The American Mathematical Monthly

ISSN: 0002-9890 (Print) 1930-0972 (Online) Journal homepage: http://www.tandfonline.com/loi/uamm20

Problems and Solutions

Gerald A. Edgar, Daniel H. Ullman & Douglas B. West

To cite this article: Gerald A. Edgar, Daniel H. Ullman & Douglas B. West (2018)
Problems and Solutions, The American Mathematical Monthly, 125:10, 944-952, DOI:
10.1080/00029890.2018.1524660

To link to this article: https://doi.org/10.1080/00029890.2018.1524660

Published online: 01 Dec 2018.

Submit your article to this journal

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=uamm20
 PROBLEMS AND SOLUTIONS
Edited by Gerald A. Edgar, Daniel H. Ullman, Douglas B. West
with the collaboration of Paul Bracken, Ezra A. Brown, Zachary Franco, Christian Friesen,
László Lipták, Rick Luttmann, Frank B. Miles, Lenhard Ng, Kenneth Stolarsky, Richard
Stong, Daniel Velleman, Stan Wagon, Elizabeth Wilmer, Fuzhen Zhang, and Li Zhou.

Proposed problems should be submitted online at

americanmathematicalmonthly.submittable.com/submit.
Proposed solutions to the problems below should be submitted by April 30, 2019 via
the same link. More detailed instructions are available online. Proposed problems
must not be under consideration concurrently at any other journal nor be posted to
the internet before the deadline date for solutions. An asterisk (*) after the number
of a problem or a part of a problem indicates that no solution is currently available.

PROBLEMS
12076. Proposed by Tibor Beke, University of Massachusetts, Lowell, MA. From each of
the three feet of the altitudes of an arbitrary triangle, produce two points by projecting this
foot onto the other two sides. Show that the six points produced in this way are concyclic.
12077. Proposed by Max A. Alekseyev, George Washington University, Washington, DC.
Let f (x) be a monic polynomial of degree n with distinct zeros a1 , . . . , an . Prove
n
ain−1
= 1.
i=1
f  (ai )
n
12078. Proposed by Hideyuki Ohtsuka, Saitama, Japan. Let m q
be the q-binomial coeffi-
cient defined by
   1 − q n−i
m−1
n
= .
m q i=0
1 − q i+1

For a positive integer s and for 0 < q < 1, prove

∞
q sn q s (1 − q s+1 )
s+n = .
n=1 s+1
1 − qs
q

12079. Proposed by Moubinoul Omarjee, Lyceé Henri IV, Paris, France. Choose x1 in
(0, 1), and let xn+1 = (1/n) nk=1 ln(1 + xk ) for n ≥ 1. Compute limn→∞ xn ln n.
12080. Proposed by Daniel Sitaru, Drobeta Turnu Severin, Romania. Let ABC be a scalene
acute triangle with semiperimeter s. Let A1 , A2 , and A3 be the points on BC such that AA1
is an altitude, AA3 is a median (i.e., A3 is the midpoint of BC), and AA2 is a symmedian
(i.e., the ray AA2 is the reflection of the ray AA3 across the angle bisector at A). Define B1 ,
B2 , B3 and C1 , C2 , C3 similarly. Prove
A2 A3 B2 B3 C2 C3 4s 2
+ + > 2 .
A1 A2 B1 B2 C1 C2 a + b2 + c2
doi.org/10.1080/00029890.2018.1524660

944 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 125
12081. Proposed by Cristian Chiser, Elena Cuza College, Craiova, Romania. Let A and
B be complex n-by-n matrices such that AB − BA is invertible and such that A2 + B 2 =
c(AB − BA) for some rational number c. Prove c ∈ {−1, 0, 1}, and show that n is a multiple
of 4 when c = 0.
12082. Proposed by Stan Wagon, Macalester College, St. Paul, MN, and Piotr Zielinski,
Boston, MA. Alice, Bob, and Charlie are prisoners in the care of a warden who lines them
up in order, Charlie in front of Bob and Bob in front of Alice. The warden has k differently
colored hats with k ≥ 3 and places one hat on each prisoner’s head, making the selection at
random and discarding the k − 3 unused hats. The prisoners know what the k colors are but
see only the hats of the prisoners in front of them (i.e., Alice sees two hats, Bob sees one,
and Charlie sees none). The prisoners then guess the colors of their hats in turn, first Alice,
then Bob, then Charlie. All prisoners hear the guesses. If the three guesses are correct, then
the prisoners will all be freed.
The prisoners know the rules and can devise a strategy in advance. No communication
other than the guesses is allowed once the hats are placed. What is the best possible strategy
for the prisoners?

SOLUTIONS

When the Nine-Point Center Lies on the Circumcircle

11958 [2017, 179]. Proposed by Kent Holing, Trondheim, Norway. (a) Find a condition on
the side lengths a, b, and c of a triangle that holds if and only if the nine-point center lies
on the circumcircle.
(b) Characterize the triangles whose nine-point center lies on the circumcircle and whose
incenter lies on the Euler line.
Solution by Koupa Tak Lun Koo, Beacon College, Hong Kong, China.
(a) Let O, H, and N be the circumcenter, orthocenter, and nine-point center, respectively.
It is well known that N is the midpoint of OH and that OH 2 = 9R 2 − (a 2 + b2 + c2 ),
where R is the circumradius. Note that N lies on the circumcircle if and only if ON = R,
or 9R 2 − (a 2 + b2 + c2 ) = OH 2 = (2R)2 . Substituting R = abc/(4K), where K is the
area of the triangle, this condition becomes
5a 2 b2 c2 5a 2 b2 c2
a 2 + b2 + c2 = 5R 2 = = ,
16K 2 (a + b + c)(−a + b + c)(a − b + c)(a + b − c)
which is equivalent to

a 6 + b6 + c6 + a 2 b2 c2 = (a 2 + b2 )(b2 + c2 )(c2 + a 2 ). (∗)

(b) It is well known that the incenter lies on the Euler line if and only if the triangle is
isosceles. Letting b = c in (∗), we have

a 6 + 2b6 + a 2 b4 = (a 2 + b2 )(2b2 )(a 2 + b2 ),

or a 2 (a 2 + b2 )(a 2 − 3b2 ) = 0. Hence the triangle satisfies the required conditions if and
only if a 2 = 3b2 = 3c2 so that the angles of the triangle are π/6, π/6, and 2π /3.
Editorial comment. For more on the triangles satisfying the condition in part (a), see
O. Bottema (2008), Topics in Elementary Geometry, New York: Springer, page 76. There
one finds the equivalent condition t = −3/8 where t = cos A cos B cos C. Also, if the
nine-point circle and the circumcircle intersect at angle θ, then t = − sin2 (θ/2). For more

December 2018] PROBLEMS AND SOLUTIONS 945

on the location of the incenter relative to the Euler line, see A. P. Guinand (1984), Euler
lines, tritangent centers, and their triangles, this Monthly, 91(5): 290–300.
Also solved by M. Bataille (France), B. S. Burdick, P. P. Dályay (Hungary), O. Geupel (Germany), S. Hito-
tumatu, (Japan), T. Horine, M. Goldenberg, M. Kaplan, O. Kouba (Syria), J. Minkus, C. R. Pranesachar
(India), A. Stadler (Switzerland), R. Stong, Z. Vőrős (Hungary), M. Vowe (Switzerland), T. Wiandt, L. Zhou,
T. Zvonaru (Romania), N. Stanciu (Romania), GCHQ Problem Solving Group (U. K.), and the proposer.

A Permanent Solution
11959 [2017, 179]. Proposed by Donald Knuth, Stanford University, Stanford, CA. Prove
that, for any n-by-n matrix A with (i, j )-entry ai,j and any t1 , . . . , tn , the permanent of A
is
 
1  
n n
σi it + σj i,j ,
a
2n i=1 j =1

where the outer sum is over all 2n choices of (σ1 , . . . , σn ) ∈ {1, −1}n .
Solution by Fredrik Ekström, University of Oulu, Oulu, Finland. We use the signs in the
choices for (σ1 , . . . , σn ) to cancel spurious contributions to the sum, leaving 2n perm(A).
Let [n] = {1, . . . , n}.  
With σ0 = 1 and ai,0 = ti , we have σi (ti + nj=1 σj ai,j ) = nj=0 σi σj ai,j . In expand-
ing the product over i, we select
 an index ϕ(i) for
each i, where ϕ is a function from
[n] to {0, . . . , n}. Hence ni=1 nj=0 σi σj ai,j = ϕ ni=1 σi σϕ(i) ai,ϕ(i) . where the sum is
over all such functions ϕ. After  interchanging
the order of summation, the sum in the
problem statement becomes ϕ σ ni=1 σi σϕ(i) ai,ϕ(i) , where σ runs through all elements
of {1} × {1, −1}n .
Consider ϕ whose image omits some k ∈ [n]. In the sum over σ for such ϕ, the terms
when σk = 1 and σk = −1 cancel. After grouping terms by the least such k, the sum over ϕ
reduces to the sum over terms where (ϕ(1), . . . , ϕ(n)) is a permutation of [n], eliminating
the dependence on t1 , . . . , tn . Summing over permutations π of [n], we then claim

n 
n 
n
σi σϕ(i) ai,ϕ(i) = σi σπ(i) ai,π(i) = 2n ai,π(i) = 2n perm(A).
ϕ σ i=1 π σ i=1 π i=1

For the central equality above, note that the product for a particular choice of σ accesses
2n
n factors in the image of σ , each one twice. Therefore, for each σ the coefficient on
n
a
i=1 i,π(i) is 1, and there are 2 choices for σ .
Also solved by R. Chapman (U. K.), P. P. Dályay (Hungary), D. Fleischman, O. Geupel (Germany), T. Horine,
O. Kouba (Syria), J. H. Lindsey II, O. P. Lossers (Netherlands), S. Navasardyan, M. Omarjee (France),
M. A. Prasad (India), J. C. Smith, R. Stong, R. Tauraso (Italy), and the proposer.

Hurwitz to the Rescue

11960 [2017, 179]. Proposed by Ulrich Abel, Technische Hochschule Mittelhessen, Fried-
berg, Germany. Let m and n be natural numbers, and, for i ∈ {1, . . . , m}, let ai be a real
number with 0 ≤ ai ≤ 1 . Define

m 
1   m
f (x) = 2 (1 + ai x) − m (1 + ai x) .
mn n
x i=1 i=1

Let k be a nonnegative integer, and write f (k) for the kth derivative of f . Show that
f (k) (−1) ≥ 0.

946 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 125
Solution by the editors. Let S be the collection of functions f that are infinitely differ-
entiable at −1 and such that f (k) (−1) ≥ 0 for all integers k ≥ 0. Note that S is closed
under sums, products, and nonnegative scalar multiples. For 0 ≤ ai ≤ 1, we claim that the
function g(x) = 1 + ai x belongs to S. Indeed, g(−1) = 1 − ai ≥ 0, g  (−1) = ai ≥ 0, and
g (k) (−1) = 0 for k ≥ 2.
For m = 1, we have f (x) = 0, so f ∈ S.
For m = 2, we have
 
f (x) = x −2 (1 + a1 x)2n + (1 + a2 x)2n − 2(1 + a1 x)n (1 + a2 x)n
 n 2
2 (1 + a1 x) − (1 + a2 x)
n
= (a1 − a2 )
(1 + a1 x) − (1 + a2 x)

n−1 2

= (a1 − a2 ) 2
(1 + a1 x) (1 + a2 x)
i n−1−i
.
i=0

Thus, by the closure properties, f ∈ S.

Finally, let m ≥ 3. We cite the 3-page note on the AM–GM inequality due to
Hurwitz, A. (1891), Uber den Vergleich des arithmetischen und des geometrischen
Mittels, J. Reine Ange. Math., 108: 266–268, in which it is proved that the polynomial
X1m + X2m + · · · + Xm
m
− mX1 X2 · · · Xm has a representation in the form

(Xi − Xj )2 Pi,j (X1 , . . . , Xm ),
1≤i<j ≤m

where Pi,j are polynomials with all nonnegative coefficients. So

 2
 (1 + ai x)n − (1 + aj x)n  
f (x) = 2
Pi,j (1 + a1 x)n , . . . , (1 + am x)n .
1≤i<j ≤m
x

Now x −2 ((1 + ai x)n − (1 + aj x)n )2 ∈ S by the case m = 2, and

Pi,j ((1 + a1 x)n , . . . , (1 + am x)n ) ∈ S
by the closure properties, so finally f ∈ S again by the closure properties.
No solutions were received for m ≥ 3.

A Radical Distribution
11962 [2017, 180]. Proposed by Elton Hsu, Northwestern University, Evanston, IL. Let
{Xn }n≥1 be a sequence of independent and identically distributed random variables each
taking the values ±1 with probability 1/2. Find the distribution of the random variable
 


 1 X1 1 X2 1
+ + + ··· .
2 2 2 2 2

Solution by Li Zhou, Polk State College, Winter Haven, FL. The probability that the random
variable is at most x is (2/π ) arcsin x, for x ∈ [0, 1].
Let Y0 , Y1 , . . . and Z0 , Z1 , . . . be random variables defined by
  
  

1 1 X1 1  1 X1 1 X2 1
Y0 = , Y1 = + , Y2 = + + , ...
2 2 2 2 2 2 2 2 2

December 2018] PROBLEMS AND SOLUTIONS 947

and
  
  

1 1 X2 1  1 X2 1 X3 1
Z0 = , Z1 = + , Z2 = + + , ....
2 2 2 2 2 2 2 2 2
For n ≥ 0, the variable Zn is √ independent of X1 and has the same distribution as Yn .
Furthermore, note that Yn+1 = (1 + X1 Zn )/2.
We first prove inductively that Yn and Zn take the value cos(kπ /2n+2 ) with probability
1/2n , for k ∈ {1, 3, 5, . . . , 2n+1 − 1}. The claim is immediate for n = 0. If it holds for
 value n, then for k ∈ {1, − 1} the random variable Yn+1 takes the value
n+1
some 3, 5, . . . , 2 
1 + cos(kπ/2n+2 ) /2 or 1 − cos(kπ/2n+2 ) /2 with probability 1/2n+1 each. Since
 
1 1 kπ kπ 1 1 kπ kπ (2n+2 − k)π
+ cos n+2 = cos n+3 and − cos n+2 = sin n+3 = cos ,
2 2 2 2 2 2 2 2 2n+3
this completes the induction.
Therefore, the probability that Yn is at most x is |An (x)|/2n , where
 
  kπ
An (x) = k ∈ 1, 3, 5, . . . , 2n+1 − 1 : 0 ≤ cos n+2 ≤ x .
2
As n → ∞,
2n+1 2n+1
(arccos 0 − arccos x) =
|An (x)| ∼ arcsin x,
π π
which yields the claimed limiting distribution.
Also solved by R. Agnew, E. Bojaxhiu (Albania) & E. Hysnelaj (Australia), R. Chapman (U. K.), P. J. Fleis-
chman, N. Grivaux (France), J. A. Grzesik, J. C. Keiffer, O. Kouba (Syria), J. H. Lindsey II, M. A. Prasad
(India), K. Schilling, N. C. Singer, J. C. Smith, R. Stong, D. B. Tyler, E. I. Verriest, T. Wiandt, and the pro-
poser.

A Summation Inequality
11963 [2017, 180]. Proposed by Gheorghe Alexe and George-Florin Serban, Braila,
Romania. Let a1 , . . . , an be positive real numbers with nk=1 ak = 1. Show that

n
(ai + ai+1 )4
≥ 12n,
i=1
ai2 − ai ai+1 + ai+1
2

where an+1 = a1 .
Solution by Leonard Giugiuc, Drobeta Turnu Severin, Romania. We first observe that if a
and b are positive real numbers, then
(a + b)4
≥ 12ab.
a 2 − ab + b2
This follows from
(a + b)4 (a 2 − 4ab + b2 )2
− 12ab = ≥ 0.
a 2 − ab + b2 a 2 − ab + b2
Applying this fact to each term of the summation and then using the AM–GM inequality,
we obtain


n 2
 n
(ai + ai+1 ) 4 n 
≥ 12 a a ≥ 12n 
n
ai = 12n.
i i+1
a 2 − ai ai+1 + ai+1
i=1 i
2
i=1 i=1

948 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 125
Editorial comment. Several solvers noted that the inequality in the problem is √an equality
√
when√n is even
√ and the terms of the sequence a1 , . . . , an alternate between ( 6 + 2)/2
and ( 6 − 2)/2. When n is odd, the inequality is always strict.
Also solved by A. Ali (India), A. Alt, R. Boukharfane (France), P. Bracken, E. Braune (Austria), R. Chapman
(U. K.) P. P. Dályay (Hungary), D. Fleischman, N. Ghosh, T. Horine, K. T. L. Koo (China), O. Kouba (Syria),
J. H. Lindsey II, O. P. Lossers (Netherlands), V. Mikayelyan (Armenia), R. Nandan, P. Perfetti (Italy), M. Reid,
E. Schmeichel, J. C. Smith, A. Stadler (Switzerland), R. Stong, R. Tauraso (Italy), L. Zhou, GCHQ Problem
Solving Group (U. K.), and the proposer.

Condition for a Certain Point on a Triangle

11965 [2017, 274]. Proposed by Leonard Giugiuc, Drobeta Turnu Severin, Romania. Let
ABC be a triangle with circumradius R. Prove that there exists a point M on side BC such
that MA · MB · MC = 32R 3 /27 if and only if 2 cot B cot C = 1.
Solution by Dain Kim, Seoul Science High School, Seoul, South Korea. We begin with the
following result.
Lemma. Let ABC be inscribed in the circle . If AD is the diameter of  through A, and
M is the point where BC and AD meet, then
MB · MC
cot B · cot C = .
MA2
Proof. Because AD is a diameter, ∠ABD = ∠ACD = π/2, and by the inscribed angle
theorem it follows that ∠BAD = ∠BCD = π/2 − C and ∠CAD = ∠CBD = π/2 − B.
The law of sines gives
cos C sin( π2 − C) MB cos B sin( π2 − B) MC
= = and = = .
sin B sin B MA sin C sin C MA
Multiplying these two equations together yields the desired result. 
We now address separately the necessity and sufficiency of 2 cot B cot C = 1.
(Necessity) Assume there exists a point M on side BC such that MA · MB · MC = 32R 3 /27.
Let line MA intersect the circumcircle of triangle ABC again at D. Let MA = x and MD =
y. Since 2R is the diameter of the circumcircle, x + y = AD ≤ 2R. By the power-of-
the-point theorem, MB · MC = MA · MD. Multiplying this on both sides by MA yields
MA · MB · MC = MA2 · MD = x 2 y. The AM–GM inequality then implies
 
1 x + x + 2y 3 (2(2R))3 32R 3
MA · MB · MC = x y ≤
2
≤ = .
2 3 54 27
Equality holds only when x = 2y and x + y = 2R, in which case x = (4/3)R and y =
(2/3)R. Thus AD is a diameter.
Applying the lemma,
MB · MC MA · (MD)2 xy 2 y 1
cot B · cot C = = = 2 = = .
MA 2 MA · MB · MC x y x 2
This is the required condition 2 cot B · cot C = 1.
(Sufficiency) Assume 2 cot B cot C = 1. Let AD be the diameter of the circumcircle
through A, and let M be the point where BC and AD meet. We must show that MA · MB ·
MC = 32R 3 /27. As before, MA · MB · MC = x 2 y, with x = MA and y = MD. From the
lemma,
MB · MC x3
x 2 y = MA · MB · MC = 2
· MA3 = cot B cot C · MA3 = .
MA 2

December 2018] PROBLEMS AND SOLUTIONS 949

Therefore y = x/2, so x + x/2 = 2R. It follows that x = 4R/3 and x 3 /2 = 32R 3 /27.
Also solved by M. Bataille (France) D. Fleischman, T. Horine, O. Kouba (Syria), O. P. Lossers (Netherlands),
R. Nandan, V. Schindler (Germany), J. C. Smith, R. Stong, E. I. Verriest, L. Zhou, and the proposer.

Evaluate an Integral
11966 [2017, 274]. Proposed by Cornel Ioan Vălean, Teremia Mare, Timiş, Romania.
Prove
 1
x ln(1 + x) π2 (ln 2)2
dx = + .
0 1 + x2 96 8

Solution by Juan Manuel Sánchez (student), Universidad de Antioquia, Medellin, Colom-

 1 x dt
bia. Since 0 1+xt = ln(1 + x),
  1  1 1
1
x ln(1 + x) 1
x2 x2
dx = dt dx = dx dt
0 1 + x2 0 0 (1 + x )(1 + xt)
2
0 0 (1 + x )(1 + xt)
2
 1  1 
1 (tx − 1) 1
= + dx dt
0 1+t 1 + x2 1 + tx
2
0
 1  
1 t ln 2 π ln(1 + t)
= − + dt
0 1+t
2 2 4 t
   1
ln 2 1 t π 1 1 ln(1 + t)
= dt − dt + dt
2 0 1 + t2 4 0 1 + t2 0 t (1 + t 2 )
 1  
(ln 2)2 π2 1 t
= − + ln(1 + t) − dt
4 16 0 t 1 + t2
 1  1
(ln 2)2 π2 ln(1 + t) t ln(1 + t)
= − + dt − dt.
4 16 0 t 0 1 + t2
1 x ln(1+x) dx
Solving this for 0 1+x 2
yields
 
1
x ln(1 + x) (ln 2)2 π2 1 1
ln(1 + t)
dx = − + dt.
0 1+x 2 8 32 2 0 t
Since
  ∞
1 ∞  (−1)i−1
1
ln(1 + t) (−1)i t i
dt = dt =
0 t 0 i=0
i+1 i=1
i2

∞ ∞ ∞
1 1 1 1 π2
= − 2 = = ,
i=1
i2 i=1
(2i)2 2 i=1 i 2 12

we obtain
 1
x ln(1 + x) (ln 2)2 π2 1 π2 π2 (ln 2)2
dx = − + · = + .
0 1 + x2 8 32 2 12 96 8

Also solved by T. Amdeberhan & V. H. Moll, K. F. Andersen (Canada), A. Arenas & M. Bello & M. Benito
& Ó. Ciaurri & E. Fernández & E. Labarga & L. Roncal (Spain), M. Bataille (France), A. Berkane (Algeria),
R. Boukharfane (France), K. N. Boyadzhiev, P. Bracken, B. Bradie, R. Chapman (U. K.), H. Chen, P. P. Dályay
(Hungary), B. E. Davis, D. Fritze (Germany), S. Gao, C. Georghiou (Greece), N. Ghosh, M. L. Glasser (Spain),

950 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 125
H. Grandmontagne (France), J. M. Groah, J. A. Grzesik, T. Hakobyan & S. Navasardyan (Armenia), A. Han-
nan (India), E. A. Herman, T. Horine, M. Ivan (Romania), S. Kaczkowski, K. T. L. Koo (China), O. Kouba
(Syria), A. Kourdouklas (Greece), K.-W. Lau (China), H. Lee (South Korea), O. P. Lossers (Netherlands),
V. Lucic (U. K.), L. Kempeneers & J. Van Casteren (Belgium), P. Magli (Italy), V. Mikayelyan (Armenia),
R. Nandan, G. Negri (Italy), M. Omarjee (France), P. Perfetti (Italy), R. Poodiack, F. A. Rakhimjanovich
(Uzbekistan), H. Ricardo, S. Seales, A. N. Sharma (India), S. Sharma (India), S. Silwal & N. Taylor, J. Singh
(India), J. C. Smith, A. Stadler (Switzerland), J. Steier, R. Stong, R. Tauraso (Italy), E. I. Verriest, H. Wang
& J. Wojdylo, T. Wiandt, M. Wildon (U. K.), M. R. Yegan (Iran), Y. Zhao, L. Zhou, GCHQ Problem Solving
Group (U. K.), Get Stoked Student Problem Solving Group, and the proposer.

Zeros of a Truncated Riemann Zeta Function

11970 [2017, 275]. Proposed by Albert Stadler, Herrliberg, Switzerland. Let

ζ5 (z) = 1 + 2−z + 3−z + 4−z + 5−z ,

where z is a complex number. Prove that ζ5 (z) = 0 when the real part of z is greater than
or equal to 0.9.
Solution by Li Zhou, Polk State College, Winter Haven, FL. Let z = x + iy with x ≥ 0.9.
Since
 |3−z + 5−z −z −z −x
 | ≤ |3 | + |5 | = 3 + 5 ≤ 3
−x −0.9
+ 5−0.9 < 0.61, it suffices to show
1 + 2−z + 4−z  > 0.61. To this end, we compute
 2  2
 
1 + 2−z + 4−z 2 = 1 + cos(y ln 2) + cos(2y ln 2) + sin(y ln 2) + sin(2y ln 2)
2x 4x 2x 4x
 
1 1 cos(y ln 2) cos(2y ln 2) cos(y ln 2)
=1+ x + x +2 + +
4 16 2x 4x 8x
  2  
cos(y ln 2) 1 1 3 1 2
=4 + 1 + + 1 −
2x 4 4x 4 4x
 
3 1 2
≥ 1 − 0.9 > 0.612 ,
4 4
completing the proof.
Editorial comment. A discussion, with plots, of the zeros of truncated zeta functions ζk (z)
can be found in Borwein, P., Fee, G., Ferguson, R., van der Waall, A. (2007), Zeros of
partial sums of the Riemann zeta function, Experiment. Math. 16(1): 21–39.
Also solved by P. Bracken, T. Horine, O. P. Lossers (Netherlands), V. Mikayelyan (Armenia), R. Stong,
E. I. Verriest, and the proposer.

A Mean Inequality
11971 [2017, 369]. Proposed by Spiros P. Andriopoulos, Third High School of Amaliada,
Eleia, Greece. For n ≥ 2, let a1 , . . . , an be positive real numbers. Prove

n n−1 ⎛  ⎞
 2
  2a a
≥⎝ ⎠ .
i j
(1 + ai ) 1+
i=1 i<j
a i + a j

Solution
by Koopa Tak Lun Koo, Beacon College, Hong Kong, China. In the product
i<j (1 + ai )(1 + aj ), there are precisely n − 1 occurrences of (1 + ak ) for each k in

December 2018] PROBLEMS AND SOLUTIONS 951

{1, . . . , n}. Therefore, using both the AM–GM and GM–HM inequalities, we can write

n n−1
  
(1 + ai ) = (1 + ai )(1 + aj ) = (1 + ai + aj + ai aj )
i=1 1≤i<j ≤n 1≤i<j ≤n
 √  √
≥ (1 + 2 ai aj + ai aj ) = (1 + ai aj )2
1≤i<j ≤n 1≤i<j ≤n

2 ⎛ ⎞
    2
2 2ai aj ⎠
≥ 1+ =⎝ 1+ .
1≤i<j ≤n
1
ai
+ 1
aj 1≤i<j ≤n
ai + aj

Equality occurs if and only if all the real numbers ai are equal.
Also solved by M. Bataille (France), A. Berkane (Algeria), R. Boukharfane (France), R. Chapman (U. K.),
P. P. Dályay (Hungary), S. Dubey, D. Fleischman, O. Geupel (Germany), N. Ghosh, L. Giugiuc (Romania),
M. Goldenberg & M. Kaplan, T. Hakobyan & S. Navasardyan (Armenia), A. Hannan (India), E. A. Herman,
T. Horine, S. Hwang (South Korea), J. Kim (South Korea), O. Kouba (Syria), H. Kwong, W. Lai & J. Risher,
J. H. Lindsey ll, O. P. Lossers (Netherlands), D. Marinescu (Romania), R. Martin (Germany), V. Mikayelyan
(Armenia), R. Nandan, M. Omarjee (France), P. Perfetti (Italy), F. A. Rakhimjanovich (Uzbekistan), E. Schme-
ichel, J. C. Smith, A. Stadler (Switzerland), N. Stanciu & T. Zvonaru (Romania), R. Stong, R. Tauraso (Italy),
D. B. Tyler, E. I. Verriest, H. Widmer (Switzerland), L. Zhou, GCHQ Problem Solving Group (U. K.), NSA
Problem Solving Group, and the proposer.

Inradius and Exradii of a Tetrahedron

11972 [2017, 369]. Proposed by Yun Zhang, Xi’an Senior High School, Xi’an China. Let
r be the radius of the sphere inscribed in a tetrahedron whose exscribed spheres have radii
r1 , r2 , r3 , and r4 . Prove
√ √ √ √  √
r 3 r1 + 3 r2 + 3 r3 + 3 r4 ≤ 2 3 r1 r2 r3 r4 .

Solution by Hansruedi Widmer, Baden, Switzerland. By the AM–GM inequality,


1 1 1 1 1
+ + ≥ 3
3 ri rj rk ri rj rk

for 1 ≤ i, j, k ≤ 4. Summing over the four triplets {i, j, k} yields

4 4 √ 3 r
1 k
≥ √ .
k=1
r k k=1
3 r 1 r 2 r3 r4

The result follows from the identity 4k=1 1/rk = 2/r. (See Toda, A. A. (2014). Radii of
the inscribed and escribed spheres of a simplex. Inter. J. Geom. 3(2): 5–13.)

 only if r1 = · · · = r4 , e.g., for isosceles tetrahe-

Editorial comment. Equality occurs if and
dra. The n-dimensional analogues are n+1 k=1 1/rk = (n − 1)/r (Toda, 2014) and
√ √ √ √
r( n r1 + n r2 + · · · + n rn+1 ) ≤ (n − 1) n r1 r2 · · · rn+1 .

Also solved by M. Bataille (France), R. Boukharfane (France), R. Chapman (U. K.), P. P. Dályay (Hungary),
O. Geupel (Germany), L. Giugiuc (Romania), T. Horine, K. T. L. Koo (China), O. Kouba (Syria), O. P. Lossers
(Netherlands), D. Marinescu (Romania), R. Nandan, J. C. Smith, A. Stadler (Switzerland), R. Stong, R. Tauraso
(Italy), T. Wiandt, L. Zhou, and the proposer.

952 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 125