Problems and Solutions

Reviewed work(s):
Source: The American Mathematical Monthly, Vol. 119, No. 3 (March 2012), pp. 247-254
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/10.4169/amer.math.monthly.119.03.247 .
Accessed: 21/09/2012 22:58

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
 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, Bog-
dan Petrenko, Richard Pfiefer, Cecil C. Rousseau, Leonard Smiley, Kenneth Stolarsky,
Richard Stong, Walter Stromquist, Daniel Ullman, Charles Vanden Eynden, Sam Van-
dervelde, 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 July 31, 2012. 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
11628. Proposed by Jeffrey C. Lagarias and Michael E. Zieve, University of Michigan,
Ann Arbor, MI. Define the Lenstra constant L(R) of a commutative ring R to be the
size of the largest subset A of R such that a − b is a unit (invertible element) in R
for any distinct elements a, b ∈ A. Show that for each positive integer N , the Lenstra
constant of the ring Z[1/N ] is the least prime that does not divide N .
11629. Proposed by Olivier Oloa, University of Versailles, Rambouillet, France. Let
Z 1  2
σ 1 1
f (σ ) = x + d x.
0 log x 1−x
(a) Show that f (0) = log(2π) − 3/2.
(b) Find a closed form expression for f (σ ) for σ > 0.

11630. Proposed by Constantin Mateescu, High School ‘Zinca Golescu’, Pitesti, Ro-
mania. For triangle ABC, let H be the orthocenter, I the incenter, O the circumcenter,
and R the circumradius. Let b and c be the lengths of the sides opposite B and C,
respectively, and let l be the length of the line segment from A to BC along the angle
bisector at A. Let α be the radian measure of angle BAC. Prove that
bc
+ max{b, c} ≤ 4R cos(α/4),
l
with equality if and only if rays AH, AI, and AO divide angle BAC into four equal angles.
11631. Proposed by Pál Péter Dályay, Szeged, Hungary. A quasigroup (Q, ∗) is a set
Q together with a binary operation ∗ such that for each a, b ∈ Q there exist unique x
and unique y (which may be equal) such that ax = b and ya = b. The Cayley table of
a finite quasigroup is its ‘times table’. A quasigroup has property P if each row of the
table is a rotation of the first row.

March 2012] PROBLEMS AND SOLUTIONS 247

 Find all positive integers n for which there exists a quasigroup ({1, . . . , n}, ∗) with
property P in which all elements are idempotent. (For instance, the Cayley table below
defines a binary operation on {1, . . . , 5} with property P in which each element is
idempotent.)

* 1 2 3 4 5

1 1 5 4 3 2
2 3 2 1 5 4
3 5 4 3 2 1
4 2 1 5 4 3
5 4 3 2 1 5

11632. Proposed by Cezar Lupu, University of Pittsburgh, Pittsburgh, PA, and Dan
Schwarz, Bucharest, Romania. Let n be a positive integer, and write a vector x ∈ Rn
as (x1 , . . . , xn ). For x, y, a, b ∈ Rn let
X
[x, y]a,b = xi y j min(ai , b j ).
1≤i, j≤n
n
Show that for x, y, z, a, b, c in R with nonnegative entries,

[x, x]a,a · [y, z]2b,c + [y, y]b,b · [z, x]2c,a

 
1/2 1/2
a,a · [y, y]b,b · [z, z]c,c · [x, x]a,a · [y, y]b,b + [x, y]a,b .
≤ [x, x]1/2 1/2

11633. Proposed
P by Anthony Sofo, Victoria University, Melbourne, Australia. For real
a, let Hn(a) = nj=1 j −a . Show that for integers a, b, and n with a ≥ 1, b ≥ 0, and
n ≥ 1,
k(Hk2 + Hk(2) ) + 2(k + b)a Hk(1) Hk+b−1
(a)
n
(a)
(Hn2 + Hn(2) ).
X
= Hn+b
k=1
k(k + b)a

11634. Proposed by D. M. Bătineţu-Giurgiu, Matei Basarab National College,

Bucharest, Romania, and Neculai Stanciu, George Emil Palade, Buzău, Roma-
nia. Let (x1 , . . . xn ) be an n-tuple of positive numbers, and let X = nk=1 xk . Let
P
a and m be nonnegative numbers, and let b, c, d be positive. Suppose p ≥ 1 and
cX p > d max1≤k≤n xk . Show that
p

n
X a X + bxk (an + b)n mp 1−mp
≥ X .
k=1
p
cX p − d xk (cn p − d)m

SOLUTIONS

Another Hankel Determinant

11475 [2010, 86]. Proposed by Ömer P
Eğecioğlu, University of California Santa Bar-
bara, Santa Barbara, CA. Let h k = kj=1 1j , and let Dn be the determinant of the
(n + 1) × (n + 1) Hankel matrix with (i, j) entry h i+ j+1 for 0 ≤ i, j ≤ n. (Thus,
D1 = −5/12 and D2 = 1/216.) Show that for n ≥ 1,

248 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 119

 n
(−1) j (n + j + 1)!(n + 1)h j+1
Qn 4 X
i=1 i!
Dn = Q2n+1 · .
i=1 i! j=0
j!( j + 1)!(n − j)!

Solution by Richard Stong, Center for Communications Research, San Diego, CA. The
desired value is the determinant of the matrix B that results from the given matrix by
subtracting row n − 1 from row n, row n − 2 from row n − 1, and so on to row 0 from
row 1. This produces b0, j = h j+1 , and bi, j = 1/(i + j + 1) for i ≥ 1. In particular, the
bottom n rows of B agree with the bottom n rows of the (n + 1) × (n + 1) Hilbert ma-
trix H , where Hi, j = 1/(i + j + 1) for i, j ∈ {0, . . . , n}. In the 19th century, Hilbert
computed
Qn 4
i=1 i!
det H = Q2n+1
i=1 i!

and
   
n+i +1 n+ j +1 i + j
Hi,−1j = (−1) i+ j
(i + j + 1) .
n− j n−i i
Hence expanding det B along the top row yields
n
X
Dn = det B = (det H ) −1
b0, j H j,0
j=0
n
(−1) j (n + j + 1)!(n + 1)h j+1
Qn 4
i=1 i!
X
= Q2n+1 · .
i=1 i! j=0
j!( j + 1)!(n − j)!

Also solved by R. Chapman (U. K.), O. Kouba (Syria), O. P. Lossers (Netherlands), K. McInturff GCHQ
Problem Solving Group, and the proposer.

Short Runs from an Urn

11485 [2011, 182]. Proposed by Neetu Badhoniya, K. S. Bhanu, and M. N. Deshpande,
Institute of Science, Nagpur, India. An urn contains a white balls and b black balls, and
a ≥ 2b + 3. Balls are drawn at random from the urn and placed in a row as they are
drawn. Drawing halts when three white balls are drawn in succession. Let X be the
number of isolated pairs of white balls in the lineup produced during play, and let Y
be the number of isolated white balls. Show that
b b(a + b + 1)
E[X ] = , and E[Y ] = .
a+1 (a + 1)(a + 2)

Solution by Bob Tomper, Mathematics Department, University of North Dakota. Let

E[X a,b ] and E[Ya,b ] be the expected values of X and Y respectively when starting
with an urn containing a white balls and b black balls, where a ≥ 2b + 3.
For a ≥ 3, we have E[X a,0 ] = E[Ya,0 ] = 0, which agrees with the given formulae.
We use induction on b, the number of black balls. Assume the formulae are correct up
to b − 1, where b ≥ 1 (and all appropriate a values), and consider the case of a white
balls and b black balls with a ≥ 2b + 3.
The lineup begins with B, WB, WWB, or WWW, where W and B indicate drawing
a white or a black ball, respectively. The last also ends the lineup; after each of the
others, the number of white balls remaining in the urn is at least three more than

March 2012] PROBLEMS AND SOLUTIONS 249

twice the number of black balls, so the conditions for the induction hypothesis apply.
Accounting for the isolated W in WB and the paired WW in WWB,


E[X a,b ] = Pr[B]E[X a,b−1 ] + Pr[WB]E[X a−1,b−1 ] + Pr[WWB] 1 + E[X a−2,b−1 ]
which gives this expression for E[X a,b ]:
 
b b−1 a b b−1 a a−1 b b−1
+ + 1+ .
a+ba+1 a+ba+b−1 a a+ba+b−1a+b−2 a−1
b
This simplifies to a+1
. Similarly,


E[Ya,b ] = Pr[B]E[Ya,b−1 ] + Pr[WB] 1 + E[Ya−1,b−1 ] + Pr[WWB]E[Ya−2,b−1 ]
whereby
b (b − 1) (a + b) (b − 1)(a + b − 1)
 
a b
E[Ya,b ] = + 1+
a + b (a + 1) (a + 2) a + b a + b − 1 a(a + 1)
a a−1 b (b − 1)(a + b − 2) b(a + b + 1)
+ = .
a+ba+b−1a+b−2 (a − 1)a (a + 1)(a + 2)

Also solved by M. Andreoli, D. Beckwith, R. Chapman (U. K.), W. J. Cowieson, P. P. Dályay (Hungary),
C. González-Alcón & Á. Plaza (Spain), N. Grivaux (France), S. J. Herschkorn, O. Kouba (Syria), J. Lobo
(Costa Rica), O. P. Lossers (Netherlands), R. Martin (Germany), K. McInturff, M. A. Prasad (India), R. Pratt,
K. Schilling, D. Senft, J. Simons (U. K.), R. Stong, R. Tauraso (Italy), S. Xiao, CMC 328, GCHQ Problem
Solving Group (U. K.), and the proposers.

A Weighted Fermat Triangle Problem

11491 [2010, 278]. Proposed by Nicolae Anghel, University of North Texas, Denton,
TX. Let P be an interior point of a triangle having vertices A0 , A1 , and A2 , opposite
sides of length a0 , a1 , and a2 , respectively, and circumradius R. For j ∈ {0, 1, 2}, let
r j be the distance from P to A j . Show that
r0 r1 r2 1
2
+ 2+ 2 ≥ .
a0 a1 a2 R

Solution I by Marian Dinca, Romania . Let E = r0 /a02 + r1 /a12 + r2 /a22 . Let P j be the
projection of P to the side of length a j , and let d j = P P j . Now P, P1 , A0 , and P2 lie
on the circle with diameter A0 P (of length r0 ), so ∠P1 P P2 = A1 + A2 and P1 P2 =
r0 sin A0 . Thus, P1 P22 = d12 + d22 − 2d1 d2 cos(A1 + A2 ) = (d1 sin A2 + d2 sin A1 )2 +
(d1 cos A2 − d2 cos A1 )2 and P1 P2 = r0 sin A0 ≥ d1 sin A2 + d2 sin A1 . By the law of
sines, r0 a0 ≥ d1 a2 + d2 a1 . Similarly, r1 a1 ≥ d0 a2 + d2 a0 and r2 a2 ≥ d0 a1 + d1 a0 . Thus
d1 a2 + d2 a1 d2 a0 + d0 a2 d0 a1 + d1 a0
E≥ 3
+ 3
+
a0 a1 a23
     
a1 a2 a2 a0 a0 a1
= + d0 + + d 1 + + d2
a23 a13 a03 a23 a13 a03
2(a0 d0 + a1 d1 + a2 d2 )
     
2 2 2 4K
≥ d0 + d1 + d2 = = ,
a1 a2 a2 a0 a0 a1 a1 a2 a3 4K R
where K is the area of 1A0 A1 A2 .

250 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 119

Solution II by Zoltán Vőrős, Tiszavasvári, Hungary . With ri and di as above, the
weighted Erdős–Mordell inequality states that r0 x 2 + r1 y 2 + r2 z 2 ≥ 2(d0 yz + d1 zx +
d2 x y) for real x, y, z. Letting x = 1/a0 , y = 1/a1 , and z = 1/a2 , we get
 
r0 r1 r2 d0 d1 d1
+ + ≥2 + + ,
a02 a12 a22 a1 a2 a2 a0 a0 a1
from which the result follows as above.
Editorial comment. This problem was on the 2000 US Olympiad Team Selection Test.
It is a weighted Fermat triangle problem whose general solution is known (see Yu-
jin Shen & Juan Tolosa, “The weighted Fermat triangle problem,” Int. J. Math. Math.
Sci. (2008), 16 pp., http://dx.doi.org/10.1155/2008/283846). The weighted
Erdős–Mordell inequality can be found in Am. Math. Monthly 108 (2001) 165–168;
available at http://dx.doi.org/10.2307/2695531. The notation of R1 and r1 as
the distance from P to A and BC helps relate Erdős–Mordell’s R1 + R2 + R3 ≥
2(r1 + r2 + r3 ) to Euler’s R > 2r.
Also solved by R. Bagby, P. P. Dályay (Hungary), J. Fabrykowski & T. Smotzer, J. Hamilton & T. Smotzer, P.
Nüesch (Switzerland), M. Tetiva (Romania), L. Zhou, and the proposer.

A linear transformation and Hermite polynomials

11493 [2010, 279]. Proposed by Johann Cigler, Universität Wien, Vienna, Austria.
Consider the Hermite polynomials Hn , defined by
X n
Hn (x, s) = (2k − 1)!! (−s)k x n−2k ,
0≤k≤n/2
2k

where m!! = i<m/2 (m − 2i) for positive m, with (−1)!! = 1. Let L be the lin-
Q

ear transformation from Q[x, s] to Q[x] determined by L1 = 1, L x k s j = x k Ls j

for j, k ≥ 0, and L H2n (x, s) = 0 for n > 0. (Thus, for example, 0 = L H2 (x, s) =
L(x 2 − s) = x 2 − Ls, so Ls = x 2 .) Define the tangent numbers T2n+1 by tan z =
z 2n
2n+1
/(2n + 1)!, and the Euler numbers E 2n by sec z = n≥0 E 2n (2n)!
P P
n≥0 T2n+1 z .
(a) Show that
LH2n+1 (x, s) = (−1)n T2n+1 x 2n+1 .
(b) Show that
E 2n
Ls n = x 2n .
(2n − 1)!!

Solution by BSI Problems Group, Bonn, Germany. We work in the ring of formal power
series in the indeterminate t withPcoefficients in the ring Q[x, PWe extend L
to this

s].
ring coefficient-wise: if F(t) = i f i (x, s)t i , then L F(t) = i L f i (x, s) t i . It is
then easy to see that, for G(t) ∈ Q[x][[t]],
L G(t)F(t) = G(t)L F(t) .

Using (2k − 1)!! = (2k)!/2k k!, we find that

∞ X (−s)k t 2k (xt)n−2k
tn
= e xt−st /2 .
2
X
Hn (x, s) =
n=0
n! k,n
2k k! (n − 2k)!

March 2012] PROBLEMS AND SOLUTIONS 251

Extracting even and odd powers of t gives
∞
t 2n
= cosh(xt)e−st /2 and
2
X
H2n (x, s) (1)
n=0
(2n)!
∞
t 2n+1
= sinh(xt)e−st /2 .
2
X
H2n+1 (x, s) (2)
n=0
(2n + 1)!

Let [t n ]F(t) denote the coefficient of t n in F(t). Applying L to (1) gives

 
1 = cosh(xt)L e−st /2 ,
2
(3)
so
  E 2n
Ls n = (−1)n 2n n! [t 2n ]L e−st /2 = (−1)n 2n n! [t 2n ] sech(xt) =
2
x 2n ,
(2n − 1)!!
which proves (b).
Applying L to (2), and using (3), we have
 
LH2n+1 (x, s) = (2n + 1)! [t 2n+1 ] sinh(xt)L e−st /2
2

sinh(xt)
= (2n + 1)! [t 2n+1 ] = (−1)n T2n+1 x 2n+1 ,
cosh(xt)
which proves (a).
Also solved by D. Beckwith, E. H. M. Brietzke (Brazil), R. Chapman (U. K.), P. P. Dályay (Hungary), G. C.
Greubel, O. P. Lossers (Netherlands), J. Matysiak and W. Matysiak (Poland), K. McInturff, R. Stong, S. Xiao,
L. Zhou, GCHQ Problem Solving Group (U. K.), and the proposer.

Zeros of Symmetric Functions

11495 [2010, 370]. 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.
Composite solution by Jim Simons, Cheltenham, U. K., and Richard Stong, Center for
Communications Research, San Diego, CA. We may assume that a, b, and c are not
all 0. Since multiplication by a constant does not affect the statements, we may also
assume that they are integers and have no common factor.
For each prime p, write p α kX to mean p α | X and p α+1 - X . We use the fact that
if X + Y + Z = 0, p α kX , p β kY , and p γ kZ , then the smallest two of α, β, and γ are
equal.
Suppose that a 2 b + b2 c + c2 a = 0. Let p be a prime divisor of abc, with p α ka,
p kb, and p γ kc. Since gcd(a, b, c) = 1, at least one of α, β, and γ is 0. If α = 0,
β

then p β ka 2 b, p 2β+γ kb2 c, and p 2γ kca ; it follows that β = 2γ . Similarly, β = 0 implies

γ = 2α, and γ = 0 implies α = 2β. Thus there are pairwise relatively prime numbers
X, Y, and Z such that a = XY 2 , b = YZ 2 , and c = ZX 2 . In fact, X (respectively, Y and
Z ) is the product of all prime powers dividing gcd(c, a) (respectively, gcd(a, b) and
gcd(b, c)).
Substituting these values into a 2 b + b2 c + c2 a yields
0 = X 2 Y 5 Z 2 + Y 2 Z 5 X 2 + Z 2 X 5 Y 2 = (XYZ)2 (X 3 + Y 3 + Z 3 ).

252 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 119

If XYZ 6= 0, then X 3 + Y 3 + Z 3 = 0; this is the cubic case of Fermat’s Last Theorem,
which has no nontrivial solutions. Hence at least one of {X, Y, Z } is 0, so two of
{a, b, c} equal 0, and hence ab2 + bc2 + ca 2 = 0. By symmetry, ab2 + bc2 + ca 2 = 0
implies a 2 b + b2 c + c2 a = 0.
However, by hypothesis exactly one of the three given expressions equals 0, so it
must be the third expression. Letting ω = e2πi/3 , we may write
0 = a 3 + b3 + c3 + 6abc = 3(a 3 + b3 + c3 + 6abc)
= (a + b + c)3 + (a + bω + cω2 )3 + (a + bω2 + cω)3 .
This is a solution to Fermat’s equation in the ring Z[ω], but the usual proof of Fermat’s
Last Theorem in the cubic case shows that, in fact, there are no nontrivial solutions in
Z[ω]. Thus, one of the three terms must be zero. If a + bω + cω2 = 0, then all three
of a, b, c must be zero and so a + b + c = 0; the same is true if a + bω2 + cω = 0.
What remains is that a + b + c = 0, as claimed.
Also solved by R. Chapman (U. K.), S. Chatadus (Poland), O. P. Lossers (Netherlands), C. R. Pranesachar
(India), R. Prasad (India), R. E. Prather, M. Tetiva (Romania), the GCHQ Problem Solving Group (U. K.), and
the proposer.

A Series with Harmonic Numbers

11499 [2010, 371]. Proposed by Omran Kouba, Higher Institute for Applied Science
Pn Technology, Damascus, Syria. Let Hn be the nth harmonic number, given by Hn =
and
k=1 1/k. Let

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

Prove that for k ≥ 2,

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

Solution by Douglas B. Tyler, Raytheon, Torrance, CA . Fix k ≥ 2. Let an = log k −

(Hkn − Hn ). Since Hn = log n + γ + o(1/n) as n → ∞, we have an = o(1/n), so
an → 0. Thus
∞ ∞ k 
∞ X 
X X X 1 1
Sk = (−1)n−1 an = (a2n−1 − a2n ) = −
n=1 n=1 n=1 m=1
(2n − 1)k + m 2nk
∞ X
X k Z 1
= (x 2nk−k+m−1 − x 2nk−1 ) d x.
n=1 m=1 0

The integrands are nonnegative, so we may sum first:

k
∞ X
X xk 1 kx 2k−1
(x 2nk−k+m−1 − x 2nk−1 ) = · +
n=1 m=1
1 + xk 1 − x x 2k − 1

kx 2k−1 1 xk − 1
= + − .
x 2k − 1 2(x − 1) 2(x − 1)(x k + 1)

March 2012] PROBLEMS AND SOLUTIONS 253

Let
1
xk − 1
Z
Ik = d x.
0 (x − 1)(x k + 1)
Since
1
kx 2k−1
Z  
1 1
2k
+ dx = log(2k),
0 x − 1 2(x − 1) 2
it remains to prove that
[k/2]
1 π X (2l − 1)π
Ik = log 2 + 2 (k + 1 − 2l) cot .
k k l=1 2k

x k −1 2
P ζ 1
Expanding the integrand yields (x−1)(x k +1) = k ζ ζ −1 · x−ζ , where we sum over
the kth roots of −1. Except for ζ = −1 when k is odd, the summands occur in
ζ ζ̄
conjugate pairs ζ −1 1
· x−ζ + ζ̄ −1 · x−1 ζ̄ = (x−cosx+1
θ )2 +sin2 θ
, where ζ = eiθ . With J =
R1 x+1
0 (x−cos θ)2 +sin2 θ
d x, we have

1 + cos θ x − cos θ 1
  
1
J= log(x − 2x cos θ + 1) +
2
tan −1
2 sin θ sin θ 0

θ 1 − cos θ cos θ
    
1
= log(2 − 2 cos θ) + cot tan−1
+ tan −1
2 2 sin θ sin θ
θ θ π −θ
   
= log 2 sin + cot .
2 2 2
Now if θ = (2l − 1)π/k, then
bk/2c 
(2l − 1)π (k + 1 − 2l)π (2l − 1)π
  
2X
Ik = log 2 sin + cot
k l=1 2k 2k 2k
bk/2c bk/2c
(2l − 1)π π X (2l − 1)π
 
2X
= log 2 sin + 2 (k + 1 − 2l) cot .
k l=1 2k k l=1 2k
R 1 1/2
If k is odd, then l = k+1 2
in the first sum corresponds to ζ = −1 and k2 0 x+1 dx =
π (2l−1)π
log(2 sin 2 ). The first sum thus equals k l=1 log(2 sin 2k ) = k log 2, from the
1 1
Pk 1
k
sin (2l−1)π . This completes the proof.
Qk 1
identity l=1 2k
= 2k−1
Editorial comment. Some solvers derived the equivalent Sk = k−1
2k
log 2 + 12 log k −
π
j=1 cot 2k by applying the digamma identity 9(x) = 9(1 − x) − π cot π x to the
P k jπ
4k
sum Sk = − 2k1 k−1j=1 9( 2 + 2k ) − 2k γ .
P 1 j k−1

Also solved by P. Bracken, R. Chapman (U. K.), H. Chen, E. A. Herman, O. P. Lossers (Netherlands), A.
Stenger, R. Stong, M. Tetiva (Romania), GCHQ Problem Solving Group (U. K.), and the proposer.

254 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 119