)(x + y + x y) x 2 y 2
√ √ 
+(6 3 − 12)x y + (6 3 − 10)(x + y) .
√ √
Observe x + y ≥ 2 x y and substitute t = x y to reduce the inequality to p(t) ≥ 0
√ √
for all t ≥ 0, where p(t) = t√
4
+ (6 3 − 12)t√ 2
+ (12 3 − 20)t. This follows from the
factorization p(t) = t (t − ( 3 − 1))2 (t + 2 3 − 2).

In triangle ABC, the numbers a = cot(A/2), b = cot(B/2), and c = cot(C/2) sat-

isfy a + b + c = s/r = abc, ab + bc + ca = (4R + r )/r , and c = min{a, b, c}. By
the lemma,
r 
s = (a − 1)(b − 1)(c − 1) + ab + bc + ca + 1
2
 √
r √ ( 3 − c)3 4R + r
≥ 6 3 − 10 − + +1
2 2c r
 √
√ ( 3 − c)3
= 2R + r 3 3 − 4 − .
4c

Equality holds if and only if the triangle is equilateral.

November 2010] PROBLEMS AND SOLUTIONS 839

This content downloaded from 128.197.27.9 on Thu, 21 Mar 2013 00:35:27 AM

All use subject to JSTOR Terms and Conditions
Editorial comment. Richard Stong proved the stronger inequality
 √ √
√ 9(2 − 3 ) ( 3 − c)2
s ≤ 2R + r 3 3 − 4 − .
8 c2

Also solved by J. H. Lindsey II, C. R. Pranesachar (India), R. Stong, GCHQ Problem Solving Group (U. K.),
and the proposer.

Closed-Form Definite Integral

11416 [2009, 180]. Proposed by Yaming Yu, University of California Irvine, Irvine,
CA. Let f be the decreasing function on (0, ∞) that satisfies
f (x)e− f (x) = xe−x .
(To visualize, draw a graph of the function xe−x and a horizontal line that is tangent to
it or crosses it at two points; if one of these points is x, then the other is f (x).) Show
that
 ∞
2π 2
x −1/6 ( f (x))1/6 d x = .
0 3

Solution by O. P. Lossers, Eindhoven University of Technology, Eindhoven, The

Netherlands. From the definition of f , we have f (x)/x = e f (x)−x . Writing u =
f (x)/x and eliminating f (x) gives x = log u/(u − 1), so that as x increases from
0 to ∞, u decreases from ∞ to 0. The integral to be computed, call it A, can then
∞
be written as A = 0 u α (x) d x (with α = 1/6). Integrating first by parts and then
changing variables from x to u in the resulting integral gives
 ∞ ∞  ∞ α−1
α α  u log u
A= u (x) d x = xu (x) +α du.
x=0 x=0 u=0 u − 1
Here we could refer to Gradshteyn & Ryzhik (formula 4.254.1) and Abramowitz &
Stegun (formula 6.4.7). In this special case, though, there is a simpler solution. For
0 < α < 1 the integral converges. The first term on the right-hand side is zero because
it is equal to u α log(u)/(u − 1)|0u=∞ . Split the second term into two parts:
 1 α−1  ∞ α−2
u log u u log u
−α du + α du.
0 1−u 1 1 − 1/u
Expand (1 − u)−1 and (1 − 1/u)−1 as geometric series, then integrate:

∞
1  ∞
1
α + α .
n=0
(n + α) 2
n=1
(n + 1 − α)2

Using the Hurwitz zeta function notation ζ (s, a) = ∞ −s
n=0 (n + a) , for arbitrary
α in (0, 1) this can be written as α(ζ (2, α) + ζ (2, 1 − α)). Starting with the known
fact that ζ (2) = ζ (2, 1) = π 2 /6, elementary calculations give ζ (2, 1/2) = 3ζ (2) and
ζ (2, 1/3) + ζ (2, 2/3) = 8ζ (2), so that ζ (2, 1/6) + ζ (2, 5/6) is given by

6 
2
ζ (2, k/6) − ζ (2, k/3) − ζ (2, 1/2) − ζ (2, 1) = (36 − 8 − 3 − 1)ζ (2) = 4π 2 .
k=1 k=1

The required sum is thus 4απ 2 = 2π 2 /3.

840 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 117

This content downloaded from 128.197.27.9 on Thu, 21 Mar 2013 00:35:27 AM

All use subject to JSTOR Terms and Conditions
Also solved by R. Bagby, D. Beckwith, P. Bracken, B. S. Burdick, P. Corn, L. Gerber, M. L. Glasser, J. Grivaux
(France), E. A. Herman, F. Holland & T. Carroll (Ireland), K. McInturff, O. G. Ruehr, V. Rutherfoord, R. Stong,
J. B. Zacharias, BSI Problems Group (Germany), GCHQ Problem Solving Group (U. K.), Microsoft Research
Problems Group, and the proposer.

An Integral-Derivative Inequality
11417 [2009, 180]. Proposed by Cezar Lupu (student), University of Bucharest,
Bucharest, Romania, and Tudorel Lupu, Decebal High School, Constanţa, Roma-
nia. Let f be a continuously differentiable real-valued function on [0, 1] such that
 2/3 1  2
1
1/3
f (x) d x = 0. Show that 0
( f (x)) 2
d x ≥ 27 0
f (x) d x .

Solution by Moubinool Omarjee, Paris, France. Let h(x) be the continuous, piecewise
linear function given by
⎧
⎪
⎨−x, 0 ≤ x ≤ 1/3,
h(x) = 2x − 1, 1/3 ≤ x ≤ 2/3,
⎪
⎩1 − x, 2/3 ≤ x ≤ 1.

Integrating by parts gives

 1  1  2/3  1

h(x) f (x) d x = f (x) d x − 3 f (x) d x = f (x) d x,
0 0 1/3 0

and we compute that

 1
1
h(x)2 d x = .
0 27

Hence the Cauchy-Schwarz inequality applied to h and f reads

 1  1 2

( f (x)) d x ≥ 27 2
f (x) d x ,
0 0

as desired.
Editorial comment. Several solvers remarked that this problem generalizes with
essentially the same proof. In the simplest form, suppose that φ(x) is an inte-
1 x
grable function with 0 φ(x) d x = 1, and define h(x) = −x + 0 φ(t) dt and C =
1
h(x)2 d x. For any continuously differentiable real-valued function f on [0, 1] such
0 
1
that 0 f (x)φ(x) d x = 0, one has
 1  1 2

C ( f (x)) d x ≥ 2
f (x) d x .
0 0

More generally, this holds with φ(x) d x replaced by a signed Borel measure.
Also solved by K. F. Andersen (Canada), R. Chapman (U. K.), P. P. Dályay (Hungary), P. J. Fitzsimmons,
O. Geupel (Germany), J. Grivaux (France), G. Keselman, O. Kouba (Syria), J. H. Lindsey II, O. P. Lossers
(Netherlands), D. S. Ross, R. Tauraso (Italy), P. Venkataramana, E. I. Verriest, FAU Problem Solving Group,
GCHQ Problem Solving Group (U. K.), St. John’s University Problem Solving Group, and the proposers.

November 2010] PROBLEMS AND SOLUTIONS 841

This content downloaded from 128.197.27.9 on Thu, 21 Mar 2013 00:35:27 AM

All use subject to JSTOR Terms and Conditions
 Gamma Products
11426 [2009, 365]. Proposed by M. L. Glasser, Clarkson University, Potsdam, NY.
Find
(1/14)(9/14)(11/14)
,
(3/14)(5/14)(13/14)
∞
where  denotes the usual gamma function, given by (z) = 0
t z−1 e−t dt.
Solution by Matthew A. Carlton, Cal Poly State University, San Luis Obispo, CA. The
multiplication formula for the gamma function may be written as
√ (2z)
(z) = 2 π · 2−2z · .
(z + 1/2)
Apply this with z equal to each of the six values in the original expression, e.g.
√ (1/7)
(1/14) = 2 π · 2−1/7 · .
(4/7)
The numerator of the original expression can then be written
√ (1/7)(9/7)((11/7)
(2 π )3 · 2−1/7−9/7−11/7 ·
(4/7)(8/7)(9/7)
1 (1/7) · 4/7(4/7)
= 8π 3/2 · · = 4π 3/2 .
8 (4/7) · 1/7(1/7)
Similarly, the denominator simplifies to 2π 3/2 . Thus the quotient is 2.
Editorial comment. Some solvers provided generalizations. The most interesting and
complete was from Albert  Stadler
(Switzerland). Let p be an odd prime, and denote
the Legendre symbol by kp . Then
 ⎧1, if p ≡ 1 (mod 8),
  ⎪
⎨

p 2k−1
2k − 1 p
= ε( p) , if p ≡ 5 (mod 8),
h( p)
 (∗)
k=1
2p ⎪
⎩ − k=1
p−1 k k
2 p p, if p ≡ 7 (mod 8),
where ε( p) denotes the fundamental unit and h( p) the class number of the real
√
quadratic field Q( p ). The case p ≡ 3 (mod 8) was not resolved. The fundamental
√
unit ε( p) = (x + y p )/2 is a solution of Pell’s equation x 2 − py 2 = 4 with the
property that both x and y are positive and y is minimal. The result asked for here is
the case p = 7. Other examples ( p = 5, 13, 17):
√
(1/10)(9/10) 3+ 5
= ,
(3/10)(7/10) 2
√
(1/26)(3/26)(9/26)(17/26)(23/26)(25/26) 11 + 3 13
= ,
(5/26)(7/26)(11/26)(15/26)(19/26)(21/26) 2
(1/34)(9/34)(13/34)(15/34)(19/34)(21/34)(25/34)(33/34)
= 1.
(3/34)(5/34)(7/34)(11/34)(23/34)(27/34)(29/34)(31/34)
Since the values in (∗) are algebraic numbers, we have a corollary: If p is an odd prime

≡ 3 (mod 8), then the p − 1 numbers ((2k − 1)/(2 p)), 1 ≤ k ≤ p, k

= ( p − 1)/2,
are algebraically dependent.

842 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 117

This content downloaded from 128.197.27.9 on Thu, 21 Mar 2013 00:35:27 AM

All use subject to JSTOR Terms and Conditions
Also solved by Z. Ahmed & M. A. Prasad (India), K. F. Andersen (Canada), R. Bagby, B. Bauldry, D. Beck-
with, P. Bracken, M. A. Carlton, R. Chapman (U. K.), H. Chen, C. K. Cook, P. Costello, P. P. Dályay (Hungary),
F. Flores & F. Mawyer, M. R. Gopal, D. Gove, G. C. Greubel, D. Grinberg, J. Grivaux (France), J. A. Grze-
sik, C. C. Heckman, E. A. Herman, D. Hou, R. Howard, E. Hysnelaj (Australia) & E. Bojaxhiu (Albania), G.
Keselman, T. Konstantopoulis (U. K.), O. Kouba (Syria), V. Krasniqi (Kosova), H. Kwong, G. Lamb, O. P.
Lossers (Netherlands), R. Martin (Germany), K. McInturff, A. Nijenhuis, O. Padé (Israel), R. Padma (India),
C. R. Pranesachar (India), H. Riesel (Sweden), I. Rusodimos, O. A. Saleh & S. Byrd, A. S. Shabani (Kosova),
M. A. Shayib, N. C. Singer, A. Stadler (Switzerland), R. Stong, T. Tam, R. Tauraso (Italy), Z. Vörös (Hun-
gary), M. Vowe (Switzerland), Z. Wenlong (China), Con Amore Problem Group (Denmark), GCHQ Problem
Solving Group (U. K.), Microsoft Research Problems Group, NSA Problems Group, and the proposer.

An Equilateral Condition
11427 [2009, 365]. Proposed by Viorel Bǎndilǎ, C.A. Rosetti High School, Bucharest,
Romania. In a triangle ABC, let m be the length of the median from A, l the length
of the angle bisector from B, and h the length of the altitude from C. Let a, b, and
c be the lengths of the edges opposite A, B, and C, respectively. Show that ABC is
equilateral if and only if a 2 + m 2 = b2 + l 2 = c2 + h 2 .
Solution by Bianca-Teodora Iordache, student,“Carol I” High School, Craiova, Ro-
mania. If ABC is equilateral, then a = b = c and m = l = h, so the equations hold.
We must prove the converse. Let m a , la , and h a denote the lengths of the median, angle
bisector, and altitude, respectively, corresponding to the edge a, and define similar no-
tation for edges b and c. We must prove that
a 2 + m 2a = b2 + lb2 = c2 + h 2c ⇒ a = b = c.
Claim 1. a + 2
m 2a ≤b +2
m 2b ⇐⇒ a ≤ b. Indeed,
2(b2 + c2 ) − a 2 3a 2 + 2b2 + 2c2
a 2 + m 2a = a 2 + = .
4 4
Hence a 2 + m 2a ≤ b2 + m 2b ⇐⇒ 3a 2 + 2b2 + 2c2 ≤ 3b2 + 2a 2 + 2c2 ⇐⇒ a 2 ≤
b2 ⇐⇒ a ≤ b.
Claim 2. a 2 + h 2a ≤ b2 + h 2b ⇐⇒ a ≤ b. Using h a = 2S/a, where S is the area of
ABC, we have a 2 + h 2a = a 2 + 4S 2 /a 2 , so
a 2 b2 − 4S 2
a 2 + h 2a ≤ b2 + h 2b ⇐⇒ (b2 − a 2 ) ≥ 0 ⇐⇒ b ≥ a.
a 2 b2
Also recall that h a ≤ la ≤ m a and similarly for b, c. Next suppose that a 2 + m 2a =
b2 + lb2 = c2 + h 2c . We have a 2 + m 2a = c2 + h 2c ≤ c2 + m 2c , so a ≤ c from Claim 1.
We have c2 + h 2c = b2 + lb2 ≥ b2 + h 2b , so b ≤ c by Claim 2. From  the Heron
 formula,
16S

2
= (a + b + c)(−a + b + c)(a − b + c)(a + b − c) = 2 a 2 2
b − a 4 , using
for sums over cyclic permutations of the triangle. Now a + m a = c + h 2c so
2 2 2
 
3a 2 + 2b2 + 2c2 2 a 2 b2 − a 4
=c +2
,
4 4c2
so c2 (3a 2 + 2b2 − 2c2 ) = 2(a 2 b2 + b2 c2 + c2 a 2 ) − (a 4 + b4 + c4 ) and thus
c2 (c2 − a 2 ) = (b2 − a 2 )2 . (1)
Since c2 ≥ b2 > b2 − a 2 and c2 − a 2 ≥ b2 − a 2 , for equality in (1) we must have
c2 − a 2 = b2 − a 2 = 0. This shows a = c and a = b as required.
Also solved by R. Bagby, M. Bataille (France), H. Caerols (Chile), R. Chapman (U. K.), G. Crandall, P. P.
Dályay (Hungary), D. Fleischman, D. Gove, J. Grivaux (France), O. Kouba (Syria), J. H. Lindsey II, J.
McHugh, J. Minkus, M. A. Prasad (India), A. Stadler (Switzerland), R. Stong, M. Tetiva (Romania), Z. Vörös
(Hungary), GCHQ Problem Solving Group (U. K.), Microsoft Research Problems Group, and the proposer.

November 2010] PROBLEMS AND SOLUTIONS 843

This content downloaded from 128.197.27.9 on Thu, 21 Mar 2013 00:35:27 AM

All use subject to JSTOR Terms and Conditions