449

THE ACADEMY CORNER

No. 37
Bruce Shawyer
All communications about this column should be sent to Bruce
Shawyer, Department of Mathematics and Statistics, Memorial University
of Newfoundland, St. John's, Newfoundland, Canada. A1C 5S7

In this issue, we have the Memorial University Undergraduate Mathematics

Competition, written in September 2000.
1. (a) Prove that the sum of a positive real number and its reciprocal is
greater than or equal to two.
(b) Let a be a positive real number and let x, y and z be real numbers
such that x + y + z = 0.
Prove that log (1 + ax ) + log (1 + ay ) + log (1 + az )  3.
2 2 2

2. Given any ve points inside an equilateral triangle with side length 2,

prove that you can always nd two points whose distance apart is less
than 1 unit.
3. A conical drinking glass is 12 cm deep and has a top diameter of 10 cm.
The glass contains some liquid. A sphere of diameter 8 cm is placed in
the glass, and it turns out that the liquid exactly lls the volume below
the points where the sphere touches the cone. Find the depth of the
lowest point of the sphere beneath the highest point of the liquid.
4. Triangle PQR is isosceles, with PQ = PR = 3 and QR = 2 as shown
below. The tangent to the circumcircle at Q meets (the extension of)
PR at X as shown. Find the length RX .
Q
3 2
X
P 3 R

5. Sketch y ; x = 1, being careful to clearly classify all extrema, in ec-

3 3

tion points and straight line asymptotes in your analysis of the curve.
6. Let a , a , : : : , a be natural numbers such that 0 < a < a <



< a  125. Prove that at least one of the di erences di = ai ; ai,
1 2 44 1 2

i = 1, : : : , 43, occurs at least ten times.

44 +1

Send me your nice solutions!

450

Next, we present the 2000 Atlantic Provinces Council on the Sciences

Mathematics Competition, written at Dalhousie University, Halifax, Nova
Scotia, on 20 October 2000.

2000 APICS Math Competition

Time: 3 hours.

1. N people labelled 1, : : : , N visit a mathematician's home to play the

following game. They are given coins labelled 1, : : : , N , respectively,
where coin i lands \tails", when tossed, with probability pi, pi > 0.
The players throw their coin (in order 1, : : : , N ) and the winner is
simply the rst player to get \tails". Now the mathematician has, under
the assumption of independence, chosen the pi 's so that the game is
\fair", in that every player has the same chance of winning (that is N , 1

of course).
(a) If p = 0:10 and N = 5, determine values of p , p , p , and
p such that all ve players have the same chance of winning the
1 2 3 4

5
game.
(b) Show that it is possible to have a fair game if and only if p  N . 1
1

2. Let S = f1, 2, 3, : : : , 3ng. We de ne a sum{3{partition of S to be a

collection of n disjoint 3{subsets of S , Ai = fai , bi , ci g, i = 1, : : : , n,
such that the union A [ A [


[ An is S , and within each triple Ai ,
some element is the sum of the other two. For example: ff1, 5, 6g,
1 2

f2, 9, 11g, f3, 7, 10g, f4, 8, 12gg is a sum{3{partition of f1, 2, 3, : : : ,

12g.
(a) Find a sum{3{partition for f1, 2, 3, : : : , 15g.
(b) Prove that there exists no sum{3{partition for n = 1999.
3. Two polite but vindictive children play a game as follows. They start
with a bowl containing N candies, the number known to both con-
testants. In turn, each child takes (if possible) one or more candies,
subject to the rule that no child may take, on any one turn, more than
half of what is left. The winner is not the child who gets most candy,
but the last child who gets to take some.
Thus, if there are 3 candies, the rst player may take only one, as two
would be more than half. The second player may take one of the re-
maining candies; and the rst player cannot move and loses.
 451

(a) Show that if the game begins with 2000 candies the rst player
wins.
(b) Show that if the game begins with 999


999 (2000 9's) candies,
the rst player wins.
4. Show that, if a, b, c, d, e, f are integers with absolute value less than
or equal to 7, and the parabolae
y = x + ax + b
2

y = x + cx + d
2

y = x + ex + f
2

enclose a region R of the plane, then the area of R is

(a) rational,
(b) with denominator less than 2000.
5. The three-term geometric progression (2; 10; 50) is such that
(2 + 10 + 50)  (2 ; 10 + 50) = 2 + 10 + 50 .
2 2 2

(a) Generalize this (with proof) to other three-term geometric progres-

sions.
(b) Generalize (with proof) to geometric progressions of length n.
6. Solve for all real x, y > 0: 2xy ln(x=y ) < x ; y .
2 2

7. Without calculator or elaborate computation, show that

3 2701
 3 (mod 2701) .
NOTE: 2701 = 37  73 .
8. An isosceles triangle has vertex A and base BC . Through a point D
on AB , we draw a perpendicular to meet BC extended at E , such that
AD = CE. If DE meets AC at F , show that the area of triangle ADF
is twice that of triangle CFE .
Send me your nice solutions.

We also welcome contributions to this corner.

452

THE OLYMPIAD CORNER

No. 210
R.E. Woodrow
All communications about this column should be sent to Professor R.E.
Woodrow, Department of Mathematics and Statistics, University of Calgary,
Calgary, Alberta, Canada. T2N 1N4.
As a rst set of problems, we give the questions of the 1996-1997
Estonian Mathematical Olympiads, Final Round of the National Olympiad.
Thanks go to Richard Nowakowski, Canadian Team Leader, for collecting
them at the IMO in Argentina.
ESTONIAN MATHEMATICAL OLYMPIADS
1996-1997
Final Round of the National Olympiad
1. On a square table of size 3n  3n each unit square is coloured either
red or blue. Each red square not lying on the edge of the table has exactly
ve blue squares among its eight neighbours. Each blue square not lying on
the edge of the table has exactly four red squares among its eight neighbours.
Find the number of red and blue squares on the table.
2. Find
(a) all quadruples of positive integers (a; k; l; m) for which the equality
ak = al + am holds;
(b) all 5{tuples of positive integers (a; k; l; m; n) for which the equality
ak = al + am + an holds.
3. Prove that, for any real numbers x and y, the following inequality
holds: p p
x + y + 1 > x y + 1+ y x +1.
2 2 2 2

4. In a triangle ABC the values of tan \A, tan \B and tan \C relate
to each other as 1 : 2 : 3. Find the ratio of the lengths of the sides AC
and AB .
5. There are n points (n  3) in the plane, no three of which are
collinear. Is it always possible to draw a circle through three of these points
so that it has no other given points
(a) in its interior? (b) in its interior nor on the circle?
 
6. For positive integers m, n denote T (m; n) = gcd m; nm;n .gcd( )
 453

(a) Prove that there exist in nitely many pairs of integers (m; n) such that
T (m; n) > 1 and T (n; m) > 1.
(b) Does there exist a pair of integers (m;n) such that
T (m; n) = T (n;m) > 1?
7. A function f satis es the condition:
f (1) + f (2) +


+ f (n) = n f (n)
2

for any positive integer n. Given that f (1) = 999, nd f (1997).

Next we give the \exercises" of the Composition de Mathematiques

(Classe terminale S) from France. My thanks go to Richard Nowakowski,
Canadian Team Leader to the IMO in Argentina, for obtaining them for our
use.

COMPOSITION DE MATHEMATIQUES 1997
Classe Terminale S
Duree : 5 heures
1. On a place un jeton sur chaque sommet d'un polygone regulier a
1997 c^otes. Sur chacun de ces jetons est inscrit un entier relatif, la somme
de ces entiers relatifs e tant e gale a 1. On choisit un sommet de depart et on
parcourt le polygone dans le sens trigonometrique en ramassant les jetons au
fur et a mesure tant que la somme des entiers inscrits sur les jetons ramasses
est strictement positive.
Peut-on choisir le sommet de depart de facon a ramasser tous les
jetons? Si oui, combien y a-t-il de choix possibles?
2. Une capsule spatiale a la forme du solide de revolution delimite par
une sphere de centre O, de rayon R, et un c^one de sommet O qui rencontre
cette sphere selon un cercle de rayon r.
Quel est le volume maximal d'un cylindre droit contenu dans cette
capsule, le cylindre et la capsule ayant le m^eme axe de revolution?
3. C est un cube d'ar^ete 1 et p est la projection orthogonale sur un
plan. Quelle est la valeur maximale de l'aire de p(C )?
4. Etant donne un triangle ABC , on note a, b, c les longueurs de ses
c^otes et m, n, p les longueurs de ses medianes. Pour tout reel strictement
positif, on de nit le reel ( ) par la relation:
a + b + c = (( )) (m + n + p ) .
(1 ) Calculer (2).
454

(2 ) Calculer la limite de ( ) lorsque tend vers 0.

(3 ) A quelle condition portant sur a, b, c le reel ( ) est-il independant
de ?
5. Dans le plan, soient A et B deux points distincts. Pour tout point
C exterieur a la droite (AB), on note G l'isobarycentre du triangle ABC et
I le centre de son cercle inscrit.
(1 ) Soit un reel tel que 0 < <  . Quel est l'ensemble ; des points
C tels que ;! CA; ;! CB = + 2k, k e tant un entier? Lorsque C decrit ;,
montrer que G et I decrivent deux arcs de cercle que l'on precisera.
(2 ) On suppose desormais =3 < <  . Comment doit-on choisir C dans
; pour que la distance GI soit minimale?
(3 ) On note f ( ) la distance minimale GI de la question precedente. Ex-
pliciter f ( ) en fonction de a = AB et . Determiner la valeur maximale
de f ( ) lorsque decrit  ;  . [Ed. A more familiar notation is (  ;  ).]
3 3

Next we turn to readers' solutions to problems of the Turkish Team

Selection Examination for the 37th IMO [1999 : 73].
1. Let Qn (1 + nx n) = 1 + a xk1 + a xk2 +


+ amxkm where
1996 3

a , a , : : : , am are non-zero and k < k <


< km . Find a .

=1 1 2

1 2 1 2 1996

Solution by Mohammed Aassila, Strasbourg, France.

In general, if k = 2 1 + 2 2 +


+ 2 p then ak = + : : : + p . In
particular a = 10 + 9 + 8 + 7 + 6 + 3 + 2 = 45.
1

1996

2. In a parallelogram ABCD with m(A^) < 90, the circle with diam-
eter [AC ] intersects the lines CB and CD at E and F besides C , and the
tangent to this circle at A intersects the line BD at P . Show that the points
P , F , E are collinear.
Solution by Toshio Seimiya, Kawasaki, Japan.
C

D
B
F
E

Y A X P
 455

Since AC is a diameter we get \AEC = \AFC = 90 . Since AP is

tangent to the circle with diameter AC we have AP ? AC .
Let X and Y be the intersections of AP with BC and CD, respectively.
Since \XAC = 90 and AE ? XC , we have AX = XE
XC and 2

AC = EC
XC , so
2

AX : AC = XE : EC .
2 2
(1)
Similarly, we have
AY : AC = Y F : FC .
2 2
(2)
It follows that, from (1) and (2),
XE
CF = AX
AC = AX . 2 2 2
(3)
EC FY AC AY AY 2 2 2

Since AB k CY and AD k XC , we get

AX = XB and AX = CD .
AY BC AY DY
Hence, we have
AX = XB
CD .
2
(4)
AY BC DY
2

By Menelaus' Theorem for triangle CXY , we have

Y P
XB
CD = 1 . (5)
PX BC DY
Thus, we obtain from (3), (4) and (5),
Y P
XE
CF = 1 .
PX EC FY
Therefore, P , F , E are collinear, by the converse of Menelaus' Theorem.
3. Given real numbers 0 = x < x <


< x n < x n = 1 with
xi ; xi  h for 1  i  2n, show that
1 2 2 2 +1

+1

1;h < X n
x ( x ; x )  1+h.
2 i i 2 i ; 2 +1
2 2 1
i =1

Solutionby Mohammed Aassila, CRM, Universite de Montreal, Montreal,

Quebec.
The problem is equivalent to showing that
X
n
x i(x i ; x i; ) ; 21 < h2 .
2 2 +1 2 1
i=1
456

y=x
1

x2i

x2i;1 x2i+1 x2n+1

Pn
Now x i(x i ; x i; ) ; is the di erence between the area
1

i
2 2 +1 2 1 2

of the rectangles formed by the four lines x = x i; , x = x i , y = 0

=1

and y = x i and the triangle formed by the three lines x = 0, y = 1,

2 1 2 +1

x = y. The area contained in the rectangles but not in the triangle (respec-
2

tively contained in the triangle but not in the rectangles) is a union of trian-
gles of total base less than 1 and height  h. Hence, we have the required
inequality.
4. In a convex quadrilateral ABCD, Area(ABC ) = Area(ADC ) and
[AC ] \ [BD] = fE g. The parallels from E to the line segments [AD], [DC ],
[CB], [BA] intersect [AB ], [BC ], [CD], [DA] at the points K , L, M , N ,
respectively. Compute the ratio
Area(KLMN ) .
Area(ABCD)
Solution by Toshio Seimiya, Kawasaki, Japan.
We denote the area of polygon A A


An by [A A


An ]. Let B 0 ,
0
D be the feet of perpendiculars from B, D to AC , respectively. (See gure
1 2 1 2

below.)
Since [ABC ] = [ADC ], we get BB 0 = DD0 , so that
BE : ED = BB0 : DD0 = 1 : 1 .
Thus, we have BE = ED. Since EK k DA, we have
BK : KA = BE : ED = 1 : 1 .
Hence, BK = KA. Similarly, we have
BL = LC , CM = MD and DN = NA .
In triangle ABD, note that K , E and N are the mid-points of AB , BD and
DA, respectively. Thus,
[ENK ] = [ABD] .1
4
 457

N M
=

A B0 C
E D0

K
=

Similarly, we have
[EKL] = [ABC ] , [ELM ] = [BCD] and [EMN ] = [CDA] .
1
4
1
4
1
4

Hence,
[KLMN ] = [ENK ] + [EKL] + [ELM ] + [EMN ]
= [ABD] + [ABC ] + [BCD] + [CDA]
1
4
1
4
1
4
1
4

= ([ABD] + [BCD]) + ([ABC ] + [CDA])

1
4
1
4

= [ABCD] + [ABCD]
1
4
1
4

= [ABCD] .
1
2

Therefore, we obtain
[KLMN ] = 1 .
[ABCD] 2
6. For which ordered pairs of positive real numbers (a; b) is the limit
of every sequence (xn ) satisfying the condition
!1(axn+1 ; bxn) = 0
nlim (1)
zero?
Solutions by Mohammed Aassila, CRM, Universite de Montreal,
Montreal, Quebec; by Michel Bataille, Rouen, France; and by Pierre
Bornsztein, Courdimanche, France. We give the solution by Aassila.
458

If b > a, then, for xn = (b=a)n, we have axn ; bxn = 0, but

limn!1(b=a)n = 1.
+1

If b = a, we have the well-known example:

xn = 1 + 12 +


+ n1 ,
for which we have limn!1(xn ; xn ) = 0, but fxn g does not converge to
+1
a nite limit.
Let us assume that b < a. We shall prove that
limxn = limxn = 0 .
Denote by m (resp. M ), the lim (resp. lim) of fxn g. By (1), we have
M  ab m, and since m  M , we deduce that M  ab M , and conse-
quently that m, M  0. Similarly, by (1), we have that M ab  m, and
since M  m, we deduce that m  ab m, and consequently that m, M  0;
whence, m = M = 0 and limn!1 xn = 0.

Next we turn to solutions to problems of the Australian Mathematical

Olympiad 1996 [1999 : 74{75].
1. Let ABCDE be a convex pentagon such that BC = CD = DE
and each diagonal of the pentagon is parallel to one of its sides. Prove that
all the angles in the pentagon are equal, and that all sides are equal.
Solution by Toshio Seimiya, Kawasaki, Japan.
A

B S R E

T Q
= =
P

C D
=

As shown in the gure, we label the intersections of diagonals.

Since BE k CD and AC k ED, SCDE is a parallelogram, so that
CS = DE = CB.
 459

Hence, \CBE = \CBS = \CSB = \DEB .

Since \CBE = \DEB and BC = ED, it follows that BCDE is an
isosceles quadrilateral, so that B , C , D, E are concyclic. Since AB k CE ,
AC k DE, we have
\BAC = \ACE = \CED = \CBD = \BDC .
Thus, A, B , C , D are concyclic, giving that A, B , C , D, E are concyclic. Since
BC k AD, we get AB = CD, and since AC k ED, we have AE = CD.
Therefore, AB = BC = CD = DE = EA.
Consequently, corresponding minor arcs AB , BC , CD, DE and EA
are equal, and also corresponding inscribed angles are equal. We put
\BAC = . Then we have
\EAB = \ABC = \BCD = \CDE = \DEA = 3 .
2. Let p(x) be a cubic polynomial with roots r , r , r . Suppose that
1 2 3

p ;
+ p ;;
= 1000 .
1
2
1
2
p(0)
Find the value of r1 r2 + r2 r3 + r3 r1 .
1 1 1

Solutions by Mohammed Aassila, CRM, Universite de Montreal,

Montreal, Quebec; by Michel Bataille, Rouen, France; by Andrew Blinn,
Western Canada High School, Calgary, Alberta; by Pierre Bornsztein, Cour-
dimanche, France; and by Edward T.H. Wang, Wilfrid Laurier University,
Waterloo, Ontario. All the solutions were essentially the same. We give
Bataille's exposition.
Note that p(0) is supposed non-zero, so that r , r , r are non-zero.
Let p(x) = ax + bx + cx + d.
1 2 3
3 2

;
;

The hypothesis is: a + b + c + d + ; a + b ; c + d = 1000d ;
that is, b = 1996d. Now,
8 4 2 8 4 2

1 1 1 r + r + r = ;b=a = b = 1996 .
r r +r r +r r = r rr
3 1 2

1 2 2 3 3 1;d=a d 1 2 3

3. A number of tubes are bundled together into a hexagonal form:

f f f f
f f f f f
f f f f f f
f f f f f f f
f f f f f f
f f f f f
f f f f
460

The number of tubes in the bundle can be 1, 7, 19, 37 (as shown), 61, 91; : : : .
If this sequence is continued, it will be noticed that the total number of tubes
is often a number ending in 69. What is the 69th number in the sequence
which ends in 69?
Solution by Edward T.H. Wang, Wilfrid Laurier University, Waterloo,
Ontario.
The number is 1417969. Note rst that the sequence fan g is given by
the formula
an = 1 + 6(1 + 2 + 3 +


+ (n ; 1)) = 1 + 3n(n ; 1)
where n  1. Clearly, an ends in 69 if and only if 100 j an ; 69, and
an  69 ; that is,
100 j 3n(n ; 1) ; 68 (1)
where n  6.
In particular, 5 j 3n(n ; 1) ; 68 and so 3n(n ; 1)  68  3 (mod 5).
Since (3; 5) = 1, we have n(n ; 1)  1 (mod 5), which holds if and only if
n  3 (mod 5). Hence, n = 5k + 3 for some integer k  1. Then
n(n ; 1) = (5k + 3)(5k + 2) = 25k + 25k + 6 2

and (1) becomes 100 j 75k + 75k ; 50, or 4 j 3k + 3k ; 2.

2 2

Thus, we have 3k(k + 1)  2  6 (mod 4). Since (3; 4) = 1, we have

k(k + 1)  2 (mod 4), which holds if and only if k  1, 2 (mod 4). Thus,
k = 4t + 1 or 4t + 2, and n = 20t + 8 or 20t + 13 for some non-negative
integer t.
Conversely, if n = 20t +8, then 3n(n ; 1) ; 68 = 1200t +900t +100, 2

and if n = 20t + 13, then 3n(n ; 1) ; 68 = 1200t + 1500t + 400. In

2

both cases, (1) holds. Therefore, we conclude that an ends in 69 if and only
if n = 20t + 8 or 20t + 13 for t = 0, 1, 2, : : : . To nd the 69th such
number, we put t = 34 into n = 20t + 8 to obtain n = 688 and
a = 1 + 3  688  687 = 1417969.
688

4. For which positive integers n can we rearrange the sequence 1, 2,

: : : , n to a , a , : : : , an in such a way that jak ; kj = ja ; 1j =
6 0 for k = 2,
3, : : : , n?
1 2 1

Solutions by Pierre Bornsztein, Courdimanche, France; and by Edward

T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario. We use Wang's
solution.
The required permutations exist if and only if n is even. First of all,
since a ; 1 6= 0, we have a > 1 and so ja ; 1j = a ; 1. Partition
S = f1, 2, : : : , ng has S = A [ B where A = fi 2 S j ai ; i  0g
1 1 1 1

and B = fj 2 S j aj ; j < 0g. Since ai ; i = a ; 1 for all i 2 A and

1
 461

aj ; j = 1 ; a for all j 2 B, we have

1

X X
(ai;i)+ (aj ;j ) = jAj(a ;1)+jB j(1;a ) = (jAj;jB j)(a ;1) .
1 1 1
i2A j 2B
Since X X X X
(ai ; i) + (aj ; j ) = ak ; k = 0,
i2A j 2B k2S k2S
we conclude that
(jAj ; jBj)(a ; 1) = 0 , 1

and so, that jAj = jB j. Hence, n = jS j = jAj + jB j = 2jAj, showing that n

must be even.
Conversely, if n = 2k is even, then the permutation  below clearly
has the described property:
 1 2 3 4


2k ; 1 2k 
 = 2 1 4 3


2k 2k ; 1 ;
that is,  = (1; 2)(3; 4)


(2k ; 1; 2k) is the product of k disjoint transpo-
sitions.
Remark. Permutations with the described property are not unique in
general; for example, when n = 6, we could take  = (14)(25)(36).
5. Let a , a , : : : , an be real numbers and s, a non-negative real
1 2
number, such that
(i) a  a 


 an ;
1 2 (ii) a + a +


+ an = 0;
1 2

(iii) ja j + ja j +


+ jan j = s.
1 2

Prove that
an ; a  2ns .
1

Solution by Michel Bataille, Rouen, France.

The result is clear when s = 0, so we will suppose s > 0. This implies
that at least one of the ai 's is non-zero. Since a + a +


+ an = 0, the
ai's cannot all be non-negative, or all be non-positive. Thus:
1 2

a = min(ai) < 0 and an = max(ai) > 0 .

1

There exists k 2 f1; 2; : : : ; n ; 1g such that a  a 


 ak  0 <

ak 


 an . Then ak +


+ an = ;(a + a +


+ ak) =
1 2

ja j + ja j +


+ jak j = s ; s(jak j +


+ janj) = s ; (ak +


+ an).

+1 +1 1 2

Hence, ak +


+ an = = ;(a + a +


+ ak ). For i 2 f1, 2,

1 2 +1 +1

: : : , kg, j 2 fk + 1, : : : , ng, we have: aj ; ai  an ; a (= , say) so

+1 2 1 2

that an ; a   , an ; a   , : : : , an ; ak   . Adding up, we get:

1

1 2
462

kan + s  k. Substituting successively an; , : : : , ak for an, we get

1 +1
similarly:
2

ka + s  k , : : : ,ka + s  k .
n;1 2 k+1 2
Adding up again, we obtain: k(an +


ak ) + (n ; k) s  (n ; k)k or
k s +(n ; k) s  (n ; k)k, which leads to   k ns
n;k . But (n ; k)+ k = n;
+1 2

2 2 2 ( )
2
hence, k(n ; k)  n and k ns
4 2 (
s
n;k  n . Thus,   n .
)
s2 2

6. Let ABCD be a cyclic quadrilateral and let P and Q be points on

the sides AB and AD, respectively, such that AP = CD and AQ = BC .
Let M be the point of intersection of AC and PQ. Show that M is the
mid-point of PQ.
Solution by Toshio Seimiya, Kawasaki, Japan.
D

o C
M
A 
o
o
P 

T B

Let T be a point on AD produced beyond A such that

AT = BC .
Since AT = BC , AP = CD and \TAP = \TAB = \BCD, we get
4ATP  4CBD, so that
\ATP = \CBD .
Since \CBD = \CAD, we have
\ATP = \CAD .
Thus, TP k AC ; that is, TP k AM .
Hence, we get PM : MQ = TA : AQ = BC : AQ = 1 : 1.
Therefore, PM = MQ.
 463

7. For each positive integer n, let (n) denote the sum of all positive
integers that divide n. Let k be a positive integer and n < n <


be an
in nite sequence of positive integers with the property that  (ni) ; ni = k
1 2

for i = 1, 2, : : : . Prove that ni is a prime for i = 1, 2, : : : .

Solutions by Pierre Bornsztein, Courdimanche, France; and by Edward
T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario. We give
Bornsztein's solution.
It is well known that
n is a prime if and only if (n) ; n = 1 .
Let k be a positive integer.
We have  (ni) ; ni = k, for i = 1; 2; : : : . Suppose that none of
the ni are p prime. Then we have ni = piqi with 2  qi  pi < ni and
ni > pi  ni.
We deduce that k =  (ni) ; ni  1+ pni for i = 1; 2; : : : . pBut (ni) is
an in nite increasing sequence of positive integers, so limi!1 1+ ni = 1,
a contradiction to k  1 + pni.
Thus, there is i with ni0 a prime. Then, k =  (ni0 ) ; ni0 = 1, and,
for each i,  (ni) ; ni = 1, making ni a prime.
0

Remark by Edward T.H. Wang, Wilfrid Laurier University, Waterloo,

Ontario.
It is possible to have  (m) ; m =  (n) ; n for m 6= n. For example
(26) ; 26 = 16 = (12) ; 12. However, it is by no means clear whether
it is true that, given any k 2 N, there are k distinct natural numbers n , n ,
: : : , nk such that (ni) ; ni = (nj ) ; nj for all i, j = 1, 2, : : : , k.
1 2

8. Let f be a function that is de ned for all integers and takes only the
values 0 and 1. Suppose f has the following properties:
(i) f (n + 1996) = f (n) for all integers n;
(ii) f (1) + f (2) +


+ f (1996) = 45.
Prove that there exists an integer t such that f (n + t) = 0 for all n for which
f (n) = 1 holds.
Solutions by Michel Bataille, Rouen, France; and by Pierre Bornsztein,
Courdimanche, France. We give Bornsztein's write-up.
Let C be a circle with radius  . The real line can be wrapped around
1996

the circle in the counterclockwise sense so that the integer i is mapped to the
2

point Ai of the circle.

464
2
3 1
1996

Since the circumference of the circle is 1996, if two numbers di er by

1996, they map to the same point of the circle, (that is, Ai = A for
i 2 Z.)
+1996 0

From (i), the integers are represented on C by the points A , : : : ,

A , and we may colour those Ai such that f (i) = 1. By (ii), and since
1

f (n) 2 f0, 1g, there are exactly 45 coloured points.

1996

If i < j and i, j 2 f1, : : : , 1996g, Ai and Aj determine two arcs of

lengths j ; i and i ; j +1996 (which may be equal). The possible lengths are
1, 2, 3, : : : , 1995. But there
;
are 45 coloured points, and distinct coloured
points determine at most 2 = 1980 di erent lengths.
45
2

Then, for all i, there is t 2 f1, : : : , 1995g such that if Ai is coloured

then Ai t is not coloured (with subscripts read modulo 1996).
+

That is, for all i, if f (i) = 1, then f (i + t) = 0.

That completes the Olympiad Corner for this issue. Send me your
Olympiad contest materials and your nice solutions!

Quickie
In rectangle ABCD with side a, arcs of circles of radii a are centred at A and
B, and meet at E outside the rectangle. Find the circumradius of triangle
CDE. (Thanks to Ed Barbeau for the idea!)
Solution
Thus, the circumradius is a.
Then EF = BC = FC = FD = a, giving that F is the circumcentre.
Draw 4CDF congruent to 4BAE (BE kCF , AE kDF ).
 465

BOOK REVIEWS
ALAN LAW
Twenty Years Before the Blackboard,
by Michael Steuben with Diane Sandford,
published by the Mathematical Association of America, 1998,
ISBN 0-88385-525-9, softcover, 174+ pages, $23.50 (U.S.)
Reviewed by Nicholas Buck, College of New Caledonia, Prince George, BC.
What makes a good mathematics teacher? The answer is not clear to me, but
most students and teachers could make reasonable suggestions about what they think
makes a good mathematics teacher. In my fourteen years teaching, many ideas |
from a wide variety of sources | about how to improve (my) teaching have oated
across my desk. The amusing thing is that (not surprisingly) these suggestions are not
always consistent with each other. This just illustrates what I believe to be the case,
and that is that there is no single model of a good mathematics (or any other kind of)
teacher and we (or at least certain education theorists) should stop trying to construct
one. One student may enjoy and respond positively to one style of teaching, while
the next student may be completely turned o by the same style. One style may be
appropriate for certain subjects or levels of teaching and quite inappropriate for other
subjects and levels.
Take the question of how much history of mathematics to include in a math-
ematics course. I know mathematicians who think the history should be left out
completely; and still others who think that it is important to place the material in
historical context and connect it with other intellectual activities of the same era. I
am in the second group, and Steuben's book will appeal to anyone of like mind, but
especially those who have recently begun teaching at the high school level. That is not
to say that more experienced teachers at other levels could not bene t from and enjoy
reading (at least parts of) this book, but they probably already have their favourite
tried-and-true anecdotes, puns, etc., (but don't oblique references to Monty Python
get some strange looks these days?).
Steuben's book is a treasure trove of ideas, suggestions, problems, examples,
anecdotes, humour, and other miscellanea of a mathematical nature that he has suc-
cessfully incorporated into his teaching. Not many will agree with or use everything
in the book, but there is something here for anyone who thinks that good teaching
should be entertaining and thought provoking and not just a sequence of compart-
mentalized topics with intended learning outcomes, course objectives and skill sets.
We all know how so many modern students | products (victims?) of the video and
computer age | need to be entertained while they are learning (as Gary Trudeau has
said, \they insist on a certain comfort level."). This book is full of little ideas one
could try in order to catch the attention of these modern students. Of course, as one
presses ahead beyond high school and lower division post-secondary mathematics
courses, the lecturer necessarily focuses on the careful formulation of de nitions and
on proofs of theorems. This book becomes less useful at that level.
In his famous book How To Solve It, Georg Polya  comments amusingly on the
\traditional mathematics professor of popular legend." He describes the eccentric,
doddering, absent-minded professor we all have (or at least should have) experienced
466

(enjoyed?) somewhere along the way. This professor would surely not meet the
prototype of the good maths teacher promoted by many. But, as Polya  remarks,
\After all, you can learn something from this traditional mathematics professor. Let
us hope that the mathematics teacher from whom you cannot learn anything will not
become traditional."

Archimedes | What Did He Do Besides Cry Eureka,

by Sherman Stein,
published by the Mathematical Association of America, 1999, Classroom
Resource Materials Series.
ISBN 0-88385-718-9, softcover, 155+x pages, $24.95 (U.S.).
Reviewed by C.L. Kaller, Kelowna, BC.
Professor Stein states in the introductory remarks that his goal in writing the
book is \to make what I view as Archimedes' most mathematically signi cant discov-
eries accessible to the busy people of the mathematical community, whom I think of
as anyone who recognizes the equation of a parabola."
In pursuing his goal, Stein gives an admirable presentation, in very attractive
and readable form, of the accomplishments of Archimedes during the Golden Age of
Greece. Without going into any extensive peripheral details which could otherwise
distract the reader from the essential results of the original Archimedean investiga-
tions, the book presents the logical steps used by this great Greek mathematician
when presenting his results so long ago. Starting with some brief notes on the life
and times of Archimedes, the author outlines the geometric reasoning underlying the
theories of the lever, centres of gravity and oating bodies, as well as nding the sums
of series, areas of plane gures, volumes of solids and the very important estimation
of . The reader cannot help but be fascinated by the simple, yet powerful, ingenuity
exhibited by Archimedes in his research; even as we have today much more techni-
cally advanced tools to arrive at many of these same results based on calculus, what
must grip the admiration of the reader is, as Stein states, \how much Archimedes
accomplished with the limited tools at his disposal some 22 centuries ago."
This small book certainly meets the MAA criterion of one of the `Classroom
Resource Materials', and should be found a useful addition to any senior high school,
college and even university library. With the appendices and references which permit
more knowledgeable readers to pursue in greater depth the presentations in the book,
the author has achieved the goal he set for himself in making better known the great
contributions of Archimedes to our mathematical culture.
The only small criticisms this reviewer has are in the usual ambiguities of geo-
metric notations. Single capital letters, P , Q, R, : : : designate points; lower case
letters designate numerical quantities. This distinction is not always followed. And
when a line joining points P and Q is denoted by PQ, it would be more clear if over-
bars were used to indicate unambiguously the distance from P to Q; that is, PQ.
This would overcome the ambiguity when encountering PQ2 or PQ
A. But most of
us have learned to tolerate and work within such inconsistencies as we do geometric
work.
 467

Letter to the Editor

Regarding: Edward T. H. Wang's note \Some Bounds for (n) (n)"
[2000 : 280].
It may be of interest for the audience of Crux Mathematicorum with
Mathematical Mayhem to learn that some of the results of this note can be
found elsewhere, already. (This indeed does no harm to their neat presen-
tation within this note!)
 In [1], Theorem 329, the inequality 2 < xn < 1, where xn =  nn2 n
6 ( ) ( )

is given (n > 1).

n!1 xn = 2 and limsup xn = 1.
Furthermore, liminf 6

n!1
 By the way, the inequality (n) (n)  n ; 1 (n > 1) was again posed
2

as a problem in [2].
 Finally I would like to draw the attention to the recent referential source
[3] where in the rst few chapters there are collected very man inequal-
ities for almost all types of arithmetical functions. (Many of the results
stem from sources not easily accessible at all. This fact indeed increases
the value of this book.)
Let me quote two examples from x I.6. (Inequalities by J. Sandor):
With (n) = n

Q 1 + ; that is, Dedekind's arithmetical function,
1

pjn p
the following hold:
I (n) n < nn for n > 1, and
( )

I (n) n > nn if all prime factors of n are greater than or equal

( )

to 5.
References.
[1] G.H. Hardy and E.M. Wright, An Introductionto the Theory of Numbers,
Oxford, New York 1996, 1965.
[2] A. Makowski, Problem 3932, Mathesis 69 (1960), 65.
[3] D.S. Mitrinovic, J. Sandor and B. Crstici, Handbook of Number Theory,
Dordrecht, Boston, London.
Walther Janous
Ursulinengymnasium
A 6020 Innsbruck, Austria
[email protected]
468

A Spatial Problem Solved with

Stereographic Projection
Shay Gueron and Oran Lang

Stereographic projection is a transformation in space, used frequently

in complex analysis. It can also be interpreted as the spatial analog of inver-
sion. In this paper, we demonstrate how the elementary properties of the
stereographic projection can be used for solving a geometric problem.
The following problem was proposed to the 1999 International Mathematical
Olympiad (IMO) Jury:
Problem: A nite set F of n  3 points in space (the plane) is called
completely symmetric if it satis es the following condition: for every two
distinct points A, B from F , the perpendicular bisector plane (the perpen-
dicular bisector) of the segment AB is a plane (an axis) of symmetry for F .
Find all completely symmetric sets.
The IMO Jury looked for a relatively simple geometric question for the IMO
paper, and decided to use only the planar version of the problem. The
answer in the planar variant is not surprising: any completely symmetric set
consists of the vertices of a regular polygon. The straightforward generaliza-
tion to space would read: a completely symmetric set consists either of the
vertices of a regular polygon, or the vertices of a regular polyhedron. Surpris-
ingly however, this is not the correct answer for the 3D{version: a completely
symmetric nonplanar set consists of the vertices of a regular tetrahedron
or a regular octahedron. The other regular polyhedrons, namely the cube,
the regular dodecahedron and the regular icosahedron, are ruled out. This
counterintuitive result is not easy to see, particularly when looking at the
problem strictly as one of space geometry. We show here how stereographic
projection helps to reduce the problem to a planar one, and thus, makes it
easier to understand.
Stereographic Projection | De nitions and Properties
Let  be a sphere, S a point on  ,  the tangent plane at S , and NS
a diameter of  . A stereographic projection through N is the one-to-one
transformation from  n fN g to  in which every point A on  is mapped to
the intersecting point A0 of NA and  (see Figure 1).
Interpretation: Stereographic projection can be interpreted as the spatial
analog of inversion in the plane.
Copyright c 2000 Canadian Mathematical Society
 469
Nq


q
q
A
q
S
 A0 q
Figure 1.

Explanation: Let A be a point and A0 its stereographic projection. Since

\NAS = \NSA0 = 90 we have 4NAS  4NSA0 , and therefore
jNAj
jNA0 j = jNSj . Since NS is a diameter of , the product0 jNAj
jNA0 j
2

does not depend on the choice of A. As, by de nition, N; A; A are collinear,

the stereographic projection can be viewed as an inversion in space with
centre N and radius jNS j.
Two elementary properties of the stereographic projection are used below
(for proofs see [1]).

P1. For every two points A, B , and their projections A0 , B 0 , we have

jA0B0 j = jABj
jNAjNS j .

2

j
jNBj
P2. A circle on  , passing through N , is mapped under the stereographic
projection to a line in  . A circle on  , not passing through N , is
mapped under the stereographic projection to a circle in  .

Inversion in the Plane | De nitions and Properties

In this paper we use the following properties of inversion in the plane (for
proofs see, for example, [1]):
De nition: Let (O; r) be a circle in the plane  , with centre O and radius r.
Inversion of the plane, in (O;r) is the transformation I :  nfOg !  nfOg
that maps every point A 6= O in  to the point A0 = I (A) lying on the ray
OA, and satisfying jOAj
jOA0 j = r . The point O is called the centre of the
2

inversion.
470

Basic Properties of Inversion

I1. I (I (A)) = A.
I2. A line passing through O is invariant under the inversion.
I3. A circle passing through O is mapped to a line that does not pass
through O. A line not passing through O is mapped to a circle pass-
ing through O.
I4. A circle not passing through O is mapped to a circle not passing
through O.
I5. Let D be a disc whose boundary is the circle C , and suppose that C
passes through the inversion centre O. Denote the image of C under the
inversion by l = I (C ) (the line l does not pass through O). Then I (D)
is the half-plane, with respect to l, that does not contain O. Conversely,
if D is a half-plane with respect to a line l not passing through O,
then I (D) is the interior (disc) of the circle C = I (l).
Lemma (the stereographic projection of great circles on  ) [2]:Let c be a
great circle on  , not passing through N . Let ; be the cylinder touching
the sphere  along c. Draw a line through N , parallel to the axis of ;, and
let P 0 be its intersection with  . Then the stereographic projection of c is
the circle c0 lying in  , with the centre P 0 and the radius jNP 0 j.
Qr rN

Qq Nq
1 c

q r
A Pq A
q
q qS q
 A0 r r
0r P0
(a) (b) A
Figure 2.
Proof: Draw the plane  parallel to  through N , as in Figure 2(a). Clearly,
 is tangent to  at N . Let A be an arbitrary point on c. Draw the line
1

NAA0 and a line parallel to the generator of ; through A. Denote its

1

intersection with  by Q. Since NP 0 kAQ, the points N , P 0 , A, Q and

1
 471

A0 are coplanar, 0and0 therefore QN k0 A0P00 (see Figure 2(b)). It follows that
\QNA = \NA P , \NAQ = \A NP . Since QA and QN are two tan-
gents to  , they have equal lengths and therefore \QNA = \QAN , imply-
ing that \NA0 P 0 = \A0 NP 0 and thus, jA0 P 0 j = jNP 0 j.
Solution of the Problem by Stereographic Projection
We are now ready to tackle our problem of nding all completely sym-
metric sets in space, and will do this by using stereographic projection.
Suppose that F is a completely symmetric set. We rst show that all the
points in F lie on a sphere. Let O be the barycentre of F . Since for every
A; B 2 F , the re ection through the perpendicular bisector plane of AB,
A;B , maps F to itself, O is invariant under this re ection and therefore lies
in A;B . It follows that jOB j = jOAj; that is, the distance jOAj is the same
for all points A 2 F . Denoting jOAj = a, it follows that all the points of F
lie on a sphere  whose centre is O and whose radius is a. Every plane of
symmetry of F passes through O, and thus, cuts a great circle from  .
We now choose a point N in F . For each point P 2 F , P 6= N , the per-
pendicular bisector plane N;P is a plane of symmetry of F . Therefore, for
every other point A 2 F , there exists a point B symmetric to A with respect
to N;P . If A lies in N;P , then B = A. Further, for every pair of points
A; B 2 F , the plane A;B is a plane of symmetry of F , and thus, there ex-
ists a point P 2 F symmetric to N with respect of this plane (it is possible
that P = N ).
Now consider the stereographic projection from N that maps F to F 0
in  . Using the lemma, we see that the image of the great circle which is
cut from  by the plane of symmetry N;P , is a circle in  whose centre is P 0
and whose radius is NP 0 . Since AB kNP , the points A, B , N , P are copla-
nar, and therefore by Property P2 the projections A0 , B 0 , P 0 are collinear.
Furthermore, A0 and B 0 lie on the same side of P 0 , jPB j = jNAj and
jPAj = jNBj (see Figure 3). Using Property P1 and noting that jNSj = 2a,
we obtain

jP 0A0 j
jP 0 B0j = jPAj
jNA(2j a
)jNP j
jPBj
jNB(2j a
)jNP j

2 2

= j(2a) = jNP 0j .
4
2
NP j 2

Therefore, the points of  in the plane of symmetry N;P are mapped to a

circle cP in the plane, and for every point A0 in F 0 there exists a point B 0 in
F 0 such that the inversion with respect of cP takes A0 to B0.
The conclusion is that every plane of symmetry of F de nes a circle of in-
version of F 0 with centre in F 0. In other words, F 0 is invariant under the
inversion with respect to this circle.
472
Nq
 qN
c o
B Bq
q
q q Aq o q
A P
q q0 qP 0
P0 q
q 0S B0 q A
 B0 q A
(a) (b)
Figure 3.

Our problem is therefore reduced to nding all sets F 0 in the plane that
satisfy:
Property 1: For every two points A0 , B 0 2 F 0 , there exists a circle of in-
version of F 0 that maps A0 to B 0 . If this circle passes through S , it is the
perpendicular bisector of A0 B 0 , and considered as a circle with in nite radius.
In all other cases, the centre of the circle of inversion belongs to F 0.
Property 2: For every point A0 2 F 0, there exists a circle of inversion of
F 0 with centre A0 . Using the lemma, we see that the radius of this circle is
jNA0 j = p(2a) + jSA0 j , where a is a constant and S is a given point in
2 2

the plane (not necessarily belonging to F ).

Note that for every nite set F 0 in the plane satisfying Properties 1 and 2, we
can construct a completely symmetric set F in space by constructing a sphere
with radius a, tangent to the plane at S and performing a stereographic pro-
jection.
If the points of F 0 are collinear, the points of F lie on a circle, and this implies
that we are dealing with the planar case of the problem. It is now easy to
show that the points of F are the vertices of some regular polygon (for exam-
ple, by proving that F is invariant under appropriate rotation about the cen-
tre of the circle). We now assume that the points of F 0 are not collinear and
denote the vertices of their convex hull by A0 , A0 , : : : , A0n . Property 1 im-
plies that there exists a circle of inversion that maps A0 to A0 . By de nition,
1 2

no points of F 0 on the line A0 A0 lie outside the segment A0 A0 . Therefore,

1 2

the circle of inversion is the perpendicular bisector of A0 A0 passing through

1 2 1 2

S, and it is a line of symmetry of F 0. Similarly, for every 1  i  n, the circle

1 2

of inversion that maps Ai to A0i is the perpendicular bisector of the seg-

0
ment A0i A0i and passes through S . Recall now that the composition of two
+1

re ections through the lines l and l is a rotation centred at the intersecting

+1

point of l and l . It follows that F 0, as well as its convex hull, is invariant

1 2

under some rotation about S . Thus, A0 A0


A0n is a regular polygon with

1 2

centre S .
1 2
 473

We now denote the circle of inversion of F 0 centred at A0i by ci , the radius of

this circle by ri, and the inversion of P 0 with respect of this circle by Ii (P 0).
Clearly, ri  jA0i A0i j since otherwise Ii (A0i ) falls outside the segment
A0iA0i . The solution of our problem is now obtained by considering the
+1 +1

following three cases for n.

+1

A. The case n  5. A02 A01

r r
B0 r
I1 (A03 ) r

A03 r r A0n

r
A04
Figure 4.
Since A0 A0


A0n is a regular polygon, it is inscribed in a circle passing
through A0 . Therefore (by I3) I (A0 A0


A0n ) are n;1 collinear points, and
1 2

the line connecting them intersects the segments A0 A0 and A0 A0n . By Prop-
1 1 2 3

erty 1, there exists a circle of inversion that maps A0 to I (A0 ). The perpen-
1 2 1

dicular bisector of A0 I (A0 ) does not pass through S , and since A0 A0


A0n
4 1 3

is the convex hull of F 0, no point on the ray I (A0 )A0 lies beyond A0 . It
4 1 3 1 2

follows that the centre of this circle lies on the ray A0 I (A0 ) beyond I (A0 ).
1 3 4 4

We denote this point by B 0 ( by Property 1, B 0 2 F ). Consequently (by I5),

4 1 3 1 3

I (B0) lies outside the convex hull, which is a contradiction. Therefore, we

must have n < 5.
1

B. The case n = 4.
We rst prove that ri = jA0i A0i j. Assume that ri < jA0i A0i j (we showed
above that ri  jA0i A0i j). This implies that Ii (A0i ) lies on the seg-
+1 +1

ment A0i A0i . By Property 1, there exists a circle of inversion of F 0 that

+1 +1

maps Ii(A0i ) to A0i . Since there are no points outside the segment
+1

Ii(A0i )A0i , and their perpendicular bisector does not pass through S ,
+1 +2

we arrive at a contradiction, which implies that ri = jA0i A0i j.

+1 +2

+1
A02 q 0
rA1
q
B0
q
S

q r0
A03 A4
Figure 5.
474

Denote jA0i S j by r. Then jA0i S j
jA0i A0i j = 2r = jA0i A0i j , which im- 2 2

plies that S = Ii (A0i ). Hence S 2 F 0 . Suppose that F 0 contains a point B 0

+2 +1

di erent from A0 , A0 , A0 , A0 and S . With no loss of generality, we assume

+2

that B 0 is inside the 4A0 A0 S . It follows (by I5) that I (B 0) is outside the
1 2 3 4

line A0 A0 , and this contradicts the de nition of A0 A0 , : : : , A0n as the de n-

1 2 1

ing points of a convex hull. Consequently, the set F 0 consists of the four
2 3 1 2

vertices of the square A0 A0 A0 A0 and its centre S .

p p 1 2 3

Now, by Property 2, 2r = r + (2a) , which implies that r = 2a. Fi-

4

2 2

nally, simple calculations for the original set F , (for example, calculating
jNA j, jSA j, jA A j,


) show that the points in F are the vertices of a
1 1 1 2
regular octahedron.
C. The case n = 3.
As in the previous case, ri = jA0i A0i j. Assume that there is another point
B0 in F 0. Then (by I5) I (B0) lies outside the circle (and hence the0 triangle)
+1

A0 A0 A0 , and we arrive at a contradiction.0 We conclude that F consists

1

1 2 3
of the vertices of an equilateral triangle A A 0 A0 . We denote r = jSAi j,
p p
and then 3r = r + (2a) , which implies that r = 2a. By simple
2 2
p 1 2 3

calculations we can now conclude that F consists of the vertices of a regular

tetrahedron.
Acknowledgments
We thank an anonymous referee for a careful review and helpful suggestions.
References
[1] R.A. Johnson, Advanced Euclidean Geometry. Dover (1960).
[2] N.N. Yaglom, Geometric Transformations III, MAA series (1973).
Shay Gueron Oran Lang
Department of Mathematics Ramat Gan
University of Haifa Israel
Haifa, 31905
Israel
 475

THE SKOLIAD CORNER

No. 50
R.E. Woodrow
Last issue we gave some of the solutions to the problems of the Preliminary
Round of the British Columbia Colleges Senior High School Contest for 2000.
Here, thanks to Jim Totten, The University College of the Cariboo, one of the
organizers, are the rest of the \ocial" solutions.
6. While 10 pin bowling, Sam left 3 pins standing which formed the
vertices of an equilateral triangle. How many such equilateral triangles are
possible?
(a) 15 (b) 14 (c) 12 (d) 10 (e) none of these
Solution. The answer is (a). Number the
pins as shown in the diagram on the left.
1 There is then one large equilateral triangle
with 4 pins on a side, namely the one with
2 3
vertices numbered (1; 7; 10). There are also
three equilateral triangles with 3 pins on a
side, namely the ones whose vertices are
4 5 6 numbered (1; 4; 6), (2; 7; 9), and (3; 8; 10).
The equilateral triangles with 2 pins on a
side come in two distinct orientations, one
7 8 9 10 with a single vertex above the horizontal
base and one with a single vertex below the
horizontal base. For the rst type we have
six such: (1; 2; 3), (2; 4; 5), (3; 5; 6), (4; 7; 8), (5; 8; 9) and (6; 9; 10). For the
second type we have only three such: (2; 3; 5), (4; 5; 8), and (5; 6; 9). This
gives us a total of 13 equilateral triangles so far. However, there are two oth-
ers which are skewed somewhat to the edges of the outer triangle: (2; 6; 8)
and (3; 4; 9), which gives us a total of 15 equilateral triangles.
7. If I place a 6 cm 6 cm square on a triangle, I can cover up to 60%
of the triangle. If I place the triangle on the square, I can cover up to of 2

the square. What is the area, in cm , of the triangle?

3
2

(a) 224
5
(b) 24 (c) 36 (d) 40 (e) 60
Solution. The answer is (d). The critical idea here is to recognize that
when the square covers as much of the triangle as possible, the triangle will
also cover as much of the square as possible, and that at this point the amount
of triangle covered is the same as the amount of square covered. Let A be
the area of the triangle. Then 0:6A =
36, or A = 40 cm .
2
3
2
476

8. Two circles, each with radius 10 cm, are placed so that they are
tangent to each other and a straight line. A smaller circle is nestled between
them so that it is tangent to the larger circles and the line. What is the radius,
in centimetres, of the smaller circle?
p
(a) 10 (b) 2:5
p
(c) 2 (d) 1 (e) none of these
Solution. The answer is (b). Let A be the centre of one of the large
circles, let B be the point of contact between the two large circles and let C
be the centre of the small circle. Then AB ? BC , AC = 10 + r and
BC = 10 ; r. From the Theorem of Pythagoras, we have
(10 + r) = 10 + (10 ; r) ,
2 2 2

100 + 20r + r = 100 + 100 ; 20r + r ,

2 2

40r = 100 ,
r = 2:5 .

A 10
B
10
+
r r
C
9. Arrange the following in ascending order:
2 5555
, 3 3333
, 6 2222
.
(a) 2 5555
, 3 , 6
3333 2222
(b) 2 , 6 , 3 5555 2222 3333

(c) 6 2222
, 3 , 23333 5555
(d) 3 , 6 , 2 3333 2222 5555

(e) 3 3333
, 2 , 6
5555 2222

Solution. The answer is (e). See #10 of the Junior paper { solution on
[2000 : 347].
10. Given that 0 < x < y < 20, the number of integer solutions
(x; y ) to the equation 2x + 3y = 50 is:
(a) 25 (b) 16 (c) 8 (d) 5 (e) 3
Solution. The answer is (e). Clearly y must be even in order to get
integer solutions. The largest possible value for y is 16 since we must have
x > 0. When y = 16, we have x = 1. Thus, (x; y) = (1; 16) is a solution.
Let us consider successively smaller (even) values for y : (x; y ) = (4; 14),
(7; 12), (10; 10), etc. However, the solution (10; 10) and any further ones
do not satisfy y > x. Thus, we are left with the solutions (x; y ) = (1; 16),
(4; 14), and (7; 12).
 477

11. Suppose A, B, and C are positive integers such that

24 = A + 1 .
5 B+ C 1
+1

The value of A + 2B + 3C equals:

(a) 9 (b) 12 (c) 15 (d) 16 (e) 20
Solution. The answer is (c). Since B and C are positive integers, we see
that B + 1=(C + 1) > 1, whence its reciprocal is smaller than 1. Therefore,
A must represent the integer part of 24=5; that is, A = 4. Then we have
4 = 1 or
5 = B+ 1 .
5 B+C 1
+1
4 C+1
For exactly the same reason as above we see that B must be the integer part
of 5=4; that is, B = 1. Then 1=4 = 1=(C + 1), which implies that C = 3.
Then
A + 2B + 3C = 4 + 2(1) + 3(3) = 15 .
12. A box contains m white balls and n black balls. Two balls are
removed randomly without replacement. The probability one ball of each
colour is chosen is:
(a) m n mn
( + m n;
)( + 1)
(b) mmnn 2 ( +
(c) m n; mnm n;
) ( +
2
1)( + 1)

(d) m n mn
( m n;
+
2
)( + 1)
(e) m mn mm; n; (
(
+ )( +
1)
1)

Solution. The answer is (d). The number of ways ; of
choosing 2 balls

(without replacement) from a box with m + n balls is m n . The number of +

ways of choosing 1 white ball is m, the number of ways of choosing 1 black

2

ball is n. Thus, the probability that one ball of each colour is chosen is:
;mmnn
=
+
mn mn
m n)! = (m+n)(m+n;1) = (m + n)(m + n ; 1) .
( +
2mn
2 m n;2)!
2( + 2

13. If it takes x builders y days to build z houses, how many days

would it take q builders to build r houses? Assume these builders work at
the same rate as the others.
(a) qry
xz (b) ryz
qx
qz
(c) rxy (d) xyr
qz
rz
(e) qxy
Solution. The answer is (d). Since x builders build z houses in y days,
we see that x builders build z=y houses per day, or 1 builder builds z=(xy )
houses per day. Hence q builders build qz=(xy ) houses per day. Let s be
the number of days needed for q builders to build r houses. Then q builders
build r=s houses per day. Equating these we get qz=(xy ) = r=s, whence
s = rxy=(qz) days.
478

14. If x + xy + x = 14 and y + xy + y = 28, then a possible value

2 2

for the sum of x + y is:

(a) ;7 (b) ;6 (c) 0 (d) 1 (e) 3
Solution. The answer is (a). Adding the two given equations gives:
x + 2xy + y + x + y = 42 ;
2 2

that is, (x + y ) + (x + y ) ; 42 = 0 ,
2

(x + y ; 6)(x + y + 7) = 0 .
Thus, x + y = 6 or x + y = ;7.
15. Two congruent rectangles each measuring 3 cm 7 cm are placed
as in the gure. The area of overlap (shaded), in cm , is: 2

(a) 87
7
(b) 29
7
(c) (d)20
7
(e) none of these
21
2

Solution. The answer is (a). All the unshaded triangles in the diagram
below are right-angled and thus are congruent. By the Theorem of Pythago-
ras we have
x = (7 ; x) + 3 = 49 ; 14x + x + 9 ,
2 2 2 2

14x = 58 or x = 29 . 7
7 ;x
3 cm

7;x x
3 cm x

The area of the shaded parallelogram is 3x = 87

7
cm . 2

That completes the Skoliad Corner for this issue. Send me suitable
contest materials and suggestions for the future of the Corner.
 479

MATHEMATICAL MAYHEM
Mathematical Mayhem began in 1988 as a Mathematical Journal for and by
High School and University Students. It continues, with the same emphasis,
as an integral part of Crux Mathematicorum with Mathematical Mayhem.
All material intended for inclusion in this section should be sent to
Mathematical Mayhem, Department of Mathematics, University of Toronto,
100 St. George St., Toronto, Ontario, Canada. M65 3G3. The electronic
address is
[email protected]
The Assistant Mayhem Editor is Cyrus Hsia (University of Western On-
tario). The rest of the sta consists of Adrian Chan (Harvard University),
Jimmy Chui (University of Toronto), Donny Cheung (University of Waterloo),
and David Savitt (Harvard University)

Editorial
This issue brings to completion another year and another volume. It also
represents my last issue as Editor of Mathematical Mayhem. I have been
at this post for a long time now, and in that time, I have seen Mayhem go
through several changes, most notably the merger with Crux Mathematico-
rum, but its spirit and purpose has remained the same.
At this point, I would like to thank all the current members of the
Mayhem sta : Cyrus Hsia, Adrian Chan, Donny Cheung, Jimmy Chui, and
David Savitt, as well as all the other sta members who are no longer present.
Your hard work and dedication over the years have been greatly appreciated.
I would also like to thank Bruce Shawyer, who has had to put up with
his share of garbled LATEX and missed deadlines, for his patience and guidance,
and in addition the CMS, without whom Mayhem would not be where it is
today.
Most of all, however, I would like to thank all the people who have
read and contributed to Mayhem. We greatly value your input and support,
and you are the reason that Mayhem exists.
Taking over from me next year will be Shawn Godin, a high school math
and physics teacher in Ottawa. His enthusiasm and vision will be a welcome
addition to Mayhem, and I wish him the best of luck. Thanks again to all.
I once said I believed that Mayhem was a good thing. I still do.
Naoki Sato
Mayhem Editor
480

Shreds and Slices

Another Proof of the Ellipse Theorem
Nikolaos Dergiades, Thessaloniki, Greece, has kindly provided a short
proof of the nal theorem concerning the mid-point ellipse in \Ellipses in
Polygons" [2000 : 361]. We begin by recalling the theorem.
Theorem. Let ABC be a triangle in the plane, and let z , z , and z be
the complex numbers corresponding to the vertices A, B and C , respectively.
1 2 3

Let p(z ) = (z ; z )(z ; z )(z ; z ) .

1 2 3

Then the foci of the mid-point ellipse ! are the roots of the equation
p0 (z) = 0, and the centre of ! is the root of the equation p00 (z) = 0.
Proof. Let F be one of the foci of the mid-point ellipse ! , and let z be
the complex number corresponding to it. Let = z ; z , = z ; z and
=z;z .
1 2

Let K , M and N be the mid-points of sides BC , AC and AB , re-

spectively (See Figure 1). It is known (see [1]) that the median AK passes
through the centre of ! . Similarly, so do medians BM and CN . Thus, the
centre of ! is the centroid of ABC , with corresponding complex number
(z + z + z )=3, the root of the equation p00 (z ) = 0. In order to prove that
the foci are the roots of the equation p0 (z ) = 0, it is sucient to prove that
1 2 3

+ + = 0.
A It is known (see [2], [3]) that AF is
the bisector of \MFN , and hence,
z1 z2
Arg z ;;z z = Arg z1z z; z
+

3 ;z ,
2 1
+
1 2

Nq qM or
Arg 2+ = Arg 2+ ,
F which implies that
q
= ( + )( + )
+
B K C
Figure 1
2
+

is real, or the number

+ + is real. Similarly, the numbers
2

+ + and
+ +
2 2

are real. Hence, + + = 0, because otherwise, two of the numbers

2
, and would have the same argument, and hence, two of , and
2 2
 481

would have arguments that di er by 180, which would mean that F lies on
a side of ABC , which is impossible.
The previous method can be applied to the general case. As above, let
an inscribed ellipse be tangent to the sides BC , AC and AB at K , M and
N , respectively, and let
k = BKKC , m =
CM and n = AN .
MA NB
Then we know that AK , BM and CN are concurrent, and Ceva's Theorem
gives kmn = 1. The complex numbers corresponding to K , M and N are
z + kz , z + mz and z + nz ,
2 3 3 1 1 2

1+k 1+m 1+n

respectively, and since AF is the bisector of \MFN , we have
z1 nz2 ; z
Arg z n; z = Arg z3 zmz;1 z; z , or
+
1+ 1
+
1
m 1+

Arg (1 ++ nn) = Arg (1m+ m +

) .
As in our previous calculations, we have that the complex numbers
+k + km ,
+k + km and
+k + km
2 2 2

are real. Hence, + k + km = 0, so that the foci of the

ellipse are the roots of the equation
;
(1 + k + km)z ; z + z + k(z + z ) + km(z + z ) z
2


1 2 1 3 2 3

+z z + kz z + kmz z = 0 , 1 2 1 3 2 3

and the centre of the ellipse is

z + z + k(z + z ) + km(z + z ) .
1 2 1 3 2 3

2(1 + k + km)
References
[1] Cours de Mathematiques Elementaires { Exercices de Geometrie par
F.G{M, Theoreme 872{2081, Cinquieme Edition.
[2] Cours de Mathematiques Elementaires { Exercices de Geometrie par
F.G{M, Theoreme 889 II{2113, Cinquieme Edition.
[3] Poncelet, Traite de properties projective des gures, vol. I no 461, 469.
482

Mayhem Problems
The Mayhem Problems editors are:
Adrian Chan Mayhem High School Problems Editor,
Donny Cheung Mayhem Advanced Problems Editor,
David Savitt Mayhem Challenge Board Problems Editor.
Note that all correspondence should be sent to the appropriate editor |
see the relevant section. In this issue, you will nd only solutions | the
next issue will feature only problems.
We warmly welcome proposals for problems and solutions. With the
schedule of eight issues per year, we request that solutions from the previous
issue be submitted in time for issue 8 of 2001.

High School Solutions

Editor: Adrian Chan, 1195 Harvard Yard Mail Center, Cambridge, MA,
USA 02138-7501 <[email protected]>
H261. Solve for x:
q p x q7 + p48x = 14 .
7 ; 48 +

Solution by Andrei Simion,

p Brooklyn
p Technical HS, Brooklyn, NY, USA.
First, we notice that 7 ; 48 = p p . 1

p p x 7+ 48

Thus, let t = 7 ; 48 , and our equation becomes:

t + 1=t = 14
==) t ; 14t + 1 = 0 p
2

==) t = 7  48 .
p
If t = 7 ; 48, then x=2 = 1, leading to x = 2.
p
If t = 7 + 48, then x=2 = ;1, leading to x = ;2.
Also solved by EDWARD T.H. WANG, Wilfrid Laurier University, Waterloo, Ontario.
Edward Wang generalized the question to solving
q p x q x
a ; a2 ; 1 + a + pa2 ; 1 = 2a, showing that x = 2 are the only
solutions.
 483

H262. Proposed by Mohammed Aassila, CRM, Universite de

Montreal, Montreal, Quebec.
Solve the equation x ; p
x = 91 .
x ;1
2 60
Solution. Let x = sec A, so that x= x
p ; 1 = csc A. Our equation
2

becomes
sec A ; csc A = 91

==) 60(sin A ; cos A) = 91 sin A cos A

60

==) 3600(1 ; 2 sin A cos A) = 8281 sin A cos A . 2 2

Letting p = 2 sin A cos A = sin 2A,

14400(1 ; p) = 8281p 2

==) 8281p + 14400p ; 14400 = 0

2

==) (169p ; 120)(49p + 120) = 0

==) p = . 120
169

The value of p = ;120=49 is rejected, since p = sin 2A  ;1. Continuing,

sin 2A = q 120

==) cos 2A =  1 ;
;
=169
120 2 119
169 169

==) 2 cos A ; 1 =  2 119

169

==) cos A = 2
or 25
169
144
169

==) cos A =  ,  . 5
13
12
13

Since x = sec A, then x = 13=5 or 13=12. Checking reveals that

only x = ;13=12 and x = 13=5 are valid solutions.
Also solved by EDWARD T.H. WANG, Wilfrid Laurier University, Waterloo, Ontario; and
MURRAY S. KLAMKIN, University of Alberta, Edmonton, Alberta.
Note: Both of the other solutions consisted of rationalizing the equation and solving a
quartic.
H263. Let ABC be an acute-angled triangle such that a = 14,
sin B = 12=13, and c, a, b form an arithmetic sequence (in that order).
Find tan A + tan B + tan C .
Solution. Let c = 14 ; t and b = 14 + t. Since ABC is acute-angled
and sin B = 12=13, then cos B = 5=13. Using the Cosine Law,
(14 + t) = (14 ; t) + 14 ; 2(14 ; t)(14) cos B
2 2 2

==) 196 + 28t + t = 196 ; 28t + t + 196 ; 28(14 ; t)(5=13)

2 2

==) 56t ; 196 = ;28(14 ; t)(5=13)

==) 13(2t ; 7) = 5(t ; 14)
==) 26t ; 91 = 5t ; 70
==) 21t = 21
==) t = 1.
484

Thus, ABC is a 13-14-15 triangle.

Let D be the foot of the altitude from A. Since AD is an altitude,
sin B = AD=13 = 12=13, so that AD = 12. Thus, BD = 5 and DC = 9.
Also, tan B = 12=5, and tan C = 12=9 = 4=3.
Now,
tan A = tan(180 ; B ; C )
= ; tan(B + C )
= ; ; BBtan +tan
1 tan tan
C
C
= ; 125; 16543 =
1
+ 16
11

==) tan A + tan B + tan C = 856

165
.

Also solved by ANDREI SIMION, Brooklyn Technical HS, Brooklyn, NY, USA.
H264. Find all values of a such that x ; 6x 3
+ 11x + a ; 6 = 0 has
2

exactly three integer solutions.

Solution by Lino Demasi, student, St. Ignatius HS, Thunder Bay,
Ontario; and Andrei Simion, Brooklyn Technical HS, Brooklyn, NY, USA.
Let d, e, and f be the roots of the cubic x ; 6x + 11x + a ; 6 = 0.
3 2

Hence,
d + e +f = 6,
de + ef + fd = 11 ,
def = 6 ; a .
Notice that
(d + e + f ) = d + e + f + 2(de + ef + fd)
2 2 2 2

==) 36 = d + e + f + 22
2 2 2

==) d + e + f = 14 .
2 2 2

Clearly, ;3  d, e, f  3. But since d + e + f = 6, the only solution

is: fd, e, f g = f1, 2, 3g, which implies that 6 ; a = 1(2)(3) = 6, so that
a = 0.
Also solved by EDWARD T.H. WANG, Wilfrid Laurier University, Waterloo, Ontario; and
MURRAY S. KLAMKIN, University of Alberta, Edmonton, Alberta.
 485

Advanced Solutions
Editor: Donny Cheung, c/o Conrad Grebel College, University of
Waterloo, Waterloo, Ontario, Canada. N2L 3G6 <[email protected]>
A237. Show that for any sequence of decimal digits that does not be-
gin with 0, there is a Fibonacci number whose decimal representation begins
with this sequence. (The Fibonacci sequence is the sequence Fn generated
by the initial conditions F = 0, F = 1 and Fn = Fn; + Fn; for n  2.)
0 1 1 2

Solution. Let fxg denote the fractional part of x; that is,

fxg = x ; bxc. We begin by proving a useful lemma.
Lemma: If is irrational, then given any range [a; b]  [0; 1), there
exist in nitely many natural numbers n such that fn g 2 [a; b].
Proof: Let c = b ; a. It suces to prove that there is a natural number
m such that 0 < fm g < c, since then there must be an integer multiple of
fm g inside every interval of the form [k + a; k + b], where k is a natural
number. This gives us our in nitely many n.
We may partition the interval [0; 1) into a nite number of intervals
of size < c. Thus, by the pigeonhole principle, we must have two distinct
natural numbers m and m such that fm g and fm g lie within the
same interval. Then, 0  fjm ; m j g = jfm g ; fm gj < c. If
1 2 1 2

fjm ; m j g = 0, then jm ; m j must be an integer, meaning that

1 2 1 2

is rational, contradiction. Thus, 0 < fjm ; m j g < c, as desired.

1 2 1 2

1 2

Now, we proceed to prove the main result. We use Binet's formula,

1 1 +
p5 !n 1 1 ; p5 !n
Fn = p5 2 ; p5 2 ,

which, for suciently large n, is the integer nearest to

1 1 +
p5 !n
p5 2 .

Given a positive integer s, represented by a sequence of decimal digits

not beginning with 0, we want to nd an integer k  0 such that
s
10k  Fn < (s + 1)
10k .
Since both s
10k and (s + 1)
10k are integers, this is equivalent to
1 1 +
p5 !n
k
s
10  p5 < (s + 1)
10k .
2
486

On taking base 10 logarithms and some rearranging, we obtain the equivalent

condition
p p!
1 + 5 < k + log ((s + 1)p5) .
k + log (s 5)  n
log
10 10
2 10

By our lemma, we can nd an in nite number of n such that

p ( p !) p
flog (s 5)g  n
log 1+ 5
< flog ((s + 1) 5)g .
10 10
2 10

p
(In the case that there is an integer between log s 5 and
 p p 10

log (s + 1) 5 , we can replace the lower bound, flog s 5 g, with 0.)

10 10

This means that there will be one such n which is suciently large so that
p 1 +
p5 !  p5
k + log s 5  n
log
10
2
10 < k + log ( s + 1) 10

for a non-negative integer k, and we are done.

A238. Two circles C and C intersect at P and Q. A line through P
intersects C and C again at A and B , respectively, and X is the mid-point
1 2

of AB . The line through Q and X intersects C and C again at Y and Z ,

1 2

respectively. Prove that X is the mid-point of Y Z .

1 2

(1997 Baltic Way)

Solution by Michel Bataille, Rouen, France.
Since APY Q and BPZQ are both concyclic, we have \AY Q = \APQ
and \QZB = \QPB . We can see that \APQ + \QPB =  , since A, P and
B are collinear, and \QZB + \BZY = , since Q, Z and Y are collinear.
Thus, we have \BZY = \APQ = \AY Q.
From the fact that APY Q and BPZQ are both concyclic again, we also
have \Y AP = \Y QP = \ZQP = \ZBP . Finally, since jAX j = jXB j,
we conclude that AXY and BXZ are congruent triangles. Thus,
jY X j = jXZ j, and X is the mid-point of Y Z .
A239. Proposed by Mohammed Aassila, CRM, Universite de
Montreal, Montreal, Quebec.
Let a , a , : : : , an be n distinct numbers, n  3. Prove that
1 2

0 1
X
n
@ Y 1 A = 0.
ai

i=1 j 6=i ai ; aj
 487

Solution by Michel Bataille, Rouen, France.

Consider the rational function
0 1
X n a Y 1
R(x) = @ x ;ia
a ; a A .
i i j6 i i
=1
j =

We may rewrite the entire sum over a common denominator to get

0 1
X
n
@Y x ; aj A
ai
a ; a
R(x) = i =1
i
Qn j(6 xi ; a )
j= P (x) ,
= Q
i =1i (x)
where P (x) is a polynomial of degree  n ; 1 and Q(x) is a polynomial of
degree exactly n. However, for 1  k  n,
0 1
Y ak ; aj X @ ak ; ak Y ak ; aj A
P (ak) = ak
a ; a + ai
a ; a
a ; a = ak .
j 6=k k j i6=k i k j;k6=i i j

Q P )  n ; 1, P (x) must
Since P (x) = x for n distinct values of x and deg(
be identically equal to x, and, letting ui = ai
j 6 i ai ;aj , for convenience,
1
=
we have
X
n ui
R(x) = = Qn x .
i=1 x ; ai i=1(x ; ai)
Since xui = ui (x ; ai ) + ui ai , we have
x2 = x
X n ui
=
Xn
u +
Xn ui ai
,
i
Q(x) i=1 x ; ai i=1 i=1 x ; ai
so that n !
X X
n
x =
2
ui Q(x) + uixa;iQa(x) .
i=1 i=1 i
Each xQ;xai is a polynomial of degree n ; 1, but we have only one poly-
( )

nomial of degree n. Thus, inPorder for their sum to be a polynomial of degree

less than 3, we must have ni ui = 0, as desired.
=1

Also solved by JOSE LUIS DIAZ, Universitat Politecnica de Catalunya, Terrassa, Spain;
VEDULA N. MURTY, Dover, PA, USA; and CATHERINE SHEVLIN, Wallsend, England.
Murray S. Klamkin of the University of Alberta points out that this result is a special
case of the identity
Xn 0 Y 1 1
@ari A = 0.
i=1 j 6=i ai ; aj
The proof given above can easily be adapted to prove this more general result.
488

A240. Proposed by Mohammed Aassila, CRM, Universite de Montreal,

Montreal, Quebec.
Let a, b and c be integers, not all equal to 0. Show that
1  p3 4a + p3 2b + c .
4a + 3b + 2c
2 2 2

Solution. First, note that

p p p
( 3 4a + 3 2b + c)(2 3 2a + 3 4b + c ; 2ab ; 3 4ac ; 3 2bc)
2
p 2 2
p p
= 4a + 2b + c ; 6abc . 3 3 3

p p q3 p3 p3
By the AM{GM Inequality, [( 4a) + ( 2b) + c ]  p( 4 p
3 1
3
3 3
2abc) = 3 3 3

2abc, so that 4p3a +2b p

3
+3 c ;6abc  0, with equality when 3 4a = 3 2b = c.
3 3

In fact, since 2 and 4 are irrational, the only time we can have integers
a, b and c satisfy p3 4a = p3 2b = c is when a = b = c = 0. Thus, when a, b
and c are integers, not all equal to 0, since 4a + 2b + c ; 6abc > 0 must 3 3 3

also be an integer, we have 4a + 2b + c ; 6abc  1 . 3 3 3

p
Now, since 64 > 54 ==) 4 ;!3 3 2 > 0 and 27 > ==) 3 ; 3 4 > 0,
p 27 3

p p3 2 1 p  3p  2 2

2a + p b + p c + (4 ; 3 2)a + 3 ; 2 3 4 b + 12 c
2

0 6 3 2 2 2

2 2
p p
==) 0  (4 ; 2 2)a + (3 ; 3 4)b + c + 2ab + 3 4ac + 3 2bc
3 2
p p 2 2

p 2
p 2
p 2
p
==) 2 3 2a + 3 4b + c ; 2ab ; 3 4ac ; 3 2bc  4a + 3b + 2c . 2 2 2

Similarly, p3
 p3 
0  (a ; b) + 4a ; 2 c + 2b ; 2 c + 3a + 2b + 52 c
1 1 2 2
2 2 2 2

p p p p
==) 0  (4 + 2 3 2)a + (3 + 3 4)b + 3c ; 2ab ; 3 4ac ; 3 2bc
2 2 2

p p 2
p 2
p
==) ;2 3 2a ; 3 4b ; c + 2ab + 3 4ac + 3 2bc  4a + 3b + 2c .
2 2 2 2

Thus,
p
j2 2a + p3 4b + c ; 2ab ; p3 4ac ; p3 2bcj  4a + 3b + 2c
3 2 2 2 2 2 2

3 3 3
p p
==) j4a + 2a + c ; 6abcj  (4a + 3b + 2c ) 3 4a + 3 2b + c 2 2 2

2
p p
==) 1  (4a + 3b + 2c ) 3 4a + 3 2b + c
2 2

p p
==) 4a + 31b + 2c  3 4a + 3 2b + c ,
2 2 2

since 4a + 3b + 2c > 0.
2 2 2
 489

Challenge Board Solutions

Editor: David Savitt, Department of Mathematics, Harvard University,
1 Oxford Street, Cambridge, MA, USA 02138 <[email protected]>
C89. Proposed by Tal Kubo, Brown University.
Show that the formal power series (in x and y )
X
1
(xy)n
n=0
cannot be expressed as a nite sum
X
m
fi(x)gi(y) ,
i=1
where fi(x) and gi (y ) are formal power series in x and y , respectively,
1  i  m.
Solution.
To an in nite matrix (aij )i2N; j 2N one can still associate the idea of
\rank": the smallest number q such that there exist q rows so that every
other row is a linear combination of these q rows. (Of course, the rank might
be in nite.) Note that rank(A + B )  rank(A) + rank(B ).
Any power series in two variables x and y can be represented as such
a matrix: the power series X
aij xiyj
corresponds to the matrix (aij ). Then any matrix corresponding to a product
f (x)g(y) of power series in one variable has rank 1, Thus, a nite sum of
such matrices has nite rank, whereas
X n
(xy)
n
has in nite rank.
C90. Proposed by Noam Elkies, Harvard University.
Let S , S and S be three spheres in R whose centres are not
3

collinear. Let k  8 be the number of planes which are tangent to all three
1 2 3

spheres. Let Ai, Bi and Ci be the points of tangency between the i suchth

tangent plane, 1  i  k, and S , S and S , respectively, and let Oi be

the circumcentre of triangle Ai Bi Ci. Prove that all the Oi are collinear. (If
1 2 3

k = 0, then this statement is vacuously true.)

Solution by Michel Bataille, Rouen, France.
Let us rst solve a two-dimensional analogue of this problem. Let C
and C be two circles whose centres do not coincide. Let Qj and rj denote
1

2
490

the centre and radius of Cj for j = 1, 2. There are up to four lines tangent
to both circles. Let Ai and Bi be the points of tangency between the ith line
and C and C , respectively, i  4. We wish to prove that the mid-points
Oi of the AiBi are collinear.
1 2

To see this, we recall the notion of the radical axis of two circles: the
radical axis of C and C is the collection of points P such that
1 2

(Q P ) ; r = (Q P ) ; r .
1
2 2
1 2
2 2
2

It is not dicult to check that the solutions to this equation form a line. To
solve our two-dimensional question, we claim that each Oi lies on the radical
axis of C and C , so that they are certainly collinear. We check this claim:
since AiOi and BiOi are tangent to C and C , respectively, the triangles
1 2

AiOiQ and BiOiQ are right triangles, and so

1 2

1 2

(Q Oi) ; r = (AiOi) = (BiOi) = (Q P ) ; r ,

1
2 2
1
2 2
2
2 2
2

as desired.
We now turn to the original three-dimensional problem. For each pair
of spheres Sj , Sk , with centres Qj and Qk and radii rj and rk , we de ne the
radical plane to be the plane of points P such that
(Qj P ) ; rj = (QkP ) ; rk .
2 2 2 2

Notice that any point on the radical plane of S and S , and which is also
on the radical plane of S and S , is automatically on the radical plane of S
1 2

and S (by transitivity).

2 3 1

Moreover, the radical plane of two spheres is perpendicular to the line

between their centres; since the three spheres under consideration have cen-
tres which are non-collinear, we can conclude that the three radical planes
between them must jointly intersect in a single radical axis. Since the circum-
centre Oi of the triangles AiBi Ci is equidistant from each of Ai , Bi and Ci,
and since each line segment OiAi , OiBi and Oi Ci is tangent, respectively,
to S , S and S , it follows (exactly as in the two-dimensional case) that each
Oi must lie on the radical axis. Thus, the Oi are collinear.
1 2 3

Also solved by LAURENT LESSARD, 2nd year engineering student, University of Toronto,
 FRESNEDA, Cuba.
Toronto, Ontario; and ROMAN
 491

Problem of the Month

Jimmy Chui, student, University of Toronto
Problem. The equations x +5x+6 = 0 and x +5x;6 = 0 each have integer
2 2

solutions whereas only one of the equations in the pair x + 4x + 5 = 0 and

2

x + 4x ; 5 = 0 has integer solutions.

2

(a) Show that if x + px + q = 0 and x + px ; q = 0 both have in-

2 2

teger solutions, then it is possible to nd integers a and b such that

p =a +b .
2 2 2

(b) Determine q in terms of a and b.

(1998 Euclid, Problem 10)
Solution. First we show that p and q must be integers.
Let p ; 4q = m and p + 4q = n , for positive reals m and n. Now,
2 2 2 2

the solutions to the rst quadratic equation x + px + q = 0 are (;p  m)/2.

2

The sum of both solutions must be an integer. Thus, ;p must be an integer,

meaning that p must be an integer. The di erence of the solutions must
also be an integer. Thus, m must be an integer. (Similarly, from the second
quadratic equation, n must be an integer as well.) Now, since x + px + q = 0
2

has an integer solution, we must have q being an integer (this last point can
be seen by substituting one of the solutions into the equation.) Thus, p and q
must be integers.
Examining the equations p ; 4q = m and p + 4q = n modulo 4,
2 2 2 2

we note that m and n have the same parity as p.

Adding, we get 2p = m + n . Thus,
2 2 2

p = (m + n )=2 = [(n + m)=2] + [(n ; m)=2] ,

2 2 2 2 2

which is a sum of two squares of integers, since m and n have the same
parity. We can let a = (n + m)=2 and b = (n ; m)=2.
To nd q in terms of a and b, we also know that 8q = n ; m , or, 2 2

equivalently, q = [(n + m)=2][(n ; m)=2]=2 = ab=2.

Any reader who wishes to contribute to the Problem of the Month section
can feel free to do so by emailing a problem and any nice solution(s) to
[email protected], or sending hard copy, care of the Mayhem Editor.
492

Constructive Geometry | Part II

Cyrus Hsia

In part I of this series [2000 : 231] we laid the foundations of building

basic geometric objects. We started with two simple tools: a straightedge
and a collapsible compass. Though deceptively simple, their usefulness is
demonstrated over and over again. Using only these objects we were able
to construct all the tasks given in part I. We continue here with a selection
of problems from various sources that require only these two tools. We also
noted in part I that many geometric construction problems can be readily
done by using results from other constructions that we have done. Here,
we will discuss one result named Apollonius' Theorem that is useful in many
geometric constructions.
We begin with a nice and easy problem that appeared on the Internet.
Problem 1
Given a circle with centre O and a point P outside the circle, construct
a point Q on the circle so that angle PQO is 60 .
Solution
Here is the construction:
1. Construct an equilateral triangle on edge OP . Label the third vertex
R and the centroid S. Remember, we can use the results that we have
obtained in previous exercises. Constructing an equilateral triangle was
problem 6 in part I of this series. Now how can you construct the cen-
troid of an equilateral triangle?
2. Construct the circumcircle of triangle OSP . Again refer to results in the
previous article that you can use to help do this. First, nd the centre
of the circle.
3. This circle intersects the circle with centre O in two points. Take Q
to be the point on the opposite side of the line OP from point S .
See Figure 1.
Proof that angle PQO is 60 : Since O, P , Q and S are concyclic we
have \OQP = 180 ; \OSP = 60 . Why is \OSP = 120 ?
The following is a problem that appeared in a Hungarian Contest.
Problem 2
Given a right triangle ABC , construct a point N inside the triangle
such that the angles NBC , NCA and NAB are equal.
Copyright c 2000 Canadian Mathematical Society
 493
R

S
120
O P

60
Q

Figure 1

(Hungarian Problem Book I, Problem 1895/2)

Solution
Again, the idea is to work the problem backwards to get a sense of the
construction that is needed. Consider Figure 2a.
A

F
D
N

B C
Figure 2a

Suppose N in Figure 2b satis es the conditions. Angle ABC is the right

angle. Let \NBC = \NCA = \NAB = . Now, \ABN = \ABC ;
\NBC = 90 ; . Thus, \ANB = 180 ; \NAB ; \ABN = 90 . We
conclude from this that N must be on the circumference of the circle with
diameter AB . Thus, when we come to constructing point N , we will likely
construct a circle with diameter AB .
494

Now extending CN to the other side of the circle to point D, we see

that \NDA = \NBA = 90 ; .
This implies that \DAC = 180 ; \ADC ; \ACD = 90 . Let AC
intersect the circle at F . Angles DAF and DAC are both right angles. Thus,
DF is a diameter of the circle as well. From this we see that by extending F
through the centre O to the other side of the circle and connecting DC we
would have N on the intersection of DC and the circle.
Here is the construction:
1. Using AB as diameter, construct a circle.
2. Let F be the intersection of the circle with AC . Extend F through the
centre to D. Connect DC , and let N be the intersection of the circle
with DC extended. See Figure 2b.
A

r
F
D
N

B C
Figure 2b
Now, here is the proof that this point N satis es the conditions
in the problem. Let \NBC = . Angle NBA is then 90 ; . Since
\ANB is then a right angle, this implies that \NAB is also . Then
\NDA = \NBA = 90 ; . Since \DAF is right, \DAC is a right
angle in triangle DAC . Thus, \ACD = 180 ; \DAC ; \NDA = .
Thus, \NBC , \NCA and \NAB are equal as required.
APOLLONIUS
Here is a problem that comes in many disguises and is used in many
problems.
Problem 3A
Construct the locus of all points P such that AP : PB = 1: 2 for given
points A and B . Describe this locus.
If we had to guess what the locus would look like, the two simplest
ones would be a circle or a line. And that is indeed what the answers are in
the general case. This result is known as Apollonius' Theorem. See [1] and
[2] for more on this.
 495

Apollonius' Theorem
Given a line segment AB , the locus of all points C such that the ratio
AC to CB is a constant is either a straight line or a circle.
See if you can spot the use of this theorem in the following problems.
Problem 3B
A segment AB is given and a line m crossing it. Determine the point
C on m such that m bisects angle ACB.
Problem 3C
Given three disjoint circles in the plane, construct a point on the plane
such that all three circles subtend the same angle at the point.
Correspondence Program, Geometry Problem Set, Problem 19
Problem 3D
Four points A, B , C and D are given on a straight line. Construct a
pair of parallel lines through A and B , and another pair through C and D,
so that these pairs of parallel lines intersect in the vertices of a square.
Hint
Draw in the four circles with diameters AB , BC , CD and AD. Notice
that two opposite vertices of the square lie on the circles with diameters BC
and AD. Now look to see if you can apply Apollonius' Theorem.

A B C
D

Figure 3
496

Now for some exercises. All of the following constructions should be

performed using straight edge and compass only.
Exercises
1. Given 3 three parallel lines, construct an equilateral triangle with a ver-
tex on each of the three lines.
2. Given 3 concentric circles, construct an equilateral triangle with a vertex
on each of the three circles.
Proposed for the 26th IMO
3. Let P be one of the two points of intersection of two intersecting circles.
Construct the line l through P , not containing the common chord, such
that the two circles cut o equal segments on l.
4. An army captain wishes to station an observer equally distant from two
speci ed points and a straight road. Can this always be done? Locate
any possible stations. In other words, how many points are there in
the Euclidean plane which are equidistant from two given points and a
given line? Find them with straightedge and compass, if possible.
5. Show how to divide a circle into nine regions of equal area, using a
straight-edge and compass.
6. Construct the following special centres of a triangle: the centroid, incen-
tre, orthocentre, circumcentre, Gergonne Point, and the Fermat Point.
We saw in problem 1 the construction of all these points in an equilat-
eral triangle because they are all the same point! Now do it for any
given triangle.
References
1. Barbeau, Edward J., ATOM (A Taste of Mathematics), Volume 1, pp. 47{
48. Canadian Mathematics Society, 1997.
2. Grossman, J.P., \Ye Olde Geometry Shoppe { Part II", Mathematical
Mayhem, Volume 6 Issue 3, p. 10.

Cyrus Hsia
21 Van Allen Road
Scarborough
Ontario
M1G 1C3
[email protected]
 497

PROBLEMS
Problem proposals and solutions should be sent to Bruce Shawyer, Department
of Mathematics and Statistics, Memorial University of Newfoundland, St. John's,
Newfoundland, Canada. A1C 5S7. Proposals should be accompanied by a solution,
together with references and other insights which are likely to be of help to the editor.
When a proposal is submitted without a solution, the proposer must include sucient
information on why a solution is likely. An asterisk (?) after a number indicates that
a problem was proposed without a solution.
In particular, original problems are solicited. However, other interesting prob-
lems may also be acceptable provided that they are not too well known, and refer-
ences are given as to their provenance. Ordinarily, if the originator of a problem can
be located, it should not be submitted without the originator's permission.
To facilitate their consideration, please send your proposals and solutions
on signed and separate standard 8 12 "11" or A4 sheets of paper. These may
be typewritten or neatly hand-written, and should be mailed to the Editor-in-
Chief, to arrive no later than 1 May 2001. They may also be sent by email to
[email protected]. (It would be appreciated if email proposals and solu-
tions were written in LATEX). Graphics les should be in epic format, or encapsulated
postscript. Solutions received after the above date will also be considered if there
is sucient time before the date of publication. Please note that we do not accept
submissions sent by FAX.

2589. Proposed by Joaqun Gomez

 Rey, IES Luis Bu~nuel, Alcorcon,
Spain.
X2c n
bn= n .
For n = 2, 3, : : : , evaluate
k=1 k k;1
2590. Proposed by Joaqun Gomez  Rey, IES Luis Bu~nuel, Alcorcon,
Spain. n n   n
Y
 1 2n .
2

For n = 1, 2, : : : , prove that

k=1 k n+1 n
2591. Proposed by Joaqun Gomez  Rey, IES Luis Bu~nuel, Alcorcon,
Spain.
Two players, A and B , each toss n fair coins, and two other players, C
and D, toss n ; 1 and n + 1 fair coins, respectively.
For each n = 2, 3, : : : , prove that the two events:
A gets exactly one head more than B
and
C and D get exactly the same number of heads
are equally likely.
Find the probability of these events.
498

2592. Proposed by Nairi M. Sedrakyan, Yerevan, Armenia.

Describe all numbers, which can be represented in the form of
a +b ,
3 3

c +d
3 3

where a, b, c, d are natural numbers.

2593. Proposed by Nairi M. Sedrakyan, Yerevan, Armenia.
Let S (a) denote the sum of the digits of the natural number a. Let k
and n be natural numbers with (n; 3) = 1. Prove that there exists a natural
number m which is divisible by n and S (m) = k if either
(a) k > n ; 2; or
(b) k > S (n) + 7S (n) ; 9.
2

2594. Proposed by Nairi M. Sedrakyan, Yerevan, Armenia.

Given a point M inside the triangle ABC (see diagram), prove that
min(MA; MB; MC ) + MA + MB + MC < AB + BC + AC .
B

A C
2595. Proposed by Nairi M. Sedrakyan, Yerevan, Armenia.
Given that M and N are points inside the triangle ABC such that
\MAB = \NAC and \MBA = \NBC , prove that
AM
AN + BM
BN + CM
CN = 1 .
AB
AC BA
BC CA
CB
B

M N

A C
2596. Proposed by Clark Kimberling, University of Evansville, Evansville,
IN, USA.
Write r << s if there is an integer k satisfying r < k < s. Find, as a
function of n (n  2) the least positive integer k satisfying
k << k << k << : : : << k << k .
n n;1 n;2 2
 499

2597. Proposed by Michael Lambrou, University of Crete, Crete,

Greece.
Let P be an arbitrary interior point of an equilateral
 j\PAB ;triangle ABC
 .
Prove that j\PBC ; \PCB j  arcsin 2 sin
\ PAC j ;
 j\PAB ; \PAC j  2
2  j\PAB ; \PAC j.
Show that the left inequality cannot be improved in the sense that there
is a position Q of P on the ray AP giving an equality.
(Thus, the inequality in 2255 is improved.)
2598. Proposed by D.J. Smeenk, Zaltbommel, the Netherlands.
Suppose that AD, BE and CF are the internal angle bisectors of 4ABC ,
with D on BC , E on CA and F on AB . Write a = BC , b = CA, c = AB ,
x = AE and y = AF . We are given that x + y = a. Prove that:
(a) a = bc;
2

(b)
1 ; 1 = 1 ; 1;
x y b c
1 1 1 1 2

(c) + = p + p ;
x y c b
(d) AD < c.
2599. Proposed by Ho-joo Lee, student, Kwangwoon University,
Kangwon-Do, South Korea.
Let P be a point inside the triangle ABC and let AP , BP , CP meet
the sides BC , CA, AB at L, M , N , respectively. Show that the following
two conditions are equivalent:
1 1 1 1 1 1
AP + PL = BP + PM = CP + CN ;
\APN = \NPB = \BPL = \LPC = \CPM = \MPA = 60 .
2600. Proposed by Svetlozar Doichev, Stara Zagora, Bulgaria.
Find all real numbers x such that, if a and b are the lengths of sides of a
triangle with medians from the mid-points of these sides of lengths ma and
mb, respectively, then the equalities a + xma = b + xmb and a = b are
equivalent.
500

SOLUTIONS
No problem is ever permanently closed. The editor is always pleased to
consider for publication new solutions or new insights on past problems.

2468. [1999 : 367] Proposed by Walther Janous, Ursulinengymnas-

ium, Innsbruck, Austria.
For c > 0, let x, y , z > 0 satisfy
xy + yz + zx + xyz = c . (1)
Determine the set of all c > 0 such that whenever (1) holds, then we
have
x + y + z  xy + yz + zx .
Solution by Heinz-Jurgen Sei ert, Berlin, Germany.
The positive reals c asked for are those satisfying c  4.
From (1), we have xy < c and
z = x +c ;y +xyxy .
Substituting this value of z into the given inequality, we see that we must
determine those positive reals c such that
x + y + x +c ;y +xyxy  xy + (xx++y)(y c+;xyxy) ,
or
(x + y )(x + y + xy) + (c ; xy )  xy (x + y + xy) + (x + y )(c ; xy ) ,
which simpli es to
A(x; y) := (x + y) ; (xy) + (c ; xy)(1 ; x ; y)  0 ,
2 2
(2)
whenever x, y > 0 such that xy < c.
Taking x = c=2 and letting y approach zero, we obtain
c + c 1 ; c   0 ,
2

4 2
which means that c  4.
Now suppose that c  4. Let x, y > 0 such that xy < c.
 501

If (x ; 1)(y ; 1)  0, then (2) follows from

 
A(x; y) = (x ; y) + 4xy ; (xy) + (c ; xy) (x ; 1)(y ; 1) ; xy
2 2

= (x ; y) + (4 ; c)xy + (c ; xy )(x ; 1)(y ; 1)  0 .

2

If (x ; 1)(y ; 1) < 0, then it follows from

A(x; y) = (x + y ; 2) + 4(x + y ; 1) ; (xy) + (c ; xy)(1 ; x ; y)
2 2

= (x + y ; 2) + (c ; 4 ; xy )(1
2
 ; x ; y) ; (xy) 
2

= (x + y ; 2) + (c ; 4 ; xy ) (x ; 1)(y ; 1) ; xy ; (xy)
2 2

= (x + y ; 2) + (4 ; c)xy + (c ; 4 ; xy )(x ; 1)(y ; 1)  0 .

2

This proves the above statement. [Ed. see also [2000 : 337]]
Also solved by RICHARD I. HESS, Rancho Palos Verdes, CA, USA; MICHAEL LAMBROU,
University of Crete, Crete, Greece; KEE-WAI LAU, Hong Kong, China; and the proposer. There
was one incorrect solution received.
The problem, by the same proposer, had also been published as Aufgabe 61 in the
Austrian journal Wissenschaftliche Nachrichten, Vol. 107 (1998) p. 36, with solution in
Vol. 110 (1999) pp. 25{26.
The case c = 4 was a problem in the 1996 Vietnamese Mathematical Olympiad; see
problem 6 on [1999 : 8].

2477. [1999 : 429] Proposed by Walther Janous, Ursulinengymnas-

ium, Innsbruck, Austria.
Given a non-degenerate 4ABC with circumcircle ;, let rA be the in-
radius of the region bounded by BA, AC and arc(CB ) (so that the region
includes the triangle).
C

A B

Similarly, de ne rB and rC . As usual, r and R are the inradius and

circumradius of 4ABC .
Prove that
(a)
64 r  r r r  32 Rr ;
27
3
A B C 27
2

16 8
(b) r  rB rC + rC rA + rA rB  Rr ;
2

3 3
502

(c) 4r  rA + rB + rC  (R + r) ,
4
3
with equality occurring in all cases if and only if 4ABC is equilateral.
Solution by Heinz-Jurgen Sei ert, Berlin, Germany.
It is known ([1], [2] and [3]) that
A B  C 
rA = r sec 2 ; rB = r sec 2 , rC = r sec 2 .
2 2 2

If s denotes the semiperimeter of 4ABC , then [1996 : 130]

X X
cos A = R R+ r , cos A cos B = s ; 44R +r , 2 2 2

R 2

and Y
cos A = s ; 4R 4;R 4Rr ; r ,
2 2 2

where sums and products are cyclic over the angles A, B , C . Hence,
Y   Y
cos A2 = 81 (1 + cos A)
2

 X X Y 
= 81 1 + cos A + cos A cos B + cos A
= 16sR ,
2

giving
rA rB rC = 16Rs r .
2 3

2
(1)
On [1996 : 78], Seimiya showed that
X    
sec A2 = 1 + 4Rs+ r ,
2
2

which implies that

 4R + r  ! 2

rA + rB + rC = r 1 + s . (2)
From the above identities, we also have
X A  X
cos2
2 = 21 (1 + cos A) = 4R2R+ r ,
so that
X A B  Y A X A 
sec 2 sec = sec 2
 cos 2
2 2 2 2

2  4R
= 8
2
+ Rr .
s
2
 503

Hence,

rB rC + rC rA + rArB = 8(4R s+ Rr)r .

2 2

2
(3)
Now recall Gerretsen's inequalities (see [1996 : 130]);
16Rr ; 5r  s  4R + 4Rr + 3r ,
2 2 2 2

where, on both sides, equality holds only when the triangle ABC is equi-
lateral. The desired inequalities (a), including the conditions for equality,
follow from (1) and the estimates
 27   27 
2 Rr  s  4 R ,
2 2

which, by Euler's inequality 2r  R, are weaker than Gerretsen's inequali-

ties. Similarly, (b) follows from (3) and
3
12Rr + 3r  s  6R + 2 Rr ,
2 2 2

and (c) follows from (2) and

12Rr + 3r  s  16R + 38Rr + r ,

2 2
2 2

both including the conditions for equality.

References:
[1] L. Banko , A Mixtilinear Adventure, Crux Mathematicorum 9 (1983) 2{7.
[2] C. V. Durell and A. Robson, Advanced Trigonometry, 23.
[3] P. Yiu, Mixtilinear Incircles, Amer. Math. Monthly 106 (1999) 952{955.

Also solved by SEFKET ARSLANAGI C,  University of Sarajevo, Sarajevo, Bosnia and
Herzegovina; MICHEL BATAILLE, Rouen, France; NIKOLAOS DERGIADES, Thessaloniki,
Greece; MICHAEL LAMBROU, University of Crete, Crete, Greece; G. TSINTSIFAS, Thessaloniki,
Greece; and the proposer.
Using Gerretsen's inequalities themselves in (1), (2) and (3) would give stronger (though
more complicated) bounds than are asked for in this problem.
The expressions for rA , rB , rC given at the beginning of the above solution also
appeared in Solution II of Crux 1224 [1988: 147], as pointed out by Tsintsifas (who had
proposed 1224). In fact they are most recently in Solution II of Crux with Mayhem
problem 2464 [2000 : 432]; in particular, see pages 435{436. Moreover, the lower bound of
part (c) of the current problem occurs in the statement of 1224(b), and also in Solution II of
2464.
504

2480. [1999 : 430] Proposed by JoaqunGomez  Rey, IES Luis Bu~nuel,

Alcorcon, Spain.
Writing (n) for Euler's totient function, evaluate
X X (k)(d=k)
d k .
djn kjd
Solution by Kee-Wai Lau, Hong Kong. X
Making use of the well-known relation (d) = n, we see that
djn
X X (k)(d=k) X
d k = t(t)(k); (set d = kt)
djn kjd k;t
ktjn
X X X
= t(t) (k) = t(t)(n=t)
tjn
X kj n=t tjn ( )

= n (t) = n . 2

tjn
Also solved by MICHEL BATAILLE, Rouen, France; DAVID DOSTER, Choate Rosemary
Hall, Wallingford, CT, USA; WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria;
MICHAEL LAMBROU, University of Crete, Crete, Greece; HEINZ-JURGEN  SEIFFERT, Berlin,
Germany; ACHILLEAS SINEFAKOPOULOS, student, University of Athens, Greece; SOUTHWEST
MISSOURI STATE UNIVERSITY PROBLEM SOLVING GROUP; KENNETH M. WILKE, Topeka,
KS, USA; and the proposer.
Sei ert observes that the problem can be generalized to the Jordan totient function de-
ned as Y
Js (n) = ns (1 ; p;s )
p jn
p prime
(note that J1 = .) In this case the new problem would be to evaluate:
X s X Js (k)Js (d=k)
d ks .
djn kjd
Sei ert shows that the value must be n2s using the fact that
X J (d) = ns .
s
djn

2481. [1999 : 430] Proposed by Mih a ly Bencze, Brasov, Romania.

Suppose that A, B , C are 2  2 commutative matrices. Prove that
det ;(A + B + C )(A + B + C ; 3ABC )
 0 .
3 3 3

I. Solution by Michael Lambrou, University of Crete, Crete, Greece;

and by Kee-Wai Lau, Hong Kong (independently).
 505

We show that the inequality is in fact true for n  n real commutative

matrices A, B , C .
Let P and Q be any two real n  n commutative matrices.
We have
det ;P + Q
= det ;(P + iQ)(P ; iQ)

2 2

= det(P + iQ) det(P ; iQ) = det(P + iQ)det(P + iQ)

= jdet(P + iQ)j  0 . 2

p
Putting P = (A ; C ) and Q = (A ; 2B + C ), we have
2
3 1
2

det(A + B + C ; AB ; BC ; CA) = det(P + Q )  0 .

2 2 2 2 2

Hence
;
det (A + B + C )(A + B + C ; 3ABC )
3 3 3


= det ;(A + B + C )
det(A + B + C ; AB ; BC ; CA)
2 2 2 2

= ; det(A + B + C )
det(A + B + C ; AB ; BC ; CA)  0

2 2 2 2

as claimed.
II. Solution by Heinz-Jurgen Sei ert, Berlin, Germany.
We prove that the inequality holds for real commutative n  n matrices
A, B, C .
p p
Let D = ( 3 + i)(A ; B ) + ( 3 ; i)(B ; C )
p p
and E = ( 3 ; i)(A ; B ) + ( 3 + i)(B ; C ) .
Then it is easily veri ed that
DE = 4(A + B + C ; AB ; BC ; CA) .
2 2 2

Note that E = D, where D denotes the n  n matrix obtained from D

when every entry is replaced by its conjugate.
Hence det(DE ) = det(D) det(D) = det(D) det(D) = jdet(D)j  0. 2

Therefore,
det ;(A + B + C )(A + B + C ; 3ABC )

3 3 3

;

= det (A + B + C ) det(A + B + C ; AB ; BC ; CA)
2 2 2 2

;
;
= det(A + B + C ) det DE

2 1

=
;
n ; det(A + B + C )
det(DE)  0 .
1 2
4

Also solved by MICHEL BATAILLE, Rouen, France; NIKOLAOS DERGIADES, Thessa-

loniki, Greece; and WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria.
506

Bataille, Lambrou and Sei ert all pointed out that the given inequality need not hold if
A, B , C are not necessarily real matrices. Both Lambrou and Sei ert gave the simple coun-
terexample: A = B = 0 and C = wIn where In denotes the n  n identity matrix and w is
a complex number such that w8 = ;1.

2482. [1999 : 430] Proposed by Mih a ly Bencze, Brasov, Romania.

Suppose that p, q , r are complex numbers. Prove that
jp + qj + jq + rj + jr + pj  jpj + jqj + jrj + jp + q + rj .
Editor's comment.
Most solvers, and others who submitted only comments (Mohammed Aassila,
Joe Howard, Walther Janous, Murray S. Klamkin, Heinz-Jurgen Sei ert) noted that
this is a version of Hlawka's Inequality, and that further generalizations of it may be
found in D. S. Mitrinovic, J. E. Pecaric and A. M. Fink, Classical and New Inequal-
ities in Analysis, Kluwer Academic Publishers, 1993, pp. 521-534 and 544-551; and
D. S. Mitrinovic, Analytic Inequalities, Springer-Verlag, Heidelberg, 1970, pp. 171-
173. Aassila commented that this problem is equivalent to one proposed by France,
but not used by the jury, at IMO 1987, and that Murray Klamkin's solution appeared
in [1989 : 102]. Bataille remarked that a generalization to n complex numbers is
given as Problem 1550 in Mathematics Magazine, Vol. 72, No. 3, June 1991, p. 239.
Woo gave a geometric argument. Several solvers raised the question of when equality
occurs, and in that regard, Romero refers us to p. 23 of L. Missotte, 1850 exercises de
mathematiques pour l'oral du CAPES de mathematiques et des concours des Grandes
Ecoles, Dunod University, Paris, 1978. Finally, Janous remarked that Prof. Hlawka is
still alive and, though now almost blind, is one of the still quite productive Nestors
[Ed. an elderly and distinguished wise person, a wise counsellor] of Austrian math-
ematics.
Solved by MICHEL BATAILLE, Rouen, France; G.P. HENDERSON, Garden Hill, Camp-
bellcroft, Ontario; MICHAEL LAMBROU, University of Crete, Crete, Greece; VEDULA
N. MURTY, Visakhapatnam, India; JUAN-BOSCO ROMERO M ARQUEZ,  Universidad de Val-
ladolid, Valladolid, Spain; ANDREI SIMION, student, Brooklyn Technical HS, Brooklyn, NY,
USA; PETER Y. WOO, Biola University, La Mirada, CA, USA; and the proposer. There was one
incorrect solution.

2483. [1999 : 430] Proposed by Vaclav Konecny, Ferris State Uni-

versity, Big Rapids, MI, USA.
Suppose that 0  A, B , C and A + B + C   . Show that
0  A ; sin A ; sin B ; sin C + sin(A + B ) + sin(A + C )   .
There are, of course, similar inequalities with the angles permuted cyclically.
[The proposer notes that this came up during an attempt to generalize
problem 2383.]
 507

Solution by Heinz-Jurgen Sei ert, Berlin, Germany.

Let S (A; B; C ) denote the middle term of the desired inequalities.
From the known trigonometric identities
   
sin x ; sin y = 2 cos x +2 y sin x ;2 y
and    
cos x + cos y = 2 cos x +2 y cos x ;2 y ,
it follows that
sin(A + B) ; sin A +CA) ; sin C  A + 2C   A 
 AB++2sin(
= 2 cos B sin + 2 cos sin
 A 2  A + 22B   A +22C  2
= 2 sin 2 cos + cos
 A + B + C  2  B ; C  2 A 
= 4 cos 2 cos 2 sin 2 :
Hence
     
S(A;B; C ) = A ; sin A + 4 cos A + B2 + C cos B ;2 C sin A2 .
Under the given conditions, we thus have
S(A; B; C )  A ; sin A  0
and
A A
S(A; B; C )  A ; sin A + 4 cos 2 sin 2
= A + sin A = A + sin( ; A)
 A + ( ; A) =  .
This proves the desired inequalities. From the proof, we see that there is
equality on the left hand side if and only if A = 0, and on the right hand side
only when A =  (and B = C = 0).

Also solved by SEFKET ARSLANAGI C,  University of Sarajevo, Sarajevo, Bosnia and
Herzegovina; NIKOLAOS DERGIADES, Thessaloniki, Greece; RICHARD I. HESS, Rancho Palos
Verdes, CA, USA; THOMAS JANG, Southwest Missouri State University, Spring eld, Missouri,
USA; WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria; KEE-WAI LAU, Hong Kong,
China; and the proposer. Another solver sent in a solution similar to Sei ert's, but containing
an error, small and correctible, but fatal nevertheless!
508

2484. [1999: 430] Proposed by Toshio Seimiya, Kawasaki, Japan.

Given a square ABCD, suppose that E is a point on AB produced be-
yond B , that F is a point on AD produced beyond D, and that EF = 2AB .
Let P and Q be the intersections of EF with BC and CD, respectively.
Prove that
(a) 4APQ is acute-angled;
(b) \PAQ  45.
Solution by Peter Y. Woo, Biola University, La Mirada, CA, USA.
Let M be the mid-point of EF . Then AM = ME = MF = AB .
Hence, M lies on the quadrant BMD of the circle with AB as radius and
A as centre and, moreover, that quadrant lies inside the square. EF must
either touch the quadrant at M or intersect it at two points M and N , both
being inside the square [since P and Q are outside the circle]. Therefore the
mid-point H of MN also lies inside the square. Then \APE > \AHE =
90 and \AQF > \AHF = 90 . Hence, 4APQ is acute-angled.
Let P 0 Q0 be the tangent to the quadrant BMD; that is, parallel to PQ,
cutting BC at P 0 and CD at Q0 , and touching the quadrant at H 0 . Then H
lies on AH 0 and therefore AH  AB . Hence, P 0 has to be between C and
P while Q0 has to be between C and Q. Hence, \PAQ  \P 0AQ0 , which
proves the claim since \P 0 AQ0 = 90 =2 = 45 : AP 0 bisects \BAH 0 [since
P 0 is the intersection0 point of the tangents to the quadrant at B and at H 0 ]
while, similarly, AQ bisects \DAH 0 .

Also solved by SEFKET  University of Sarajevo, Sarajevo, Bosnia and
ARSLANAGI C,
Herzegovina (2 solutions); CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; NIKO-
LAOS DERGIADES, Thessaloniki, Greece; RICHARD I. HESS, Rancho Palos Verdes, CA, USA;
WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria; GERRY LEVERSHA, St. Paul's
School, London, England; MICHAEL PARMENTER, Memorial University of Newfoundland,
St. John's, Newfoundland; D.J. SMEENK, Zaltbommel, the Netherlands; G. TSINTSIFAS, Thes-
saloniki, Greece; and the proposer.

2485. [1999: 431] Proposed by Toshio Seimiya, Kawasaki, Japan.

ABCD is a convex quadrilateral with AB = BC = CD. Let P be
the intersection of the diagonals AC and BD. Suppose that
AP : BD = DP : AC .
Prove that either BC k AD or AB ? CD.
Solution by D.J. Smeenk, Zaltbommel, the Netherlands.
Let AB = BC = CD = x, \BAC = and \CDB = .
Then \BCA = , \CBD = , \APB = \DPC = + ,
\ABP = 180 ; 2 ; and \DCP = 180 ; 2 ; . Also, from tri-
angles CAB and DBC , we obtain AC = 2x cos and BD = 2x cos ,
respectively. Applying the Sine Law to triangles ABP and DCP ,
sin(2 + ) and DP = x sin(2 + ) .
AP = x sin( + ) sin( + )
 509

AP = DP now implies
The condition BD AC
sin(2 + ) = sin(sin(2+ +) cos) .
sin( + ) cos
This gives
sin(2 + ) cos = sin(2 + ) cos ,
or,
sin(3 + ) = sin(3 + ) .
There are two possibilities:
either 3 + = 3 + , or (3 + ) + (3 + ) = 180.
(1) If 3 + = 3 + , then = , so that ABCD can be inscribed in a
circle. Since AB = CD, it follows that BC k AD.
C B
P

D A
(2) If (3 + ) + (3 + ) = 180 , then + = 45 . Let E be the
intersection point of AB and CD.
Then \CBE = \BCA + \BAC = 2 . Similarly, \BCE = 2 .
Therefore, \BEC = 180 ; 2 ; 2 = 90 , which shows that
AB ? CD.
E

C 2 2 B
P
A
D

Also solved by SEFKET  University of Sarajevo, Sarajevo, Bosnia and
ARSLANAGI C,
Herzegovina; MICHEL BATAILLE, Rouen, France; CHRISTOPHER J. BRADLEY, Clifton Col-
lege, Bristol, UK; NIKOLAOS DERGIADES, Thessaloniki, Greece; WALTHER JANOUS, Ursu-
linengymnasium, Innsbruck, Austria; GEOFFREY A. KANDALL, Hamden, CT, USA; HEINZ-

JURGEN SEIFFERT, Berlin, Germany; and the proposer. There was also one incorrect solu-
tion submitted.
510

2486. [1999: 431] Proposed by Joe Howard, New Mexico Highlands

University, Las Vegas, NM, USA.
It is well known that cos(20) cos(40) cos(80) = . 1
8
p
Show that sin(20 ) sin(40) sin(80 ) = . 8
3

Solution by Nikolaos Dergiades, Thessaloniki, Greece.

We have
4 sin 20 sin 40 sin80 = 2[cos 20 ; cos 60] sin 80
= 2 cos 20 sin80 ; sin80
  
= p3100 + sin60 ; sin100
sin
= 2 ,
p3
and hence, sin 20 sin 40 sin80 = 8
.

Also solved by SEFKET ARSLANAGI C,  University of Sarajevo, Sarajevo, Bosnia and
Herzegovina (5 solutions); MICHEL BATAILLE, Rouen, France; FRANK BATTLES, Massa-
chusetts Maritime Academy, MA, USA; SOUMYA KANTI DAS BHAUMIK, student, Angelo State
University, TX, USA; CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; JONATHAN
CAMPBELL, student, Chapel Hill High School, NC, USA; JENN CARRUTHERS, Burlington, On-
tario; DAVID DOSTER, Choate Rosemary Hall, Wallingford, CT, USA; IAN JUNE L. GARCES,
Ateneo de Manila University, The Philippines; DOUGLASS L. GRANT, University College
of Cape Breton, Sydney, Nova Scotia; RICHARD I. HESS, Rancho Palos Verdes, CA, USA;
JOHN G. HEUVER, Grande Prairie Composite High School, Grande Prairie, Alberta; PETER
HURTHIG, Columbia College, Vancouver, BC (2 solutions); WALTHER JANOUS, Ursulinen-

gymnasium, Innsbruck, Austria; V ACLAV KONECN  Y, Ferris State University, Big Rapids, MI,
USA; MICHAEL LAMBROU, University of Crete, Crete, Greece; R. LAUMEN, Antwerp, Belgium
(2 solutions); GERRY LEVERSHA, St. Paul's School, London, England; JUAN-BOSCO ROMERO

M ARQUEZ, 
Universidad de Valladolid, Valladolid, Spain; HEINZ-JURGEN SEIFFERT, Berlin,
Germany; TOSHIO SEIMIYA, Kawasaki, Japan; ANDREI SIMION, student, Brooklyn Tech-
nical High School, NY, USA; ACHILLEAS SINEFAKOPOULOS, student, University of Athens,
Greece; D.J. SMEENK, Zaltbommel, the Netherlands; DIGBY SMITH, Mount Royal College,
Calgary, Alberta; CHOONGYUP SUNG, Pusan Science High School, Pusan, Korea; PANOS E.
TSAOUSSOGLOU, Athens, Greece; EDWARD T.H. WANG, Wilfrid Laurier University, Waterloo,
Ontario; KENNETH M. WILKE, Topeka, KS, USA; PETER Y. WOO, Biola University, La Mi-
rada, CA, USA; JEREMY YOUNG, student, University of Cambridge, Cambridge, UK; and the
proposer.
Several solvers noted that generalizations of this problem or the problem itself
can be found in various sources, for example, Sinefakopoulos refers to D.O. Shklarsky,
N.N. Chentzov and I.M. Yaglom, The USSR Olympiad Problem Book, Dover, NY, 1993, p. 35,
problem 232, (a). Wilke mentioned the equality
(n;1)=2
Y 2 sin k = pn ,
k=1 n
found in T. Nagell, Number Theory, Chelsea, NY, 1964, p.173.
 511

2487. [1999 : 431] Proposed by Jose Luis Daz, Universitat Politecnica

de Catalunya, Terrassa, Spain.
If a, b, c, d are distinct real numbers, prove that
a +14
b +1
4

(a ; b)(a ; c)(a ; d) + (b ; a)(b ; c)(b ; d)

+ (c ; a)(cc ;+b1)(c ; d) + (d ; a)(dd ;+ b1)(d ; c) = a + b + c + d .
4 4

I. Composite solution by Vedula N. Murty, Visakhapatnam, India and

Peter Y. Woo, Biola University, La Mirada, CA, USA.
The left hand side of the given identity can be written as N=D where
D = (a ; b)(a ; c)(a ; d)(b ; c)(b ; d)(c ; d) and
N = (a + 1)(b ; c)(b ; d)(c ; d) ; (b + 1)(a ; c)(a ; d)(c ; d)
4 4

+ (c + 1)(a ; b)(a ; d)(b ; d) ; (d + 1)(a ; b)(a ; c)(b ; c) .

4 4

By the Factor Theorem, it is readily veri ed that each factor of D is also

a factor of N .
Since both D and N are antisymmetric polynomials of degrees six and
seven, respectively, the quotient N=D must be a symmetric polynomial in
a, b, c and d of degree one.
Hence N = k(a + b + c + d)(a ; b)(a ; c)(a ; d)(b ; c)(b ; d)(c ; d)
for some constant k.
Comparing the coecient of a we see that k = 1 and the conclusion
4

follows.
II. Solution by Mangho Ahuja, Southeast Missouri State University,
Cape Girardearu, MO, USA.
Let A, B , C and D denote the four terms of the left hand side of the
identity to be proved. We need to show that
A + B + C + D = a + b+ c + d.
Consider F (x) =
x + 1 ; (x ; a)(x ; b)(x ; c)(x ; d) .
4

(x ; a)(x ; b)(x ; c)(x ; d)

Using the method of partial fractions we easily nd that
X a +1 1 = X A .


4
F (x) = (a ; b)(a ; c)(a ; d) x ; a x;a
cyclic cyclic

Multiplying the equation above by (x ; a)(x ; b)(x ; c)(x ; d) we get

X
(x + 1) ; (x ; a)(x ; b)(x ; c)(x ; d) =
4
A(x ; b)(x ; c)(x ; d) .
cyclic
512

Comparing the coecient of x we get a + b + c + d = A + B + C + D,

3

as claimed.
III. Solutionand generalization by Michael Lambrou, University of Crete,
Crete, Greece (modi ed slightly by the editor).
We show more generally that if a , a , : : : , an are distinct complex
Xn k
Y ai
1 2

numbers, n  2 and if Sn(k) = , then

i=1 (ai ; aj )
j 6=i
80 if 0  k  n ; 2 ,
>
>
<1 if k = n ; 1 ,
Sn(k) = > X
n
: ai if k = n .
>
i=1
X
4

Since the left side of the given identity is S (4)+ S (0), it equals ai
4 4
i
as claimed. For 0  k  n ; 2, using partial fractions, it is easy to see that
=1

xk X
n aki
x ;1 a .
+1 +1

Y
n = Y (1)
(x ; aj ) i j 6 i(ai ; aj )
=1 i
j =1 =

Letting x = 0, we get Sn (k) = 0 as claimed.

In particular, for k = n ; 2, (1) becomes
X n
n ; Y  x ; aj 
xn;1 = ai 1
. (2)
i j 6 i ai ; aj
=1 =

Comparing the coecients of xn; on both sides of (2) then yields

1

X n n;
Y ai
1

1 = = Sn (n ; 1) .
i =1(ai ; aj )
j 6=i
X
n
It remains to show that Sn (n) = ai .
i=1
By long division, it is easily seen that
0n 1
@ X aj A xn; + f (x) 1

xn = 1+ j Y , =1
(3)
Y
n n
(x ; aj ) (x ; aj )
j =1 j =1
 513

nX
;2
where f (x) = Akxk is a polynomial of degree at most n ; 2.
k=0
For each xed i = 1, 2, : : : , n, multiplying both sides of (3) by x ; ai
and letting x = ai we get:
0n 1
@ X aj A ani ; + f (ai) 1

n
Y ai = j Y =1
. (4)
(ai ; aj ) (ai ; aj )
j 6=i j 6=i
Adding up (4) for i = 1, 2, : : : , n, and using the facts that
Sn(n ; 1) = 1 , Sn(n ; 2) = : : : = Sn(1) = Sn(0) = 0 ,
we then obtain
0n 1
X
Sn(n) = @ aj A Sn(n ; 1) + An; Sn(n ; 2) 2
j =1
+ : : : + A Sn(1) + A Sn(0)
1 0

X
n
= aj ,
j =1
as claimed.

Also solved by SEFKET ARSLANAGI C,  University of Sarajevo, Sarajevo, Bosnia and
Herzegovina; MICHEL BATAILLE, Rouen, France; CHRISTOPHER J. BRADLEY, Clifton College,
Bristol, UK; JAMES T. BRUENING, Southeast Missouri State University, Cape Girardeau, MO,

USA; OSCAR CIAURRI, Universidad La Rioja, Logro~no, Spain; NIKOLAOS DERGIADES, Thessa-
loniki, Greece; H. N. GUPTA and J. CHRIS FISHER, University of Regina, Regina, Saskatchewan;
RICHARD I. HESS, Rancho Palos Verdes, CA, USA; WALTHER JANOUS, Ursulinengymnasium,
Innsbruck, Austria; R. LAUMEN, Antwerp, Belgium; JUAN-BOSCO ROMERO M ARQUEZ,  Uni-

versidad de Valladolid, Valladolid, Spain; HEINZ-JURGEN SEIFFERT, Berlin, Germany; TOSHIO
SEIMIYA, Kawasaki, Japan; ANDREI SIMION, student, Brooklyn Technical HS, Brooklyn, NY,
USA; DIGBY SMITH, Mount Royal College, Calgary, Alberta; PANOS E. TSAOUSSOGLOU,
Athens, Greece; JEREMY YOUNG, student, Nottingham High School, Nottingham, UK; and
the proposer.
Gupta and Fisher showed that the result holds if the 1's in the denominators are re-
placed by a constant k. Using Lagrange's Interpolation !Theorem, Sei ert also obtained the
Xn
more general result that Sn (n) + cSn (0) = ai where c is any complex constant.
i=1
See notations in III above. The special case of this when c = 1 was also obtained by
Ciaurri by using the theory of residue in complex analysis. Of course, all of these general-
izations are contained in Solution III above. The most general result was given by Janous
who, also using P the theory of residue, showed that if m denotes any non-negative integer,
then Sn (m) = a1 1 a2 2 : : : ann where the summation is over all non-negative integers
1 ; 2; : : : n such that 1 + 2 +


+ n = m ; n + 1. It is easily seen that
this result implies the one given in III above. Laumen pointed out that in the article Shreds and
Slices: Cyclic Sums the Easy Way which appeared in Mathematical Mayhem (Vol. 8, issue 4, p. 3)
514

X a4
[Ed: Written by Naoki Sato], it was proved that (a ; b)(a ; c)(a ; d) = a + b + c + d.
cyclic
X 1
Hence to prove the given identity, it suces to show that
cyclic
( a ; b )( a ; c)(a ; d) = 0
which can be easily veri ed by straightforward computations.
Murray S. Klamkin, University of Alberta, Edmonton, Alberta commented that the iden-
tity is a special case of some more general and known results about cyclic sums. Using the
notation of Solution III above, this result, which can be found in A Treatise on the Theory of
Determinants by T. Muir (Dover, NY, 1960, pp. 329{331), essentially states that Sn (k) = 0 if
0  k < n ; 1 and equals the complete symmetric function of the ai 's of degree k ; n + 1 if
k  n ; 1. (For example, when k = n + 1, the sum is P a2i + P ai aj .) For k > n, this
n
i=1 i6=j
extends the results obtained by Lambrou in Solution III above.

2488. [1999 : 431] Proposed by G. Tsintsifas,

n
Thessaloniki, Greece.
Let Sn = A A : : : An be a simplex in E , and M a point in Sn .
It is known that there are real positive numbers  ,  , : : : , n such
1 2 +1

1 2 +1
nX
+1 nX
+1

that j = 1 and M = j Aj (here, by a point P, we mean the

j =1 j
;! =1

position vector OP). Suppose also that    


 n , and let

1 2 +1
X k
Bk = k1 Aj .
j =1

Prove that
M 2 convex cover of fB , B , : : : , Bn g ;1 2 +1

that is, there are real positive numbers  ,  , : : : , n such that

1 2 +1

nX +1

M = k Bk .
k=1
nP
k = 1.
+1
Note the necessary condition for a convex cover:
k=1
Solution by Walther Janous, Ursulinengymnasium, Innsbruck, Austria.
Comment. The condition given in the original problem was \ 
 


 n". This has been corrected to \   


 n ".
1

Furthermore,  ;  ; : : : ; n being \real positive numbers" has to be

2 1 2 +1

1 2 +1
weakened to \real non-negative numbers" as the following example shows.
Let n = 2 and  =  = 1=2. Then M = (A + A )=2, B = A and
B = (A + A )=2 = M , whence M = 0
B + 1
B !
1 2 1 2 1 1

2 1 2 1 2

[Ed. The editor, not the proposer, was responsible for these oversights.]
 515

Now, from Bk = (A +


+ Ak )=k, M =  A +


+ n An ,

and the required M =  B +


+ n Bn , we infer that
1 1 1 +1 +1

1 1 +1 +1

   
M =  + 2 +


+ nn+ 1 A + 2 +


+ nn+ 1 A
1
2 +1
1
2 +1
2

   
+


+ nn + nn+ 1 An + nn+ 1 An . +1 +1
+1

Whence, \looking from the back", we get:

n = (n + 1)n ,
+1 +1

n = n(n ; n ) , +1

n; = (n ; 1)(n; ; n) ,

1 1

..
.
 = 2( ;  ) , and nally
2 2 3

 =  ; .
1 1 2

Of course,
 +


+ n
1 +1 = ( ;  ) + (2 ; 2 ) +



1 2 2 3

+ (nn ; nn ) + (n + 1)n +1 +1

=  +


+ n = 1 .
1 +1

Also solved by CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; JOSE LUIS
DIAZ, Universitat Polite cnica de Catalunya, Terrassa, Spain; and the proposer.
All solvers found Janous's two corrections to the statement of the problem, as well as
the correction mentioned on [2000 : 179].

2489. [1999: 505] Proposed by Joaqun Gomez  Rey, IES Luis Bu~nuel,
Alcorcon, Spain.
The set of twelve vertices of a regular icosahedron can be partitioned
into three sets of four vertices, each being such that none of the sets have
their four vertices forming a golden rectangle. In how many di erent ways
can this be done?
Solution by Manuel Benito and Emilio Fernandez, I.B. Praxedes Mateo
Sagasta, Logro~no, Spain.
Let  be the positive solution of the equation  = 1 +  ; that is,
p
2

= 5+1
. Golden rectangles are those similar to one with side lengths 1
and  .
2

One regular icosahedron inscribed into the edges of the regular octahe-
dron of vertices (0; 0;  ), (0;  ; 0), ( ; 0; 0) (and that divides those
2 2 2
516

sections into golden sections), has the following twelve vertices (following
H.S.M. Coxeter, Fundamentos de Geometria, Spanish version of the Intro-
duction to Geometry, Limusa, 1988, pp. 193{195; see also, by the same au-
thor, Regular Polytopes, Dover, 1973, pp. 50{53):
1 (0;; 1) 1 (0; ;; ;1)
2 (1; 0;  ) 2 (;1; 0; ; )
3 (; 1; 0) 3 (;; ;1; 0)
4 (;; 1; 0) 4 (; ;1; 0)
5 (;1; 0;  ) 5 (1; 0; ; )
6 (0; ;; 1) 6 (0; ; ;1)
(Point n is the point opposite to n.)
Let us space out these twelve vertices on 4 layers, according to Euclid xiii, 16:
 Layer 1. { The point 1.
 Layer 2. { Points 2, 3, 4, 5, 6. (These are the vertices of a regular plane
pentagon.)
 Layer 3. { Points 2, 3, 4, 5, 6. (These form, with the vertices of layer 2,
a pentagonal antiprism.)
 Layer 4. { The point 1.

The distances from point 1 to any of the vertices of layer 2 are all equal to 2
(edges of the regular icosahedron), since, for example, 12 = 1 +  +
2 2

(1 ;  ) = 4.
2

The distances from point 1 to any of the vertices of layer 3 are all equal
to 2 , since, for example, 12 = 1 +  + (1 +  ) = 4 .
2 2 2 2

Further 11 = 4 + 4 = 4 + 8, so that the distance 11 is equal to

2p2 +  .
2 2

Therefore, the unique golden rectangles formed by four vertices of the

above regular icosahedron are built with two sides of length 2 (two edges)
and two sides of length 2 . But any given edge, say the 12 edge, is only
parallel to its opposite edge (the edge 12), so that the only golden rectangles
formed by four vertices of any regular icosahedron are the fteen rectangles
formed from two opposite edges of the polyhedron. For our icosahedron,
these are the following (the sequence of vertices listed is not necessarily the
geometric one):
1212 2323 3434 4545 5656
1313 2424 3535 4646
1414 2525 3636
1515 2626
1616
And nally we count the number of partitions of the twelve vertices of
the polyhedron into three sets of vertices:
 517

 Total number of such partitions:

1
12
8
4 = 5775 .
3! 4 4 4
 Partitions with three golden rectangles (Number of partitions for which
all three sets of vertices form golden rectangles):
1
15
6
1 = 15 .
3!
 Partitions with exactly two golden rectangles: 0 (there are none.)
 Partitions with exactly one golden rectangle:
1
15
8 ; 6
1 = 160 ,
3! 4
where the factor 1=3! in each expression is present because the order of
choosing the three sets is irrelevant. Thus, our answer to the proposed ques-
tion is : 5775 ; 15 ; 160 = 5600 di erent partitions.
There was one incorrect solution.

2490. [1999 : 505] Proposed by Mih a ly Bencze, Brasov, Romania.

Let > 1. Denote by xn the only positive root of the equation:
(x + n )(2x + n )(3x + n )


(nx + n ) = n n .
2 2 2 2 2

!1 xn .
Find nlim
Solution by Kee-Wai Lau,; Hong
Kong,
; China.

;

For x  0 let f (x) = 1 + nx2 1 + nx2


1 + nx
2
n2 ; . Then the
given equation is equivalent to f (x) = 0.
Since f is strictly increasing for x > 0, f (0) < 0 and f (x) ! 1 as
x ! 1, the equation f (x) = 0 has a unique positive root xn.
X
n kxn (n + 1)xn xn
We note that = f (xn) + >
k n = 2n > 2 , and so,
2

we have that xn < 2 .

=1

It is well known that

x ; x2 < ln(1 + x) < x for 0 < x < 1 .
2
(1)
kx
If n > 2 , then xn < n, and so, we have that 0 < n < 1 for all
k = 1, 2, : : : , n. n 2
518

kx
Letting x = n in (1), and summing the inequalities from k = 1 to
k = n, we obtain n
2

1 X n  kxn  X n  
2
kx n
 kxn 
;2 n < ln 1 + n ; n < 0 ,
k
2
k
2 2
=1 =1

or
; (n + 1)(2n + 1)xn < ln ; (n + 1)xn < 0 ,
2

12n 3 2n
whenever n > 2 .
Letting n ! 1, we then have
 x n

!1 ln ; 2  0 ,
0  nlim

!1 xn = 2 ln .
from which it follows immediately that nlim
Also solved by MICHEL BATAILLE, Rouen, France; MANUEL BENITO and EMILIO
FERNANDEZ, I.B. Praxedes Mateo Sagasta, Logro~no, Spain; PAUL BRACKEN, CRM, Univer-
site de Montreal, Montreal, Quebec; NIKOLAOS DERGIADES, Thessaloniki, Greece; RICHARD
I. HESS, Rancho Palos Verdes, CA, USA; WALTHER JANOUS, Ursulinengymnasium, Inns-

bruck, Austria; V ACLAV KONECN Y, Ferris State University, Big Rapids, MI, USA; MICHAEL
LAMBROU, University of Crete, Crete, Greece; HEINZ-JURGEN  SEIFFERT, Berlin, Germany;
and the proposer.
Most of the other submitted solutions are far more complicated than the one given
above. Among the methods used were: di erentiation, integration, Intermediate Value
Theorem, Mean Value Theorem, AM{GM Inequality, Jensen's Inequality, Weierstrass'
Inequality, (generalized) Bernoulli's Inequality, Bolzano's Theorem, uniform boundedness, and
majorization!
Konecny gave a one-line \proof" based on the \fact" that
n  kx 
Y x=2
n!1 k=1 1 + n2 = e
lim
which he believed \must be well known", but could not nd a reference. Neither could this
editor. Can any reader supply a reference or a proof of this?

2491. [1999: 505] Proposed by Mih a ly Bencze, Brasov, Romania.

Suppose that f : R ! R is a continuous function and that fak gk and
n
fbkgnk are two geometric sequences for which
=1
=1

Xn Xn
f (ak) < 0 < f (bk) .
k=1 k=1
Prove that there exists a geometric sequence fck gnk for which =1

X
n
f (ck) = 0 .
k=1
 519

Compilation of essentially identical solutions by Manuel Benito and

Emilio Fernandez, I.B. Praxedes Mateo Sagasta, Logro~no, Spain; Jonathan
Campbell, student, Chapel Hill High School, Chapel Hill, NC, USA; Michael
Lambrou, University of Crete, Crete, Greece; and Kee-Wai Lau, Hong Kong,
China. Pn f (xyk). This is a continuous
De ne g : R ! R by g (x;y ) =
2

k
function. Thus, there exist (x ; y ), (x ; y ) 2 R satisfying
=1
2

g(x ; y ) < 0 and g(x ; y ) > 0. By the Intermediate Value Theorem,

1 1 2 2

there exists (p ; q ) 2 R satisfying g (p ; q ) = 0. Take ck = p q k;

1 1 2 2
2 1

for k = 1, 2, : : : , n, and we are done.

1 1 1 1 1 1

Also solved by MICHEL BATAILLE, Rouen, France; NIKOLAOS DERGIADES, Thessa-

loniki, Greece; KEITH EKBLAW, Walla Walla, WA, USA; WALTHER JANOUS, Ursulinengymnas-

ium, Innsbruck, Austria; RICHARD I. HESS, Rancho Palos Verdes, CA, USA; HEINZ-JURGEN
SEIFFERT, Berlin, Germany; and the proposer. There was one incomplete solution.

2492. [1999: 506] Proposed by Toshio Seimiya, Kawasaki, Japan.

In 4ABC , suppose that \BAC is a right angle. Let I be the incentre
of 4ABC , and suppose that D and E are the intersections of BI and CI
with AC and AB , respectively. Let points P and Q be on BC such that
IP kAB and IQkAC .
Prove that BE + CD = 2PQ.
I. Solution by Walther Janous, Ursulinengymnasium, Innsbruck, Aus-
tria.
The converse is also true. More precisely, we see that
for any 4ABC , BE + CD = 2PQ if and only if A is a right angle:
Since 4IPQ  4ABC and the altitude to side PQ of 4IPQ is the inradius
of 4ABC , PQ : a = r : ha . Using [ABC ] = aha = r(a + b + c), we
deduce that PQ=a = a=(a + b + c), so that
PQ = a +ab + c .
2

Since BE : EA = a : b, we get
BE = a ac
+b.
Similarly,
CD = a ab
+c .
Therefore, the following six statements are equivalent:
BE + CD = 2PQ ,
520

ac + ab = 2a , 2

a+b a+c a+b+c

c + b ; 2a = 0 ,
a+b a+c a+b+c
(2a + b + c)(b + c ; a ) = 0 ,
2 2 2

(a + b)(a + c)(a + b + c)
b +c ; a = 0,
2 2 2

4ABC has a right angle at A .

II. Solution by Gottfried Perz, Pestalozzigymnasium, Graz, Austria.
Re ect D in CI giving D0 , and re ect E in BI giving E 0 . Since
\BIC = 180 ; \B +2 \C = 180 ; 45 = 135 ,
we have
\BIE = \CID = 45 , \EIE 0 = \DID0 = 90 .
Furthermore, since IP jjAB , we have \BIP = \IBE = \PBI , which
means that triangle IBP is isosceles, and consequently, that P , as the in-
tersection of BC with the perpendicular bisector of BI , is the circumcentre
of the right triangle IBD0 . From this, it follows that
PB = PI = PD0 .
Analogously, Q is the circumcentre of the right triangle CIE 0 , so that
QC = QI = QE0 .
Finally, this means that
BE + CD = BE0 + CD0 = (BQ ; QE0) + (CP ; PD0)
= (BQ ; PD0) + (CP ; QE0 )
= (BQ ; PB ) + (CP ; CQ)
= 2PQ .
Also solved by MIGUEL AMENGUAL COVAS, Cala Figuera, Mallorca, Spain; SEFKET 
ARSLANAGI C,  University of Sarajevo, Sarajevo, Bosnia and Herzegovina; MICHEL BATAILLE,
Rouen, France; MANUEL BENITO and EMILIO FERNANDEZ, I.B. Praxedes Mateo Sagasta,
Logro~no, Spain; CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; NIKOLAOS
DERGIADES, Thessaloniki, Greece; DAVID DOSTER, Choate Rosemary Hall, Wallingford, CT,
USA; RICHARD I. HESS, Rancho Palos Verdes, CA, USA; JOHN G. HEUVER, Grande Prairie

Composite High School, Grande Prairie, Alberta; V ACLAV KONECN Y,
 Ferris State University,
Big Rapids, MI, USA; HO-JOO LEE, student, Kwangwoon University, Kangwon-Do, South Ko-
rea; HENRY J. PAN, student, East York Collegiate Institute, Toronto, Ontario; HEINZ-JUR- 
GEN SEIFFERT, Berlin, Germany; D.J. SMEENK, Zaltbommel, the Netherlands; ANDRE LOUIS
SOUZA DE ARAUJO, Rio de Janeiro; ECKARD SPECHT, Otto-von-Guericke University, Magde-
burg, Germany; CHOONGYUP SUNG, Pusan, Korea; ALBERT WHITE, Bonaventure, NY; PETER
Y. WOO, Biola University, La Mirada, CA, USA; JEREMY YOUNG, student, University of Cam-
bridge, Cambridge, UK; and the proposer.
Only Sei ert and Janous mentioned that the converse also holds.
 521

2493. [1999: 506] Proposed by Toshio Seimiya, Kawasaki, Japan.

Suppose that ABCD is a convex cyclic quadrilateral, that
\ACB = 2\CAD, and that \ACD = 2\BAC .
Prove that BC + CD = AC .
Solution by Jeremy Young, student, University of Cambridge, Cam-
bridge, UK.

2 B

2
2
D C
Let \CAD = and \BAC = . Then \ACB = 2 and
\ACD = 2 . Since the quadrilateral ABCD is cyclic, we have that
\DAB + \DCB = 180 , which gives ( + ) + (2 + 2 ) = 180; that
is, + = 60 . Let R be the circumradius of the quadrilateral ABCD. By
the Sine Rule applied to 4ABC , BC = 2R sin . Similarly, from 4ADC ,
CD = 2R sin . Then

BC + CD = 2R sin + 2 R sin = 2 R ; sin + sin(60 ; )

 p 
= 2R sin + cos ; sin 3 1

 p  2 2

= 2R sin + cos 1
2 2
3

= 2R sin(60 + ) = 2R sin( + 2 )
= 2R sin \ABC = AC ,

as desired.
Also solved by MIGUEL AMENGUAL COVAS, Cala Figuera, Mallorca, Spain; SEFKET 
ARSLANAGI C , University of Sarajevo, Sarajevo, Bosnia and Herzegovina; MICHEL BATAILLE,
Rouen, France; MANUEL BENITO and EMILIO FERNANDEZ, I.B. Praxedes Mateo Sagasta,
Logro~no, Spain; CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; NIKOLAOS DER-
GIADES, Thessaloniki, Greece; DAVID DOSTER, Choate Rosemary Hall, Wallingford, CT, USA;
RICHARD I. HESS, Rancho Palos Verdes, CA, USA; JOHN G. HEUVER, Grande Prairie Com-
posite High School, Grande Prairie, Alberta; WALTHER JANOUS, Ursulinengymnasium, Inns-

bruck, Austria; V ACLAV KONECN  Y,
 Ferris State University, Big Rapids, MI, USA; KEE-WAI
LAU, Hong Kong; HENRY J. PAN, student, East York Collegiate Institute, Toronto, On-
tario; MICHAEL PARMENTER, Memorial University of Newfoundland, St. John's, Newfound-
land; GOTTFRIED PERZ, Pestalozzigymnasium, Graz, Austria(two solutions); HEINZ-JURGEN
SEIFFERT, Berlin, Germany; ACHILLEAS SINEFAKOPOULOS, student, University of Athens,
522

Greece; D.J. SMEENK, Zaltbommel, the Netherlands; ANDRE LOUIS SOUZA de ARAUJO,
 Insti-
tuto Militar de Engenharia, Brazil; ECKARD SPECHT, Otto-von-Guericke University, Magde-
burg, Germany; CHOONGYUP SUNG, Pusan Science High School, Pusan, Korea; PETER
Y. WOO, Biola University, La Mirada, CA, USA; and the proposer.
Most of the submitted solutions are similar to the one given above.

2494. [1999 : 506] Proposed by Toshio Seimiya, Kawasaki, Japan.

Given 4ABC with AB < AC , let I be the incentre and M be the
mid-point of BC . The line MI meets AB and AC at P and Q, respectively.
A tangent to the incircle meets sides AB and AC at D and E , respectively.
AP AQ PQ
Prove that
BD + CE = 2MI .
Solution by the proposer.
Let L and N be the mid-points of BE and CD, respectively. Since M
is the mid-point of BC , we have MLjjCE , ML = CE , and MN jjBD,
1

MN = BD. We put \APQ =  and \AQP = . Since MN jjPA, we

2
1

get \NMI = \APQ = . Similarly, we have \LMI = \AQP = . Since

2

quadrilateral BCED has an incircle with centre I , we have that L, N and

I are collinear by Newton's Theorem. By the Law of Sines for 4APQ, we
have
AP = AQ = PQ . (1)
sin  sin  sin( + )
[Editor's comment. For the next step we require that I lies between L and
N : because D and E are on the sides AB and AC , L and N are on the sides
of the mid-point triangle of 4ABC , while I is always inside that triangle.]
Since I is between L and M , we have [LMN ] = [LMI ] + [NMI ],
(where [XY Z ] denotes the area of triangle XY Z ). Therefore,
1 MN
ML sin( + ) = 1 ML
MI sin  + 1 MN
MI sin  .
2 2 2
Dividing both sides by ML
MI
MN , we get
sin( + ) = sin  + sin  . (2)
2MI 2MN 2ML
From (1) and (2), we obtain
PQ = AP + AQ .
2MI 2MN 2ML
Since 2MN = BD and 2ML = CE , we have
PQ = AP + AQ .
2MI BD CE
Also solved by MANUEL BENITO and EMILIO FERNANDEZ, I.B. Praxedes Mateo
Sagasta, Logro~no, Spain; and CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK.
 523

2495. [1999 : 506, 2000 : 238] Proposed by G. Tsintsifas, Thessa-

loniki, Greece.
Let P be the interior isodynamic point of 4ABC ; that is,
AP = BP = CP (a, b, c are the side lengths, BC , CA, AB, of 4ABC ).
bc ca ab p3
Prove that the pedal triangle of P has area F , where F is the area
2

p d 2

of 4ABC and d = 2
a +2
b + c
2
+ 2 3F .
2

2
I. Solution by D.J. Smeenk, Zaltbommel, the Netherlands.
Denote the projection of P onto BC , CA and AB by K , L and M ,
respectively. Let  > 0 be the common ratio that de nes P ; that is,
PA = bc , PB = ca , PC = ab .
Because BP is the diameter of the circumcircle of 4BKM , the Sine
Law applied to that triangle implies
KM = PB sin = ac sin = 2F (1)
(where F is the area of 4ABC ). In the same way LM = LK =p2F . Thus,
the pedal triangle KLM is equilateral and its area [KLM ] = ( 3=4)KM . 2

Moreover, 60 = \PKM + \PKL = \PBM + \PCL, so that

\BPC = 180 ; \PBC ; \PCB = 180 ; ( + ; 60 ) = + 60 .
De ne B 0 to be the point outside 4ABC for which 4ACB 0 is equilateral.
The Cosine Law for 4ABB 0 yields

BB0
= b + c ; 2bc cos( + 60p)
2 2 2

= b + c ; bc cos + 2bc( 3=2) sin

2 2

= b + c + a ; b2 ; c + 2 3F
2 2
p 2 2 2

2 p
= a + b2 + c + 2 3F
2 2

= d . 2

Since PB : PC = AB : AB 0 = c : b and \BPC = \BAB 0 ,

4PBC  4ABB0 and therefore, PB : a = c : d, so that PB = acd .
This together with (1) implies that
KM = PB sin = ac sin
d = 2F .
d
p3 ! p
F 3 , as desired. 2
Thus, [KLM ] = KM = 2
4 d 2
524

II. Solution by Murray S. Klamkin, University of Alberta, Edmonton,

Alberta.
The formula for the area [KLM ] of the pedal triangle of an isodynamic
point is derived in [2, p. 24] in the form
[KLM ] = 2(cot ! cot F
60  1) ,
where ! is the Brocard angle, \+" is used for our interior isodynamic point,
and \;" for the exterior isodynamic point. The desired form comes from
replacing cot ! by its equivalent [4, p. 266]
cot ! = a +4bF + c :
2 2 2

Also solved by MANUEL BENITO and EMILIO FERNANDEZ, I.B. Praxedes Mateo
Sagasta, Logro~no, Spain; CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; NIKOLAOS
DERGIADES, Thessaloniki, Greece; ECKARD SPECHT, Otto-von-Guericke University, Magde-
burg, Germany; PETER Y. WOO, Biola University, La Mirada, CA, USA; and the proposer.
This problem is connected with a variety of familiar results. An arbitrary point P inside
a triangle, whose distances from the vertices are x, y and z, determines a pedal triangle whose
sides are 2ax
R ; 2byR and 2czR [1, p. 23]. Thus, we deduce immediately that the pedal triangle in
this problem is equilateral. See also 2377 [1999 : 438], which deals speci cally with equilateral
pedal triangles. More formulas for the area of a pedal triangle can be found in the solutions to
1076 [1987 : 62-64]; or see [3, p. 235].
Here is one further connection. The three segments that join a vertex of 4ABC to the
remote vertex of an equilateral triangle erected externally on the opposite side all have length d
and pass through one isogonic centre | the point from which the angles subtended by the sides
are all 120 [4, pp. 218-221 and 295-296]. This point is the isogonal conjugate of our interior
isodynamic point; it is perhaps more familiar as the Fermat-Torricelli point: when no angle of
4ABC is as great as 120 , this point minimizes the sum x + y + z, thus resolving a problem
Fermat proposed to Torricelli.
References
[1] H.S.M. Coxeter and S.L. Greitzer, Geometry Revisited. Math. Assoc. of America New
Mathematical Library 19, 1967.
[2] W. Gallatly, The Modern Geometry of the Triangle. Hodgson, London, 1910.
[3] Ross Honsberger, From Erdos } to Kiev: Problems of Olympiad Caliber. (Dolciani Mathe-
matical Expositions, No. 17). Math. Assoc. of America, 1995.
[4] R.A. Johnson, Advanced Euclidean Geometry. Dover, N.Y., 1960.

2496. [1999 : 506] Proposed by Paul Yiu, Florida Atlantic University,

Boca Raton, FL, USA.
Given a triangle ABC , let CA be the circle tangent to the sides AB ,
AC and to the circumcircle internally. De ne CB and CC analogously. Find
the triangle, unique up to similarity, for which the inradius and the radii of
the three circles CA , CB and CC are in arithmetic progression.
 525

Solution by the proposer.

We denote by r, rA , rB and rC the radii of the incircle and the circles
CA , CB and CC , respectively. It is known [Leon Bankho , A Mixtilinear
Adventure, Crux Math. 9 (1983), 2{7] that
rA = r sec A = r ;1 + tan A
,
2 2

rB = r sec B = r ;1 + tan B
,
2 2
2 2
(1)
2
;
rC = r sec C = r 1 + tan C .
2

2
2

2 2

If we assume, without loss of generality, that r  rA  rB  rC , then the

condition that these radii are in arithmetic progression implies that
rA = r + d, rB = r + 2d and rC = r + 3d for some common di erence d.
Using (1), it follows that d = r tan A , 2d = r tan B and
2 2

3d = r tan C .
2 2
2
2

From this it follows that

p
tan B = 2 tan A and tan C = 3 tan A .
p (2)
2 2 2 2

If we let a, b and c be the sides opposite an- A

gles A, B and C , respectively, we know that
we can write a = x + y , b = x + z and z z
c = y + z, as shown in Athe diagram
r B
to the
r
right. We notice that tan = z , tan = y c b
C r
2 2
and tan = x . Using (2) p above,
p we
p easily
deduce that x : y : z = 2 : 3 : 6, and
2
y x
hence, it follows that the radii are in the de-
sired arithmetic progression if the three sides B y x C
of the triangle are in the ratio a
p p p p p p
a : b : c = 2 + 3 : 2+ 6 : 3+ 6. (3)
Furthermore, we know that tan A tan B + tan B tan C + tan C tan A = 1,
2 2
p p p ;
2 2 2 2

which, in combination with (2), gives us tan A = 2 + 3 + 6 .

1
2
2
This tells us that the angles in the triangle are
A = 2 tan; pp 1p p  45:83 ,
1

2 +p 3 + 6
B = 2 tan; pp p2 p  61:75 and
1

2 +p 3 + 6
C = 2 tan; pp p3 p  72:42 .
1

2+ 3+ 6
Also solved by MICHEL BATAILLE, Rouen, France; MANUEL BENITO and EMILIO
FERNANDEZ, I.B. Praxedes Mateo Sagasta, Logro~no, Spain; CHRISTOPHER J. BRADLEY,
526

Clifton College, Bristol, UK; WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria;


HEINZ-JURGEN SEIFFERT, Berlin, Germany; and PETER Y. WOO, Biola University, La Mi-
rada, CA, USA. RICHARD I. HESS, Rancho Palos Verdes, CA, USA replaced \circumcircle" in
the statement of the problem by \incircle", and gave a solution in this case.
All of the correct solutions that were submitted made use of the formula rA = r sec2 A2 ,
or of a variant of this. This formula was either assumed to be known (with suitable references
to the literature), or waspderivedas ppart ofthe solution. Sei
p ert used
  (1) pto deduce the an-
swer in the form b = 3 + 1 6 ; 2 a and c = 2 + 1 3 ; 6 a [which can
be re-scaled to givep (3)]. Woo used an inversion argument, by inverting in the circle centred
at A with radius bc. If this inversion takes B to B 0 and C to C 0 , then the circumcircle is
taken to the line B 0 C 0 and the circle CA is inverted into the excircle of 4ABC , which lies
opposite angle A. He then showed that rA = ra bc2 , where ra is the radius of this excir-
cle, and s is the semi-perimeter. Finally, letting K sbe the area of 4ABC , he showed that
rA = s2Kbc (s;a)
= s(rbc
s;a) and tan A2 = s;r a to deduce that s(sbc;a) = sec2 A2 , and hence,
A
that rA = r sec 2 . [Woo notes that he used the same argument to solve problem 2464.]
2

Bradley used a power-of-point argument to deduce that (rA ; r) csc2 A2 = rA , from which
rA = r sec2 A2 follows at once. Benito and E. Fernandez use elementary p pmeans to  show
p3 +that
rA = bcs ptanA2 . Theycomplete
p p  their proof by showing that cb = 36 2 ; 1 1
and ac = 33 2 ; 1
p 3 + 2 [which can be re-scaled to give (3)].

2497. [1999: 506] Proposed by Nikolaos Dergiades, Thessaloniki,

Greece.
Given 4ABC and a point D on AC , let \ABD =  and \DBC = .
Find all values of \BAC for which  > AD
DC .
Solution by the proposer.
First, suppose that \BAC =  . 2

Since
A
AD = AD
AB
DC AC ; AD
AB AB D
= tan( +tan ); tan  , 
B C
the given inequality can be written as
 > tan  (where 0 < +  <  ) .
tan( + ) ; tan  2

Using Lagranges's Mean Value Theorem, this is equivalent to

tan( +  ) ; tan  > tan  ; tan 0 ,
( + ) ;  ;0
or
d d tan t (0 <  <  <  < +  <  ) ;
dt tan t t=1
> dt
t=2
2 1 2
 527

that is,
1 > 1 , or cos  < cos  , which is true.
cos 
2
1 cos  2
2
1 2

Now suppose that \BAC >  . Let D0 and C 0 be points on AB such

that DD0 ? BA and CC 0 ? BA.
2

Then
CC 0 C0
tan( + ) = BC 0 (1)
D0
and A
D
1 = BD0 < BC 0 . (2)
tan  DD0 DD0 
B C
By multiplying (1) and (2) we get
tan( + ) < CC 0 = AC since 4ADD0  4ACC 0 .
tan  DD0 AD
This gives, equivalently,
tan( + ) ; tan  < AC ; AD = DC ,
tan  AD AD
or
tan  > AD .
tan( + ) ; tan  DC
Thus, to prove that >
 AD , it is sucient to prove that, for 0 < +  <  ,
DC 2

 > tan 
tan( +  ) ; tan  .
But this inequality has been proved previously, and so, the given inequality
is also valid for every \BAC >  . [Ed. remember \BAC <  .]
2

Lastly, suppose that \BAC <  . Then the inequality is not valid. To
show this, it is sucient to create a counterexample for such an angle.
2

We construct 4ABC with \BCA > \CAB , so that BA > BC . Let

BD be the bisector of \ABC , so that  = 1. But AD BA
DC = BC > 1, giving
 < AD .
DC
In conclusion, the given inequality is valid for every 4ABC with
\BAC   . 2

Also solved by MANUEL BENITO and EMILIO FERNANDEZ, I.B. Praxedes Mateo
Sagasta, Logro~no, Spain; and CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK. There
was one partial solution.
528

2498. [1999 : 507] Proposed by K.R.S. Sastry, Dodballapur, India.

A Gergonne cevian is the line segment from a vertex of a triangle to the
point of contact, on the opposite side, of the incircle. The Gergonne point is
the point of concurrency of the Gergonne cevians.
In an integer triangle ABC , prove that the Gergonne point ; bisects
the Gergonne cevian AD if and only if b, c, j3a ; b ; cj form a triangle
1

where the measure of the angle between b and c is  .

2

Solution by D.J. Smeenk, Zaltbommel, the Netherlands.

See the gure below. The Gergonne cevian through C intersects AB
at E . Point F lies on AB so that DF jjCE . Thus, A; : ;D = AE : EF .
Also, AE = s ; a, EB = s ; b = BD and DC = s ; c, where s is the
semiperimeter.
A
s;a

E I
;
s;b

F
B s;b D s;c C

Now ; is the mid-point of AD if and only if EF = AE = s ; a. Since

EF : EB = DC : BC , this is equivalent to a(s ; a) = (s ; b)(s ; c), or to
3a ; b ; c + 2bc ; 2ac ; 2ab = 0 .
2 2 2
(1)
In a triangle with sides b and c, and an angle =3 between b and c, for
the third side a0 , we have a0 = b + c ; bc. Thus, we are to show
2 2 2

4b + 4c ; 4bc = (3a ; b ; c) , or
2 2 2

3a ; b ; c + 2bc ; 2ac ; 2ab = 0 .

2 2 2
(2)
(1) and (2) are identical; that is it!
Also solved by MANUEL BENITO and EMILIO FERNANDEZ, I.B. Praxedes Mateo
Sagasta, Logro~no, Spain; CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; NIKOLAOS
DERGIADES, Thessaloniki, Greece; WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Aus-
tria; PAUL YIU, Florida Atlantic University, Boca Raton, FL, USA; and the proposer.
Most solvers noted that the triangle need not be integer-sided, as can be seen from
the above proof. Also, many noted that the absolute value signs can be dropped, since (for
example) from (1), it follows that 3a2 ; 2ac ; 2ab  0, so that 3a ; b ; c must be positive.
 529

Janous asserts that the condition of the problem is also equivalent to ra = rb + rc;
that is, one exradius is the sum of the other two, which readers may enjoy establishing for
themselves.

2499. [1999 : 507] Proposed by K.R.S. Sastry, Dodballapur, India.

A Gergonne cevian is the line segment from a vertex of a triangle to the
point of contact, on the opposite side, of the incircle. The Gergonne point is
the point of concurrency of the Gergonne cevians.
Prove or disprove:
two Gergonne cevians may be perpendicular to each other.
Solution by Toshio Seimiya, Kawasaki, Japan.
More precisely, we shall prove that each side of a triangle subtends
an obtuse angle at the Gergonne point; that is, we prove that if P is the
Gergonne point of 4ABC , then \BPC > 90 , \CPA > 90 , \APB >
90.
The incircle touches BC , CA and AB at D, E and F , respectively.
Then AE = AF , BF = BD, CD = CE , and AD, BE and CF
are concurrent at P . The proof is by contradiction. We assume that
\BPC  90 . Then
BP + CP  BC . 2 2 2
(1)
Moreover, we have \CPE  90 . Therefore, CE > CP . Similarly, we get
BF > BP . Thus,
BP + CP < BF + CE = BD + CD < (BD + CD) = BC .
2 2 2 2 2 2 2 2

That is, BP + CP < BC , which contradicts (1). Therefore \BPC  90

2 2 2

is not true. Thus, we have \BPC > 90 , as desired. Similarly, we have
\CPA > 90 and \APB > 90 .
Also solved by MICHEL BATAILLE, Rouen, France; MANUEL BENITO and EMILIO
FERNANDEZ, I.B. Praxedes Mateo Sagasta, Logro~no, Spain; CHRISTOPHER J. BRADLEY,
Clifton College, Bristol, UK; NIKOLAOS DERGIADES, Thessaloniki, Greece; RICHARD I. HESS,
Rancho Palos Verdes, CA, USA; G. TSINTSIFAS, Thessaloniki, Greece; PETER Y. WOO, Biola
University, La Mirada, CA, USA; PAUL YIU, Florida Atlantic University, Boca Raton, FL, USA;
and the proposer. There was one incorrect solution.
Sastry's solution exploited a related problem of his (#227, College Math. J. 15:2 (March,
1984) 165-166). For a convex quadrangle such as BCEF , BC 2 + EF 2 = BF 2 + CE 2 if and
only if BE ? CF .
530

2500. [1999 : 507] Proposed by G. Tsintsifas, Thessaloniki, Greece.

In the lattice plane, the unit circle is the incircle of 4ABC .
Determine all possible triangles ABC .
Solution by Nikolaos Dergiades, Thessaloniki, Greece.
Let a, b and c be the side lengths, s the semiperimeter,
the area and
r the inradius of 4ABC . Without loss of generality, let us assume
a  b  c. (1)
It is well known that
p

= sr = s(s ; a)(s ; b)(s ; c) .
Since r = 1, we obtain

= s, (2)
or,
(b + c ; a)(c + a ; b)(a + b ; c) = 4(a + b + c) . (3)
In the lattice plane, the area of a triangle is a rational number, by Pick's
Theorem; hence, (2) implies that s must also be a rational number. Since A,
B and C are lattice points, then a, b and c are square roots of integers.
Consequently, s can be rational only when a, b and c are positive integers.
Let
x = b + c ; a , y = c + a ; b and z = a + b ; c . (4)
Then
a = y +2 z , b = z +2 x and c = x +2 y . (5)
Equation (3) becomes
xyz = 4(x + y + z) . (6)
From (1), (4), (5) and (6), it follows that x, y and z are positive even integers
with z  y  x. Then
xyz = 4(x + y + z)  12x ==) yz  12 ==) z = 2 .
From (6), x = 2 + y; , which implies y = 4 and x = 6. This gives the
16

only triangle with the required property: the right triangle with sides a = 3,
2 4

b = 4 andpc = 5. The distance between the vertex of the right angle and the
origin is 2, so that the vertex of the right angle must be at one of the points
(1; 1). This gives 4 possible locations of the triangle in the lattice plane.
 531

Four other locations can be obtained by re ections about the lines through
the origin and the vertex of a right angle.
Also solved by MANUEL BENITO and EMILIO FERNANDEZ, I.B. Praxedes Mateo
Sagasta, Logro~no, Spain; RICHARD I. HESS, Rancho Palos Verdes, CA, USA; PAUL YIU, Florida
Atlantic University, Boca Raton, FL, USA; and the proposer. There was also one incorrect so-
lution submitted. Yiu pointed out that the following fact is known: The only triangle with
integer-valued sides and with inradius r = 1 is the (3; 4; 5) triangle. Yiu refers to the editor's
note to Problem 1168, American Mathematical Monthly, 63 (1956), p. 43-44.

Problems from One Hundred Years Ago

Preliminary Examination for the Army, 1888
Time allowed | 2 hours
1. Add 2 to 4 .
6
7
11
35
3. Multiply by . 11
42
28
55
2. Subtract 3 from 6 .
5
8
7
68
4. Divide by . 13
96
65
84

5. Add together 4:30726, :076428, 371:864 and 20:0472.

6. Subtract 47:063782 from 701:04681.
7. Multiply 40:637 by :028403. 8. Divide 8:31183 by 23:05.
9. Reduce 1:047 of 2 weeks 5 hours to minutes and the decimal of a
minute.
10. In 347693 inches, how many miles, furlongs, pols, yards, etc.?
11. What would a tax of 3s. 11d. in the pound amount to in $480?
12. Find the simple interest of $11175 in 2 years at 2 per cent per
1

annum.
2

How life has changed!!

Crux Mathematicorum
Founding Editors / Redacteurs-fondateurs: Leopold Sauve & Frederick G.B. Maskell
Editors emeriti / Redacteur-emeriti: G.W. Sands, R.E. Woodrow, Bruce L.R. Shawyer
Mathematical Mayhem
Founding Editors / Redacteurs-fondateurs: Patrick Surry & Ravi Vakil
Editors emeriti / Redacteurs-emeriti: Philip Jong, Je Higham,
J.P. Grossman, Andre Chang, Naoki Sato, Cyrus Hsia
532

YEAR END FINALE

Again, a year has own by! It is dicult to realize that I have now done
this job for ve years. My term has two years to go until the end of 2002,
when you will have a new Editor-in-Chief.
There are many people that I wish to thank most sincerely for par-
ticular contributions. Again, rst and foremost is BILL SANDS. Bill is of
such value to me and to the continuance of CRUX with MAYHEM. As well,
I thank most sincerely, ILIYA BLUSKOV, ROLAND EDDY, CHRIS FISHER,
CLAYTON HALFYARD, GEORG GUNTHER, BILL SANDS, JIM TOTTEN, and
EDWARD WANG, for their regular yeoman service in assessing the solutions;
BRUCE GILLIGAN, ED BARBEAU, CHRIS FISHER, RICHARD McINTOSH,
DONALD RIDEOUT, DIETER RUOFF, GARY SNEDDON, for ensuring that
we have quality articles; ALAN LAW, for ensuring that we have quality book
reviews, ROBERT WOODROW, who carries the heavy load of two corners,
and RICHARD GUY for sage advice whenever necessary.
The editors of the MAYHEM section, NAOKI SATO, CYRUS HSIA,
ADRIAN CHAN, DONNY CHEUNG, JIMMY CHUI and DAVID SAVITT, all
do a sterling job. We look forward to welcoming SHAWN GODIN, the
incoming Mayhem Editor.
I also thank all those who assist with proofreading. The quality of all
these people is a vital part of what makes CRUX with MAYHEM what it is.
Thank you one and all.
As well, I would like to give special thanks to our Associate Edi-
tor, CLAYTON HALFYARD, for continuous sage advice, and for keeping
me from printing too many typographical and mathematical errors; and to
my colleagues, YURI BAHTURIN, DAVID PIKE, ROLAND EDDY, EDGAR
GOODAIRE, MAURICE OLESON, MIKE PARMENTER, DONALD RIDEOUT,
NABIL SHALABY, in the Department of Mathematics and Statistics at
Memorial University, and to JOHN GRANT McLOUGHLIN, Faculty of Ed-
ucation, Memorial University, for their occasional sage advice. I have
also been helped by some Memorial University students, KARELYN DAVIS,
ALASDAIR GRAHAM, as well as WISE Summer students, ALLISON
BLUNDON and MEGAN CHAYTOR.
The staff of the Department of Mathematics and Statistics at Memor-
ial University deserve special mention for their excellent work and support:
ROS ENGLISH, MENIE KAVANAGH, WANDA HEATH, and LEONCE
MORRISSEY; as well as the computer and networking expertise of DWAYNE
HART and CRAIG SQUIRES.
Also the LATEX expertise of JOANNE LONGWORTH at the University of
Calgary, the MAYHEM sta , and all others who produce material, is much
appreciated.
 533

Thanks to BOB QUACKENBUSH as Managing Editor. The CMS's TEX

Editor, MICHAEL DOOB has been very helpful in ensuring that the printed
master copies are up to the standard required for the U of T Press, who
continue to print a ne product.
The online version of CRUX with MAYHEM continues to attract atten-
tion. We recommend it highly to you. Thanks are due to LOKI JORGENSON,
JUDI BORWEIN, and the rest of the team at SFU who are responsible for
this.
Finally, I would like to express real and heartfelt thanks to the Head
of my Department, HERBERT GASKILL, and to the Dean of Science, BOB
LUCAS. Without their support and understanding, I would not be able to do
the job of Editor-in-Chief.
Last but not least, I send my thanks to you, the readers of CRUX with
MAYHEM. Without you, CRUX with MAYHEM would not be what it is.
Keep those contributions and letters coming in. We need your ARTICLES,
PROPOSALS and SOLUTIONS to keep CRUX with MAYHEM alive and well.
I do enjoy knowing you all.

A Maze in Three Dimensions

Izador Hafner
Here is a maze on a tetrahedron, given as an unfolded plane plan. Can you
solve it? What is the path from one dot to the other?
[Ed. We will print a solution in a subsequent issue. Let us know if you would
like more of these amazing mazes.]

Izidor Hafner
Faculty of Electrical Engineering
University of Ljubljana
Trzaska 25
1000 Ljubljana, Slovenia
534

F.G.-M. Mystery solved?

Professor Paul Yiu has provided us with the following quotation:
Some of us often wondered who F.G.-M. was. [See, for example, Hyacinthos
Messages 46 and 50]. I have recently acquired a copy of Gabay's 1997 reprint of
F.G.-M.: Trigonometrie (4th edition), 1890.
The inside front cover of the book contains the following interesting note.
Signi cation des initiales

Lorsqu'un Frere des Ecoles Chretiennes e crivait un livre, son nom n'etait pas
mentionne, mais on indiquait les initiales du Superieur general en fonction.
C'est ainsi que les Exercices de geometrie descriptive e crits par le Frere GABRIEL-
MARIE, furent publies pour la premiere fois en 1877 sous les initiales F.I.-C.
F.I.-C.: Frere IRLIDE { Jean-Pierre CAZENEUVE (nom civil) { Superieur de
1875 a 1884.
La troisieme e dition a e te publiee en 1893 sous les initiales F.J.
F.J.: Frere JOSEPH { Jean-Marie JOSSERAND (nom civil) { Superieur de 1884
a 1897.
Par une heureuse concidence, la quatrieme et la cinquieme e dition ont e te
publiees en 1909 et 1920 sous les initiales du veritable auteur, F.G.-M.
F.G.-M.: Frere GABRIEL-MARIE { Edmond BRUNHES (nom civil), 1836-1916
{ Superieur de 1897 a 1913.
There is an endnote explaining that
Ces precisions ont e te aimablement communiquees par l'Association La Salle
a Paris et par le Centre Scolaire Jean-Baptiste de la Salle a Lyon.
ENGLISH TRANSLATION
What the initials mean
When a Christian Brother wrote a book, his name would not appear anywhere.
Instead, the initials of the superior general in oce were used.
Hence the Exercices de geometrie descriptive written by brother GABRIEL-
MARIE were rst published in 1877 under the initials F.I.-C.
F.I.-C.: Frere IRLIDE { Jean-Pierre CAZENEUVE (civil name) { Superior from
1875 to 1884.
The third edition was published in 1893 under the initials F.J.
F.J.: Frere JOSEPH { Jean-Marie JOSSERAND (civil name) { Superior from
1884 to 1897.
It is a happy coincidence that the fourth and fth editions were published in
1909 and 1920 under the initials of the real author, F.G.-M.
F.G.-M.: Frere GABRIEL-MARIE { Edmond BRUNHES (civil name), 1836-1916
{ Superior from 1897 to 1913.
There is an endnote explaining that
These explanations were kindly provided by the Association La Salle in Paris
and the Centre Scolaire Jean-Baptiste de la Salle in Lyon.
 535

Announcement
Florida Atlantic University announces the publication of
Forum Geometricorum
an electronic journal devoted to classical Euclidean geometry and related
areas, freely accessible to the internet community. The rst papers will
appear in early 2001. Please visit
http://www.math.fau.edu/ForumGeom/
and enter your free subscription, and support Forum Geometricorum by
submitting your papers in Euclidean geometry.
Thanks to Professor Paul Yiu for providing this information.

Dr. Herta Freitag

We have just learned of the passing of a Dr. Herta Freitag, a contributor
to Crux Mathematicorum from 1977 to 1993. Dr. Freitag died in January 2000
at the age of 91.
She was born and raised in Vienna, Austria. Despite having had bad
experiences with Mathematics in her early schooling, she became enamoured
of Mathematics at age 12 because \this is the one subject where I don't need
to memorize, but can just think things out myself".
She graduated in Mathematics and Physics from the University of
Vienna in 1934. Circumstances in Europe at that time resulted in a move
to England to await an immigration visa to the United States. She moved
there in 1944.
After a brief stint as a school teacher in upstate New York, she attended
Columbia University, graduating with a master's degree in 1948, and she
stayed there to complete her doctorate in 1953. She then accepted a position
at Hollins College, (now Hollins University), Roanoke, Virginia, where she
spent her entire academic career, \retiring" in 1971.
She always described herself as a \student of Mathematics, albeit, quite
an addicted one". She was very interested in problem posing and solving, as
our readers will be well aware. She never gave up, leading a very active life,
and publishing into the 1990's, including two papers in 1999.
536

INDEX TO VOLUME 26, 2000

Crux Articles
The Existence of Looped Langford Sequences
Nabil Shalaby and Tara Stuckless .......................................... 86
An Asymptotic Approximation for the Birthday Problem
S. Ejaz Ahmed and Richard J. McIntosh ................................. 151
The Devil's Dartboard
Trevor Lipscombe and Arturo Sangalli ................................... 215
Some bounds for (n)(n)
Edward T.H. Wang ....................................................... 280
A Nice COMC Problem
Daryl Tingley ............................................................. 407
A Spatial Problem Solved with Stereographic Projection
Shay Gueron and Oran Lang .............................................. 468
The Academy Corner Bruce Shawyer
February No. 30 1999 Putnam Competition ............................. 3
April No. 32 1999 APICS Annual Mathematics Competition ...... 129
May No. 33 1999 APICS Annual Mathematics Competition ...... 193
September No. 34 Memorial University Undergraduate Mathematics
Competition March 2000 ............................ 257
The Bernoulli Trials 2000 | Questions .............. 258
October No. 35 1999 APICS Annual Mathematics Competition ...... 321
November No. 36 7th International Mathematics Competition for
University Students, 2000 ............................ 385
December No. 37 Memorial University Undergraduate Mathematics
Competition, September 2000 ....................... 449
2000 Atlantic Provinces Council on the Sciences
Mathematics Competition ........................... 450
The Olympiad Corner R.E. Woodrow
February No. 203 ........................................................ 5
March No. 204 ....................................................... 65
April No. 205 ...................................................... 132
May No. 206 ...................................................... 196
September No. 207 ...................................................... 261
October No. 208 ...................................................... 324
November No. 209 ...................................................... 388
December No. 210 ...................................................... 452
The Skoliad Corner R.E. Woodrow
February No. 43 ........................................................ 26
March No. 44 ........................................................ 93
April No. 45 ....................................................... 156
May No. 46 ....................................................... 218
September No. 47 ....................................................... 286
October No. 48 ....................................................... 341
November No. 49 ....................................................... 412
December No. 50 ....................................................... 475
 537

Book Reviews Alan Law

IX and XIX International Mathematical Olympiads
by Vladimir Jankovic and Vladimir Micic
Reviewed by Richard Hoshino ............................................. 24
144 Problems of the Austrian-Polish Mathematics Competition
by Marcin E. Kuczma
Reviewed by Ed Barbeau ................................................... 25
Baltic Way, 1990{1996: Mathematical Team Competitions
by Marcus Better
Reviewed by Ed Barbeau ................................................... 25
The German Teams at the International Mathematical Olympiads, 1959{1998
by Wolfgang Engel, Hans-Dietrich Gronau, Hans-Heinrich Langmann
and Horst Sewerin
Reviewed by Ed Barbeau ................................................... 25
Notes on Geometry
by Dan Barnzei
Reviewed by Ed Barbeau ................................................... 25
Logic as Algebra
by Paul Halmos and Steven Givant
Reviewed by Maria Losada ................................................ 84
The Sensual (Quadratic) Form
by John Horton Conway, assisted by Francis Y.C. Fung
Reviewed by Richard K. Guy ............................................. 147
Conjecture and Proof
by Miklos
 Laczkovich
Reviewed by Andy Liu .................................................... 212
Winning Solutions
by Edward Lozansky and Cecil Rousseau
Reviewed by Andy Liu .................................................... 213
A Panorama of Harmonic Analysis
by Steven G. Krantz
Reviewed by Alan Law .................................................... 214
Geometry from Africa | Mathematical and Educational Explorations
by Paulus Gerdes
Reviewed by Julia Johnson ................................................ 278
Mathematical Fallacies, Flaws, and Flim am
by Edward J. Barbeau
Reviewed by Catherine Shevlin ........................................... 338
Readings in Cooperative Learning for Undergraduate Mathematics
by Ed Dubinsky, David Mathews, and Barbara E. Reynolds
Reviewed by John Grant McLoughlin ..................................... 339
Calculus Mysteries and Thrillers
by R. Grant Woods
Reviewed by G.J. Grith ................................................. 405
The Math Chat Book
by Frank Morgan
Reviewed by Sandy Graham .............................................. 405
538

Twenty Years Before the Blackboard

by Michael Steuben with Diane Sandford
Reviewed by Nicholas Buck .............................................. 465
Archimedes | What Did He Do Besides Cry Eureka
by Sherman Stein
Reviewed by C.L. Kaller .................................................. 466
Miscellaneous
Jessie Lei ...................................................................... 1
Awards of Subscriptions ...................................................... 23
ATOM Announcement ........................................................ 64
 curiosity .................................................................... 85
Letter to the Editor - David Singmaster .................................... 284
Congratulations, Paco Bellot! ...............................................320
Mathematical Poems | Eileen Tupaz .......................................337
Announcement | ATOM Volume IV ........................................384
Announce | ATOM tome IV ................................................ 384
Letter to the Editor - Walther Janous .......................................467
Problems from One Hundred Years Ago .................................... 531
Year End Finale ............................................................. 532
A Maze in Three Dimensions ............................................... 533
Mystery solved? ............................................................ 534
Announcement of Forum Geometricorum .................................. 535
Dr. Herta Freitag ........................................................... 535
Problems
February 2501{2512, .................................................... 45
March 2513{2525 .................................................... 114
April 2525{2538, 2488 ............................................. 177
May 2539{2550, 2495 ............................................. 236
September 2551{2562 .................................................... 303
October 2539, 2563{2575 ............................................. 372
November 2576{2588 .................................................... 429
December 2589{2600.................................................... 497
Solutions
February 2401{2412, 2415 .............................................. 47
March 2413{2414, 2416{2425 ....................................... 117
April 2426{2430, 2432{2437 ....................................... 180
May 2431, 2438{2450 ............................................. 239
September 2436, 2451{2459 ............................................. 306
October 2460{2463, 2465{2466 ....................................... 375
November 2464, 2467, 2469, 2471{2476, 2478{2479 .................... 432
December 2468, 2477, 2480{2500 ....................................... 500
Mathematical Mayhem
February ................................................................. 29
March ................................................................. 98
April ............................................................... 165
May ............................................................... 220
September ............................................................... 291
October ............................................................... 350
November ................................................................ 417
December ............................................................... 479
 539

Shreds and Slices

February A Fifth Way to Count .......................................... 29
March A Calculus Proof of a Trigonometry Identity ................. 98
December Another Proof of the Ellipse Theorem ....................... 480
Mayhem Articles A Do-It-Yourself Proof of the n = 3 case of Fermat's Last Theorem
Ravi Vakil .................................................................. 36
Dividing Points Equally, II
Cyrus C. Hsia ............................................................. 111
How to Solve the Cubic
Naoki Sato ................................................................ 171
Constructive Geometry, Part I
Cyrus C. Hsia ............................................................. 231
Ellipses in Polygons
Naoki Sato ................................................................ 361
Another Do-It-Yourself Proof of the n = 3 case of Fermat's Last Theorem
Andy Liu .................................................................. 422
An Interesting Application of the Sophie Germain Identity
Carl Johan Ragnarsson ................................................... 426
Constructive Geometry, Part II
Cyrus C. Hsia ............................................................. 492
Mayhem High School Problems
February H265{H268 ................................................... 31
April H269{H272 .................................................. 165
September H273{H276 .................................................. 297
November H277{H280 .................................................. 417
Mayhem High School Solutions
March H249{H252 ................................................... 99
May H253{H256 .................................................. 220
October H257{H260 .................................................. 350
December H261{H264................................................... 482
Mayhem Advanced Problems
February A241{A244 .................................................... 32
April A245{248 .................................................... 166
September A249{A252 ................................................... 298
November A253{A256 ................................................... 418
Mayhem Advanced Solutions
March A225{A228 ................................................... 102
May A229{A232 ................................................... 224
October A232{A236 ................................................... 353
December A237{A240.................................................... 485
Mayhem Challenge Board Problems
February C91{C92 ....................................................... 32
April C93{C94 ...................................................... 167
September C93{C96 ...................................................... 298
November C97{C98 ...................................................... 419
540

Mayhem Challenge Board Solutions

March C83{C84 ...................................................... 105
May C85{C86 ...................................................... 225
October C87{C88 ...................................................... 357
December C89{C90 ...................................................... 489
Problem of the Month: Jimmy Chui
February ................................................................. 33
March ............................................................... 107
April ............................................................... 167
May ............................................................... 228
September ............................................................... 302
October ............................................................... 360
November ................................................................ 420
December ............................................................... 491
Contests
J.I.R. McKnight Problems Contest 1993 ..................................... 34
J.I.R. McKnight Problems Contest 1994 ................................... 109
J.I.R. McKnight Problems Contest 1995 ................................... 168
J.I.R. McKnight Problems Contest 1996 ................................... 230
J.I.R. McKnight Problems Contest 1997 ................................... 300
J.I.R. McKnight Problems Contest 1986 ................................... 422

Proposers, solvers and commentators in the PROBLEMS and SOLUTIONS

sections for 2000 are:
Names of problem proposers
Mohammed Aassila 2426, 2461, 2474, 2476 Mih a ly Bencze 2481, 2482, 2490, 2491
Miguel Amengual Covas 2473 Vedula N. Murty 2459, 2462, 2471
Mansur Boase 2408 Victor Oxman 2466
Christopher J. Bradley 2402, 2403, 2407, 2417, 2418 Ice B. Risteski 2421
Nikolaos Dergiades 2458, 2497 Bill Sands 2413
Jose Luis Daz 2487 K.R.S. Sastry 2419, 2424, 2425, 2432, 2433, 2434, 2498, 2499
Antal E. Fekete 2452, 2453 Catherine Shevlin 2446
Joaqun Gomez
 Rey 2475, 2480, 2489 Zun Shan 2411
G.P. Henderson 2505 Toshio Seimiya 2440, 2427, 2428, 2439, 2484, 2485, 2492, 2493,
Joe Howard 2486 2494
Peter Hurthig 2438 D.J. Smeenk 2401, 2409, 2420, 2429, 2430
Walther Janous 2422, 2423, 2467, 2468, 2477 Jill Taylor 2431
Clark Kimberling 2289 G. Tsintsifas 2488, 2495, 2500
Vaclav Konecny 2404, 2410, 2416, 2435, 2436, 2460, 2463, 2472, Darko Veljan 2412
2483 Edward T.H. Wang 2411, 2414
Michael Lambrou 2442, 2443, 2444, 2445, 2451, 2464 Wu Wei Chao 2414
Gerry Leversha 2447, 2448, 2449, 2450, 2454, 2455, 2456, 2457, Albert White 2465
2457 Paul Yiu 2406, 2415, 2437, 2441, 2469, 2470, 2496
The Editors 2405
 541

Names of featured solvers

Mangho Ahuja 2487 Kathleen E. Lewis 2413
Miguel Amengual Covas 2456 Mark Lyon 2438
Michel Bataille 2403, 2408, 2412, 2450, 2454, 2459, 2461 Phil McCartney 2435
Manuel Benito 2489, 2491 Vedula N. Murty 2487
Christopher J. Bradley 2433 Jose H. Nieto 2451
Jonathan Campbell 2491 Gottfried Perz 2492
Nikolaos Dergiades 2409, 2417, 2418, 2419, 2424, 2427, 2428, K.R.S. Sastry 2425, 2432, 2434
2436, 2439, 2441, 2446, 2449, 2457, 2464, 2467, 2469, 2470, 2486, 2497, Heinz-Jurgen Sei ert 2403, 2436, 2444, 2468, 2468, 2471, 2472,
2500 2477, 2481, 2483
Charles R. Diminnie 2452, 2453, 2463 Toshio Seimiya 2409, 2415, 2429, 2454, 2458, 2464, 2466, 2494,
David Doster 2460, 2475 2499
Emilio Fernandez Moral 2489, 2491 Max Shkarayev 2438
Florian Herzig 2401 Achilleas Sinefakopoulos 2442
John G. Heuver 2447 D.J. Smeenk 2434, 2473, 2485, 2495, 2498
Walther Janous 2405, 2421, 2423, 2435, 2488, 2492 Eckard Specht 2430, 2439
Murray S. Klamkin 2462, 2495 David R. Stone 2410
Vaclav Konecny 2440, 2443 Choongyup Sung 2465
Dimitar Mitkov Kunchev 2440 Kenneth M. Wilke 2426, 2431
Michael Lambrou 2407, 2408, 2411, 2412, 2420 , 2426, 2467, Larry White 2463
2481, 2487, 2491 Peter Y. Woo 2416, 2448, 2457, 2460, 2465, 2469, 2470, 2484,
Kee-Wai Lau 2402, 2404, 2414, 2461, 2480, 2481, 2490, 2491 2487
Ho-joo Lee 2455 Paul Yiu 2496
Gerry Leversha 2408, 2415, 2431, 2475 Jeremy Young 2406, 2437, 2445, 2493

Names of other solvers

Mohammed Aassila 2482 Jose Luis Daz 2471, 2488
Miguel Amengual Covas 2415, 2424, 2440, 2454, 2462, 2492, Charles R. Diminnie 2406, 2420, 2426, 2431, 2443, 2445, 2451,
2493 2471, 2475

Sefket Arslanagic 2428, 2429, 2433, 2437, 2439, 2440, 2444, Colin Dixon 2418
2446, 2448, 2449, 2455, 2456, 2457, 2459, 2461, 2462, 2467, 2471, 2472, Mike Dowell 2411
2475, 2477, 2483, 2484, 2485, 2486, 2487, 2492, 2493 David Doster 2459, 2462, 2465, 2471, 2480, 2486, 2492, 2493
Federico Arboleda 2411, 2426 Richard B. Eden 2439
Charles Ashbacher 2411 Keith Ekblaw 2413, 2421, 2453, 2491
Sam Baethge 2428, 2434, 2440 Russell Euler 2471
Michel Bataille 2401, 2415, 2420, 2424, 2428, 2430, 2433, 2435, Emilio Fernandez Moral 2404, 2490, 2492, 2493, 2494, 2495,
2437, 2439, 2440, 2441, 2444, 2445, 2447, 2448, 2449, 2452, 2453, 2455, 2496, 2497, 2498, 2499, 2500
2456, 2457, 2458, 2462, 2463, 2464, 2465, 2471, 2472, 2473, 2475, 2477, J. Chris Fisher 2487
2480, 2481, 2482, 2485, 2486, 2487, 2490, 2491, 2492, 2493, 2496, 2499 Ian June L. Garces 2439, 2462, 2486
Frank Battles 2486 Shawn Godin 2443
Soumya Kanti Das Bhaumik 2486 Douglass L. Grant 2426, 2486
Niels Bejlegaard 2408, 2415, 2447, 2473 Karthik Gopalratnam 2459
Francisco Bellot Rosado 2401, 2407, 2409, 2417, 2424, 2428, H.N. Gupta 2487
2429, 2439, 2441, 2448 
Angel Joval Roquet 2434
Manuel Benito 2404, 2490, 2492, 2493, 2494, 2495, 2496, 2497, G.P. Henderson 2426, 2465, 2482
2498, 2499, 2500 Florian Herzig 2403, 2407
Paul Bracken 2435, 2439, 2444, 2453, 2490 Richard I. Hess 2402, 2403, 2404, 2406, 2407, 2410, 2416, 2418,
Christopher J. Bradley 2426, 2427, 2428, 2430, 2431, 2434, 2419, 2420, 2421, 2425, 2426, 2428, 2431, 2435, 2436, 2438, 2439, 2440,
2439, 2440, 2441, 2443, 2446, 2454, 2455, 2456, 2457, 2459, 2460, 2462, 2442, 2443, 2444, 2446, 2448, 2449, 2451, 2452, 2453, 2459, 2462, 2463,
2463, 2464, 2465, 2467, 2469, 2470, 2471, 2475, 2484, 2485, 2486, 2487, 2465, 2468, 2471, 2472, 2475, 2483, 2484, 2486, 2487, 2490, 2491, 2492,
2488, 2492, 2493, 2494, 2495, 2496, 2497, 2498, 2499 2493, 2496, 2499, 2500
James T. Bruening 2403, 2420, 2459, 2465, 2471, 2475, 2487 John G. Heuver 2403, 2449, 2459, 2462, 2486, 2492, 2493
Miguel Angel Cabez on  Ochoa 2459, 2462 Joe Howard 2459, 2462, 2482
Elsie Campbell 2459 Peter Hurthig 2426, 2465, 2471, 2486
Jonathan Campbell 2486 Thomas Jang 2483
Jenn Carruthers 2486 Walther Janous 2401, 2403, 2406, 2408, 2410, 2412, 2415, 2417,
Doug Cashing 2451 2418, 2420, 2424, 2426, 2428, 2429, 2430, 2431, 2433, 2434, 2437, 2439,

Oscar Ciaurri 2465, 2487 2440, 2441, 2443, 2444, 2445, 2447, 2448, 2449, 2450, 2452, 2453, 2455,
Nikolaos Dergiades 2401, 2406, 2407, 2408, 2410, 2412, 2415, 2457, 2459, 2460, 2461, 2462, 2464, 2465, 2469, 2470, 2471, 2472, 2473,
2416, 2420, 2426, 2429, 2430, 2431, 2433, 2434, 2435, 2437, 2440, 2443, 2475, 2480, 2481, 2482, 2483, 2484, 2485, 2486, 2487, 2490, 2491, 2493,
2445, 2447, 2448, 2451, 2452, 2453, 2454, 2455, 2456, 2457, 2459, 2460, 2496, 2498
2462, 2463, 2465, 2469, 2470, 2471, 2472, 2473, 2475, 2477, 2481, 2483, Michael Josephy 2443, 2445
2484, 2485, 2487, 2490, 2491, 2492, 2493, 2495, 2498, 2499 Masoud Kamgarpour 2439, 2440, 2441
542

Geo rey A. Kandall 2485 Catherine Shevlin 2422

Murray S. Klamkin 2439, 2459, 2471, 2475 Max Shkarayev 2426, 2435
Vaclav Konecny 2416, 2418, 2420, 2422, 2424, 2428, 2434, 2439, Andrei Simion 2440, 2441, 2482, 2486, 2487
2447, 2448, 2449, 2459, 2462, 2486, 2490, 2492, 2493 Achilleas Sinefakopoulos 2444, 2445, 2480, 2486, 2493
Mitko Kunchev 2441, 2447, 2462 Amrit Preet Singh 2462
Michael Lambrou 2401, 2403, 2404, 2406, 2409, 2410, 2412, D.J. Smeenk 2406, 2407, 2408, 2415, 2417, 2418, 2424, 2425, 2427,
2414, 2415, 2416, 2417, 2418, 2421, 2422, 2424, 2427, 2428, 2429, 2430, 2428, 2431, 2433, 2434, 2437, 2439, 2440, 2441, 2447, 2448, 2449, 2455,
2431, 2432, 2433, 2434, 2435, 2436, 2437, 2438, 2439, 2440, 2441, 2446, 2456, 2457, 2459, 2462, 2467, 2472, 2484, 2486, 2492, 2493
2447, 2448, 2449, 2450, 2453, 2455, 2459, 2460, 2462, 2463, 2465, 2466, Digby Smith 2403, 2426, 2443, 2459, 2462, 2465, 2471, 2475, 2486,
2468, 2468, 2471, 2472, 2473, 2475, 2477, 2480, 2482, 2486, 2490 2487
Kee-Wai Lau 2406, 2418, 2423, 2427, 2434, 2435, 2439, 2443, 2444, Trey Smith 2426, 2443, 2445, 2451, 2462
2450, 2462, 2465, 2468, 2468, 2471, 2472, 2475, 2483, 2493 Dee Snell 2451
R. Laumen 2439, 2440, 2486, 2487 Andre Louis Souza de Araujo 2492, 2493
Ho-joo Lee 2439, 2440, 2462, 2471, 2472, 2492 Eckard Specht 2415, 2422, 2441, 2443, 2462, 2472, 2492, 2493,
Gerry Leversha 2401, 2403, 2406, 2407, 2409, 2412, 2413, 2416, 2495
2417, 2419, 2420, 2421, 2424, 2428, 2429, 2430, 2432, 2433, 2434, 2437, David R. Stone 2403, 2406, 2411, 2421, 2438, 2443, 2446
2439, 2440, 2441, 2443, 2446, 2451, 2459, 2462, 2467, 2484, 2486 J. Suck 2415
Kathleen E. Lewis 2403, 2410, 2421, 2446, 2451 Choongyup Sung 2455, 2459, 2462, 2475, 2486, 2492, 2493
Michael Liaw 2462 Aram Tangboondouangjit 2459, 2462
Mara Ascensi on  L opez
 Chamorro 2415, 2417, 2424, 2441 Paragiou Theoklitos 2401, 2403, 2408, 2412, 2415, 2416, 2417,
Mark Lyon 2421, 2426, 2435 2428, 2439, 2440, 2441, 2447, 2454, 2455, 2459, 2462
Giovanni Mazzarello 2462 Panos E. Tsaoussoglou 2402, 2411, 2420, 2447, 2459, 2460,
Stewart Metchette 2462, 2472, 2486, 2487
Vedula N. Murty 2423, 2427, 2435, 2482 G. Tsintsifas 2477, 2484, 2499
Jose H. Nieto 2465 David Vella 2459
Henry J. Pan 2492, 2493 Edward T.H. Wang 2421, 2471, 2475, 2486
Michael Parmenter 2410, 2445, 2446, 2448, 2484, 2493 Albert White 2492
Gottfried Perz 2403, 2415, 2439, 2493 Kenneth M. Wilke 2403, 2406, 2410, 2411, 2418, 2420, 2440,
Carl Johan Ragnarsson 2403 2459, 2465, 2480, 2486
Bernardo Recam a n Santos 2426 Peter Y. Woo 2401, 2408, 2409, 2415, 2417, 2424, 2427, 2428,
Juan-Bosco Romero Marquez 2462, 2482, 2486, 2487 2429, 2430, 2433, 2434, 2437, 2440, 2441, 2443, 2445, 2447, 2449, 2454,
Jawad Sadek 2471 2455, 2456, 2458, 2464, 2467, 2473, 2475, 2482, 2486, 2492, 2493, 2495,
Robert P. Sealy 2451, 2459 2496, 2499
Harry Sedinger 2446, 2451, 2459, 2465 Aplakides Yiannis 2440
Heinz-Jurgen Sei ert 2401, 2406, 2407, 2410, 2422, 2426, 2427, Paul Yiu 2407, 2498, 2499, 2500
2428, 2433, 2434, 2435, 2437, 2439, 2440, 2441, 2450, 2462, 2465, 2475, Jeremy Young 2401, 2403, 2407, 2408, 2410, 2417, 2421, 2424,
2480, 2482, 2485, 2486, 2487, 2490, 2491, 2492, 2493, 2496 2440, 2441, 2446, 2448, 2449, 2451, 2454, 2455, 2456, 2459, 2462, 2465,
Toshio Seimiya 2401, 2407, 2408, 2416, 2417, 2424, 2430, 2433, 2475, 2486, 2487, 2492
2434, 2437, 2441, 2447, 2448, 2449, 2455, 2456, 2457, 2462, 2467, 2469, Roger Zarnowski 2446
2470, 2473, 2486, 2487

Angelo State Problem Group 2465 The Bookery Problem Group 2452
Ateneo Problem-Solving Group 2440 Skidmore College Problem Group 2421, 2446
Con Amore Problem Group 2446, 2450, 2459, 2465 Southwest Missouri State University Problem
Solving Group 2480
 543

Proposers, solvers and commentators in the MAYHEM PROBLEMS and

SOLUTIONS sections for 2000 are:
High School
Proposers Solvers
Mohammed Aassila H260, H262 Mohammed Aassila H262
Jose Luis Daz H273 Lino Demasi H253, H264
Ho-joo Lee H272, H271, H276, H279 Masoud Kamgarpour H254
Alexandre Tritchtchenko H254, H259 Murray S. Klamkin H253, H262, H264
Fotifo Casablanca H280 Andrei Simion H261, H263, H264
The Editors H249, H250, H251, H252, H253, H255, H256, H257, D.J. Smeenk H250
H258, H261, H263, H264, H265, H266, H267, H268, H269, H270, Edward T.H. Wang H249, H251, H252, H253, H254, H257,
H274, H275, H277, H278 H258, H259, H261, H262, H264
Luyun Zhong-Qiao H252
The Editors H255, H256, H257, H258, H259, H260, H263

Advanced
Proposers Solvers
Mohammed Aassila A227, A231, A235, A236, A239, A240, A241, Mohammed Aassila A240
A246, A249, A252, A253, A256 
Miguel Carrion Alvarez A226, A227
Ho-joo Lee A251, A255 Michel Bataille A238, A239
Baltic Way A228, A232, A234, A238, A243, A248, A254 Jose Luis Daz A239
Naoki Sato A226, A230, A233 Walther Janous A232
Ravi Vakil A244 Masoud Kamgarpour A231, A233
The Editors A225, A229, A237, A242, A245, A247, A250 Geo rey A. Kandall A232
Vishaal Kapoor A230, A231
Murray S. Klamkin A229, A230, A231, A239
Ho-joo Lee A227
Andy Liu A232
Vedula N. Murty A231, A239
Daniel Riesz A232
Catherine Shevlin A232, A239
Andrei Simion A225
D.J. Smeenk A232
Edward T.H. Wang A234
Peter Y. Woo A232
Luyun Zhong-Qiao A225
The Editors A228, A235, A236, A237

Challenge Board
Proposers Solvers
Dima Arinkin C83 Dima Arinkin C83
Edward Crane C94 Michel Bataille C90
Noam Elkies C90 Roman Fresneda C90
Tal Kubo C89 Mark Krusemeyer C87
Christopher Long C84, C85 Tal Kubo C89
Mark Krusemeyer C87 Michael Lambrou C85
Russell Mann C94 Laurent Lessard C90
The Editors C86, C88, C91, C92, C93, C95, C96, C97, C98 Christopher Long C84
Roman Muchnik C86
The Editors C86, C88
544

Renew Your Subscriptions!

The 2001 subscription rates for
Crux Mathematicorum with Mathematical Mayhem
(CRUX with MAYHEM) are given on the page
opposite. No change from last year!
You will note that all 2001 rates include online
access. We hope that subscribers will enjoy the
bene ts of this.
Renew by 1 March 2001 to avoid disappointment
of missed or delayed issues.

Renouvelez vos abonnements !

Vous trouverez ci-contre les tarifs d'abonnement a
Crux Mathematicorum with Mathematical Mayhem
(CRUX with MAYHEM) pour l'an 2001.
Vous remarquerez que les tarifs pour l'an 2001
comprennent l'acces aux versions en ligne. Nous
esperons que nos abonnes en apprecieront les
avantages.
Faites-nous parvenir votre renouvellement avant
le 1er mars 2001 pour eviter
 tout retard et ne pas
manquer de numero.