PROBLEMS AND SOLUTIONS

EDITORS
Curtis Cooper Shing S. So
CMJ Problems CMJ Solutions
Department of Mathematics and Department of Mathematics and
Computer Science Computer Science
University of Central Missouri University of Central Missouri
Warrensburg MO 64093 Warrensburg MO 64093
email: cooper@ucmo.edu email: so@ucmo.edu
This section contains problems intended to challenge students and teachers of college mathematics.
We urge you to participate actively BOTH by submitting solutions and by proposing problems that are
new and interesting. To promote variety, the editors welcome problem proposals that span the entire
undergraduate curriculum.
Proposed problems should be sent to Curtis Cooper, either by email (preferred) as a pdf, T
E
X, or
Word attachment or by mail to the address provided above. Whenever possible, a proposed problem
should be accompanied by a solution, appropriate references, and any other material that would be
helpful to the editors. Proposers should submit problems only if the proposed problem is not under
consideration by another journal.
Solutions to the problems in this issue should be sent to Shing So, either by email as a pdf,
T
E
X, or Word attachment (preferred) or by mail to the address provided above, no later than February
15, 2014.
PROBLEMS
Correction to 1002. Proposed by Mowaffaq Hajja, Yarmouk University, Irbid, Jordan.
Let ABCD be a convex quadrilateral. Prove that there exists a point P inside ABCD
such that [PAB] = [PBC] = [PCD] = [PDA] if and only if at least one diagonal bisects
the other. Here [ABC] denotes the area of triangle ABC.
1011. Proposed by Greg Oman, University of Colorado, Colorado Springs, CO.
Let G be a group in which every non-identity element has order 2. Suppose that
f : G G is an injective map with the property that if H is a subgroup of G, then
f [H] (the image of H under f ) is a subgroup of G. Must f be a group homomor-
phism? Prove or nd a counterexample.
1012. Proposed by George Apostolopoulos, Messolonghi, Greece.
Let ABC denote a triangle, I its incenter, s its semiperimeter, and R
a
, R
b
, and R
c
the
circumradii of triangles IBC, ICA, and IAB, respectively. Prove that
(a)
a
R
a
+
b
R
b
+
c
R
c
3

3, and
(b) R
a
+ R
b
+ R
c

2s

3
3
.
1013. Proposed by D. M. B atinetu-Giurgiu, Matei Basarab National College, Bucha-
rest, Romania and Neculai Stanciu, George Emil Palade Secondary School, Buz au,
Romania.
http://dx.doi.org/10.4169/college.math.j.44.5.437
VOL. 44, NO. 5, NOVEMBER 2013 THE COLLEGE MATHEMATICS JOURNAL 437
Let m and n be positive real numbers and let ABC be any triangle with inradius r and
circumradius R. Prove that
(a)

tan
A
2
tan
B
2
m tan
A
2
+n tan
B
2

s
2
(4R +r)
2
m +s
2
(n 2m)
and
(b)

tan
A
2
tan
B
2
m +n tan
2
C
2

s
2
m s
2
+n (4R +r) r
.
The sums are over all cyclic permutations of (A, B, C).
1014. Proposed by Luz M. DeAlba, Drake University, Des Moines, IA.
Let C be a circle of radius r and center O. For any point A = O, the inverse of A with
respect to C is the point A

on the ray

OA with OA OA

= r
2
. The radical axis (or
power line) of two circles is the locus of points at which tangents drawn to both circles
have the same length, and a family of circles is said to be coaxal if every pair of circles
from the family has the same line as the radical axis.
Let C
1
and C
2
be nonintersecting circles, one circle not inside the other, with centers
O
1
and O
2
, and radii r
1
, r
2
> 0, respectively. Let O

1
be the inverse point of O
1
with
respect to circle C
2
, and O

2
be the inverse point of O
2
with respect to circle C
1
. Show
that O

1
and O

2
are not the limit points of the coaxal family determined by the two
circles.
1015. Proposed by Michael Goldenberg, The Ingenuity Project, Baltimore Polytechnic
Institute, Baltimore MD and Mark Kaplan, Towson University, Towson MD.
Let ABC be a given triangle, its circumcircle, and O its circumcenter. Extend median
AD to meet at A
1
and extend A
1
O to meet at A
2
. Let E be the point on the ray
A
1
D such that DA
1
= DE. Extend A
2
E to meet at A
3
. Prove that AA
3
is symmetric
to median AD with respect to the angle bisector from A.
SOLUTIONS
Inequalities with the sine and cosine functions
986. Proposed by D. M. B atinetu-Giurgiu, Matei Basarab National College, Bucha-
rest, Romania and Neculai Stanciu, George Emil Palade Secondary School, Buz au,
Romania.
For an acute ABC, prove that
(a)

_
sin A sin B
sin C
_
2

9
4
,
(b)

_
cos A cos B
cos C
_
2

3
4
,
where the sums

are over all cyclic permutations of (A, B, C).

Solution by Michael Vowe, Therwil, Switzerland.
(a) Let x = cot A, y = cot B, and z = cot C. Then xy + yz + zx = 1 and
1
y + z
=
1
cot B +cot C
=
sin B sin C
sin(B +C)
=
sin B sin C
sin A
.
438 THE MATHEMATICAL ASSOCIATION OF AMERICA
The desired inequality can be written as
1
(x + y)
2
+
1
(y + z)
2
+
1
(z + x)
2

9
4
(1)
for xy + yz + zx = 1. Inequality (1), a special case of a problem on the 1996
Iranian Mathematical Olympiad, was proposed as 1940 in Crux Mathematico-
rum and solved in Vol. 23 (6), 1996, p. 321.
(b) Let x = tan A, y = tan B, and z = tan C. Then xyz = x + y + z and
x
y + z
=
tan A
tan B +tan C
=
cos B cos C
cos A
.
The desired inequality can be written as
x
2
(y + z)
2
+
y
2
(z + x)
2
+
z
2
(x + y)
2

3
4
.
Since by Nesbitts inequality

x
y+z

3
2
,

_
x
y + z
_
2
3
_
1
3

x
y + z
_
2
3
_
1
9
__
3
2
_
2
=
3
4
.
Editors Note. The proposers noted the following generalization of (b):
By Nesbitts inequality,

tan C
tan A +tan B

3
2
. (2)
By Radons inequality and from (2), for m N,

_
cos A cos B
cos C
_
m+1

1
3
m
_

cos A cos B
cos C
_
m+1
=
1
3
m
_

tan C
tan A +tan B
_
m+1

1
3
m
_
3
2
_
m+1
=
3
2
m+1
.
The desired inequality follows when m = 1.
Also solved by ARKADY ALT, San Jose CA; GEORGE APOSTOLOPOULOS (2 solutions), Messolonghi,
Greece; MICHEL BATAILLE, Rouen, France; CHIP CURTIS, Missouri Southern State U.; MARIAN DINC A,
Bucharest, Romania; HABIB FAR, LSCS Montgomery; DMITRY FLEISCHMAN, Santa Monica CA; JOHN
HEUVER, Grande Prairie, AB Canada; ELIAS LAMPAKIS, Kipparissia, Greece; WILLIAM SEAMAN, Albright
C.; HAOHAO WANG and JERZY WOJDYLO (jointly), Southeast Missouri State U.; TITU ZVONARU, Com aesti,
Romania; and the proposers.
VOL. 44, NO. 5, NOVEMBER 2013 THE COLLEGE MATHEMATICS JOURNAL 439
A limit converging to
1
e
987. Proposed by Ovidiu Furdui, Technical University of Cluj-Napoca, Cluj-Napoca,
Romania.
Find the value of
lim
n
1
n
2
n
_
_

0
ln
n
(1 +e
nx
)e
x
dx,
where n is a positive integer.
Solution by GWstat Problem Solving Group, George Washington University, Washing-
ton DC.
We show the following generalization: For xed nonnegative b,
lim
n
1
n
2
n
_
_

0
ln
n
(b +e
nx
)e
x
dx =
1
e
.
Since
ln(b +e
nx
) = ln(e
nx
(1 +be
nx
)) = nx +ln(1 +be
nx
),
ln(b +e
nx
) = nx + O(e
nx
),
and hence
ln
n
(b +e
nx
) = (nx + O(e
nx
))
n
= n
n
x
n
+ O(n
n
x
n1
e
nx
).
It follows that
_

0
ln
n
(b +e
nx
)e
x
dx =
_

0
(n
n
x
n
+ O(n
n
x
n1
e
nx
))e
x
dx
= n
n
_
(n +1) + O
_
(n)
(n +1)
n
__
= n
n+1
(n)
_
1 + O
_
1
n
n+1
__
.
By Stirlings approximation, (n) =
_
2
n
_
n
e
_
n
_
1 + O
_
1
n
__
, and hence
_

0
ln
n
(b +e
nx
)e
x
dx = n
n+1
_
2
n
_
n
e
_
n
_
1 + O
_
1
n
__
.
Since
lim
n
ln
_
_
n
1/2

2
_
1 + O
_
1
n
___
1/2
_
= lim
n
1
2
ln n +ln

2 + O
_
1
n
_
n
= 0,
lim
n
_
n
1/2

2
_
1 + O
_
1
n
___
1/2
= 1.
440 THE MATHEMATICAL ASSOCIATION OF AMERICA
Thus,
lim
n
1
n
2
n
_
_

0
ln
n
(b +e
nx
)e
x
dx = lim
n
1
n
2
_
n
2n+
1
2

2
e
n
_
1 + O
_
1
n
__
_
1/2
=
1
e
.
Also solved by GEORGE APOSTOLOPOULOS, Messolonghi, Greece; MICHEL BATAILLE, Rouen, France; HONG-
WEI CHEN, Christopher Newport U.; BRUCE DAVIS, St. Louis C.C. at Florissant Valley; JAMES DUEMMEL,
Bellingham WA; BILL DUNN, LSCS Montgomery; DMITRY FLEISCHMAN, Santa Monica CA; TOM JAGER,
Calvin C.; STEPHEN KACZKOWSKI, South Carolina Governors School for Sci. and Math.; OMRAN KOUBA,
H.I.A.S.T., Damascus, Syria; ELIAS LAMPAKIS, Kipparissia, Greece; GEORGE MATTHEWS, Indianapolis IN;
NORA THORNBER, Raritan Valley C.C.; HAOHAO WANG and JERZY WOJDYLO (jointly), Southeast Missouri
State U.; and the proposer.
Applications of Radon and Bergstr om inequalities
988. Proposed by Jos e Luis Daz-Barrero, Universitat Polit ecnica de Catalunya,
Barcelona, Spain.
Let a
1
, a
2
, . . . , a
n
be nonzero real numbers. Prove that
n

k=1
_
a
k
a
2
k
+a
k
a
k+1
+a
2
k+1
_
2

(a
1
+a
2
+ +a
n
)
2
9(a
4
1
+a
4
2
+ +a
4
n
)
,
where a
n+1
= a
1
.
Solutions by D. M. B atinetu-Giurgiu, Matei Basarab National College, Bucharest, Ro-
mania and Neculai Stanciu, George Emil Palade Secondary School, Buz au, Romania.
Solution 1. We shall prove the following generalization of the desired inequality: If
m > 0 and a
k
= 0 for k = 1, 2, . . . , n, then
n

k=1
_
a
k
a
2
k
+a
k
a
k+1
+a
2
k+1
_
2m

n
k=1
a
2m
m+1
k
_
m+1
9
m
_
n
k=1
a
4
k
_
m
.
First note that if x, y = 0, then
(x
2
+ xy + y
2
)
2
x
4
+ y
4
+3x
2
y
2
+2xy(x
2
+ y
2
)
x
4
+ y
4
+
3
2
(x
4
+ y
4
) +(x
2
+ y
2
)
2

7
2
(x
4
+ y
4
) +2x
2
y
2

9
2
(x
4
+ y
4
). (1)
From (1) and Radons inequality,
n

k=1
_
a
k
a
2
k
+a
k
a
k+1
+a
2
k+1
_
2m

_
2
9
_
m n

k=1
a
2m
k
(a
4
k
+a
4
k+1
)
m
=
2
m
9
m
n

k=1
_
a
2m
m+1
k
_
m+1
(a
4
k
+a
4
k+1
)
m

2
m
_

n
k=1
a
2m
m+1
k
_
m+1
9
m
_
n
k=1
a
4
k
+a
4
k+1
_
m

_

n
k=1
a
2m
m+1
k
_
m+1
9
m
_
n
k=1
a
4
k
_
m
.
VOL. 44, NO. 5, NOVEMBER 2013 THE COLLEGE MATHEMATICS JOURNAL 441
Solution 2. We shall prove the following generalization of the desired inequality: If
a, b > 0 and a
k
= 0 for k = 1, 2, . . . , n, then
n

k=1
_
a
k
aa
2
k
+ba
k
a
k+1
+a
2
k+1
_
2

_
n
k=1
a
k
_
2
(2a +b)
2

n
k=1
a
4
k
.
First note that
(ax
2
+bxy +ay
2
)
2

_
a(x
2
+ y
2
) +
b
2
(x
2
+ y
2
)
_
2
=
(2a +b)
2
4
(x
4
+ y
4
+2x
2
y
2
)

(2a +b)
2
2
(x
4
+ y
4
). (2)
By Bergstr om inequality and (2),
n

k=1
_
a
k
aa
2
k
+ba
k
a
k+1
+a
2
k+1
_
2
2
n

k=1
a
2
k
(2a +b)
2
(a
4
k
+a
4
k+1
)

2(

n
k=1
a
k
)
2
(2a +b)
2

n
k=1
(a
k
+a
4
k+1
)
=
(

n
k=1
a
k
)
2
(2a +b)
2

n
k=1
a
4
k
.
The desired inequality follows from solutions 1 and 2 when m = 1 and a = b = 1,
respectively.
Also solved by HUSAIN ALSHEHHI, Texas A&M; ARKADY ALT, San Jose CA; GEORGE APOSTOLOPOULOS,
Messolonghi, Greece; MICHEL BATAILLE, Rouen, France; MINH CAN, U. of Redlands; HONGWEI CHEN,
Christopher Newport U.; CHIP CURTIS, Missouri Southern State U.; MARIAN DINC A, Bucarest, Romania;
HABIB FAR, LSCS Montgomery; OMRAN KOUBA, H.I.A.S.T., Damascus, Syria; ELIAS LAMPAKIS, Kip-
parissia, Greece; JI HU LEE, Seoul Sci. High School, Seoul; PAOLO PERFETTI, Dipartimento di Matematica,
Universit` a degli Studi di Roma Tor Vergata; WILLIAM SEAMAN, Albright C.; HAOHAO WANG and JERZY
WOJDYLO (jointly), Southeast Missouri State U.; CONNIE XU, Montgomery Blair High School; TITU ZVONARU
and NECULAI STANCIU (jointly), Romania; and the proposer.
Inequalities for areas of triangles
989. Proposed by Yagub N. Aliyev, Qafqaz University, Absheron, Azerbaijan.
Let D and E be arbitrary points on sides BC and AC of a triangle ABC and let n be a
positive, real number.
(a) Prove that if n
2
> n, then
[BDE]
n
+[ADE]
n

_
[BDA]
n
1n
+[ADC]
n
1n
_
1n
.
(b) Prove that if n
2
< n, then
[BDE]
n
+[ADE]
n

_
[BDA]
n
1n
+[ADC]
n
1n
_
1n
.
442 THE MATHEMATICAL ASSOCIATION OF AMERICA
Solution by Michel Bataille, Rouen, France.
Letting h and h

be the distances to the line

BC from A and E, respectively,

2[ABC] = h BC,
2[BDA] = h BD,
2[ADC] = h CD,
2[BDE] = h

BD,
and
h

h
=
CE
CA
.
Similarly, letting k and k

be the distances to the line

AC from B and D, respectively,

2[ABC] = k CA, 2[ADE] = k

AE, and
k

k
=
CD
BC
.
Letting =
CD
BC
and =
AE
CA
,
[ADC] = [ABC] ,
[BDA] = (1 ) [ABC] ,
[ADE] = [ABC] ,
[BDE] = (1 )(1 ) [ABC] . (1)
(a) From (1), we need to prove
(1 )
n
(1 )
n
+
n

n
((1 )
n
1n
+
n
1n
)
1n
for n > 1.
For x [0, 1], let f (x) = (1 )
n
(1 x)
n
+
n
x
n
. Since
f

(x) = n[
n
x
n1
(1 )
n
(1 x)
n1
] > 0
_
x
1 x
_
n1
>
_
1

_
n

x
1 x
>
_
1

_ n
n1
. (2)
Solving for x, we obtain f

(x) > 0 if and only if x > , where
=
(1 )
n
n1

n
n1
+(1 )
n
n1
,
and hence the minimum of f on [0, 1] is f (). Using
n
2
n 1
= n +
n
n 1
,
VOL. 44, NO. 5, NOVEMBER 2013 THE COLLEGE MATHEMATICS JOURNAL 443
we obtain
f () =
n
(1 )
n
_

n
n1
+(1 )
n
n1
_
1n
=
n
(1 )
n
_
1

n
1n
+
1
(1 )
n
1n
_
1n
=

n
(1 )
n
[
n
1n
+(1 )
n
1n
]
1n
[
n
1n
(1 )
n
1n
]
1n
.
Thus, f () = [(1 )
n
n1
+
n
1n
]
1n
and the desired inequality follows.
(b) Since 0 < n < 1, the proof is similar with (2) reversed, so that this time f ()
f (), which yields the desired inequality.
Also solved by ARKADY ALT, San Jose CA; GEORGE APOSTOLOPOULOS, Messolonghi, Greece; CHIP CURTIS,
Missouri Southern State U.; PRITHWIJIT DE, HBCSE, Mumbai, India; OMRAN KOUBA, H.I.A.S.T., Damascus,
Syria; ELIAS LAMPAKIS, Kipparissia, Greece; and the proposer.
A limit with the number of perfect squares
990. Proposed by the late Ion Cucurezeanu, Ovidius University of Constant a, Constant a,
Romania and Cezar Lupu, University of Pittsburgh, Pittsburgh, PA.
Let f be a continuous real-valued function on [1, 1], differentiable at 0. Consider
the sequence {a
n
}
n1
dened by
a
n
= N(n)
_
1/n
1/n
f (x) dx,
where N(n) is the number of perfect squares between n
3
and (n +1)
3
. Show that
lim
n

na
n
= 3 f (0).
Solution 1 by Antonio Trusiani, Universit a di Roma Tor Vergata, Rome, Italy.
Let q
n
be the number of perfect squares less than or equal to n
3
. Then q
2
n
n
3
and
(q
n
+1)
2
> n
3
, and hence
n
3/2
1 < q
n
n
3/2
for every positive integer n. Since N(n) = q
n+1
q
n
,
3
2
= lim
n
(n +1)
3/2
1 n
3/2
n
1/2
lim
n
N(n)

n
lim
n
(n +1)
3/2
+1 n
3/2
n
1/2
=
3
2
.
Moreover,
n
_
n
1/n
f (x) dx =
F(
1
n
) F(0)
1
n
+
F(
1
n
) F(0)

1
n
= 2F

(0) = 2 f (0)
444 THE MATHEMATICAL ASSOCIATION OF AMERICA
where F(t ) =
_
t
0
f (x) dx. Finally,
lim
n

na
n
= lim
n
N(n)

n
n
_
1/n
1/n
f (x) dx = 3 f (0).
Note that the differentiability of f at 0 is not necessary.
Solution 2 by Santiago de Lus an (student) and

Angel Plaza, University of Las Palmas
de Gran Canaria, Spain.
Within k consecutive numbers located around some large integer m, we would ex-
pect to nd about
k
2

m
perfect squares. There are k = 3n
2
+3n +1 numbers between
m = n
2
and the next cube, so we may expect to nd roughly
3
2

n perfect squares
among these. This is an excellent estimate, since the actual count is always one of the
two integers that bracket that quantity. (See A sequence of bits with strange statistics
at http://www.numericana.com/answer/numbers.htm#bits and The On-
line Encyclopedia of Integer Sequences, Sequence A140074 at http//oeis.org/
A140074.) Therefore, the proposed limit can be written as
lim
n

na
n
= lim
n

n
3
2

n
_
1/n
1/n
f (x) dx =
3
2
lim
n
_
1/n
1/n
f (x) dx
1/n
=
3
2
lim
n
f (1/n) + f (1/n)
1
= 3 f (0).
Also solved by GEORGE APOSTOLOPOULOS, Messolonghi, Greece; MICHEL BATAILLE, Rouen, France;
CHARLES BURNETTE (student), Drexel U.; CON AMORE PROBLEM GROUP, U. of Aarhus, Denmark; CHIP
CURTIS, Missouri Southern State U.; JAMES DUEMMEL, Bellingham WA; DMITRY FLEISCHMAN, Santa
Monica CA; NATACHA FONTES-MERZ, Westminster C.; GWSTAT PROBLEM SOLVING GROUP, George Wash-
ington U.; AHMAD HABIL, Damascus U., Damascus, Syria; TOM JAGER, Calvin C.; STEPHEN KACZKOWSKI,
South Carolina Governors School for Sci. and Math.; OMRAN KOUBA, H.I.A.S.T., Damascus, Syria; ELIAS
LAMPAKIS, Kipparissia, Greece; NORTHWESTERN UNIVERSITY MATH PROBLEM SOLVING GROUP; ONU-
SOLVE PROBLEM GROUP, Ohio Northern U.; S K PATEL, Gujarat U., Ahmedabad, India; WILLIAM SEAMAN,
Albright C.; SKIDMORE COLLEGE PROBLEM GROUP; JOHN ZACHARIAS, Arlington VA; and the proposers.
Editors Note. The original statement Show that lim
n

na
n
= 3| f (0)|. in Problem
990 should be replaced by Show that lim
n

na
n
= 3 f (0)..
Editors Note. The Con Amore Problem Group was inadvertently omitted from the list
of solvers of CMJ Problems 968 and 970.
Editors Note. The following were inadvertently omitted from the list of solvers in the
September issue of CMJ:
Reza Akhlahgi, Big Sandy C.T.C. (Problems 983 and 984); Michel Bataille, Rouen,
France (Problems 982, 983, and 984); Paraviz Khalili, Christopher Newport U. (Prob-
lem 984); C at alin Fetoiu and Alexandru Orzan (jointly), Iancu de Hunedoara Na-
tional C., Romania (Problems 981, 982, 983, and 984); and John Zacharias, Arlington,
VA (Problem 982).
VOL. 44, NO. 5, NOVEMBER 2013 THE COLLEGE MATHEMATICS JOURNAL 445