Problems and Solutions

Source: The American Mathematical Monthly, Vol. 122, No. 9 (November 2015), pp. 899-906
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/10.4169/amer.math.monthly.122.9.899
Accessed: 02-11-2015 06:22 UTC

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 130.49.181.203 on Mon, 02 Nov 2015 06:22:51 UTC
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, Tamas Erdelyi, Zachary Franco, Christian Friesen, Ira M. Gessel, Laszlo
Liptak, Frederick W. Luttmann, Vania Mascioni, Frank B. Miles, Steven J. Miller,
Richard Pfiefer, Dave Renfro, Cecil C. Rousseau, Leonard Smiley, Kenneth Sto-
larsky, Richard Stong, Walter Stromquist, Daniel Ullman, Charles Vanden Eynden,
and Fuzhen Zhang.

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

problems address on the back of the title page. Proposed problems should never be
under submission concurrently to more than one journal, nor posted to the internet
before the due date for solutions. Submitted solutions should arrive before March
31, 2016. Additional information, such as generalizations and references, is wel-
come. The problem number and the solvers 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
11865. Proposed by Gary H. Chung, Clark Atlanta University, Atlanta, GA.  Let an  be

a monotone decreasing sequence of nonnegative
 real numbers. Prove that n=1 an /n
is finite if and only if limn an = 0 and n=1 (an an+1 ) log n < .
11866. Proposed by Arindam Sengupta, University of Calcutta, Kolkata, India. Con-
sider a finite set {1 , . . . , m } of rational numbers in (0, 1). For 0 < p < 1 and k 1,
let k be the probability space for k independent flips of a coin that comes up heads
with probability p. Show that there exists a positive integer k, a suitable p, and events
E 1 , . . . , E m in k , such that for each j with 1 j m, the probability of E j is j .
11867. Proposed by George Apostolopoulos, Messolonghi, Greece. For real numbers
a, b, c, let
 1/4
a2
f (a, b, c) = .
a 2 ab + b2
Prove that f (a, b, c) + f (b, c, a) + f (c, a, b) 3.
11868. Proposed by James Propp, University of Massachusetts Lowell, Lowell, MA.
For fixed positive integers a and b, let m = ab 1 and let R be the set {1, . . . , a}
{1, . . . , b}, indexed as p0 through pm in lexicographic order, so that p0 = (1, 1),
p1 = (1, 2), and pm = (a, b). Define T from R to R as the map that sends p0 to p0
and pm to pm , and for 1 i m 1 sends pi to p j where j ai (mod m). As a
bijection, T partitions R into orbits. Show that the center of mass of each orbit lies on
the line joining p0 and pm .
11869. Proposed by George Stoica, University of New Brunswick, Saint John, Canada.
Prove that |y log y x log x| |y x|11/e for 0 < x < y 1.
http://dx.doi.org/10.4169/amer.math.monthly.122.9.899

November 2015] PROBLEMS AND SOLUTIONS 899

This content downloaded from 130.49.181.203 on Mon, 02 Nov 2015 06:22:51 UTC
All use subject to JSTOR Terms and Conditions
11870. Proposed by Finbarr Holland, University College Cork, Cork, Ireland. Suppose
0 x 1, y = 1 x, and a and b are unimodular  complex numbers. Let c1
= 2(xa + yb) and c2 = 2(xa 2 + yb2 ). Prove that |c12 + c2 | 3|c1 | 3, with equality
if and only if x = y = 1/2 and ba = e2i/3 .
11871. Proposed by Cezar Lupu, University of Pittsburgh, Pittsburgh, PA, and Stefan
Spataru, Harvard University, Boston, MA. Let ABC be a triangle in the Cartesian
plane with vertices in Z2 (lattice vertices). Show that, if P is an interior lattice point
of ABC, then at least one of the angles P AB, P BC, and PC A has a radian measure
that is not a rational multiple of .
11872. Proposed by Phu Cuong Le Van, College of Education,
1 Hue, Vietnam. Let f be
a continuous function from [0, 1] into R such that 0 f (x) d x = 0. Prove that for all
c
positive integers n there exists c (0, 1) such that n 0 x n f (x) d x = cn+1 f (c).

SOLUTIONS

A Circumradial Inequality
11735 [2013, 854]. Proposed by Cosmin Pohoata, Princeton University, Princeton,
NJ. Let P be a point inside triangle ABC. Let d A , d B , and dC be the distances from
P to A, B, and C , respectively. Let r A , r B , and rC be the radii of the circumcircles of
P BC, PC A, and P AB, respectively. Prove that
1 1 1 1 1 1
+ + + + .
dA dB dC rA rB rC

Solution by Traian Viteam, Cape Town, South Africa. Let O A , O B , and OC be the
circumcenters of P BC, PC A, and P AB, respectively. The line O A O B is the perpen-
dicular bisector of the common chord PC, and similarly for O A OC and O B OC . Hence
P is inside O A O B OC . Also, r A , r B , and rC are the distances from P to the vertices of
triangle O A O B OC , and d A /2, d B /2, and dC /2 are the distances from P to the sides.
Therefore the requested inequality is inequality (5) from A. Oppenheim, The Erdos
Inequality and Other Inequalities for a Triangle, this M ONTHLY 68 (1961) 226-230.
Also solved by M. Aassila (France), A. Alt, M. Dinca (Romania), O. Geupel (Germany), B. Karaivanov, J.
Minkus, P. Nuesch (Switzerland), R. Stong, L. Zhou, GCHQ Problem Solving Group (U. K.), and the proposer.

Lucas numbers and roots of unity

11736 [2013, 855]. Proposed by Mircea Mirca, University of Craiova, Craiova, Ro-
mania. For n 1, let f be the symmetric polynomial in variables x1 , . . . , xn given
by

n1
f (x1 , . . . , xn ) = (1)k+1 ek (x1 + x12 , x2 + x22 , . . . , xn + xn2 ),
k=0

where ek is the kth elementary polynomial in n variables. (For example, when n = 6,

e2 has 15 terms, each a product of two distinct variables.) Also, let be a primitive nth
root of unity. Prove that
f (1, , 2 , . . . , n1 ) = L n L 0 ,
where L k is the kth Lucas number (L 0 = 2, L 1 = 1, and L k = L k1 + L k2 for k 2).

900

c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 122

This content downloaded from 130.49.181.203 on Mon, 02 Nov 2015 06:22:51 UTC
All use subject to JSTOR Terms and Conditions
Solution by Borislav Karaivanov, Lexington, SC. Let

n1 
n 
n
g(t1 , . . . , tn ) = (1) k+1
ek (t1 , . . . , tn ) = (1) n
tk (1 tk ),
k=0 k=1 k=1

and let

n1
p(x, y) = (x k y) = x n y n .
k=0

Let = (1 + 5)/2 and = (1 5)/2, so that 1 x x 2 = (1 x)(1 x),
= 1, and L n = n + n . We compute
f (1, , . . . , n1 ) = g(1 + 1, + 2 , . . . , n1 + 2(n1) )

n1 
n1
= (1)n ( k + 2k ) (1 k 2k )
k=0 k=0


n1 
n1 
n1 
n1
= ( ) k
(1 + )
k
(1 ) k
(1 k )
k=0 k=0 k=0 k=0

= p(0, 1) p(1, 1) p(1, ) p(1, )

= (1) 1 (1)n (1 n )(1 n )

= 1 + (1)n 1 n n + ()n
= n + n 2 = L n L 0.

Also solved by D. Beckwith, R. Chapman (U. K.), D. Constales (Belgium), I. Gessel, Y. J. Ionin, O. P. Lossers
(Netherlands), J. Martnez (Spain), M. Omarjee (France), M. A. Prasad (India), R. Stong, R. Tauraso (Italy),
T. Viteam (South Africa), L. Zhou, GCHQ Problem Solving Group (U. K.), and the proposer.

An Infinite Matrix Product

11739 [2013, 855]. Proposed by Fred Adams, Anthony  Bloch, and Jeffrey Lagarias,
University of Michigan, Ann Arbor, MI. Let B(x) = 1x x1 . Consider the infinite matrix
product

M(t) = B(2t )B(3t )B(5t ) = B( pt ),
p

where the product runs over the primes, taken in increasing order. Evaluate M(2).
Solution
 7 3  by Finbarr Holland, University College, Cork, Cork, Ireland. The value is
1
3
2 2 37
. Note that B(x) = Q A(x)Q , where
 
1+x 0 1 1
A(x) = and Q= .
0 1x 1 1

For each prime p and each complex number t with t > 1, let x p = p t . We compute
  1 + x 0

1
Q M(t)Q = A(x p ) = p
0 1 xp
p p

November 2015] PROBLEMS AND SOLUTIONS 901

This content downloaded from 130.49.181.203 on Mon, 02 Nov 2015 06:22:51 UTC
All use subject to JSTOR Terms and Conditions
  
p (1
+ x p)  0
=
0 p (1 x p )
  2    
1x p (t)
0 (2t)
0
= p 1x p
 = 1 ,
0
0 p (1 x p ) (t)

where, by Euler and Riemann,


1 
(t) = t
= (1 x p )1 .
n=1
n p

Since
2 4
(2) = and (4) = ,
6 90
it follows that
15  
0 1 3 7 3
M(2) = Q 2
6 Q = .
0 2 2 2 3 7

Also solved by D. Beckwith, R. Boukharfane (Canada), M. A. Carlton, M. Chamberland, R. Chapman (U.K.),

H. Chen, D. Constales (Belgium), C. Degenkolb, C. Delorme (France), E. S. Eyeson, D. Fleischman, O. Furdui
(Romania), O. Geupel (Germany), M. Goldenberg & M. Kaplan, J.-P. Grivaux (France), E. A. Herman, J.
Iliams, M. Janas, B. Karaivanov, P. Khalili, J. C. Kieffer, J. H. Lindsey II, O. P. Lossers (Netherlands), G.
Martin (Canada), R. Martin (Germany), J. Martnez (Spain), R. Molinari, R. Nandan, M. Omarjee (France), E.
Pite (France), M. A. Prasad (France), A. J. Rosenthal, C. M. Russell, M. Safaryan (Armenia), E. Schmeichel,
N. C. Singer, A. Stadler (Switzerland), A. Stenger, R. Stong, R. Tauraso (Italy), N. Thornber, T. Trif (Romania),
E. I. Verriest, C. Vignat & V. H. Moll (France & U.S.A.), J. Vinuesa (Spain), T. Viteam (South Africa), Z. Voros
(Hungary), T. Wiandt, L. Zhou, FAU Math Club, GCHQ Problem Solving Group (U.K.), GWstate Problem
Solving Group, Missouri State University Problem Solving Group, Northwestern University Math Problem
Solving Group, NSA Problems Group, TCDmath Problem Group (Ireland), and the proposers.

An Infinite Set of Prime Ideals

11740 [2013, 941]. Proposed by Cosmin Pohoata, Princeton University, Princeton NJ.
Let p and q be prime ideals in a commutative Noetherian ring R with unity. Suppose
that p q. Let I be the set of all prime ideals j in R such that p j q. Prove that I
is either empty or infinite.
Solution by Borislav Karaivanov, Lexington, SC. Passing to the quotient ring R/p and
subsequent localization (R/p)q at the prime ideal q allows us to assume, without loss
of generality, that R is a local (integral) domain, p = (0), and q is the only maximal
ideal in R.
If I is nonempty, then q is not a minimal prime ideal, and hence its height is at least
2. Let a1 q be nonzero, and let p1 be a minimal prime ideal over the principal ideal
(a1 ). By Krulls Principal Ideal Theorem, the height of p1 is 1. Therefore p1  q.
We extend the construction by induction. Suppose we have constructed distinct
prime ideals p1 , . . . , pn1 , each strictly contained in q. By Krulls Prime Avoidance
n1 n1
Lemma, l=1 pk  q. Let an q l=1 pk , and let pn be a minimal prime ideal over
(an ). Clearly, pn = pk for 1 k n 1, since only pn contains an . By Krulls Princi-
pal Ideal Theorem again, the height of pn is 1, while that of q is at least 2. Therefore,
pn  q. This completes the induction argument, showing that I is infinite.

902

c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 122

This content downloaded from 130.49.181.203 on Mon, 02 Nov 2015 06:22:51 UTC
All use subject to JSTOR Terms and Conditions
Editorial comment. Several solvers mentioned that Krulls results and, in some
cases, the stated problem have appeared in textbooks on commutative algebra by
Atiyah-Macdonald (Introduction to Commutative Algebra, 1994), Eisenbud Commu-
tative Algebra with an Eye Towards Algebraic Geometry, 1995), Isaacs (Algebra: A
Graduate Course, 1994, mentioned by the proposer himself), and Matsumura (Com-
mutative Ring Theory, 1970), and in a survey article by Wiegand and Wiegand in Ring
and Module Theory (2010).
Also solved by N. Caro (Brazil), R. Chapman (U. K.), O. Geupel (Germany), I. M. Isaacs, O. P. Lossers
(Netherlands), J. Rosoff, R. Stong, M. Wildon (U. K.), TCDmath Problem Group (Ireland), and the proposer.

A Homomorphism to the Center?

11741 [2013, 941]. Proposed by Chindea Filip-Andrei, University of Bucharest,
Bucharest, Romania. Given a ring A, let Z (A) denote the center of A, which is the set
of all z A that commute with every element of A. Prove or disprove: For every ring
A, there is a map F : A Z (A) such that f (1) = 1 and f (a + b) = f (a) + f (b)
for all a, b A.
Solution by O. P. Lossers, Department of Mathematics and Computer Science, Eind-
hoven University of Technology, Eindhoven, The Netherlands. The statement is false.
Let x and y be indeterminates, and consider the ring B = Q[x, y]; its center is Q.
Let A be the subring of B generated over Z by x, y, and 2x1 3i
for i N. Note that
Z (A) Q.
We claim that Z (A) = Z. Assuming this to be true, suppose that f : A Z (A)
satisfies the stated conditions, namely f (1) =
f (a + b) = f (a) + f (b) for
1 and
all a, b A. Observe that f (2x 1) = 3i f 2x1 3i
for all i. Since f (2x 1) and
f ((2x 1)/3 ) are both integers, we deduce that f ((2x 1)/3i ) Z for all i, which
i

is only possible if f (2x 1) = 0. It follows that

2 f (x) = f (2x) = f (2x 1) + f (1) = 1.
This contradicts f (x) being an integer. Hence Z = Z implies that there is no
f : A Z (A) satisfying f (1) = 1 and f (a + b) = f (a) + f (b) for all a, b A.
It thus suffices to prove Z (a) = Z. To show this, it suffices to show that if a poly-
nomial with integral coefficients in x, y, and 2x1
3i
is a rational number z, then it is
an integer. If F is such a polynomial, then evaluating F at any x, y Q yields z.
Since F is a polynomial, it depends on only finitely many of the values 2x1 3i
, say
3n +1
F = F(x, y, 3 , 32 , . . . , 3n ) = z. Evaluate F for y = 0 and x = 2 , an inte-
2x1 2x1 2x1

ger. We see that

   n 
2x 1 2x 1 2x 1 3 +1
z = F x, y, , ,..., =F , 0, 3 , 3 , . . . , 1 .
n1 n2
3 32 3n 2
Since all coefficients of F are integral and all variables have integer values, we con-
clude that z Z. Hence Z = Z, as claimed.
Also solved by R. Stong, M. Towers & M. Wildon (U. K.), and the proposer.

Use the Jacobi Triple Product

11742 [2013, 941]. Proposed by Alexandr Gromeko, Odessa, Ukraine. For 0 p
< q < 1, find all zeros in C of the function f given by

f (z) = (aq n , p/(aq n ); p)(z)n q n(n1)/2 ,
n=

where (u, v; w) = m=0 (1 uwm )(1 vwm ).

November 2015] PROBLEMS AND SOLUTIONS 903

This content downloaded from 130.49.181.203 on Mon, 02 Nov 2015 06:22:51 UTC
All use subject to JSTOR Terms and Conditions
Solution by Xinrong Ma, Soochow University, Suzhou, China. We prove that f (z) = 0
if and only if z = q k or z = apk q k for some k Z. Let


(z; q) = (1 zq n ), (a1 , a2 , . . . , am ; q) = (a1 ; q) (a2 ; q) (am ; q)
n=0

for 0 < |q| < 1. These infinite products converge absolutely, so they have value zero
only if one of the factors is zero. The Jacobi theta function is defined by


+
(z; q) = q (n)z n , where q (n) = (1)n q n(n1)/2 .
n=

We will use the Jacobi triple product identity:

(z; q) = (q, z, q/z; q) (I)

(for example (II.28) in G. Gasper, M. Rahman, Basic Hypergeometric Series, 2nd ed.,
Cambridge University Press, 2004) and the identities

(z; q) = (q/z; q) = q (n)z n (zq n ; q). (II)

(Lemma 2.1.4 in B. C. Berndt, Number Theory in the Spirit of Ramanujan, American

Mathematical Society, Providence, RI, 2006).
We now compute several expression for Fp (z) = ( p; p) f (z); steps citing def
are based on the definition of .
  
m
F p (z) = (aq n ; p)z n q (n) = z n q (n) aq n p (m) by (I) and def
nZ nZ mZ
 
n
= a m p (m) zq m q (n) by absolute convergence
mZ nZ
  a m p (m)
= a m p (m)(zq m ; q) = (z; q) by def and (II)
z m q (m)
mZ mZ
  a m
= (z; q) p/q (m) = (z; q)(a/z; p/q) def.
z
mZ

Now ( p; p) = 0, so f (z) = 0 if and only if (z; q)(a/z; p/q) = 0. By the Jacobi

triple product, this leads to z = q k , apk q k , k Z, as claimed.
Also solved by R. Chapman (U. K.) and the proposer.

Inverse of a Tridiagonal Toeplitz Matrix

11743 [2013, 941]. Proposed by Francois Capac`es, Nancy, France. Let n be a positive
integer, let x be a real number, and let B be the n-by-n matrix with 2x in all diagonal
entries, 1 in all sub- and super-diagonal entries, and 0 in all other entries. Compute the
inverse of B, when it exists, as a function of x.
Solution by O. Geupel, Bruhl, Germany. We will show that the (i, k)-entry of B 1 is

(1)i+k
Umin(i,k)1 Unmax(i,k)
Un (x)

904

c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 122

This content downloaded from 130.49.181.203 on Mon, 02 Nov 2015 06:22:51 UTC
All use subject to JSTOR Terms and Conditions
for (i, k) [n] [n], where Un (x) is the n-th Chebyshev polynomial of the second
kind, defined by the recursion
U0 (x) = 1, U1 (x) = 2x, and Un+1 (x) = 2xUn (x) Un1 (x) for n 1

or by Un (cos ) = sin (n + 1) /sin .

We will write B = Bn to emphasize the dependence on n. It is well-known (or
follows easily from expanding along the first row), that det Bn = Un (x). For 1 i
k n, the (i, k)-minor of Bn has the block form
 
 Bi1 0 0 

det Bn(i,k) =  0 D 0 
 0 0 Bnk 
where D is a (k i) (k i) upper triangular matrix with constant diagonal 1. There-
fore,
det Bn(i,k) = Ui1 (x)Unk (x).
Similarly, for 1 k < i n, we have det Bn(i,k) = Uk1 (x)Uni (x). Thus the (i, k)-
minor has general form
det Bn(i,k) = Umin(i,k)1 (x)Unmax(i,k) (x)
and the result follows.
Editorial comment. Finding the inverse of a tridiagonal Toeplitz matrix is a standard
result. The earliest appearance seems to be D. Moskovitz, The numerical solution of
Laplaces and Poissons equations, Quart. Appl. Math., 2 (1944) 148163. The explicit
solution above is contained in D. Kershawl, The Explicit Inverses of Two Commonly
Occurring Matrices, Math.Comp. 23 (1969) no. 105, 189-191.
Also solved by M. Bataille (France), D. Beckwith, R. Chapman (U. K.), D. Constales (Belgium), S. Falcon
Plaza (Spain), J.-P. Grivaux (France), E. A. Herman, S. Hitotumatu (Japan), R. A. Horn, B. Karaivanov,
& A.
J. C. Kieffer, O. Kouba (Syria), J. H. Lindsey II, M. Omarjee (France), R. E. Prather, M. Safaryan (Arme-
nia), N. C. Singer, R. Stong, J. L. Stuart, R. Tauraso (Italy), N. Thornber, E. I. Verriest, Z. Voros (Hungary),
T. Wiandt, Armstrong Problem Solvers, GCHQ Problem Solving Group (U. K.), and the proposer.

An Easy Inequality
11751 [2014, 83]. Proposed by Carol Kempiak, Aliso Niguel High School, Aliso Viejo,
CA, and Bogdan Suceava, California State University, Fullerton, CA. In a triangle with
angles of radian measure A, B, and C, prove that
csc A + csc B + csc C 1 1 1
+ + ,
2 sin B + sin C sin C + sin A sin A + sin B
with equality if and only if the triangle is equilateral.
Solution by Boris Karaivanov, Lexington, SC. Use the harmonicarithmetic mean
inequality. For any positive x, y, and z,

1 1 1
1
x
+ 1
y
1
y
+ 1
z
1
+ 1
2 2 2
+ + = + + z x
+ +
x y z 2 2 2 x+y y+z x+z
with equality if and only if x = y = z. Set x = sin A, y = sin B, z = sin C.
Also solved by A. Alt, G. Apostolopoulos (Greece), H. I. Arshagi, R. Bagby, M. Bataille (France),
D. M. Batinetu-Giurgiu & N. Stanciu (Romania), D. Beckwith, E. Braune (Austria), R. Chapman (U. K.),

November 2015] PROBLEMS AND SOLUTIONS 905

This content downloaded from 130.49.181.203 on Mon, 02 Nov 2015 06:22:51 UTC
All use subject to JSTOR Terms and Conditions
C. T. R. Conley, P. P. Dalyay (Hungary), A. Ercan (Turkey), E. S. Eyeson, D. Fleischman, O. Geupel
(Germany), M. Goldenberg & M. Kaplan, E. A. Herman, Y. J. Ionin, S. Kaczkowski, S. H. Kim (Korea),
O. Kouba (Syria), P. T. Krasopoulos (Greece), W.-K. Lai & A. Khristyuk, K.-W. Lau (China), O. P. Lossers
(Netherlands), R. Mabry, V. Mikayelyan (Armenia), D. J. Moore, Y. Oh (Korea), P. Perfetti (Italy), C. R.
Pranesachar (India), M. Safaryan (Armenia), A. Salgarkar (India), E. Schmeichel, C. R. Selvaraj & S. Selvaraj,
Y. Shim (Korea), Y. Song (Korea), R. Stong, T. P. Turiel, D. Vacaru (Romania), T. Viteam (South Africa),
Z. Voros (Hungary), M. Vowe (Switzerland), J. Wakem, T. Wiandt, J. Zacharias, GCHQ Problem Solving
Group (U. K.), NSA Problems Group, and the proposers.

The Product of n Distinct Derivatives

11753 [2014, 84]. Proposed by Prapanpong Pongsriiam, Silpakorn University, Nakhon
Pathom, Thailand. Let f be a continuous map from [0, 1] to R that is differentiable on
(0, 1), with f (0) = 0 and f (1) = 1. Show that for each positive integer n there exist
distinct numbers c1 , . . . , cn in (0, 1) such that nk=1 f  (ck ) = 1.
Solution by John W. Hagood, Northern Arizona University, Flagstaff, AZ. The proof
proceeds by induction on n. The case n = 1 follows from the Mean Value Theorem.
Suppose n > 1 and the statement holds for n 1. It then follows that there
exist distinct c1 , . . . , cn1 with f  (c1 ) f  (cn1 ) = 1. If there is a number cn /
{c1 , c2 , . . . , cn1 } such that f  (cn ) = 1, then the statement (for n) follows.
So, suppose f  (x) = 1 for all x / {c1 , c2 , . . . , cn1 }. By the Mean Value Theorem,

f (ci ) = 1 for some i {1, 2, . . . , n 1}. We may assume i = 1 by reordering if need
 
be, which leaves n1 k=2 f (ci ) = 1. Now let d1 < d2 < < dm be the points in (0, 1)
such that f (di ) = 1 and put d0 = 0 and dm+1 = 1. If it were the case that f  (x) 1


on (0, 1), then on each interval (di1 , di ) we should have f  (x) < 1 and thus for each
m+1f (di ) f (di1 ) < di 
i, that di1 . This leads to a contradiction: 1 = f (1) f (0)
m+1
= i=1 ( f (di ) f (di1 )) < i=1 (di di1 ) = dm+1 d0 = 1.
Thus f  assumes a value greater than 1 and f  takes on all values in some inter-
val (a, b) with a < 1 < b. Choose cn such that f  (cn ) (max{a, 1/b}, 1), but with
f  (cn ) = f  (ci ) and f  (cn ) = 1/ f  (ci ), for 2 i n 1. Since 1/ f  (cn ) (1, b),
the intermediate value property assures the existence of a value cn+1 such that f  (cn+1 )
n+1 
= 1/ f  (cn ). Now we have that i=2 f (ci ) = 1 and c2 , . . . , cn+1 are distinct. This
implies that the statement holds for case n, which completes the induction.
Also solved by M. Aassila (France), I. A. S. Aburub (Jordan), J. Boersema, M. W. Botsko, R. Boukharfane
(Canada), P. Budney, N. Caro (Brazil), R. Chapman (U. K.), P. P. Dalyay (Hungary), A. Ercan (Turkey),
D. Fleischman, O. Geupel (Germany), J.-P. Grivaux (France), E. A. Herman, S. J. Herschkorn, E. J. Ionascu,
S. Kaczkowski, B. Karaivanov, E. Katsoulis, P. T. Krasopoulos (Greece), J. H. Lindsey II, O. P. Lossers (Nether-
lands), R. Mabry, R. Martin (Germany), M. D. Meyerson, M. Omarjee (France), N. C. Overgaard (Sweden),
V. Pambuccian, S. K. Patel (India), P. Perfetti (Italy), T. Persson & M. P. Sundqvist (Sweden), R. E. Prather,
D. Ritter, M. Safaryan (Armenia), R. Stong, R. Tauraso (Italy), E. I. Verriest, T. Viteam (South Africa), GCHQ
Problem Solving Group (U. K.), NSA Problems Group, and the proposer.

906

c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 122

This content downloaded from 130.49.181.203 on Mon, 02 Nov 2015 06:22:51 UTC
All use subject to JSTOR Terms and Conditions