Problems and Solutions

Source: The American Mathematical Monthly, Vol. 119, No. 7 (August‒September 2012), pp.
608-615
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/10.4169/amer.math.monthly.119.07.608 .
Accessed: 31/03/2013 15:35

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

This content downloaded from 128.206.9.138 on Sun, 31 Mar 2013 15:35:08 PM

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 December 31, 2012. Additional information, such as gen-
eralizations and references, is welcome. The problem number and the solver’s
name and address should appear on each solution. An asterisk (*) after the num-
ber of a problem or a part of a problem indicates that no solution is currently
available.

PROBLEMS
11656. Proposed by Valerio De Angelis, Xavier University of Louisiana, New Orleans,
LA. The sign chart of a polynomial f with real coefficients is the list of successive
pairs (, σ ) of signs of ( f 0 , f ) on the intervals separating real zeros of f f 0 , together
with the signs at the zeros of f f 0 themselves, read from left to right. Thus, for f =
x 3 − 3x 2 , the sign chart is ((1, −1), (0, 0), (−1, −1), (0, −1), (1, −1), (1, 0), (1, 1)).
As a function of n, how many distinct sign charts occur for polynomials of degree n?
11657. Proposed by Gregory Galperin, Eastern Illinois University, Charleston, IL,
and Yury Ionin, Central Michigan University, Mount Pleasant,
MI. Given a set V of n
points in R2 , no three of them collinear, let E be the set of n2 line segments joining
distinct elements of V .
(a) Prove that if n 6 ≡ 2 (mod 3), then E can be partitioned into triples in which the
length of each segment is greater than the sum of the other two.
(b) Prove that if n ≡ 2 (mod 3) and e is an element of E, then E\{e} can be so parti-
tioned.
11658. Proposed by Greg Oman, University of Colorado at Colorado Springs, Col-
orado Springs, CO. Let V be the vector space over R of all (countably infinite) se-
quences (x1 , x2 , . . .) of real numbers, equipped with the usual addition and scalar
multiplication. For v ∈ V , say that v is binary if vk ∈ {0, 1} for k ≥ 1, and let B be the
set of all binary members of V . Prove that there exists a subset I of B with cardinality
2ℵ0 that is linearly independent over R. (An infinite subset of a vector space is linearly
independent if all of its finite subsets are linearly independent.)
11659. Proposed by Albert Stadler, Herrliberg, Switzerland. Let x be real with 0 <
x < 1, and consider the sequence han i given by a0 = 0, a1 = 1, and, for n > 1,
2
an−1
an = .
xan−2 + (1 − x)an−1
http://dx.doi.org/10.4169/amer.math.monthly.119.07.608

608 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 119

This content downloaded from 128.206.9.138 on Sun, 31 Mar 2013 15:35:08 PM

All use subject to JSTOR Terms and Conditions
Show that
∞
1 X
lim = (−1)k x k(3k−1)/2 .
n→∞ an
k=−∞

11660. Proposed by Stefano Siboni, University of Trento, Trento, Italy. Consider the
following differential equation: s 00 (t) = −s(t) − s(t)2 sgn(s 0 (t)), where sgn(u) denotes
u. Show that if s(0) = a and s 0 (0) = b with ab 6 = 0, then (s, s 0 ) tends to
the sign of √
(0, 0) with s 2 + s 02 ≤ C/t as t → ∞, for some C > 0.
11661. Proposed by Giedrius Alkauskas, Vilnius University, Vlinius, Lithuania. Find
every function f on R+ that satisfies the functional equation
 
1−z
(1 − z) f (x) = f f (x z)
z
for x > 0 and 0 < z < 1.
11662. Proposed by H. Stephen Morse, Fairfax, Va. Let ABCD be the vertices of a
square, in that order. Insert P and Q on AB (in the order APQB) so that each of P
and Q divides AB ‘in extreme and mean ratio’ (that is, |AB|/|BQ| = |BQ|/|QA| and
|AB|/|AP| = |AP|/|PB|.) Likewise, place R and S on CD so that CRSD is divided in
the same proportions as APQB. The four intersection points of AR, BS, CP, and DQ
are called the harmonious quartet of the square √ on its base pair (AB, CD). They form
a rhombus whose long diagonal has length ( 5 + 1)/2 times the length of its short
diagonal.
Given a cube, create the harmonious quartet for each of its six faces, using each
edge as part of a base pair exactly once, according to this scheme: label the vertices
on one face of the cube ABCD and the corresponding vertices of the opposite face
A0 B 0 C 0 D 0 . Pair AB with CD, AA 0 with BB 0 , and BC with B 0 C 0 . The rest of the pairings
are then forced: A0 B 0 with C 0 D 0 , AD with A0 D 0 , and CC 0 with DD 0 . This generates 24
points.
(a) Show that these 24 points are a subset of the 32 vertices of a rhombic triaconta-
hedron (a convex polyhedron bounded by 30 congruent rhombic faces, meeting three
each across their obtuse angles at 20 vertices, and five each across their acute angles
at 12 vertices), and find a construction for the remaining eight vertices.
(b) Show, moreover, that the 12 end points of the longer diagonals of the six con-
structed rhombi are the vertices of an icosahedron I , and these diagonals are edges of
the icosahedron.
(c) Show that the 12 end points of the shorter diagonals of the constructed rhombi,
together with the eight additional vertices of the triacontahedron, are the vertices of a
dodecahedron. Show also that these shorter diagonals are edges of that dodecahedron.

SOLUTIONS

An Inequality in Three Variables

11543 [2010, 390]. Proposed by Richard Stong, Center for Communications Research,
San Diego, CA. Let x, y, z be positive numbers with x yz = 1. Show that (x 5 + y 5 +
z 5 )2 ≥ 3(x 7 + y 7 + z 7 ).

August–September 2012] PROBLEMS AND SOLUTIONS 609

This content downloaded from 128.206.9.138 on Sun, 31 Mar 2013 15:35:08 PM

All use subject to JSTOR Terms and Conditions
Solution by Erik I. Verriest, Georgia Institute of Technology, Atlanta, GA. The problem
is equivalent to showing that the minimum of

2
x 5 + y5 + z5
,
x 7 + y7 + z7
subject to the constraints x yz = 1 and x, y, z > 0, is at least 3. Since the ratio equals
3 when x = y = z = 1, the bound can be achieved.
First suppose that x → 0. In this case or y, z tends to ∞, and

2
10
x 5 + y5 + z5 max{x, y, z}
≥ → ∞.
x 7 + y7 + z7 3 max{x, y, z}7
Similar reasoning holds for y → 0 and z → 0. Therefore the minimum value is
achieved at an interior point where x, y, z > 0.
We apply the method of Lagrange multipliers, letting

2
x 5 + y5 + z5
L= + λ(x yz − 1).
x 7 + y7 + z7
Necessary conditions for a stationary point are
∂ ∂L ∂L ∂L
λ= = = = 0.
L ∂x ∂y ∂z
Thus, at any stationary point, x yz = 1. Setting U = x 5 + y 5 + z 5 and V = x 7 + y 7 +
z 7 and computing the other partials derivatives gives

10x 5 U V − 7x 7 U 2 + λV 2 = 0,
10y 5 U V − 7y 7 U 2 + λV 2 = 0, (1)
10z 5 U V − 7z 7 U 2 + λV 2 = 0.

Adding and simplifying gives U 2 V + λV 2 = 0, and since V 6 = 0, λ = −U 2 /V . Sub-

stituting this into (1) and dividing (as we may) by U V 2 gives
7 7 10 5
x − x + 1 = 0. (2)
V U
and counterparts with y and z in place of x.
Consider now for a > 0 the function f defined by f (a) = pa 7 − qa 5 + 1, where
both p and q are positive. Note f (0) = 1. This function is minimal √ at that√a (call it
a ∗ ) such that 7 pa 6 − 5qa 4 = 0. Since a ∗ > 0, it follows that a ∗ = 5q/ 7 p. The
minimum is
 7/2  5/2 ! 7/2
2 5 5/2 q 7/2
 
5 5 q
f (a ) =
∗
− + 1 = 1 − < 1.
7 7 p 5/2 7 7 p 5/2

Thus f has two positive zeros, one (double) positive zero, or no positive zeros, when
f (a ∗ ) is less than, equal to, or larger than 0, respectively. Now return to (2) by writing
p = 7/V and q = 10/U , which we can do because U and V are positive. If f has no
positive zero, then (2) has no solution. If f has one positive zero, then the solution to
(2) is x = y = z, and since x yz = 1 this is the first case x = y = z = 1. If f has two

610 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 119

This content downloaded from 128.206.9.138 on Sun, 31 Mar 2013 15:35:08 PM

All use subject to JSTOR Terms and Conditions
positive zeros, then in the solution to (2), two of x, y, z are equal, say x = y, and the
other is z = 1/x 2 . This leads to the minimization of

(2x 15 + 1)2
x 6 (2x 21 + 1)
over x > 0. The stationary value is given by

2(x 3 )12 + 8(x 3 )5 − 9(x 3 )7 − 1 = 0,

which has a unique positive solution for X = x 3 = 1 and corresponds to a minimum.

This again yields the first solution x = y = z = 1.
Editorial comment. Oliver Geupel and the proposer independently proved the inequal-
ity using the method found in V. Cı̂rtoaje, The equal variable method, Journal of In-
equalities in Pure and Applied Mathematics 8 (2007) issue 1, article 15, corollary 1.7
(page 3), available at http://www.emis.de/journals/JIPAM/images/059_06_
JIPAM/059_06.pdf.
Also solved by R. Bagby, P. Bracken, D. Constales (Belgium), A. Cooper, P. P. Dályay (Hungary), D. Fleis-
chman, O. Geupel (Germany), J.-P. Grivaux (France), K.-W. Lau (China), J. H. Lindsey II, O. P. Lossers
(Netherlands), K. McInturff, D. J. Moore, P. Perfetti (Italy), C. R. Pranesachar (India), A. Stenger, C. Y.
Yildirim (Turkey), Y. Yu, S. M. Zemyan, GCHQ Problem Solving Group (U. K.), Texas State University
Problem Solvers Group, and the proposer.

A Rhombus From a Triangle

11547 [2011, 84]. Proposed by Francisco Javier Garcı́a Capitán, I.E.S Álvarez
Cubero, Priego de Córdoba, Spain, and Juan Bosco Romero Márquez, Univer-
sity of Valladolid, Spain. Let the altitude AD of triangle ABC be produced to
meet the circumcircle again at E. Let K , L, M, and N be the projections of D
onto the lines BA, AC, CE, and EB, and let P, Q, R, and S be the intersec-
tions of the diagonals of DKAL, DLCM, DMEN , and DNBK , respectively. Let
|XY| denote the distance from X to Y , and let α, β, γ be the radian measure
of angles BAC, CBA, ACB, respectively. Show that PQRS is a rhombus and that
|QS|2 /|PR|2 = 1 + cos(2β) cos(2γ )/sin2 α.
Solution by Robin Chapman, University of Exeter, Exeter, England. We have a cyclic
quadrilateral BACE whose diagonals are perpendicular and meet at D. Let a = |DA|,
b = |DB|, c = |DC|, and e = |DE|. (Note that a, b and c are not the side-lengths of
triangle ABC.) We use Cartesian coordinates with origin D. By aligning the coordinate
axes appropriately, we get A = (0, a), B = (b, 0), C = (−c, 0), and E = (0, −e). By
a standard property of intersecting chords of a circle, ae = bc.
The point K is the unique point on AB for which  2 DK and AB are perpendicular. A
2
routine calculation gives the ordered pair K = a 2a+bb 2 , a 2ab+b2 , and, similarly, the pairs
 2 2
  2 2
  2 2

L = − a 2a+cc 2 , a 2ac+c2 , M = − c2ce+e2 , − c2c+ee 2 , and N = b2be+e2 , − b2b+ee 2 .
The point P is the intersection of the line KL with the y-axis DA. Let X =
(x, y) and U = (u, v) with
x 6 = u. The intersection of the line XU and the y-axis
is 0, (xv − yu)/(x − u) . Performing the calculation gives
   
abc ae
P = 0, 2
= 0, ,
a + bc a+e

August–September 2012] PROBLEMS AND SOLUTIONS 611

This content downloaded from 128.206.9.138 on Sun, 31 Mar 2013 15:35:08 PM

All use subject to JSTOR Terms and Conditions
since ae = bc. Similarly,
     
bc ae bc
Q= − ,0 , R = 0, − , S= ,0 .
b+c a+e b+c
Thus PQRS is a parallelogram, with centre D and with perpendicular diagonals. Hence
PQRS is a rhombus.
Since e = bc/a, we have that
|QS|2 (a + e)2 (a 2 + bc)2
= = 2 .
|PR| 2 (b + c) 2 a (b + c)2
Considering the right-angled triangle BDA yields tan β = ab .
Therefore,
1 − tan2 β b2 − a 2
cos(2β) = = .
1 + tan2 β a 2 + b2
Similarly,
c2 − a 2
cos(2γ ) = .
a 2 + c2
Computing twice the area of triangle ABC one way yields
|AD| |BC| = a(b + c),
and another yields
p
|AB| |AC| sin α = (a 2 + b2 )(a 2 + c2 ) sin α.
Equating the right sides gives
a 2 (b + c)2
sin2 α = .
(a 2 + b2 )(a 2 + c2 )
Therefore,
cos(2β) cos(2γ ) (a 2 − b2 )(a 2 − c2 )
1+ = 1 +
sin2 α a 2 (b + c)2
a 4 + 2a 2 bc + b2 c2 (a 2 + bc) |QS|2
= = = .
a 2 (b + c)2 a 2 (b + c)2 |PR|2

Also solved by M. Bataille (France), P. P. Dályay (Hungary), A. Ercan (Turkey), M. Goldenberg & M. Ka-
plan, D. Gove, J.-P. Grivaux (France), J. G Heuver (Canada), O. Kouba (Syria), P. Nüesch (Switzerland), C. R.
Pranesachar (India), J. Schlosberg, R. Stong, M. Tetiva (Romania), Z. Vörös (Hungary), M. Vowe (Switzer-
land), GCHQ Problem Solving Group (U. K.), and the proposers.

Derivative Cauchy–Schwarz
11548 [2011, 85]. Proposed by Cezar Lupu (student), University of Bucharest,
Bucharest, Romania, and Tudorel Lupu, Decebal High School, Constanta, Romania.
Let f be a twice-differentiable real-valued function with continuous second derivative,
and suppose that f (0) = 0. Show that
Z 1 Z 1 2
( f (x)) d x ≥ 10
00 2
f (x) d x .
−1 −1

612 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 119

This content downloaded from 128.206.9.138 on Sun, 31 Mar 2013 15:35:08 PM

All use subject to JSTOR Terms and Conditions
Solution by Kee-Wai Lau, Hong Kong, China. Let g(x) = f (x) + f (−x) so that
g(0) = g 0 (0) and g is twice continuously differentiable. Integrate by parts twice:
Z 1 Z 1
1 1
Z
g(x) d x = − (x − 1)g 0 (x) d x = (x − 1)2 g 00 (x) d x.
0 0 2 0

By Cauchy–Schwarz, and abbreviating g(x) d x to g and so on, we have

Z 1 2 Z 1 Z 1 Z 1
1 00 2 1
2
(x − 1) 4
g 00 .


g ≤ g = (1)
0 4 0 0 20 0

Since (g 00 (x))2 = ( f 00 (x) + f 00 (−x))2 ≤ 2 ( f 00 (x))2 + ( f 00 (−x))2 , we have

Z 1
2
Z 1 Z 1
2
g 00 ( f 00 (x))2 + ( f 00 (−x))2 = 2 f 00 .


≤2 (2)
0 0 −1

Finally, from (1) and (2) we have

Z 1 Z 1 2 Z 1 2 Z 1 2
00
f (x) + f (−x) .



f ≥ 10 g = 10 = 10 f
−1 0 0 −1

Also solved by K. F. Andersen (Canada), R. Bagby, M. Bello-Hernández & M. Benito (Spain), G. E. Bilodeau,
M. W. Botsko & L. Mismas, C. Burnette, R. Chapman (U. K.), H. Chen, D. Constales (Belgium), P. P. Dályay
(Hungary), P. J. Fitzsimmons, P. Gallegos (Chile), J.-P. Grivaux (France), L. Han (U.S.A.) & L. Yu (China),
E. A. Herman, E. J. Ionascu, O. Kouba (Syria), J. H. Lindsey II, O. P. Lossers (Netherlands), R. Mortini & J.
Noël (France), M. Omarjee (France), P. Perfetti (Italy), Á. Plaza (Spain), K. Schilling, J. Simons (U. K.), A.
Stenger, R. Stong, R. Tauraso (Italy), T. Trif (Romania), E. I. Verriest, M. Vowe (Switzerland), H. Wang & Y.
Xia, Y. Wang, T. Wiandt, L. Zhou, Barclays Capital Problems Solving Group, NSA Problems Group, and the
proposer.

A Triple Functional Equation

11549 [2011, 85]. Proposed by Marian Tetiva, National College “Gheorghe Roşca
Codreanu”, Bı̂rlad, Romania. Determine all continuous functions f on R such that
for all x,

f ( f ( f (x))) − 3 f (x) + 2x = 0.

Solution by Barclays Capital Problems Solving Group, London (U. K.). We show that
all such functions f have the form f (x) = x + c or f (x) = c − 2x, for a real constant
c. These are readily verified to be solutions, so we need only show that there are no
others.
Fix a solution f . First note that f is injective. Indeed, if f (x) = f (y), then 2x =
3 f (x) − f ( f ( f (x))) = 3 f (y) − f ( f ( f (y))) = 2y, so x = y. It follows that f is
strictly monotone, and hence f (x) has a limit (finite or infinite) as x → ∞. If f tends
to a finite limit a as x → ∞, then 2x = 3 f (x) − f ( f ( f (x))) → 3a − f ( f (a)), a
contradiction. Similarly, we see that f (x) tends to an infinite limit as x → −∞. Thus
f is surjective.
Now write f i for the ith iterate of f , which makes sense for all integers (posi-
tive, negative, and zero). For a given real number x, we have f i+3 (x) − 3 f i+1 (x) +
2 f i (x) = 0 for all i. Solving this linear homogeneous recurrence yields real numbers
A x , Bx , C x such that f i (x) = A x + Bx i + C x (−2)i for all i.

August–September 2012] PROBLEMS AND SOLUTIONS 613

This content downloaded from 128.206.9.138 on Sun, 31 Mar 2013 15:35:08 PM

All use subject to JSTOR Terms and Conditions
 Now we consider two cases: f increasing and f decreasing. First, suppose that f
is increasing. For a given x, examine f i (x) = A x + Bx i + C x (−2)i . If C x > 0, then
f i (x) is large and positive when i is large and even, and large and negative when i is
large and odd.
Thus there are arbitrarily large positive X such that f (X ) is large and negative,
which contradicts f (x) → ∞ as x → ∞. A similar contradiction arises if C x < 0.
Thus C x = 0, so f i (x) = A x + Bx i for all i. Substituting i = 0, we have f i (x) =
x + Bx i for all i.
If f (x) = x for all x, then we are done. Suppose there is some x with f (x) 6 = x.
Now Bx 6= 0. Since f is increasing, for any integer n and real number y we have
x + Bx n ≤ y if and only if x + Bx (n + 1) ≤ f (y). Iterating yields x + Bx n ≤ y if
and only if x + Bx (n + k) ≤ f k (y) = y + B y k. Varying k thus does not affect whether
x + Bx (n + k) ≤ y + B y k. Sending k to ∞ and to −∞, we conclude that Bx = B y .
Thus f (x) = x + c for some c.
Second, suppose that f is decreasing. For a given x, examine f i (x) = A x + Bx i +
C x (−2)i . If Bx > 0, then for i large and negative, f i (x) and f i+1 (x) are both large
and negative. Thus there are arbitrarily large negative X such that f (X ) is large and
negative, which contradicts f (x) → ∞ as x → −∞. Again, a similar contradiction
arises if Bx < 0. Thus Bx = 0, and hence f i (x) = A x + C x (−2)i for all i. Note that
f i (x) → A x as i → −∞, so f (A x ) = A x . A decreasing function can have at most
one fixed point, so A x = A y for all y.
Since f (x) = A x − 2C x = 3A x − 2(A x + C x ) = 3A x − 2x, it follows that f (x) =
c − 2x for some c.
Also solved by M. Bataille (France), C. Burnette, R. Chapman (U. K.), P. P. Dályay (Hungary), C. Delorme
(France), N. Grivaux (France), E. A. Herman, M. Huibregtse, O. P. Lossers (Netherlands), J. Simons (U. K.),
R. Stong, T. Trif (Romania), E. I. Verriest, GCHQ Problem Solving Group (U. K.), and the proposer.

Angles at an Inside Point of a Triangle

11550 [2011, 85]. Proposed by Stefano Siboni, University of Trento, Trento, Italy. Let
G be a point inside triangle ABC. Let α, β, γ be the radian measures of angles BGC,
CGA, AGB, respectively. Let O, R, S be the triangle’s circumcenter, circumradius, and
area, respectively. Let |XY| be the distance from X to Y . Prove that

|GA| · |GB| · |GC|(|GA| sin α + |GB| sin β + |GC| sin γ ) = 2S(R 2 − |GO|2 ).

Solution by Michael Vowe, Therwil, Switzerland. Writing [ABC] for the area of tri-
angle ABC, we have 2[BGC] = |GB| · |GC| sin α, 2[CGA] = |GC| · |GA| sin β, and
2[AGB] = |GA| · |GB| sin γ . Let (g1 , g2 , g3 ) be the normalized barycentric coordi-
−
→
nates of G, and for points P and Q, let PQ denote the vector from P to Q. Then
[BGC] [CGA] [AGB]
g1 = , g2 = , g3 = , g1 + g2 + g3 = 1,
[ABC] [ABC] [ABC]
and
−
→ −→ −→
g1 GA + g2 GB + g3 GC = 0.

The desired equality is equivalent to

R 2 = |GO|2 + g1 |GA|2 + g2 |GB|2 + g3 |GC|2 .

614 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 119

This content downloaded from 128.206.9.138 on Sun, 31 Mar 2013 15:35:08 PM

All use subject to JSTOR Terms and Conditions
Finally, we obtain
R 2 = g1 R 2 + g2 R 2 + g3 R 2 = g1 |OA|2 + g2 |OB|2 + g3 |OC|2
−→ − → −→ − → −→ −→ −→ −→
= g1 (OG + GA) · (OG + GA) + g2 (OG + GB) · (OG + GB)
−→ −→ −→ −→
+ g3 (OG + GC) · (OG + GC)
−→ −
→ −→ −→
= |OG|2 + g1 |GA|2 + g2 |GB|2 + g3 |GC|2 + 2OG · (g1 GA + g2 GB + g3 GC)
= |OG|2 + g1 |GA|2 + g2 |GB|2 + g3 |GC|2 .

Also solved by M. Alexander & T. Smotzer, M. Bataille (France), R. Chapman (U. K.), P. P. Dályay (Hungary),
P. De (India), A. Ercan (Turkey), O. Geupel (Germany), M. Goldenberg & M. Kaplan, S. Hitotumatu (Japan),
O. Kouba (Syria), J. H. Lindsey II, C. R. Pranesachar (India), J. Schlosberg, R. Stong, M. Tetiva (Romania), Z.
Vörös (Hungary), J. B. Zacharias & K. T. Greeson, Barclays Capital Problems Solving Group, GCHQ Problem
Solving Group (U. K.), and the proposer.

Points in Figures
11551 [2011, 178]. Proposed by Gregory Galperin, Eastern Illinois University,
Charleston, IL, and Yury Ionin, Central Michigan University, Mount Pleasant, MI.
Given a finite set S of closed bounded convex sets in Rn having positive volume, prove
that there exists a finite set X of points in Rn such that each A ∈ S contains at least
one element of X and any A, B ∈ S with the same volume contain the same number
of elements of X .
s
Solution by Jim Simons, Cheltenham, U. K. Let the sets of S be {Ai }i=1 , and consider
s
the 2 − 1 sets created by taking the intersection, as i ranges from 1 to s, of either Ai
or its complement Aic , but excluding Ac1 ∩ Ac2 ∩ · · · ∩ Acs . These sets are not necessar-
ily convex or even connected (indeed, they may have infinitely many components of
positive volume), but they are measurable. Discard any that are empty or have zero
measure, leaving {B j }tj=1 . Each Ai is a finite disjoint union of some of the B j , together
with a set of measure zero.
Consider giving real “weights” P w j to the B j and adding these weights to define
weights vi for the Ai , so that vi = w j , where the sum is over all j with B j ⊆ Ai .
Let Vi be the volume (measure) of Ai . The requirement that Vi = Vi 0 imply vi = vi 0
takes the form of a set L of linear equations in the w j with rational coefficients (indeed,
with coefficients −1, 0, 1 only).
Now (since the rank may be computed by evaluating certain determinants) a matrix
with rational entries has the same rank over the reals as it has over the rationals. So
the rational solution space for L has the same dimension as the real solution space,
and therefore is dense in it. There is a positive real solution (namely, each w j is the
measure of B j ). So there is a nearby rational solution (we only need it near enough that
all w j are positive). Then we can multiply by a common denominator to get a solution
in positive integers. For the set X , choose w j points in each set B j .
Note that this solution provides more information than requested. We can insure
that Vi < Vi 0 implies vi < vi 0 , Vi = Vi 0 implies vi = vi 0 , and Vi > Vi 0 implies vi > vi 0 .
Also solved by O. P. Lossers (Netherlands), R. Stong, and the proposers.

August–September 2012] PROBLEMS AND SOLUTIONS 615

This content downloaded from 128.206.9.138 on Sun, 31 Mar 2013 15:35:08 PM

All use subject to JSTOR Terms and Conditions