Problems and Solutions

Source: The American Mathematical Monthly, Vol. 117, No. 4 (April 2010), pp. 370-377
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/10.4169/000298910X480865 .
Accessed: 21/03/2013 00:23

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 [email protected].

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 146.201.208.22 on Thu, 21 Mar 2013 00:23:09 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 inside front cover. Submitted solutions should arrive at
that address before August 31, 2010. 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
11474. Proposed by Cezar Lupu, student, University of Bucharest, Bucharest, Roma-
nia, and Valentin Vornicu, Aops-MathLinks forum, San Diego, CA. (Correction) Show
that when x, y, and z are greater than 1,
2 +2yz 2 +2zx 2 +2x y
(x)x (y) y (z)z ≥ ((x)(y)(z))x y+yz+zx .

11483. Proposed by Éric Pité, Paris, France. (Correction) The word “nonnegative”
should read “positive.”
11495. Proposed by Marc Chamberland, Grinnell College, Grinnell, IA. Let a, b, and
c be rational numbers such that exactly one of a 2 b + b2 c + c2 a, ab2 + bc2 + ca 2 , and
a 3 + b3 + c3 + 6abc is zero. Show that a + b + c = 0.
11496. Proposed by Benjamin Bogoşel, student, West University of Timisoara, Timi-
soara, Romania, and Cezar Lupu, student, University of Bucharest, Bucharest, Roma-
nia. For a matrix X with real entries, let s(X ) be the sum of its entries. Prove that if A
and B are n × n real matrices, then
 
n s(A A T ) + s(B B T ) − s(AB T )s(A T B) ≥
 
s(A A T )(s(B))2 + s(B B T )(s(A))2 − s(A)s(B) s(AB T ) + s(A T B) .

11497. Proposed by Mihály Bencze, Brasov, Romania. Given n real numbers x1 , . . . , xn

and a positive integer m, let xn+1 = x1 , and put

n
 m 
n
A= xk2 − xk xk+1 + xk+1
2
, B=3 xk2m .
k=1 k=1

Show that A ≤ 3m B and A ≤ (3m B/n) . n

doi:10.4169/000298910X480865

370 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 117

This content downloaded from 146.201.208.22 on Thu, 21 Mar 2013 00:23:09 AM

All use subject to JSTOR Terms and Conditions
11498. Proposed by Y. N. Aliyev, Qafqaz University, Khyrdalan, Azerbaijan. Let
ABC D be a convex quadrilateral. A line through the intersection O of the diagonals
AC and B D intersects the interior of edge BC at L and the interior of AD at N .
Another line through O likewise meets AB at K and C D at M. This dissects ABC D
into eight triangles AK O, K B O, B L O, and so on. Prove that the arithmetic mean
of the reciprocals of the areas of these triangles is greater than or equal to the sum of
the arithmetic and quadratic means of the reciprocals of the areas of triangles AB O,
BC O, C DO, and D AO. (The quadratic mean is also known as the root mean square;
it is the square root of the mean of the squares of the given numbers.)
11499. Proposed by Omran Kouba, Higher Institute for Applied Science and
Technol-
ogy, Damascus, Syria. Let Hn be the nth harmonic number, given by Hn = nk=1 1/k.
Let

∞
Sk = (−1)n−1 (log k − (Hkn − Hn )) .
n=1

Prove that for k ≥ 2,

  
k−1 1 π k/2 (2l − 1)π
Sk = log 2 + log k − 2 (k + 1 − 2l) cot .
2k 2 2k l=1 2k

11500. Proposed by Bhavana Deshpande, Poona College, Camp Pune, Maharashtra,

India, and M. N. Deshpande, Institute of Science, Nagpur, India. We have n balls,
labeled 1 through n, and n urns, also labeled 1 through n. Ball 1 is put into a randomly
chosen urn. Thereafter, as j increments from 2 to n, ball j is put into urn j if that urn is
empty, otherwise, it is put into a randomly chosen empty urn. Let the random variable
X be the number of balls that end up in the urn bearing their own number. Show that
the expected value of X is n − Hn−1 .
11501. Proposed by Finbarr Holland, University College Cork, Cork, Ireland. Let
3
g(z) = 1 − .
1− 1
1−az
+ 1−i1 z + 1
1+i z

Show that the

√ coefficients in the Taylor series expansion of g are all nonnegative if and
only if a ≥ 3.

SOLUTIONS

An Unusual GCD/LCM Relationship

11346 [2008, 167]. Proposed by Christopher Hillar, Texas A&M University, College
Station, TX, and Lionel Levine, University of California, Berkeley, CA. Let n be an
integer greater
 than 1, and let S = {2, . . . , n}. For each nonempty subset A of S, let
π(A) = j ∈A j. Prove that when k is a positive integer and k < n,

n
lcm({1, . . . , n/i}) = gcd({π(A) : |A| = n − k}).
i=k
n
(In particular, setting k = 1 yields i=1 lcm({1, . . . , n/i}) = n!.)

April 2010] PROBLEMS AND SOLUTIONS 371

This content downloaded from 146.201.208.22 on Thu, 21 Mar 2013 00:23:09 AM

All use subject to JSTOR Terms and Conditions
Solution by Richard Stong, Center
 for Communications Research,
n
San Diego,CA. We
prove that both sides equal p p e p (n,k) , where e p (n, k) = i=k log p (n/i) and the
product runs over all primes (only finitely many primes contribute). Let v p (n) denote
the maximum r such that pr divides n.

For the left side, letting l(x) = lcm({1, . . . , x}), we have v p (l(x)) = log p x ,
 
since pr divides l(x) if and only if x ≥ pr . Hence i=k n
l(n/i) = p p e p (n,k) .
For the right side, let (b1 , . . . , bn−1 ) be the result of putting (v p (2), . . . , v p (n)) in
nonincreasing order. The number of terms with v p (k) ≥ r equals the number of mul-
tiples
of pr in S,namely n/ pr . Thus bk ≥ r if and only if k ≤ n/ pr , and hence
bk = log p (n/k) . The smallest value of v p (π(A)) such that |A| = n − k will be
achieved when A consists of exactly the elements of S corresponding to bk , . . . , bn−1 .
Hence
n−1
v p (gcd({π(A) : |A| = n − k})) = bi = e p (n, k),
i=k

using the fact that the term for i = n in the summation for e p (n, k) always
 equals 0.
Applying this formula over all primes shows that the right side also equals p p e p (n,k) .
Also solved by D. R. Bridges, J. H. Lindsey II, O. P. Lossers (Netherlands), M. A. Prasad (India), T. Rucker,
K. Schilling, A. Stadler (Switzerland), M. Tetiva (Romania), S. Vandervelde, B. Ward (Canada), GCHQ Prob-
lem Solving Group (U. K.), NSA Problems Group, and the proposers.

Some Triangle Inequalities

11363 [2008, 461]. Proposed by Oleh Faynshteyn, Leipzig, Germany. Let m a , m b ,
and m c be the lengths of the medians of a triangle T . Similarly, let Ia , Ib , Ic , h a , h b ,
and h c be the lengths of the bisectors and altitudes of T , and let R, r , and S be the
circumradius, inradius, and area of T . Show that
Ia Ib Ib Ic Ic Ia
+ + ≥ 3(2R − r ),
Ic Ia Ib
and
m a Ib m b Ic m c Ia √
+ + ≥ 35/4 S.
hc ha hb

Solution by GCHQ Problem Solving Group, Cheltenham, U. K. We write a, b, c for

the lengths
 of  the three sides, and s = (a + b + c)/2 for the semiperimeter. We will
write or for a three or six term sum or product, respectively, over permutations
of the triangle, with three terms ifthe sum is formally independent  of the direction
of the cycle, and six if not. Thus, ab denotes ab + bc + ca while a 2 b = a 2 b +
b2 c + c2 a + ab2 + bc2 + ca 2 . We use several results from (or easily deduced from)
Geometric Inequalities by Bottema et. al. (Nordhoff, Groningen, 1969), including:
2S r A
Ia = , abc = 4Rr s, = sin ,
(b + c) sin(A/2) 4R 2
 
a = 2(s − 4Rr − r ),
2 2 2
a b = 2s(s − 2Rr + r 2 ),
2 2


a 2 b2 c = 4Rr s(s 2 + 4Rr + r 2 ),

a 3 b2 = 2s(s 4 + r 4 + 6Rr 3 + 8R 2 r 2 + 2r 2 s 2 − 10Rr s 2 ),

a 4 b = 2s(s 4 − 3r 4 − 14Rr 3 − 8R 2 r 2 − 2r 2 s 2 − 6Rr s 2 ).

372 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 117

This content downloaded from 146.201.208.22 on Thu, 21 Mar 2013 00:23:09 AM

All use subject to JSTOR Terms and Conditions
The first inequality must be reversed. In fact, we will show that
16 Ia Ib Ib Ic Ic Ia
(2R − r ) < + + ≤ 3(2R − r ).
9 Ic Ia Ib
We begin with
 Ia Ib  2S 2S
(b+c) sin(A/2) (c+a) sin(B/2)
= 2S
Ic (a+b) sin(C/2)

2S  C
=  (a + b)2 sin2 .
(a + b) sin(A/2) 2
Now
 C 
2 =
(a + b)2 sin2 (a + b)2 (1 − cos C)
2
   
=2 a2 + 2 ab − a 2 cos C − 2 ab cos C.
  
But 2 ab cos C = (a 2 + b2 − c2 ) = a 2 , so
   
2 a2 + 2 ab − 2 ab cos C = ( a)2 = 4s 2
and
 1  3 1  2 2
a 2 cos C = a bc cos C = a c(a + b2 − c2 )
abc 2abc
1  4  
= a c+2 a 2 b2 c − a 2 c3
2abc
1  4
= s − 3r 4 − 14Rr 3 − 8R 2 r 2 − 2r 2 s 2 − 6Rr s 2 + 4Rr (s 2 +
4Rr

4Rr + r 2 ) − (s 4 +r 4 +6Rr 3 +8R 2 r 2 +2r 2 s 2 −10Rr s 2 )
2Rs 2 − 4Rr 2 − r 3 − r s 2
= .
R
Therefore
 C 2Rs 2 + 4Rr 2 + r 3 + r s 2
(a + b)2 sin2 =
2 .
2 R
  2 
Furthermore, (a + b) = a b + 2abc = 2s(s 2 + 2Rr + r 2 ) and sin(A/2) =
r/(4R). Hence
Ia Ib Ib Ic Ic Ia 2(2Rs 2 + 4Rr 2 + r 3 + r s 2 )
+ + = . (∗)
Ic Ia Ib s 2 + 2Rr + r 2
Now by Geometric Inequalities (5.9), 4R 2 + 4Rr + 3r 2 ≥ s 2 ≥ r (16R − 5r ). For our
lower bound: 2Rs 2 + 36Rr 2 + 17r s 2 + 17r 3 ≥ 32R 2 r + 26Rr 2 + 17r s 2 + 17r 3 >
32R 2 r , so 9(2Rs 2 + 4Rr 2 + r s 2 + r 3 ) > 8(2Rs 2 + 4R 2 r − r s 2 − r 3 ) = 8(s 2 +
2Rr + r 2 )(2R − r ). Hence
Ia Ib Ib Ic Ic Ia 16
+ + > (2R − r ).
Ic Ia Ib 9

April 2010] PROBLEMS AND SOLUTIONS 373

This content downloaded from 146.201.208.22 on Thu, 21 Mar 2013 00:23:09 AM

All use subject to JSTOR Terms and Conditions
For our upper bound: R ≥ 2r , so 0 ≤ (R − 2r )(24R + 10r )r = 24R 2 r − 38Rr 2 −
10r 3 , and hence 44R 2 r − 10Rr 2 ≥ 20R 2 r + 28Rr 2 + 20r 3 . Therefore 2Rs 2 + 12R 2 r
≥ 44R 2 r − 10Rr 2 ≥ 20R 2 r + 28Rr 2 + 20r 3 ≥ 8Rr 2 + 5r s 2 + 5r 3 , and 3(2R −
r )(s 2 + 2Rr + r 2 ) = 6Rs 2 + 12R 2 r − 3r s 2 − 3r 3 ≥ 4Rs 2 + 8Rr 2 + 2r 3 + 2r s 2 .
This inequality, in combination with (∗), gives
Ia Ib Ib Ic Ic Ia
+ + ≤ 3(2R − r ).
Ic Ia Ib
Now consider the second inequality. By elementary calculus, a function of the form
f (x) = x 2 + 2λ/x
 achieves its minimum at x = λ , so f (x) ≥ 3λ .
1/3 2/3

Letting λ = m a Ib / h c , we have
    m2 I 2 
m a Ib 2  m 2a Ib2  m a Ib m b Ic hc
= +2 = a b
+ 2λ ≥ 9λ2/3 .
hc h 2c hc ha h 2c m a Ib
Denote the exradii of T by ra , rb , and rc . By Geometric Inequalities (8.21) and (6.27),
we have m a m b m c ≥ ra rb rc = S 2 /r = Ss. By (8.7) we have

8a 2 b2 c2 A 8a 2 b2 c2 s(s − a)
Ia Ib Ic =  cos = 
(a + b) 2 (a + b) bc
8a 2 b2 c2 Ss 8abcSs 32Rs S 2
= = = ,
(a + b) abc (a + b) (a + b)
2S 8S 3 2S 3
ha hbhc = = = .
a abc Rr s
Now
 2  2/3
32Rs S 2 Rr s 16R 2 r s 3 m a Ib 16R 2 r s 3
λ = Ss  =  and ≥9  .
(a + b) 2S 3 (a + b) hc (a + b)

√
By (5.5) and (5.1), s 2 ≥ 3r (4R + r ) ≥ 3r (9r ) = 27r 3 , so s ≥ 3 3r . By (5.8) s 2 ≤
4R 2 + 4Rr + s 2 + 2Rr + r 2 ≤ 4R 2 + 6Rr + 4r 2 ≤ 4R 2 + 3R 2 + R 2
3r 2 , and thus  =
8R . Hence (a + b) =
2
a 2 b + 2abc = 2s(s 2 −√ 2Rr + r 2 ) + 8Rr s = 2s(s 2 +

2Rr + r ) ≤ 2s(8R ) = 16R s. This leads to 3 3( (a + b))2 ≤ s(16R 2 s)2
2 2 2
=
256R 4 s 3 . Now 315/2 S 3 = 315/2r 2 s 3 , and
 2/3  
256R 4 r 2 s 6 16R 2 r s 3 m a Ib 2
315/2r 3 s 3 ≤ 729  2 ⇒ 35/2
S ≤ 9  ≤ ,
(a + b) (a + b) hc
√ 
so that finally 35/4 S ≤ m a Ib / h c .
Also solved by V. V. Garcı́a (Spain) and R. Stong.

A Multiple of a Prime
11364 [208, 461]. Proposed by Pál Péter Dályay, Szeged, Hungary. Let p be a prime
greater than 3, and let t be the integer nearest p/6.
(a) Show that if p = 6t + 1, then

2t−1  
1 1
( p − 1)! (−1) j
+ ≡ 0 (mod p).
j =0
3j + 1 3j + 2

374 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 117

This content downloaded from 146.201.208.22 on Thu, 21 Mar 2013 00:23:09 AM

All use subject to JSTOR Terms and Conditions
(b) Show that if p = 6t − 1, then
 
 (−1) j
2t−1  (−1) j
2t−2
( p − 1)! + ≡ 0 (mod p).
j =0
3j + 1 j =0
3j + 2

Solution by Robin Chapman, University of Exeter, Exeter, U. K. The desired congru-

ence in both cases is

p−1
χ(k)
( p − 1)! ≡0 (mod p), (1)
k=1
k
where
⎧
⎪
⎨ 0 if k ≡ 0, 3 (mod 6),
χ(k) = 1 if k ≡ 1, 2 (mod 6),
⎪
⎩−1 if k ≡ 4, 5 (mod 6).
√ √
Note that χ(k) = (ζ k − ζ −k )/ −3, where ζ = eπi/3 = 12 (1 + −3). Letting
 p−1 k
F(z) = k=1 z /k, we have

p−1
χ(k) F(ζ ) − F(ζ −1 )
√ = . (2)
k=1
k −3
 p−1 k−1 p−1
For the value on the right, note that F (z) = k=1 z = 1−z , so F (1 − z) =
 p−2     1−z

k=0 (−1)
k+1 p−1
k+1
z k . Note also that j ≡ (−1) j (mod p). Hence F (1 − z) =
p−1

pG(z) + F (z) (mod p), where G is a polynomial having integer coefficients and
degree at most p − 2. We conclude that
d
(F(z) − F(1 − z)) = − pG(z). (3)
dz
 p−1
Let G(z) = k=1 bk z k−1 with each bk ∈ Z. Integrating (3) from 0 to z gives

p−1
bk
F(z) − F(1 − z) + F(1) = − p zk .
k=1
k

Setting z = ζ and using 1 − ζ = ζ −1 yields


p−1
bk
F(ζ ) − F(ζ −1 ) = −F(1) − p zk .
k=1
k
( p−1)/2 ( p−1)/2 p
Since p is odd, F(1) = k=1 ( 1k + p−k
1
) = k=1 k( p−k)
. It follows that
( p − 1)! F(1) is a multiple of p. We conclude that in the context
√ of algebraic inte-
gers, ( p − 1)! (F(ζ ) − F(ζ −1 )) ≡ 0 (mod p). Multiplying by −3 yields a rational
integer, and dividing by −3 (justified by p > 3) and invoking (2) yields the desired
congruence (1).
Editorial comment. Stong showed also that ( p − 1)! F(ζ ) ≡ ( p − 1)! F(ζ −1 ) ≡ 0
 p−1 χ(k+s)
(mod p), which leads to ( p − 1)! k=1 k
≡ 0 (mod p) for every integer s.
Also solved by J. H. Lindsey II, M. A. Prasad (India), A. Stadler (Switzerland), R. Tauraso (Italy), M. Tetiva
(Romania), A. Wyn-Jones, GCHQ Problem Solving Group (U. K.), and the proposer.

April 2010] PROBLEMS AND SOLUTIONS 375

This content downloaded from 146.201.208.22 on Thu, 21 Mar 2013 00:23:09 AM

All use subject to JSTOR Terms and Conditions
 Relating Two Integer Sequences
11365 [2008, 462]. Proposed by Aviezri S. Fraenkel,√Weizmann Institute of Science,
Rehovot, Israel. Let t be a positive integer. Let γ = t 2 + 4, α = 12 (2 + γ − t), and
β = 12 (2 + γ + t). Show that for all positive integers n,
nβ = (nα + n(t − 1))α + 1 = (nα + n(t − 1) + 1)α − 1.

Solution I by Donald R. Bridges, Woodstock, MD. Letting

= (γ − t)/2, we have
α = 1 +
and β = 1 + t +
. Note that t 2 < γ 2 < (t + 2)2 , so γ and
are irrational
and 0 <
< 1.
We write the expressions in terms of
. For the first, nβ = n + nt + n
. For the
second,
nα + n(t − 1) = nt + n
 ,
(nα + n(t − 1))α = nt + n
 + nt
+ n

.
√
Squaring both sides of t 2 + 4 = t + 2
yields t
+
2 = 1, so nt
+ n
2 = n. Also,
nt
+ n

> nt
+ (n
− 1)
, so the floor of the last displayed expression is nt +
n
 + n − 1, since 0 <
< 1. This proves the first equality.
To compute the rightmost expression in the problem statement, begin with
(nα + n(t − 1) + 1)α = nt + n
 + 1 + nt
+ n
+ 1
.
Since nt
+ n
+ 1
≤ nt
+ n
2 +
< n + 1, we obtain the desired equality
(nα + n(t − 1) + 1)α = nβ + 1.

Solution II by the proposer. First, observe that α and β are irrational numbers satisfy-
ing 1 < α < β and α + β = αβ, and that as a result, β > 2. It is well known that under
these conditions, A ∪ B = N, where A = {nα : n ≥ 1} and B = {nβ : n ≥ 1}.
Since β > 2, the set B does not contain consecutive integers. Hence each term of
B lies between two consecutive terms of A. That is, for each positive integer n there
exists m such that mα, nβ, and (m + 1)α are consecutive integers. Given n, the
problem is to determine m.
Among the integers from 1 to nβ, exactly n lie in B, so nβ − n lie in A. There-
fore, m = nβ − n. Thus
(nβ − n) α , nβ, ((nβ − n) + 1) α

only toshow that nβ − n = nα + n(t − 1).
are consecutive
integers. It remains
This reduces to 12 n(γ + t) = 12 n(γ − t) + nt, which is true.
Editorial comment. The claim that A ∪ B = N in Solution II is well known; the pro-
poser cited A. S. Fraenkel, How to beat your Wythoff games opponent on three fronts,
Amer. Math. Monthly 89 (1982) 353–361. The result is so astonishing and yet easily
proved that we include a short proof for the reader’s pleasure.
First note that a + b = ab is equivalent to a1 + b1 = 1. Also, a, b > 1. For any
k ∈ N, the number of terms less than k in A ∪ B is k/a + k/b, since a and b are
irrational. We compute
            
k k k 1 k −k
+ = + k 1− =k+ + = k − 1.
a b a a a a
Similarly, A ∪ B contains k terms less than k + 1. Hence there is exactly one term less
than k + 1 but not less than k; it equals k.

376 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 117

This content downloaded from 146.201.208.22 on Thu, 21 Mar 2013 00:23:09 AM

All use subject to JSTOR Terms and Conditions
Also solved by R. Chapman (U. K.), P. Corn, C. Curtis, J. H. Lindsey II, O. P. Lossers (Netherlands),
M. A. Prasad (India), A. Stadler (Switzerland), R. Stong, GCHQ Problem Solving Group (U. K.), and the
proposer.

An Exponential Inequality
11369 [2008, 567]. Proposed by Donald Knuth, Stanford University, Stanford, CA.
Prove that for all real t, and all α ≥ 2,
 α
eαt + e−αt − 2 ≤ et + e−t − 2α .

Solution by Knut Dale, Telemark University College, Bø, Norway. For t ∈ R and
α ≥ 0, let f (t, α) = ((et + e−t )α − 2α ) − (eαt + e−αt − 2). Since f (0, α) = 0 and
f (−t, α) = f (t, α), we need only consider t > 0. Write
 t 
−x α sinh x αx −αx
f (t, α) = α (e + e )
x
− (e − e ) d x
0 cosh x
 t
 
= α (e x + e−x )α g(x, 1) − g(x, α) d x,
0

where g(x, α) = (eαx − e−αx )/(e x + e−x )α . Let x > 0 and observe that g(x, α) ≥ 0,
g(x, 2) = g(x, 1) > 0, and g(x, 0) = g(x, ∞) = 0. Note that
∂g(x, α) ln(e x + e−x ) + x
>0 ⇐⇒ > e2αx . (∗)
∂α ln(e x + e−x ) − x
Likewise, equivalence holds if we replace “>” with “=” or with “<” throughout (∗).
Since e2αx is an increasing function of α,
ln(e x + e−x ) + x
= e2αx
ln(e x + e−x ) − x
has a unique solution α in the interval (1, 2). Thus, as a function of α, g(x, α) increases
from 0 to a maximum in (1, 2) and then decreases towards 0. Hence f (t, α) > 0 for
α ∈ (0, 1) ∪ (2, ∞), f (t, α) < 0 for α ∈ (1, 2), and f (t, α) = 0 for α ∈ {0, 1, 2}.
Editorial comment. Grahame Bennett (Indiana University) provided an instructive so-
lution including a general context for this inequality. That solution is now incorporated
into a paper, appearing in the current issue of this M ONTHLY (see p. 334).
Also solved by F. Alayont, K. Andersen (Canada), R. Bagby, G. Bennett, D. & J. Borwein (Canada), P. Bour-
don, P. Bracken, R. Chapman (U. K.), H. Chen, P. P. Dályay (Hungary), K. Endo, G. C. Greubel, J. Grivaux
(France), J. A. Grzesik, S. J. Herschkorn, M. Hildebrand, F. Holland (Ireland), A. Incognito & T. Mengesha,
V. K. Jenner (Switzerland), O. Kouba (Syria), K.-W Lau (China), W. R. Livingston, O. P. Lossers (Nether-
lands), K. McInturff, K. Nagasaki (Japan), T. Nakata (Japan), O. Padé (Israel), P. Perfetti (Italy), Á. Plaza
& J. M. Pacheco (Spain), D. S. Ross, V. Rutherfoord, B. Schmuland (Canada), A. Stadler (Switzerland), R.
Stong, R. Tauraso (Italy), M. Tetiva (Romania), M. Thaler (Australia), J. Vinuesa (Spain), Z. Vörös (Hungary),
T. Wilkerson, Y. Yu, BSI Problems Group (Germany), GCHQ Problem Solving Group (U. K.), Microsoft
Research Problems Group, and the proposer.

April 2010] PROBLEMS AND SOLUTIONS 377

This content downloaded from 146.201.208.22 on Thu, 21 Mar 2013 00:23:09 AM

All use subject to JSTOR Terms and Conditions