Problems and Solutions

Source: The American Mathematical Monthly, Vol. 120, No. 2 (February 2013), pp. 174-181
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/10.4169/amer.math.monthly.120.02.174
Accessed: 17-05-2015 02:26 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 152.13.18.10 on Sun, 17 May 2015 02:26:16 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 Mike Bennett, Itshak Borosh, Paul Bracken, Ezra A. Brown,
Randall Dougherty, Tamás Erdélyi, Zachary Franco, Christian Friesen, Ira M. Ges-
sel, László Lipták, Frederick W. Luttmann, Vania Mascioni, Frank B. Miles, Richard
Pfiefer, Dave Renfro, 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. Proposed problems should never
be under submission concurrently to more than one journal. Submitted solutions
should arrive before June 30, 2013. Additional information, such as generaliza-
tions 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
11691. Proposed
R ∞ by M. L. Glasser, Clarkson University, Potsdam, NY. Show that the
2nth moment 0 x 2n f (x) d x of the function f given by
 
d sinh x
f (x) = arctan
dx cos x
is zero when n is an odd positive integer.
11692. Proposed by Cezar Lupu, University of Pittsburgh, Pittsburgh, PA, and Ştefan
Spătaru, International Computer High School of Bucharest, Bucharest, Romania. Let
a1 , a2 , a3 , a4 be real numbers in (0, 1), with a4 = a1 . Show that
3 3
3 X 1 X 1 1
+ 3
≥ 2
+ 2
.
1 − a1 a2 a3 k=1 1 − ak k=1
1 − ak ak+1 1 − ak ak+1

11693. Proposed by Eugen Ionascu, Columbus State University, Columbus, GA, and
Richard Stong, CCR, San Diego CA. Let T be an equilateral triangle inscribed in the
d-dimensional unit cube [0, 1]d , with d ≥ 2. As a function of d, what is the maximum
possible side length of T ?
11694. Proposed by Kent Holing, Trondheim, Norway. Let g(x) = x 4 + ax 3 + bx 2 +
ax + 1, where a and b are rational. Suppose g is irreducible over Q. Let G be the
Galois group of g. Let Z4 denote the additive group of the integers mod 4, and let D4
be the dihedral group of order 8. Let α = (b + 2)2 − 4a 2 and β = a 2 − 4b + 8.
(a) Show that G is isomorphic to one of Z4 or D4 if and only if neither α nor β is the
square of a rational number, and G is cyclic exactly when αβ is the square of a rational
number.
http://dx.doi.org/10.4169/amer.math.monthly.120.02.174

174 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 120

This content downloaded from 152.13.18.10 on Sun, 17 May 2015 02:26:16 UTC
All use subject to JSTOR Terms and Conditions
(b) Suppose neither α nor β is square, √ but αβ is. Let r be one of the roots of g.
(Trivially, 1/r is also a root.) Let s = αβ, and let

t = ((s + (b − 6)a)r 3 + (as + (b − 8)a 2 + 4(b + 2))r 2 +

((b − 1)s + (b2 − b + 2)a − 2a 3 )r + 2(b + 2)b − 6a 2 )/(2s).

Show that t ∈ Q[r ] is one of the other two roots of g. Comment on the special case
a = b = 1.
11695. Proposed by Ovidiu Furdui, Technical University of Cluj-Napoca, Cluj-
Napoca, Romania. The Stirling numbers of the first kind, denoted P s(n, k), can be
defined by their generating function: z(z − 1) · · · (z − n + 1) = nk=0 s(n, k)z k . Let
m and p be nonnegative integers with m > p − 4. Prove that

log x · logm (x y) · log y

Z 1Z 1
d x dy
0 0 (1 − x y) p
(
(−1)m 16 (m + 3)!ζ (m + 4), if p = 1;
=
(−1)m+ p−1 6((m+3)! (−1) (m if p > 1.
P p−1 k
p−1)! k=1 s( p − 1, k)ζ + 4 − k)

11696. Proposed by Enkel Hysnelaj, University of Technology, Sydney, Australia, and

Elton Bojaxhiu, Kriftel, Germany. Let T be a triangle with sides of length a, b, c,
inradius r , circumradius R, and semiperimeter p. Show that
 
1 1X 1 4X 1
+ ≥ .
2(r 2 + 4Rr ) 9 cyc c( p − c) 9 cyc 9Rr − c( p − c)

11697. Proposed by Moubinool Omarjee, Lycée Henri IV, Paris, France. Let n and q
be integers, with 2n > q ≥ 1. Let
2n
e−t (x1 +·+xq )
Z 2n

f (t) = d x1 · · · d xq .
Rq 1 + x12n + · · · xq2n

Prove that limt→∞ t q/2n f (t) = n −q (0(1/2n))q .

SOLUTIONS

Some Inequalities for Triangles

11569 [2011, 372]. Proposed by M. H. Mehrabi, Nahavand, Iran. Let a, b, and c be
the lengths of the sides of a triangle, and let s, r , and R be the semi-perimeter, inradius,
and circumradius, respectively, of that triangle. Show that
(a + b)(b + c)(c + a)
   
1 1 1
2 < log < (a + b + c) + + −6
abc a b c
and
       
r b+c c+a a+b 2r
8 < log log log < .
p a b c R

February 2013] PROBLEMS AND SOLUTIONS 175

This content downloaded from 152.13.18.10 on Sun, 17 May 2015 02:26:16 UTC
All use subject to JSTOR Terms and Conditions
Solution by Oliver Geupel, Brühl, NRW, Germany. In fact, the first two inequalities
hold for any three positive numbers, even if they are not the sides of a triangle. The
first inequality can be sharpened: for positive a, b, c we have
(a + b)(b + c)(c + a)
8≤ . (1)
abc
Indeed, by the Arithmetic Mean-Geometric Mean Inequality,
√ √ √
2 ab · 2 bc · 2 ca (a + b)(b + c)(c + a)
8= ≤ .
abc abc
The second inequality can also be sharpened: for positive a, b, c we have
(a + b)(b + c)(c + a)
   
1 1 1
log < (a + b + c) + + + log 8 − 9.
abc a b c
In fact, let u = (a + b)(b + c)(c + a)/(abc) = (a + b + c)(1/a + 1/b + 1/c) − 1.
By (1),
 
1 1 1
8 − log 8 ≤ u − log u = (a + b + c) + + − 1 − log u.
a b c
For the second pair of inequalities, we will use three different formulas for the area
F of the triangle: Heron’s formula F 2 = s(s − a)(s − b)(s − c), F = r s, andF =
abc/(4R). Note that for 0 < x < 1 we have 2x < log((1 + x)/(1 − x)).
Write x = (s − a)/s to get 2(s − a)/s < log((b + c)/a). Let a 0 = b + c, b0 = c +
a, c0 = a + b. Now
 r 2 8F 2 8(s − a)(s − b)(s − c)
 0
a
 0
b
 0
c
8 = 4 = < log log log .
s s s3 a b c
Finally, for positive real x we have log(1 + x) < x.
Write x = 2(s − a)/a to get 2(s − a)/a > log((b + c)/a). Hence,
 0  0  0
a b c 8(s − a)(s − b)(s − c) 2F 4F 2r
log log log < = · = .
a b c abc s abc R

Also solved by M. Bataille (France), P. P. Dályay (Hungary), D. Fleischman, E. Hysnelaj & E. Bojaxhiu
(Australia & Germany), O. Kouba (Syria), J. Minkus, R. Stong, Z. Vörös (Hungary), M. Vowe (Switzerland),
J. B. Zacharias, GCHQ Problem Solving Group (U.K.), and the proposer.

Euclidean Construction
11572 [2011, 463]. Proposed by Sam Sakmar, University of South Florida, Tampa, FL.
Given a circle C and two points A and B outside C, give a Euclidean construction to
find a point P on C such that if Q and S are the second intersections with C of AP and
BP respectively, then QS is perpendicular to AB. (Special configurations, including the
case that A, B, and the center of C are collinear, are excluded.)
Solution by Robert A. Russel, New York, NY. Construct the circle C 0 through A and B
orthogonal to C. Let P1 and P2 be the two intersections of C and C 0 . We claim that P1
and P2 each satisfy the conditions of the problem. The circle C 0 is the circumcircle of
4ABA0 , where A0 is the inverse of A with respect to C. On the other hand, from the
Inscribed Angle Theorem and the fact that an angle formed by a chord and a tangent

176 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 120

This content downloaded from 152.13.18.10 on Sun, 17 May 2015 02:26:16 UTC
All use subject to JSTOR Terms and Conditions
is equal to the inscribed angle, we deduce that the circumcircle C 0 intersects C at the
same angle as the angle between the lines AB and QS.
Editorial comment. Several solvers noted that A and B need not be outside C.
Also solved by R. Bagby, M. Bataille (France), C. T. R. Conley, P. P. Dályay (Hungary), O. Geupel (Hungary),
M. Goldenberg & M. Kaplan, O. Kouba (Syria), J. H. Lindsey II, J. McHugh, R. Murgatroyd, R. Stong, and
the proposer.

A Three-Variable Inequality
11575 [2011, 463]. Proposed by Tuan Le (student), Worcester Polytechnic Institute,
Worcester, MA. Prove that if a, b, and c are positive, then
 3  1/3
16 a b c abc 5
+ + + ≥ .
27 b + c c + a a + b (a + b)(b + c)(c + a) 2

Solution by Michel Bataille, Rouen, France. The required

√ inequality
√ has the form
√
16 3
27
A + B 1/3
≥ 5
2
. Note that (a + b)(b + c)(c + a) ≥ 2 ab · 2 bc · 2 ca = 8abc >
0. Thus, 0 < B 1/3 ≤ 1/2. Also,
a 3 + b3 + c3
A= + 1 + B ≥ 2(1 − 2B),
(a + b)(b + c)(c + a)
where we have used a 3 + b3 + c3 ≥ (a + b)(b + c)(c + a) − 5abc, a rewriting of
Schur’s inequality a(a − b)(a − c) + b(b − c)(b − a) + c(c − a)(c − b) ≥ 0. As a
result, it suffices to prove that φ(x) ≥ 5/2 for any x ∈ (0, 1/2], where φ is given by
φ(x) = 128 27
(1 − 2x 3 )3 + x. Compute

256x 2 512x
φ 0 (x) = 1 − (1 − 2x 3 )2 , φ 00 (x) = − (1 − 2x 3 )(1 − 8x 3 ).
3 3
Since φ 00 (x) < 0 for x ∈ (0, 1/2), the function φ 0 decreases from φ 0 (0) = 1 to
φ 0 (1/2) < 0. Thus there exists α ∈ (0, 1/2) such that φ is increasing on [0, α]
and
decreasing on [α, 1/2]. Therefore min{φ(x) : x ∈ [0, 1/2]} = min φ(0), φ(1/2) =
5/2, as required.
Also solved by G. Apostolopoulos (Greece), D. Beckwith, E. Braune (Austria), P. P. Dályay (Hungary), D.
Fleischman, O. Geupel (Germany), E. A. Herman, F. Holland (Ireland), E. Hysnelaj & E. Bojaxhiu (Australia
& Germany), S. Kaczkowski, K.-W. Lau (China), J. H. Lee (Korea), J. H. Lindsey II, J. Loverde, P. Perfetti
(Italy), E. A. Smith, R. Stong, E. I. Verriest, M. Vowe (Switzerland), H. Wang & J. Wojdylo, T. R. Wilkerson,
J. Zacharias, GCHQ Problem Solving Group (U. K.), Mathramz Problem Solving Group, and the proposer.

An unusual functional equation

11578 [2011, 464]. Proposed by Roger Cuculière, Clichy la Garenne, France. Let E
be a real normed vector space of dimension at least 2. Let f be a mapping from E to
E, bounded on the unit sphere {x ∈ E : kxk = 1}, such that whenever x and y are in
E, f (x + f (y)) = f (x) + y. Prove that f is a continuous, linear involution on E.
Solution by Nicholás Caro, Universidade Federal dePernambuco, Recife, Brazil. We
have f ( f (z)) = f (0) + z for all z, so f (z) = 0 implies z = 0. Taking z = − f (0),
we obtain f ( f (z)) = 0, and hence z = 0. Thus f (0) = 0 and f ( f (z)) = z for all z.
Hence for all x and z we have f (x + z) = f (x + f ( f (z))) = f (x) + f (z). Therefore
f is additive and hence is a Q-linear involution on E.

February 2013] PROBLEMS AND SOLUTIONS 177

This content downloaded from 152.13.18.10 on Sun, 17 May 2015 02:26:16 UTC
All use subject to JSTOR Terms and Conditions
 Let M > 0 be such that f (z) ≤ M for all z ∈ E withkzk = 1. By Q-linearity, for
any x 6= 0 in E with kxk ∈ Q, we have k f (x)k ≤ Mkxk. For each x 6 = 0in E, let
z ∈ E be a vector with x and z being R-linearly independent. For every rational r with
0 < r < kxk, consider the map Pr : R2 \{(0, 0)} → R given by

ax + bz
Pr (a, b) = x + r .


kax + bzk
Since Pr is continuous, Im(Pr ) is an interval in R containing the points kxk ± r (cor-
responding to b = 0 and a = ±1). Hence there exists yr = x + r (ar x + br z)/kar x +
br zk ∈ E such that kxk − r < kyr k < kxk + r and kyr k is rational.
Since we also have kx − yr k = r rational, it follows that k f (x)k ≤ k f (yr )k +
k f (x − yr )k ≤ M(kyr k + r ) ≤ M(kxk + 2r ). Letting r tend to zero, we have
k f (x)k ≤ Mkxk for all x ∈ E.
Thus f is Lipschitz continuous, which implies R-linearity. Indeed, if rn → r with
rn rational, then f (rn x) = rn f (x) → f (r x), so f (r x) = r f (x).
Also solved by R. Bagby, J. Boersema, C. Burnette, T. Castro & M. Velasquez (Colombia), R. Chapman
(U. K.), W. J. Cowieson, P. P. Dályay (Hungary), P. J. Fitzsimmons, R. Ger (Poland), N. Grivaux (France),
E. A. Herman, J. C. Kieffer, O. Kouba (Syria), O. P. Lossers (Netherlands), R. Mortini (France), M. Omarjee
(France), K. Schilling, J. Simons (U. K.), R. Stong, M. Tetiva (Romania), and the proposer.

Optimally Nested Regular Polygons

11579 [2011, 557]. Proposed by Hallard Croft, University of Cambridge, Cambridge,
U. K., and Sateesh Mane, Convergent Computing, Shoreham, NY. Let m and n be
integers, with m, n ≥ 3. Let B be a fixed regular n-gon, and let A be the largest regular
m-gon that does not extend beyond B. Let d = gcd(m, n), and assume d > 1. Show
the following:
(a) A and B are concentric;
(b) If m | n, then A and B have m points of contact, these being the vertices of A;
(c) If m - n and n - m, then A and B have 2d points of contact;
(d) A and B share exactly d common axes of symmetry.
Editorial comment. The proofs can be found in the paper: S. J. Dilworth & S. R. Mane,
“On a problem of Croft on optimally nested regular polygons.” Journal of Geometry
99 (2010) 43–66. Claim (a) is Proposition 4.2 in the paper; statements (b) and (c)
follow from Corollary 4.7; part (d) is Corollary 4.6.
Also solved by O. Geupel (Germany), R. Simon (Chile), J. Simons (U. K.), R. Stong, Ellington Management
Problem Solving Group, and the proposers.

That’s Sum Minimum!

11580 [2011, 557]. Proposed by David Alfaya Sánchez, Universidad Autónoma de
Madrid, Madrid, Spain, and José Luis Dı́az-Barrero, Universidad Politécnica de
Cataluña, Barcelona, Spain. For n ≥ 2, let a1 , . . . , an be positive numbers that sum to
1, let E = {1, . . . , n}, and let F = {(i, j) ∈ E × E : i < j}. Prove that
X (ai − a j )2 + 2ai a j (1 − ai )(1 − a j ) X (n + 1)a 2 + nai n 2 (n + 2)
+ i
≥ .
(i, j)∈F
(1 − ai ) (1 − a j )
2 2
i∈E
(1 − ai ) 2 (n − 1)2

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

Netherlands. A reshuffling of the terms on the left side leads to
X (ai − a j )2 X ai a j X ai2 X a 2 + ai
+ 2 + + n i
.
i< j
(1 − ai ) 2 (1 − a )2
j i< j
(1 − ai )(1 − a j ) (1 − ai ) 2 (1 − ai ) 2

178 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 120

This content downloaded from 152.13.18.10 on Sun, 17 May 2015 02:26:16 UTC
All use subject to JSTOR Terms and Conditions
The first sum is nonnegative; we leave it out completely. The rest can be simplified to
X 2 X ai + a 2
ai
+n i
.
1 − ai (1 − ai )2
The functions x/(1 − x) and (x + x 2 )/(1 − x)2 are products ofPpositive, increasing,
convex functions, so they are convex. Under the assumption that ai = 1, these sums
attain their minimum for ai all equal, that is ai = 1/n. It follows that the minimum is
n2 n+1
(n−1)2
+ n 2 (n−1)2.

Also solved by M. Bataille (France), M. A. Carlton, M. Cipu (Romania), P. P. Dályay (Hungary), D. Fleis-
chman, M. Goldenberg & M. Kaplan, E. Hysnelaj & E. Bojaxhiu (Australia & Germany), D.-H. Kim (Korea),
O. Kouba (Syria), J. C. Linders (Netherlands), R. Stong, Z. Vörös (Hungary), Ellington Management Problem
Solving Group, GCHQ Problem Solving Group (U. K.), and the proposer.

Inequality for an Integral

11581 [2011, 557]. Proposed by Duong Viet Thong, National Economics University,
RHanoi, Vietnam. Let f be a continuous, nonconstant function from [0, 1] to R such that
1
0
f (x) d x = 0. Also, let m = min0≤x≤1 f (x) and M = max0≤x≤1 f (x). Prove that
Z 1 
(x)  ≤ −m M .
 

 x f d x  2(M − m)
0

Solution by Katie Elliott (student), Westmont College, Santa Barbara, CA. Since f is
nonconstant and integrates to zero, m < 0 < M. Multiplying f by a scalar, we may
R1
assume M − m = 1 and 0 x f (x) d x > 0. Consider the function F defined by

m 0≤x ≤M
F(x) =
M M <x ≤1
R1
and note that 0
F(x) d x = 0. Using −m = 1 − M, we get
Z 1 Z M Z 1
−m M
x F(x) d x = mx d x + Mx dx = .
0 0 M 2(M − m)
R1 R1
Now we show that 0 x F(x) d x is an upper bound for 0 x f (x) d x; equivalently,
R1
that 0 x( f (x) − F(x))d x ≤ 0. Since f (x) ≥ F(x) on [0, M] and F(x) ≥ f (x) on
(M, 1], it follows that
Z 1 Z M Z 1
x( f (x) − F(x))d x = x( f (x) − F(x))d x + x( f (x) − F(x))d x
0 0 M
Z M Z 1
≤M ( f (x) − F(x))d x + M ( f (x) − F(x))d x
0 M
Z 1
=M ( f (x) − F(x))d x = 0.
0

Therefore
1 1
Z  Z
−m M
x f (x) d x  ≤ .
 
 x F(x) d x =

0 0 2(M − m)

February 2013] PROBLEMS AND SOLUTIONS 179

This content downloaded from 152.13.18.10 on Sun, 17 May 2015 02:26:16 UTC
All use subject to JSTOR Terms and Conditions
Also solved by U. Abel (Germany), R. Bagby, R. Chapman (U. K.), D. Constales (Belgium), W. J. Cowieson,
P. P. Dályay (Hungary), N. Eldredge, P. J. Fitzsimmons, O. Geupel (Germany), W. R. Green, E. A. Herman,
J. C. Kieffer, O. Kouba (Syria), K.-W. Lau (China), J. C. Linders (Netherlands), J. H. Lindsey II, O. P. Lossers
(Netherlands), P. R. Mercer, J. Noël (France), J. M. Pacheco, Á. Plaza & K. Sadarangani (Spain), P. Perfetti
(Italy), I. Pinelis, P. Rodrı́guez-Chavez (Mexico), K. Schilling, J. Simons (U. K.), A. Stenger, R. Stong, M.
Tetiva (Romania), M. Wildon (U. K.), W. C. Yuan (Singapore), Ellington Management Problem Solving Group,
GCHQ Problem Solving Group (U. K.), and the proposer.

A Line and a Triangle Generate Three Convergent Point Sequences

11586 [2011, 653]. Proposed by Takis Konstantopoulos, Uppsala University, Uppsala,
Sweden. Let A0 , B0 , and C0 be noncollinear points in the plane. Let p be a line that
meets lines B0 C0 , C0 A0 , and A0 B0 at A∗ , B ∗ , and C ∗ respectively. For n ≥ 1, let An
be the intersection of B ∗ Bn−1 with C ∗ Cn−1 , and define Bn and Cn similarly. Show that
all three sequences converge, and describe their respective limits.
Editorial comment. As pointed out by several readers, we must assume that p does not
pass through any of the three initial points. Also, this must be viewed in the projec-
tive plane, since it may happen that a point in one of the sequences is defined by the
intersection of two parallel lines.
Solution I by George Apostopoulos, Messolonghi, Greece. Let φ be a projective trans-
formation taking line p to the line at infinity. Since points A0 , B0 , and C0 are non-
collinear, so are their images under φ. Since A∗ , B0 , C0 are collinear, so are their im-
ages, which is to say that line φ(B0 )φ(C0 ) is parallel to the direction of φ(A∗ ), a point
at infinity. Similarly for the other two cases. The image of line A∗ A0 is the line through
φ(A0 ) parallel to the direction of φ(A∗ ), that is, parallel to φ(B0 )φ(C0 ), and similarly
for the other two cases. Therefore, triangle φ(A0 )φ(B0 )φ(C0 ) is the median triangle
of triangle φ(A1 )φ(B1 )φ(C1 ). By induction, for every n, triangle φ(An )φ(Bn )φ(Cn )
is the median triangle of triangle φ(An+1 )φ(Bn+1 )φ(Cn+1 ).
Thus all φ(An ) are collinear, and—since the triangles increase regularly in size—
limn→∞ φ(An ) is the infinite point of line φ(A0 )φ(A1 ), which is the intersection of line
φ(A0 )φ(A1 ) and the line at infinity φ(A∗ )φ(B ∗ )φ(C ∗ ). Removing the transformation
φ, we see that limn→∞ An is the intersection of line A0 A1 and line A∗ B ∗ C ∗ = p.
Similarly for limn→∞ Bn and limn→∞ Cn .
Solution II by Á. Montesdeoca, Univ. de la Luna, Spain, and Á. Plaza, Univ. de Las Pal-
mas de Gran Canaria, Spain. Corresponding sides of triangles A0 B0 C0 and A1 B1 C1
meet at points on line p, so by Desargues’s Theorem the two triangles are perspective
from a point, say P. By induction, all subsequent triangles An Bn Cn are also mutually
perspective from this same point P. Thus all the points An are collinear, all Bn are
collinear, and all Cn are collinear. We claim that the limits of the respective sequences
are the intersection points with p of the lines through the sequences. We show this for
sequence {An }.
We use homogeneous coordinates (x : y : z) with A0 B0 C0 as reference triangle,
so that A0 = (1 : 0 : 0), B0 = (0 : 1 : 0), and C0 = (0 : 0 : 1). We write U V for the
line through U and V . Let the equation of p be q x + r y + sz = 0 for some choice of
coefficients q, r, s. Then p intersects B0 C0 (x = 0) at A∗ = (0 : −s : r ). It intersects
C0 A0 (y = 0) at B ∗ = (s : 0 : −q). It intersects A0 B0 (z = 0) at C ∗ = (−r : q : 0).
Since point P is the intersection of A0 A1 (r y = sz) and C0 C1 (q x = r y), we have
P = (1/q : 1/r : 1/s). Therefore A1 = (1/q : −1/r : −1/s) = (r s : −sq : −qr ),
B1 = (−r s : sq : −qr ), and C1 = (−r s : −sq : qr ).
Since A2 is the intersection of B ∗ B1 (q x + 2r y + sz = 0) with C ∗ C1 (q x + r y +
2sz = 0), it follows that A2 = (3r s : −sq : −qr ), and similarly for B2 and C2 . The

180 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 120

This content downloaded from 152.13.18.10 on Sun, 17 May 2015 02:26:16 UTC
All use subject to JSTOR Terms and Conditions
next points are A3 = (5r s : −3sq : −3qr ), A4 = (11r s : −5sq : −5qr ), and in gen-
eral An = (an r s : −an−1 sq : −an−1 qr ) = ((an /an−1 )r s : −sq : −qr ), where {an } is
defined recursively by a0 = 1, an = 2an−1 + (−1)n , the Jacobsthal sequence. Take the
limit as n → ∞ to find (since an /an−1 → 2) that An → A∗ = (2r s : −sq : −qr ),
so A∗ is the intersection of p with the line A0 P (r y = sz, same as A0 A1 ), since its
coordinates satisfy both equations.
Also solved by R. Chapman (U. K.), M. Goldenberg & M. Kaplan, J.-P. Grivaux (France), A. Habil (Syria),
M. E. Kidwell & M. D. Meyerson, L. R. King, O. Kouba (Syria), J. C. Linders (Netherlands), O. P. Lossers
(Netherlands), J. Minkus, R. Stong, GCHQ Problem Solving Group (U. K.), University of Louisiana at
Lafayette Math Club, and the proposer.

Where Are the Zeros?

11589 [2011, 653]. Proposed by Catalin Barboianu, Infarom Publishing, Craiova, Ro-
mania. Let P be a polynomial over R given by P(x) = x 3 + a2 x 2 + a1 x + a0 , with
a1 > 0. Show that P has a least one zero between −a0 /a1 and −a2 .
Solution by William J. Cowieson, Fullerton College, Fullerton, CA. Note that
P(−a0 /a1 ) = (a02 /a13 )(a1 a2 − a0 ) and P(−a2 ) = a0 − a1 a0 , so that

P(−a0 /a1 ) = −(a02 /a13 )P(−a2 ). (1)

There are three cases.
(1) If a0 − a1 a2 = 0, then the interval reduces to a single point, and that point is a
zero of P.
q (2) If a0 = 0, then P(x) = x(x + a2 x + a0 ) has zeros at 0 and at (−a2 ±
2

a22 − 4a1 )/2. If a22 − 4a1 < 0, then 0 is the only real zero of P. Otherwise,
q
(−a2 ± a22 − 4a1 )/2 are both strictly between −a0 /a1 = 0 and −a2 , since a1 > 0.
(3) Both a0 − a1 a2 6 = 0 and a0 6 = 0. In this case, from (1) we see that P(−a0 /a1 )
and P(−a2 ) are nonzero and of opposite sign when a1 > 0. Hence the Intermediate
Value Theorem implies that there is a zero between −a0 /a1 and −a2 .
Also solved by B. K. Agarwal (India), G. Apostolopoulos (Greece), S. J. Baek & D.-H. Kim (Korea), B. D.
Beasley, M. W. Botsko, D. Brown & J. Zerger, V. Bucaj, P. Budney, H. Caerols (Chile), E. M. Campbell &
D. T. Bailey, M. Can, M. A. Carlton, T. Castro, J. Montero & A. Murcia (Colombia), R. Chapman (U. K.),
H. Chen, W. ChengYuan (Singapore), J. Christopher, D. Constales (Belgium), W. J. Cowieson, C. Curtis, P. P.
Dályay (Hungary), C. Degenkolb, C. R. Diminnie, K. Farwell, J. Ferdinands, D. Fleischman, V. V. Garcı́a
(Spain), O. Geupel (Germany), W. R. Green & T. D. Lesaulnier, J.-P. Grivaux (France), M. Hajja (Jordan),
E. A. Herman, G. A. Heuer, S. Kaczkowski, B. Kalantari, B. Karaivanov, T. Keller, L. Kennedy, J. C. Kieffer,
O. Kouba (Syria), P. T. Krasopoulos (Greece), R. Lampe, K.-W. Lau (China), J. C. Linders (Netherlands), J. H.
Lindsey II, O. López (Colombia), O.P. Lossers (Netherlands), J. Loverde, Y.-H. McDowell & F. Mawyer, F. B.
Miles, S. Mosiman, K. Muthuvel, M. Omarjee (France), Á. Plaza & K. Sadarangani (Spain), P. Pongsriiam &
T. Pongsriiam (U. S. A. & Thailand), V. Ponomarenko, C. R. Pranesachar (India), R. E. Prather, R. Pratt, D.
Ritter, A. J. Rosenthal, U. Schneider (Switzerland), C. R. Selvaraj & S. Selvaraj, A. K. Shafie & S. Gholami
(Iran), J. Simons (U. K.), N. C. Singer, E. A. Smith, N. Stanciu & T. Zvonaru (Romania), J. H. Steelman, A.
Stenger, R. Stong, M. Tetiva (Romania), N. Thornber, V. Tuck & A. Stancu, D. B. Tyler, D. Vacaru (Roma-
nia), E. I. Verriest, J. Vinuesa (Spain), T. Viteam (Germany), Z. Vörös (Hungary), M. Vowe (Switzerland), T.
Wiandt, H. Widmer (Switzerland), R. Wieler, S. V. Witt, N. Youngberg, J. Zacharias, Z. Zhang, Fejéntaláltuka
Szeged Problem Solving Group (Hungary), GCHQ Problem Solving Group (U. K.), University of Louisiana
at Lafayette Math Club, Missouri State University Problem Solving Group, NSA Problems Group, and the
proposer.

February 2013] PROBLEMS AND SOLUTIONS 181

This content downloaded from 152.13.18.10 on Sun, 17 May 2015 02:26:16 UTC
All use subject to JSTOR Terms and Conditions