Crux

Mathematicorum
Published by the Canadian Mathematical Society.

http://crux.math.ca/

The Back Files

The CMS is pleased to offer free access to its back file of all
issues of Crux as a service for the greater mathematical
community in Canada and beyond.

Journal title history:

➢ The first 32 issues, from Vol. 1, No. 1 (March 1975) to
Vol. 4, No.2 (February 1978) were published under the
name EUREKA.
➢ Issues from Vol. 4, No. 3 (March 1978) to Vol. 22, No.
8 (December 1996) were published under the name
Crux Mathematicorum.
➢ Issues from Vol 23., No. 1 (February 1997) to Vol. 37,
No. 8 (December 2011) were published under the
name Crux Mathematicorum with Mathematical
Mayhem.
➢ Issues since Vol. 38, No. 1 (January 2012) are
published under the name Crux Mathematicorum.
 CRUX MATHEMATICORUM
April / avril
Volume 19 U
1993

CONTENTS I TABLE PES MATIERES

Prime Pyramids Richard K. Guy 97

The Olympiad Corner: No. 144 R.E. Woodrow 100

Problems: 1831-1840 112

Solutions: 1621, 1741-1748, 1750 114

Canadian Mathematical Society

Q> Soci6t6 math&natique du Canada
 Founding Editors / R£dacteurs-fondateurs: Leopold Sauv6, Frederick G.B. Maskell
Editors-in-chief / Rldacteurs en chef: G.W. Sands, R.E. Woodlrow
Managing Editor / R£dacteiur-g£rant: G.P. Wright

EDITORIAL BOARD / CONSEEL DE REDACTION

G.W. Sands (Calgary)
R.E. Woodrow (Calgary)
G.P. Wright (Ottawa)
R. Guy (Calgary)
C. Fisher (Regina)
D. Hanson (Regina)
A. Liu (Alberta)
R. Nowakowski (Dalhousie)
E. Wang (Wilfrid Laurier)

GENERAL INFORMATION RENSEIGNEMENTS GENERAUX

Crux Mathematicorum is a problem-solving journal at the Crux Mathematicorum est une publication de resolution de
senior secondary and university undergraduate levels for problemes de niveau secondaire et de premier cycle
those who practice or teach mathematics. Its purpose is universitaire. Bien que principalement de nature educative,
primarily educational but it serves also those who read it for elle sert aussi a ceux qui la lisent pour des raisons
professional, cultural or recreational reasons. professionnelles, culturelles ou recreative.

Problem proposals, solutions and short notes intended for Les propositions de problemes, solutions et courts articles
publication should be sent to the appropriate member of the a publier doivent etre envoyes au membre approprie du
Editorial Board as detailed on the inside back cover. conseil de redaction tel qu'indiqu6 sur la couverture arrfere.

SUBSCRIPTION INFORMATION RENSEIGNEMENTS SUR L'ABONNEMENT

Crux Mathematicorum is published monthly (except July Crux Mathematicorum est publie mensuellement (sauf juillet
and August). The subscription rates for ten issues are et aout). Les tarifs d'abonnement pour dix numeros
detailed on the inside back cover. Information on Crux figurent sur la couverture airriere. On peut egalement y
Mathematicorum back issues is also provided on the inside retrouver de plus amples renseignements sur les volumes
back cover. anterieurs de Crux Mathematicorum

ACKNOWLEDGEMENTS REMERCIEMENTS

The support of the Department of Mathematics & Statistics Nous rendons hommage a l'appui du departement de
of the University of Calgary and of the Department of math&natique et statistique de l'Universite de Calgary et du
Mathematics of the University of Ottawa is gratefully departement de mathematiques de l'Universite d'Ottawa.
acknowledged.

Canadian Mathematical Society / Societe mathematique du Canada 1993

Printed at / imprime a: Ottawa Laser Copy

ISSN 0705-0348 Publication Mail Registration No. 5432

 97

PRIME PYRAMIDS
Richard K. Guy

Margaret J. Kenney of Boston College, in the Student Math Notes enclosed with the
Nov. 1986 NCTM News Bulletin, proposed the following p r i m e p y r a m i d as a classroom
activity.

*
1 2
1 2 3
1 2 3 4
1 4 3 2 5
1 4 3 2 5 6
1 - - - - - 7

Row n contains the numbers 1,2, . . . , n . It begins with 1 and ends with n. The
sum of two consecutive entries in a row is prime.
There are various questions we may ask.

1. Can the pyramid be continued?

2. How many ways are there of arranging the numbers in row n? (Check that the
arrangements shown above are unique.)

3. Can you find a pattern that will serve for an infinity of rows?

4. Will there ever be a row that cannot be completed?

The n — 1 sums add to

1 + 2(2 + 3 + • • • + (n - 1)) + n = n 2 - 1,

so the average sum of two consecutive members is n + 1, The probability that a number
of this size is prime is about 1 / l n n . So, if we wrote the numbers down at random, our
chance of success would only be about
n-l

Inn
However, there are (n — 2)! ways in which the numbers may be arranged and when we
estimate
1 N n-n(*-2)!
1 -
Inn
(which is our chance of failure if we try every order) we find that we're pretty certain to
98

succeed. Of course, this isn't a proof, but if you start answering question 2, you find

n= 1 2 3 4 5 6 7 8 9 10 11 12
# of solutions * 1 1 1 1 1 2 4 8 28 84 216

and it seems likely that t h e answer grows super-exponentially, at a factorial-type rate.

Now we could answer question 3 if we knew that there were infinitely many twin
primes. If n and n + 2 are both prime, then

1 (n-1) 3 (n-3) 5 ... (n - 2) 2 n

will work.
Here is a more convincing algorithm which it may be possible to formalize into a
rigorous proof. It's clear that we have to arrange the numbers alternately odd and even.
This reduces our number of arrangements from (n — 2)! to

n - 1 ! n-1
. 2 . L 2 J
but this is still a pretty huge number.
Assume for the moment that n is odd. You can see how the argument is easily
adapted for n even. Take n = 11 for an example. Consider all 4! ways of arranging the
odd numbers 3, 5, 7, 9 between 1 and 11, leaving spaces for the even numbers. Then we
can fill in the even numbers, subject only to the constraints that

the number 2 4 6 8 10
mustn't sit next to 7 5 or 11 3 or 9 1 or 7 5 or 11

But this is almost' exactly Lucas's classical probleme des menages, to seat couples
round a table with no-one sitting next to their own spouse. The difference is that some
even numbers have as many as two spouses, but the number of legal arrangements is still
a substantial fraction of the ((n — l ) / 2 ) ! arrangements, and this grows rapidly. For n = 11
you could be unlucky. You can't seat the even numbers among the arrangements

1 . 3 . 9 . 5 . 7 . 11, 1 . 7 . 3 . 5 . 9 . 11,
1 . 7 . 9 . 5 . 3 . 11, 1 . 9 . 3 . 5 . 7 . 11,

but the other 20 all allow solutions, some of them,

1 . 3 . 9 . 7 . 5 . 11, 1 . 7 . 3 . 9 . 5 . 11,
1 . 7 . 9 . 3 . 5" . 11, 1 . 7 . 9 . 3 . 5 . 11,

as many as eight.
What is going on? The constraints above are only needed to avoid the divisor 3.
The odd and even arrangement avoids the divisor 2. And since our sums are at most
 99

11 -f 10.< 25 we don't have to worry about divisor 5. For larger n we need watch out only
for divisors < y/2n — 1 and the number of constraints is less than
n n n
+ + +
6~ 10 14 '"

where the denominators are twice the odd primes less than y/2n. Unfortunately this series,
even when added only thus far, is divergent. Its sum is presumably en In In V ^ , where c
is something like 0.43, though one needs a much more powerful telescope than the one I've
used. A better estimate, possibly good enough to formalize into a proof, could be obtained
by using the inclusion-exclusion principle.
Although there is no necessary connexion between one row and another, note that
any row can probably be used as the initial segment of a later row, provided we leave
ourselves enough room to manoeuvre. We couldn't extend 1 6 5 2 3 4 7 to a row of length
8, 9 or 10, but it can be the beginning of

1 6 5 2 3 4 7 10 9 8 11 12 or 1 6 5 2 3 4 7 12 11 8 9 10 13,

not the beginning of rows of length 14 or 15, but

1 6 5 2 3 4 7 12 11 8 15 14 9 10 13 16
1 6 5 2 3 4 7 12 11 8 9 10 13 16 15 14 17
1 6 5 2 3 4 7 10 9 8 11 12 17 14 15 16 13 18
1 6 5 2 3 4 7 10 9 8 11 18 13 16 15 14 17 12 19

and probably any longer row. Note how bits of the 'twin prime' construction can be
patched together.

Has anyone any better answers?

Department of Mathematics and Statistics

The University of Calgary
Calgary, Alberta T2N 1N4

Editor's note. Crux readers are encouraged to check out the long and interesting
interview with Professor Guy, complete with photos, which graces a recent issue of the
College Mathematics Journal (Volume 24, Number 2, March 1993). Look for Richard and
Louise mountaineering on the cover!
100

T H E OLYMPIAD C O R N E R
No. 144
R.E. WOODROW

All communications about this column should be sent to Professor R.B. Woodrow,
Department of Mathematics and Statistics, The University of Calgary, Calgary, Alberta,
Canada, T2N IN4.
The first set of problems we give are "pre-Olympiad" even though the title seems to
say otherwise. They are the 9th form problems of the XXV Soviet Mathematical Olympiad,
written at Smolensk, Russia, April 17-24, 1991. My thanks go to Georg Gunther, Sir
Wilfred Grenfell College, Corner Brook, Newfoundland, leader of the Canadian LM.O.
team, for collecting and forwarding these to me.

X X V SOVIET MATHEMATICAL OLYMPIAD

9th Form

1 . Find all integer solutions of the system

xz •— 2yt = 3, xt + yz = 1.
(Ju. Nesterenko)
2 . On the blackboard are written n numbers. One may erase any pair of them,
say a and 6, and write down the number (a + 6)/4 to replace them. After this procedure is
repeated n — 1 times, only one number remains on the blackboard. Prove that given that
the n numbers at the beginning are all 1, the last number will not be less than 1/n. (B.
Berlov)
3 . ' There are four straight lines, each two of them intersecting, and no three of
them having a common point. Each line is divided into four pieces with two finite length
segments among them. The total number of segments is eight. Is it possible that the
lengths of these segments are equal to:
(a) 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 ?
(b) pairwise different natural numbers? (A. Berzinsh)
4 . A ticket for a lottery is a card which has 50 empty cells in a line. Each participant
writes down in the cells the numbers 1 , 2 , . . . , 50 without repetitions. The organizer of the
lottery has his own card with the numbers written on it according to the rule. A ticket
wins if at least one number in it coincides with the number in the corresponding cell of the
organizer's card. What is the least possible number of cards which the participant must
fill to guarantee a win? (A. Berzinsh)

For an Olympiad set this issue we give the problems of the 22nd Austrian Math-
ematical Olympiad (2nd Round). Many thanks to Walter Janous, Ursulinengymnasium,
Innsbruck, Austria, for sending them to me.
 101

22nd A U S T R I A N M A T H E M A T I C A L O L Y M P I A D 1991
2nd Round — May 7, 1991 (Time allowed: 4 hours)

1 . Let a, b be rational numbers such that ^fa + y/b is a rational number c ^ 0.

Show that ^fa and \fb themselves are rational numbers.
2 . Determine all real solutions of the equation

1 1 1 . 1 1 1 1 1
-+ . .+ + . = o.
x x +2 x +4 x +6 x +8 z + 10 x + 12 x + U
3* Determine the number of all square numbers contained in the sequence
{a 0 , ai, a2, •. •} where a 0 = 91 and a n + i = 10a n -f ( — l ) n , n > 0.
4o Let A, B be two points on a circle k of radius r such that Ai? = c.
(i) Give a construction of all triangles ABC having k as their circumcircle and such
that one of the medians sa or Sb (through vertices A or £?, resp.) is of given length d.
(ii) How have r, c and d to be chosen such that AABC is uniquely determined?

Final Round, 1st Day — June 11, 1991 (Time allowed: 4 hours)
1 . We are given a convex solid K in R 3 (i.e. for any two points of K the segment
joining them belongs to K), and two parallel planes S\ and 62 with mutual distance 1 both
tangent to K. Let e be parallel to £1 and £2? between S\ and e2 and at distance d\ from
S\. Determine the values of d\ such that the part of K between £1 and e has (a) at least,
(b) at most half of the volume of K1
2 . Determine all functions / : Z \ { 0 } — > Q satisfying the functional equation

x. + y\ f{x) + f{y)
?m
3 . (a) Show that 91 \ n37 - n for all n G IN.
(b) Determine the greatest integer k such that k \ n37 — n for all n £ N .

Final Round, 2nd Day — June 12, 1991 (Time allowed: 4 hours)
4 . The sequence {an} is defined by a\ = 1, a 2 = 0 and a2k+i = a>k + ak+i,
&2k+2 = 2afc+i, k > 1. Determine a m , where m = 2 1 9 + 91.

5 . Show that for all natural numbers n > 1 the inequality

n + i 1
^ l + (n + l ) y (l+nr>
\ n + 2 J U + l

is valid.
6* Determine the number of numbers (ag . . . ao)io which have no initial zeros and
do not contain the block of digits 1991, when written in decimal notation.
102

* * *

First we give an alternate solution to one given in the January number.

8 . [1991: 197; 1993: 8] Proposed by Ireland.
Let ABC be a triangle and £ the line through C parallel to the side AB. Let the
internal bisector of the angle at A meet the side BC at D and the line £ at E. Let the
internal bisector of the angle at B meet the side AC at F and the line £ at G. If GF = Di£
prove that AC = 5 C .

Alternate solution by Geoffrey A. Randall, Hamsden, Connecticut.

E b C a G

Let AD — i a , £?JF = £&. Using similar triangles, we obtain

GF _a DE __b
h c ' ta c

Therefore GF = JDJB iff ai& = 6t 0 . So the problem really is to show that a£& = bta => a = b.
Suppose a ^ &, say a < b. Then it is well known that i& < i a , and hence ati < bta,
a contradiction.
Even if one does not use the "well-known" fact, it is possible to give an elementary
proof of it from atf, = bta. Suppose at^ = bta. Suppose a ^ 6, say a < b. Then LlAB <
LIB A, hence 5 / < AL Since B / / L F = (a + c)/6 and AI/ID = (6 + c)/a, it follows that

^ a + 6+ c ia a + 6+ c

and from £ 7 < A / this gives

a+ c f b+ c f .
1 e
/7 \
, L ,
a + b+ c fr < a +, Lb+, c ^ -» ( a + c )** < (& + c)*a-

Since atfc = bta it follows that ci& < ct a , and i& < ta.
5}C * 5[C

Next a typo spotted by an ever vigilant reader, Edward T.H. Wang, Wilfrid Lau-
rier University, Waterloo, Ontario. Problem 4 of the Singapore Interschool Mathematics
Competition, Part A [1993: 65] should replace a - f f e + 1 by a + b + c= 1, to read
 103

4 . Let a, 6, c be real numbers satisfying a + b + c = 1, a 2 + 62 + c2 = 1/2. Find the

maximum value of c.
* *

Shortly after the March number of the Corner was off to the printers, the mail
brought a package of solutions to the five Selection Questions For the 1990 Irish I.M.O.
Team, as well as to problem 1 of the 1988 Chinese Olympiad Training Camp. Thanks to
Michael Selby, University of Windsor, for the solutions.

The next solutions we give deal with responses to the problems More Selection
Questions For the 1990 Irish I.M.O. Team [1992: 65].
1 . Let ABC be a right-angled triangle with right angle at A. Let X be the foot of
the perpendicular from A to BC and Y the midpoint of XC. Let AB be extended to D
so t h a t AB — BD. Prove that DX is perpendicular to AY.
Solutions by Seung-Jin Bang, Seoul, Republic of Korea; by Dieter Bennewitz,
Koblenz, Germany; by C Bradley, Clifton College, Bristol, United Kingdom; by Bob
Prielipp, University of Wisconsin-Oshkosh; and by Pavlos Maragoudakis, student, Uni-
versity of Athens, Greece. Most solvers used coordinates in one form or another. We use
the (slightly) more geometric proof of Maragoudakis.
Let BC be extended to X' so that BX = BX'. Then AX' and XD are parallel
and of equal length. It is enough to prove that triangle YAX' is right-angled at A. Let
BC = a, AC = 6, AB = c. We have AX = ha = (BC)/a. Also, by similar triangles,
b = XC • a. Therefore XC = b /a, and XY = b /2a. Now c = XB • a so XB = c2/a.
2 2 2 2

From Pythagoras,
h2c2 b4
2 2 2
AY = AX + XY =°-^ + i-. (1)
a2 4a 2
Because XB is the median of triangle AXD,

AX2 + XD2 = 2XB2 + ^ - ,

,r ^2 2C 4 0 2 b2C2
2 2
XD = —-
a2 + 2c - a2
and rt o
2r4 h2c2
(AX')2 = 4 + 2c' - -T • W
a2 a2
We also have
(X'Y)2 = {XY + 2XB)2 =
which gives
/)4 4r 4 2b2 c2
X Y2
' = TI + -^ + ±
T-- (3)
4a 2 a2 a2
104

Finally lYAX' = 90° iff (X'Y)2 = AY2 + {AX')2 which by (1), (2) and (3) is equivalent
to
2c 4 2b2c2 n 2
—1 + — j2,- = 2c 2 ,
a a
i.e. c2 + 62 = a 2 , which is true.
2. The real number x satisfies all the inequalities

2k <xk + xk+1 < 2k+1

for fc = l , 2 , . . . , n . What is the greatest possible value of n?

Solutions by Christopher J. Bradley, Clifton College, Bristol, U.K.; by Pavlos
Maragoudakis, student, University of Athens, Greece; and by Edward T.H. Wang, Wil-
frid Laurier University, Waterloo, Ontario. We use Bradley's solution.
We cannot have n > 4 because 2 < x + x < 4 and 16 < x4 + x5 < 32 would
2

mean x3 > 4 or x > 2 2 / 3 . But 2 2 / 3 + 2 4 / 3 > 4, by the Arithmetic Mean - Geometric Mean
inequality, so we have a contradiction.
On the other hand n = 3 is possible, for example x = 3/2 will do, since

2< +
2 U) = T < 4 ' 4<
U ) + (2) = T < 8 '
and
3 4
0 /3\ /3\ 135 .,„

3 . Three sides of a quadrilateral are given, of lengths a, 6, c, respectively. If the

area of the quadrilateral is as large as possible prove that the length x of the remaining
side satisfies the equation

x3 - (a 2 + b2 + c2)x - 2abc = 0.

Solution by Christopher J. Bradley, Clifton College, Bristol, U.K.

We have

d2 = a2 + b2 — 2a6 cos y> = c2 + £ 2 — 2cx cos 0 (1)

and 2A = absimp + cxsmO, where A is the area of the

quadrilateral. It is perfectly possible, but cumbersome,
to prove purely algebraically that the maximum area
occurs when the quadrilateral is cyclic. Using Lagrange
multipliers is shorter, using three variables 0, <p, x and
the constraint (1). One obtains

(ex cos 0, ab cos 9?, c sin 0) = \{2cx sin 0, —2ab sin <p, 2x— 2c cos 0).
 105

So, 2A = cot 6 = - cot 9?, showing that 0 + <^ = 180°. Also 2\(x - ccos 9) = csin0, from
which x cos (9 = c, and cos c/p = —c/z. Substituting back in (1) gives

x3 - (a2 + b2 + c2)x - 2abc = 0.

4 . Let ABC be a triangle and let the internal bisectors of the angles at A and B
meet the sides BC and AC at D and D, respectively. Let CF and C G be the perpendiculars
from C to the lines BE and AD, respectively. Prove that the line FG is parallel to AB.
Solutions by Christopher J. Bradley, Clifton College, Bristol, U.K.; and by Pavlos
Maragoudakis, student, University of Athens, Greece. We use the solution of Maragoudakis.
Let / denote the incentre of the triangle, Q
i.e. the point of intersection of AD and BE. The
quadrilateral IFCG can be inscribed in a circle
since LlFC + LlGC = 180°. So

LIFG = LICG. (1)

But

LlCG = LACG- LICA

= 90° - LGAC - LIC A = 90° - l-LA - ]-LC. (2)

From (1), (2) we have LlFG = \LB = ZADF. Thus AD is parallel to FG as required.
5 . Let n = 2& — 1 where k > 6 is an integer. Let T be the set of all n-tuples
(xi,x2,..-.,xn) where :rt-is 0 or 1 (i = 1 , 2 , . . . ,ra). F o r x = ( x l 5 . . . , x n ) a n d y = ( j / i , . . . ,y n )
in T, let d!(x,y) denote the number of integers j with 1 < j < n such that Xj ^ j/j. (In
particular, d ( x , x ) = 0). Suppose that there exists a subset S of T with 2k elements which
has the following property: given any element x in T, there is a unique element y in S
with d ( x , y ) < 3. Prove that n — 23.
Solution by Christopher J. Bradley, Clifton College, Bristol, U.K.
It is not clear whether such a set S exists, but if it does, then it clearly partitions
T into disjoint subsets Tt- such that the elements of Tt- are linked to the element x in S
(uniquely) by the specification that the distance from any element in Tt- to x is less than
or equal to three. Furthermore these subsets will each have the same number of elements,
M + M + Q + Q . The conditions of the problem require this to be equal to 2( n " 1 )/ 2 .
This is only true if n = 23.

Next come solutions to some of the problems from the Second, Third, and Fourth
Tests of the 1988 Chinese Olympiad Training Camp [1992: 66,67].
2 . Second Test. For a finite sequence A of 0's and l's, let f(A) denote the sequence
obtained from A by replacing each 1 by 0 , 1 , and each 0 by 1,0; e.g., / ( ( I , 0,0,1)) =
106

( 0 , 1 , 1 , 0 , 1 , 0 , 0 , 1 ) . Let fn(A) denote the nth iterate of / on A. Determine the number in

/ n ( ( l ) ) of two consecutive terms which are 0, 0.
Solution by Christopher J. Bradley, Clifton College, Bristol, U.K.
Let an be the number in / n ( ( l ) ) of two consecutive terms which are 0, 0 and bn the
number in / n ( ( l ) ) of two consecutive terms that are 1,1. The following facts about the
set / n ( ( l ) ) are easy to establish by induction, and the proofs are omitted.
(i) / n ( ( l ) ) is a sequence of 2n digits, all zero or one.
(ii) The last digit is always 1.
(iii) The first digit alternates 0 , 1 , 0 , 1 , . . . , with the first digit 1 for n even.
(iv) The first 2n~l digits in / ( n ) ( ( l ) )
are the complements with respect to 1 of the
71 1
last 2 " digits.
These facts taken together, imply that
bn = a n _i + 6 n _i, . n > 2 (1)

a 2 n +i = a2n + b2n, n>\ (2)

and
a
#2n+2 = 2 n + l + &2n+l + 1, n > 0 (3)

The extra 1 creeps into a2n because of a "010" created in the centre, which occurs only on
even rows.
From these equations one may eliminate bn to obtain

a 2n +2 = 2a 2 n +i + 1, n > 0, (4)

and
a 2 n +i = 2a 2 n - 1, n > 1. (5)
Also, for the start ax = 0 and a2 — 1. Observe that there is a different recurrence
relation for odd and even n, so readjusting to take account of this, from (4) and (5), one
has
a 2 n+i = 4 a 2 n - i + 1, n > 1 (6)
and
a 2n +2 = 4a 2 n - 1, n > 1. (7)
The solution of these difference equations provides the answer to the problem:

a 2 n + i = ^ ( 4 n - 1) and a2n = » ( 4 n + 2).

[Editor's note: this problem is the same as Crux 1790 [1992: 275].]
3 . Second Test. A mathematics teacher wants her two intelligent students S and P
to derive the exact value of a 2-digit natural number n by revealing the number of positive
divisors of n to S and the sum of the digits of n to P. A brief conversation between S and
P goes as follows:
 107

P: I can not determine n.

S: I can't either but I know whether n is even or not.
P: Now I know what n is.
S: So do I now.
Suppose both students are honest and have perfect logical reasoning for whatever they say.
Determine n and justify your answer.

Solutions by Christopher J. Bradley, Clifton College, Bristol, U.K.; and by Pavlos

Maragoudakis, student, University of Athens, Greece.
First we make a table which shows the number of positive divisors of all 2-digit
natural numbers:

2 3 4 5 6 7 8 9 10 12
11, 13 25 10, 14, 15 16 12, 18 64 24 36 48 60
17, 19 49 21,22, 26 81 20,28 30 80 72
23,29 27, 33, 34 32,44 40 84
31, 37 35, 38, 39 45, 50 42 90
41, 43 46,51, 55 52, 63 54 96
47, 53 57, 58, 62 68, 76 56
59, 61 65, 69, 74 92,98 66
67, 71 75, 77, 82 99 70
73, 79 85, 86, 87 78
83, 89 93, 94, 95 88
91,97

P can not determine n, so n is neither 10 nor 99, and the sum of the digits lies between 2
and 17 inclusive. S can't either, so, from the table n is neither 36 nor 64. But he knows
whether n is even or not. So n can have only 2, 3, 8, 10 or 12 divisors because from the
table these are the columns which have only even or only odd number entries.
Then P says that he knows n. This means that among the 2-digit numbers with 2,
3, 8, 10 or 12 divisors there is exactly one with the sum of its digits equal to the sum of
the digits of n. We therefore make a second table which shows the sum of the digits of the
numbers with 2, 3, 8, 10 or 12 divisors:

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
11 30 13 23 42 25 17 54 19 29 48 67 59 78 79 89
31 41 60 43 53 72 37 47 66 49 96 88
40 24 61 71 90 73 56 84 97
70 91 83

From this table S can see that n is one of the numbers 11, 59, 89, 30. Then he says he
also knows n. This means that among 11, 59, 89, 30 there is exactly one with its number
of divisors the same as for n. The numbers 11, 59, 89 have 2 divisors, so n must be the
number 30.
108

1 . Third Test. Suppose the inscribed circle of AAiA2A3 touches the sides A2A3,
A3A1 and A\A2 at 7 \ , T2 and T 3 respectively. From the midpoints M i , M2 and M 3 of A2A3,
A3A1 and AiA2, draw lines perpendicular to T2T3, I3T1, and T1T2 respectively. Prove that
these three perpendicular lines are concurrent at a point P and determine the location of
P.
Solution by Christopher J. Bradley, Clifton College, Bristol, U.K.
We make use of the following known lemma, which can be called the distance lemma.
Given its usefulness in problems in geometry it is surprising it is not more widely known.
Lemma. Given any triangle ABC with sides a, 6, c, then it is possible to choose an
origin 0 such that with OA = x, OB = y, and OC = z and areal coordinates ( a , / ? , 7 ) ,
any point P in the plane of the triangle has OP = a x + /3y + 7Z with a + ft + 7 = 1.
Furthermore 0 can be chosen so that if P i ( a i , / ? i , 7 i ) and P 2 (^i? ^2,72) axe given then

(p7Plf . (ai.a2? p±^) + ( A _ A ) . (^!) +(7l _ 72) * p±£^).

Let O be as given in the lemma. The point T\ has then areal coordinates

/ a + b—c a — 6 + c\

since TiA3 = s — c and T1A2 = 3 — b. T2 and T3 may similarly be represented as

(a + b-c -a + b + c\ fa-b +c -a + b + c \
(,—26 '°' 26 J ^ ( — 2 ^ — ' — 2 T — ' °j
respectively. M x , M 2 , M 3 have coordinates ( 0 , 1 / 2 , 1 / 2 ) , ( 1 / 2 , 0 , 1 / 2 ) and ( 1 / 2 , 1 / 2 , 0 )
respectively. We have
T2T3 = — — — ( c - 6, 6, - c ) .

Consider now the point K with areal coordinates

2>rW 6 + c , c + a , a + 6 ) -
We have

From the distance lemma Mi if and T2T3 are perpendicular since

M^K • 2VT3 = 0,b.+ C~* A(c2 - b2){b2 + c2 - a 2 ) - 6 2 (c 2 + a 2 - 62) + c 2 (a 2 + i 2 - c2)] = 0.

ooc(a + o + cj

It follows that the perpendiculars from Mi to T2T3, M 2 to T3T1 and M3 to TiT 2 are
concurrent at K. It is easily verified that K is the centre of mass of a uniform wire in the
 109

shape of triangle AiA2As and it lies on the line joining the incentre to the centroid of the
triangle. It is also the radical centre of the three excircles.
2 . Third Test. Consider a quadrilateral ABCD inscribed in a circle. Suppose we
fix A and C and move B and D along the arcs AC and CA in the clockwise direction in
such a way that BC — CD. Let M denote the point of intersection oi AC and BD. Find
the locus of the circumcenter of triangle AMB.
Comment and solution by Christopher J. Bradley, Clifton College, Bristol, U.K.
There appears to be something mildly wrong with the wording of the question,
for in order to maintain BC = CD then if D moves clockwise along CA then B must
move anticlockwise along AC. However, I shall answer this amended question rather than
suppose the question should read AB = CD, which is rather a different question altogether.
Set up a coordinate system with O the centre of the circle, and having A(cos a, sin a ) ,
C ( l , 0) where a is fixed. Let B have coordinates (cos 0, sin 9). Then for BC = CD we have
D(cos9, — sin#). We require the locus of the circumcentre of triangle AMB as 9 varies,
and shall prove it is a circle. The equation of AC is x s i n a + j/(l — cos a ) = sin a. The
equation of BD is x = cos 9. So the coordinates of M are

/ . s i n a ( l — cos 9)\
cos 9 , '- .
\ I — cos a J
Let (X,Y) be the coordinates of the circumcentre S of triangle ABM. The condition
SA = SB gives X cos a + Y sin a = X cos 9 + Y sin #, and the condition SB = SM gives,
after some algebra,
,. . a . 9 . a +9
Y sin — — sin - sin ——-
2 2 2
giving the coordinates as

( . 9 a +9 a . 9 . a +9 a\
sin - cos —-— esc — , sin - sin esc —
I 2 2 2 ' 2 2 2
or _T/ . a . fa n\ . a
2A sin - = sin I ~ + 0J - sin - ,

2r sin — = — cos — h 9 } + cos

2 V2 J 2
So the locus of 5 is
/ ex a\^ ( a a\^
( 2 X s i n - + s i n - J + ( 2Y sin - - cos - J = 1

which is a part of a circle centred at ( —1/2, l / 2 c o t ( a / 2 ) ) , of radius l / 2 c s c ( a / 2 ) and

passing through 0 , the centre of the given circle.
are
1 . Fourth Test. Suppose Xi, #2, • • • 5 #n positive reals with sum equal to 1.
Determine the minimum value of the function
110

/ ( x i , x2, . . . , ! „ = 7 - ; ; + 7-: : ; +
1 + x2 H \- xn 1 + xi + x3 H h x„

4"
1 + xi + x2 H + x n _i
and find all n-tuples (xi, x2l • • • ? #n) which yield this minimum value.
Solutions by Christopher J. Bradley, Clifton College, Bristol, U.K.; and by Murray
S. Klamkin, University of Alberta. We give Klamkin's solution and generalization.
The sum can be rewritten as
n n
x- 2
/(*!,..., xn) = £ - f - = -n + £ r—- .
Applying Cauchy's inequality

w X
t= l z=l * \i=l /

Since ^ = 1 ( 2 - ^ ) = 2n - 1,

2n 2 n
m i n / ( x i , x 2 , . . - , a : n ) . = -2n
—-—1 - ra = 2n - 1

and with equality just in case each X{ = 1/ra.

Generalization. Let (?(xi,£2, •. •, £ n ) = X3?=i F(xi/(a — re,-)) where the a;t-'s are
nonnegative and with sum s and F is a non-increasing convex function over the range of
X{/(a — X{) where a > s. Since x/(a — x) is a convex function in [0, a), F(x/(x — a)) is also
convex. Hence by the majorization inequality (see [1988: 120] or [1]),

s \ , x „,„^ ^, , n
F ( - L - ) + ( n - 1)F(0) > G(xux2,...,xn) >
\a — s/ as — 1

Here the maximum value is achieved for the n-tuple (5, 0, 0 , . . . , 0) and the minimum value
is achieved for the n-tuple (1/s, 1/s,.. - , 1/s).
Reference:
[1] A.W. Marshall, I. Olkin, Inequalities: Theory of Majorization and its Applications,
Academic Press, N.Y., 1979.
2 . Fourth Test, (a) Prove that there exist positive real numbers A such that [Xn]
and n have the same parity for all positive integers n. ([•] denotes the greatest integral
part function.)
(b) Find one such number A.
 Ill

Solution by Christopher J. Bradley, Clifton College, Bristol, U.K.

Consider the sequence 7 , 5 7 , 4 2 7 , . . . defined by u\ = 7, u2 = 57, and u n + 2 = 7un+1 +
4u n , n > 1. First observe that with u x , u 2 odd it follows that un is odd for all positive
integral n. Secondly by standard methods on recurrence relations, it is readily shown that
Un = An + Hn where A, \L are the roots of the quadratic equation x2 — 6x — 4 = 0. We
take A = | ( 7 + V65) and // = | ( 7 — \/65). The claim is that A satisfies the conditions of
the problem, namely [An] — n = 0 mod 2, for all positive integral n. The reason for this
is that /I is small and negative, that is — 1 < \i < 0, from which it follows that fin lies
between —1 and 0 when n is odd, and between 0 and 1 when n is even. Since un is odd for
all n, it follows that [An] is even when n is even, and odd when n is odd. Part (a) of the
problem asks for the existence of such numbers (plural) and the same construction holds
with different sequences. For example, A = | ( 3 + \/T7) is another case.
3o Fourth Test. Suppose convex hexagon ABCDEF can be inscribed in a circle.
Prove that a necessary and sufficient condition for the three diagonals A D , BE and CF
to be concurrent is AB • CD • EF = BC • DE • FA.
Solution by Christopher J. Bradley, Clifton College, Bristol, U.K.
Suppose first that AD, BE, and CF
are concurrent at X. Now triangles ABX and
EDX are similar with LB AX = LDEX, etc.
So AB/DE = BX/XD. Similarly CD/AF =
XD/FX and EF/BC = FX/BX. It follows
immediately that AB • CD • EF = BC • DE • FA.
Suppose next that AB -CD • EF = BC • DE • FA.
Let AD and BE meet at X. Produce CX to meet
the circle again at F'. We want to show that F
and F' coincide. By the first part

AB-CD-EF'=:BC-DE-FfA.

Hence EF/EFf = FA/FfA.(*)

Now in the case with F' nearer to A than F , we would have F'A < FA and
EF < EF' and so {F'A){EF) < (FA)(EF') and (*) could not hold. It follows that F'
must lie on arc FE. Similarly it must lie on the arc FA. Hence F' and F coincide.

That concludes the file of solutions to problems from the March 1992 number as
well as this month's Corner. Send me your Olympiad and pre-Olympiad contests as well
as vour nice solutions.

* *
112

PROBLEMS
Problem proposals and solutions should be sent to B. Sands, Department of
Mathematics and Statistics, University of Calgary, Calgary, Alberta, Canada T2N 1N4-
Proposals should, whenever possible, be accompanied by a solution, references, and other
insights which are likely to be of help to the editor. An asterisk (*) after a number indicates
a problem submitted without a solution.
Original problems are particularly sought. But other interesting problems may also
be acceptable provided they are not too well known and references are given as to their
provenance. Ordinarily, if the originator of a problem can be located, it should not be
submitted by somebody else without permission.
To facilitate their consideration, your solutions, typewritten or neatly handwritten
on signed, separate sheets, should preferably be mailed to the editor before N o v e m b e r 1,
1993, although solutions received after that date will also be considered until the time when
a solution is published.

1 8 3 1 * . Proposed by Walther Janous, Ursulinengymnasium, Innsbruck, Austria.

Let x, y, 2 be any real numbers and let A be an odd positive integer. Prove or
disprove that
x{x + y)x + y(y + z)x + z(z + x)x > 0.

1 8 3 2 . Proposed by K.R.S. Sastry, Addis Ababa, Ethiopia.

An old unsolved problem is: "is it possible that a box can have its sides, face
diagonals, and space diagonal all of integer lengths, i.e., are there positive integers a, 6, c
such that
a 2 + 62, 62 + c 2 , c2 + a2, and a2 + b2 + c2
are all perfect squares?" What if we replace the squares by triangular numbers? For n a
positive integer, let tn — n(n + l ) / 2 be the n t h triangular number.
(a) Find positive integers a, 6, c such that

are all triangular numbers.

(b)* Is there such a solution so that ta + £& + tc is also a triangular number?
1 8 3 3 « Proposed by Toshio Seimiya, Kawasaki, Japan.
E and F are points on sides BC and AJD, respectively, of a quadrilateral ABCD.
Let P = AE n BF and Q = CF n DE. Prove that E and F divide BC and AD (or BC
and DA) in the same ratio if and only if

[FPQ] = [EQP]
[EDA] [FBC] '

where [XYZ] denotes the area of triangle XYZ.

 113

1834. Proposed by Marcin E. Kuczma, Warszawa, Poland.

Given positive numbers A, G and H, show that they are respectively the arithmetic,
geometric and harmonic means of some three positive numbers x,y,z if and only if

- +- +1^(1 +-) .

1835. Proposed by Joaquin Gomez Rey, LB. Luis Bunuel, Alcorcdn, Madrid,
Spain.
Evaluate

for n = 1 , 2 , 3 , . . . .
1 8 3 6 * Proposed by Jisho Kotani, Akita, Japan.
Let ABCD be a quadrilateral inscribed in a circle T, and let ACDBD = P. Assume
that the center of T does not lie on AC or BD. Draw circles with diameters AB, BC',
C D , DA, and let the areas of the moon-shaped regions inside these circles and outside T
be Fu F 2 , F 3 , F4. Mu M 2 , M 3 , M4 are the midpoints of the sides of ABCD, and Hu H2,
^ 3 , H4 are the feet of the perpendiculars from P to the sides of ABCD. Prove that, if
F1 + F2 + F3 + F4 = area(ADCD), then Mi, M 2 , M 3 , M 4 , # i , # 2 , # 3 , # 4 are concyclic.
1837. Proposed by Andy Liu, University of Alberta.
A function / : 1R —> I R + is said to be strictly log-convex if

2
/ ( x , ) / ( x 2 ) > ( / ( ^ ) )

for all £i, x 2 6 IR, with equality if and only if xi = x 2 . / is said to be strictly log-concave
if the inequality is reversed.
(a) Prove that if / and g are strictly log-convex functions, then so is / + g.
(b)* Does the same conclusion hold for strictly log-concave functions?
1838. Proposed by Stoyan Kapralov and Iliya Bluskov, Technical University,
Gabrovo, Bulgaria.
Find all sequences a\ < a 2 < • • • < an of positive integers such that

ax + a2 + - • • + an = 26, a\ + a\ + • • • + a\ = 62, a* + a 2 + • • • + a\ = 164.

1839. Proposed by N. Kildonan, Winnipeg, Manitoba.

Notice that
122 = l l 2 + 1 = 1 2 2 - 2 2 ,
i.e., the (base 10) integer TV = 122 can be partitioned into two parts (1 and 22), so that
the first part is the difference between TV and the greatest square less than N, and the
second part is the difference between N and the least square greater than N. Find another
positive integer with this property.
114

1 8 4 0 * Proposed by Jun-hua Huang, The J^th Middle School of Nanxian, Hunan,

China,
Let AABC be an acute triangle with area F and circumcenter O. The distances
from O to P C , C A, AB are denoted da, d&, dc respectively. AA1B1C1 (with sides e&i, 61,
c\) is inscribed in A ABC, with A\ G P C etc. Prove that

4 ^ i + 4&i + dcci > F.

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

1 6 2 1 * . [1991: 78; 1992: 88] Proposed by Murray S. Klamkin, University of Alberta.

(Dedicated to Jack Garfunkel.)
Let P be a point within or on an equilateral triangle and let ci < c2 < C3 be the
lengths of the three concurrent cevians through P. Determine the minimum value of C2/C3
over all P ,
II. Comment by the proposer.
In Baron's solution [1992: 89], he uses calculus to determine the minimum of

(l + 3^2)2
1 + 9s2

More elementarily, by letting 1 + 9s 2 = t, we find that

(l+3.2)2_(* + 2 ) 2
_ l ^ , 4 , ^ _ 1 (Q , (n 2\z\
! > 8
1 + 9s 2
9< £0*H-§('+(*-*)
9V t J 9\ V y/ij J~ 9'
with equality when t = 2, i.e. s = 1/3.

1 7 4 1 . [1992: 138] Proposed by Toshio Seimiya, Kawasaki, Japan.

ABCD is a convex cyclic quadrilateral, and P is an interior point of ABCD such
that LB PC = LBAP + IP DC. Let E, F and G be the feet of the perpendiculars from P
to AB, AD and DC. Prove that AFEG is similar to APBC.
 115

Combination of solutions by Christopher J. Bradley, Clifton College, Bristol, U.K.,

and the proposer.
We denote the circumcircles of
ABCB, APAB and APCD by I \ I \ and
T2 respectively. Draw a ray PT in LB PC
such that LBPT = LBAP; then we get

LTPC = LBPC - LBPT

= LBPC - LBAP = LPDC.

Therefore PT is tangent to both Fi and r 2 .

As AB and CD are the common chords of
I \ , r and F 2 , r respectively, AB, CD and
PT are concurrent at the radical center Q of
Ti, T 2 and X\ As PQ is a common tangent
to Ti and r 2 at P , we get

LQPA = LPBA. (1)

Since LAPP = LAPP = 90°, A,E,P,F are

concyclic cind we have

LFEP = LFAP = ZDAP. (2)

Similarly, F, D, G, P are concyclic, and

LBPC = LBAP + LPDC = Z £ F P + Z P P G = LEFG. (3)

Since Z P J B Q = LPGQ = 90°, Q, F , P, G are concyclic and we have

LGEP = LGQP = LDQP. (4)

From (2) and (4), and LDAP + LQPA = LQDA + LDQP, we get

LFEG = LFEP ~ LGEP = LDAP - LDQP = LQDA - LQPA. (5)

As A, P , C , D are concyclic, we have LQDA = LQBC. Then we obtain from (1) and (5)
that
LFEG = Z Q £ C - Z P P A = LPBC. (6)
By (3) and (6), AFEG is similar to APBC.
When Q i s a point on A S produced beyond J3, the proof is similar. And when
AB\\CD, we get PT\\AB\\CD and then J5, P , G are collinear, so we have APAD =
APBC. From A F £ G - A P A D we get A F £ G - APBC.
Also solved by D.J. SMEENK, Zaltbomrnel, The Netherlands.
Smeenk in fact notes that the locus of point P is an arc of the circle with centre Q
which is orthogonal to T.
116

1 7 4 2 . [1992: 139] Proposed by Murray S. Klamkin, University of Alberta.

Let 1 < r < n be integers and x r + i , x r + 2 , . . -, x n be given positive real numbers.
Find positive Xi, X 2 , . . . , x r so as to minimize the sum

taken over all z, j G {1, 2 , . . . , n} with i -=/=• j .

(This problem is due to Byron Calhoun, a high school student in McLean, Virginia.
It appeared, with solution, in a science project of his.)
Solution by Jun-hua Huang, The 4th Middle School of Nanxian, Hunan, China.
Let
1 1 1
x = xi + x2 H h xr, y = - 1 !-••• + — ,
1 1 1
a — x r +i + xr+2 + - - + xn, b— • • + — + ••• + — .
Then
X{ •% -\ Xi vr—^ X{ \ -\ X{

X
i^j J l<i^j<r ^3 T<ijij<n ^3 l<i<r *^J r<i<n ^3
r<j<n l<J<r

= x1 (y - •) + x 2 (y - — ) + ••• + xr (y - — )
V X\J \ x2J V xrJ
+xr+i b — —— + xr+2 6- — H h xn ( 6
\ Xr+1J \ Xr+2J \ Xn
+{xx + x2 + • • • + xr)b + ( — + — + ••• + — ) a
\X\ X2 Xr J
= xy — r + ab — (n — r) + xb + j/a = (x + a)(y + 6) — n
> (y/xy + vab)2 —n [true by Cauchy's inequality]
2
> (r + \/o6) - 7v
[equivalent to £ ^ = 1 x; • X^=i I/31* ~ r 2 ' w hich is again true by Cauchy's inequality],
which is independent of x i , . . . , xr.
Equality holds in the first inequality if and only if x/y = a/b, and in the second
inequality if and only if x\ — x2 = • • • = x r ( = x / r = r/y). So 5 attains its minimum
value if and only if
#1 — #2 ~ ' " * — xr — y T •

A/so solved by SEUNG-JIN BANG, University of California, Berkeley; C. J.

BRADLEY, Clifton College, Bristol, U.K.; WALTHER JANOUS, Ursulinengymnasium,
Innsbruck, Austria; MARGIN E. KUCZMA, Warszawa, Poland; P. PENNING, Delft,
The Netherlands; IGNASI MUNDET I RIERA, Barcelona, Catalunya, Spain; and the
proposer.
The proposer gave a generalization to the sum ^ ^ / x ^ where p^q > 0.
 117

1 7 4 3 * . [1992: 139] Proposed by Walther Janous, Ursulinengymnasium, Innsbruck,

Austria.
Let 0 < 7 < 180° be fixed. Consider the set A(7) of all triangles ABC having
angle 7 at C, whose altitude through C meets AB in an interior point D such that the
line through the incenters of AADC and ABCD meets the sides AC and BC in interior
points E and F respectively. Prove or disprove that

/ a r e a ( A £ F C ) \ _ / c o s ( 7 / 2 ) - sin( 7 /2) -f l \ 2
rjJJ Varea( A A B C ) J 2cos(7/2)

(This would generalize problem 5 of the 1988 IMO [1988: 197].)

Solution by Toshio Seimiya, Kawasaki, Japan.
We shall prove that the given relation does hold.
Let the incenters of AADC and ABCD C
be P and Q, and put 7/2 = e, CD = h, and
LACP = LPCD = 9, LDCQ = LQCB = y>.
Then 0 + <p = e, where 0° < e < 90°.
Let [XYZ] denote the area of triangle
XYZ. Because

[CEP] + [CPQ] = [CEQ],

we have

CE-CPsinfl C P - C g s i n e _ CE-CQsm(e + 9)
+
2 2 ~ 2
which implies
sine sin(£ + 0) sin#
(1)
CE CP CQ '
By the law of sines for ACPD,

CP
= \/2CP,
sin(45° + 0) sin 45°

thus we get

1 V2smU5° + 6) , . ., , 1 V2sin(45° + y )
— —- and similarly ••-- (2)
CP h CQ h

From (1) and (2) we have

_sine
_ =
x/2sin(e + 0)sin(45°
__ + 0) y/2 sin 6 sin(45°
_ + if)
118

Therefore

^ ' m £
= 2 sin(£ + 9) sin(45° + 9) - 2 sin 9 sin(45° + <p)
CE
= [cos(45° - e) - cos(45° + e + 20)} - [cos(45° + <p - 9) - cos(45° + 9 + ^)]
= [cos(45° - e) + cos(45° + e)\ - [cos(45° + £ + 20) + cos(45° + y - 9)\
= 2 cos 45° cos e - 2 cos(45° + e) cos 29
— A/2 cos e — A/2(COS £ — sin e) cos 20.

Thus we have
C£= ,.fe5in£ x „,. (3)
COS 6 + (sill £ — COS 5) COS 26
Similarly we get
CF=- _Ai^£___. (4)
cos £ + (sin £ — cos e) cos 2<£>
We have CAcos2# = h = CB cos2<p, therefore we obtain from (3) and (4)

[EFC] CE-CF sin 2 e cos 29 cos 2y (5)

[AJBC] CM • CB [cos£ + (sin£ — cos£) cos2^][cos£ + (sin£ — COSE) cos2y?]

and note

cos 29 cos 2(p = -[cos(2<? + 2(f) + cos(20 - 2<p)] = -[cos 2£ + cos 2(9 - 97)]

= Ul - 2 s m 2 £ + 2 c o s 2 ( 0 - < ^ ) - l ] = cos2(9-ip) - sin 2 e. (6)

Case I: e = 45°. Then sine — cose = 0, and therefore we get from (5) and (6)

[EFC] sin 2 e[cos 2 (0 - <p) - sin 2 e] 2//1 N 1 ^ 1

2
[ABC] cos £
(equality holding when 6 = ip). Also

/cos(7/2)-sin(7/2) + l \ 2 ^ 1 = 1
\ 2cos(7/2) J 4 cos 2 45° 2 '.

and hence the given relation holds. [Editor's note: this case is the IMO problem mentioned
above.]
When e 7^ 45°, we have sine — cose 7^ 0. We put
cos e 7 9 . 9
a = __ ^ 6 = a — sin e ;
sin £ — cos e
then (5) becomes

[EFC] sin 2 e cos2(0-^)-sin2e

[ABC] (sine-cose)2 (a + cos20)(a + cos2y?) 7
 119

where by (6)

(a + cos 29)(a + cos 2<p) = a2 + a(cos 20 + cos 2c/?) + cos 26 cos 2y>
= a 2 + 2a cos(# + ip) cos(0 — 9?) + cos 2 (0 — ^ ) — sin 2 e
= a 2 + 2a cos e cos(9 — ip) + cos 2 (0 — <£>) — sin 2 £
= cos 2 (0 — y>) + 2a cos e cos(# — <p) + b.

Therefore

[EFC] _ sin 2 e cos2(fl-y)-sin2e

[ABC] (sine — cose) 2 cos 2 (# — tp) + 2acosecos(0 — tp) + 6 (7)

Put x = cos(# — 9?), where cose < x < 1. We consider

x 2 — sin 2 e
/(*) x2 + 2a cos e - x + h

for which

, \ __ 2x(x 2 + 2a cos e - x + b) — 2(x + a cos e)(x2 — sin 2 e)

* ^ ~ (x 2 + 2a cos e • x + b)2 "
2[a cos e • x2 + (6 + sin 2 e)x + a sin 2 e cos e]
(x 2 + 2acose • x + b)2
2acos e ( 9 # . 2 \ /ox
= — — x2 + — + sm 2 £ . (8)
(x 2 + 2a cos e • x + b)2 V sin £ — cos e /
Case II: e > 45°. Then sine — cose > 0, so we have f'(x) > 0. Hence / ( x ) is
increasing in cose < x < 1 and has maximum at x = 1, i.e., when 6 = (/p. Therefore by (7)
the maximum value of [EFC]/[ABC] is

sin 2 e 1 — sin 2 e sin 2 e cos 2 e

(sine — cose) 2 1 + 2a cose + b (sine — cose) 2 (l + 2a cose + a 2 — sin 2 e)
sin 2 e cos 2 e sin 2 e cos 2 e
sine — cose) 2 (a + cose) 2 [cose + (sine — cose) cose] 2
sine \2 (\ — sine + c o s e \ 2 _ / s i n e ( l — sine + cose)
, 1 + sin e — cos e / \ 1 — sin e + cos e / \ 2 sin e cos e
- fcos(7/2)-sin(7/2) + l \ 2
~ V 2cos( 7 /2) ) '
as claimed.
Case III: 45° > e > 0°. Then sin£ - cose < 0. We put
X
#(x) = x 2 + — h sin 2 e. (cose < x < 1)
sin e — cos e
120

Then
/ \ 2 cose . 2 sine
g(cos £) = cos £ + — h sin e = — — < 0,
and sin e — cos e sin 6 — cos £
,., 1 .9 sin£ — cos£ + 1 + sin 2 £(sin£ — cos£)
0(1) = 1 + _ + sin2£ = _ _ _ i . L
sin £ — cos e sin £ — cos e
where
sin e — cos e + 1 + sin 2 £(sin £ — cos e) > sin 2 e — cos e + 1 + sin 3 £ — sin 2 £ cos £
= (1 - cos £)(1 + sin 2 e) + sin 3 e > 0,

therefore g ( l ) < 0. Because g{x) is a convex function, we have g{x) < 0 for cos£ < x < 1.
Since a < 0, we get from (8) that /'(a;) > 0. Hence / ( # ) is increasing in cos£ < x < 1 and
has maximum at x = 1, i.e., when 8 = Lp. Therefore as in Case II the maximum value of
[EFC]/[ABC] is again
/cos(7/2)-sin(7/2) + l \ 2
V 2cos( 7 /2) | "
[Editor's note: in all cases, the given maximum occurs when CA — CB, i.e., when
EF\\AB]

1 7 4 4 . [1992: 139] Proposed by Vaclav Konecny, Ferris State University, Big

Rapids, Michigan.
Find the number of points of intersection of the graphs of y = ax and y = log a x for
any 0 < a < 1.
Solution by Christopher J. Bradley, Clifton College, Bristol, U.K.
First observe that
y = l°gax & x = ay,
so clearly y = ax, y = log a x has at least one solution with x = y, for all a satisfying
0 < a < 1. In fact when a = l / c c (c > 1) we have a value of a satisfying 0 < a < 1, and
since the function l / c c decreases continuously from 1 as c increases from 1, and tends to
0 as c —> oc, every a corresponds to just one value c. Since (l/c0)1/0 = 1/c, there is thus
only one solution of the type x — y, namely x = y = 1/c when a = l / c c .
Now note that x*2 = xx^ has solutions with X\ ^ £2, the totality of which may be
identified as follows. Consider the graph of
lnX
X
(capitals used here so as not to cause confusion with the x,y of the question). For any
value k of Y in the interval 0 < Y < 1/e draw a line parallel to the X-axis. This meets
the graph in exactly 2 points (#1, k) and (#2, &) satisfying
lnzi In X2
Xi X2
 121

for which x\2 = xXl. The value k = 1/e is excluded because Y — 1/e is tangent to
Y — (InX)lX at X = e where the curve has its (maximum) turning point. Furthermore,
for any pair of such values Xj,x2 we have xX2 = xx^ > e e , and conversely, if c > e e unique
x1,x2 exist with xX2 = x2l — c.
In terms of the original problem, this implies that for 0 < a < l / e e there are two
more solutions of the simultaneous equations y — ax and x = ay. For taking such an a we
can then find corresponding x1,x2 with xX2 = x2x = 1/a and set either x = l/a?i, y = l/x2
or x — l/x2, y = l/a:i, and then we have

i /1 y/x2 , i /1 y/xi ., i i
= a =
— I ~^2 I — ( ~~^f J ' with a = —^ = —£- .
X\ \%l / *^2 V *^ 2 / *^ 1 *^2
,r y
It is now claimed that all solutions of the equations y = a , x = a derive from one
of the above two methods. For if the equations hold then In 3/ = x In a and In x — y In a so
xhix = ylnj/. Now put a; = l / s i and y = l / x 2 ; then

1 , ,( 1 \ 1 , ,f 1
— In — ) = — In => ^1 In 2:2 = o ^ l n ^ i ==> x^1 = x%2
x
Xi 'V x i / X2 'Kx2

So for 1 > a > l / e e there is one solution and for 0 < a < l / e e there are three.
(I suppose some might say that a = l / e e leads to a triple root.)
Also solved by H.L. ABBOTT, University of Alberta; CHARLES ASHBACHER,
Cedar Rapids, Iowa; FRANCISCO BELLOT ROSADO, LB. Emilio Ferrari, Valladolid,
Spain; DAVID DOSTER, Choate Rosemary Hall, Wallingford, Connecticut; HANS
ENGELHAUPT, Franz-Ludwig-Gymnasium, Bamberg, Germany; MARCIN E. KUCZMA,
Warszawa, Poland; P. PENNING, Delft, The Netherlands; UNIVERSITY OF ARIZONA
PROBLEM SOLVING GROUP, Tucson; and the proposer. Four incorrect solutions were
sent in.
Ashbacher points out that the problem appeared as problem 4334 in School Science
and Mathematics (proposed by its editor Richard A. Gibbs), with solution on pages 292-293
of the May/June 1992 issue.
The equation xy = yx has been frequently studied. See for instance Mathematics
Magazine, Vol. 63 (1990) 30-33, or Mathematical Gazette, Vol. 71 (1987) 131-135.
>jc ^ * * *

1 7 4 5 . [1992: 139] Proposed by Juan Bosco Romero Mdrquez, Universidad de

Valladolid, Spain.
Let A = A0Ai A2A3 be a square and X an arbitrary point in the plane of A. Define
the fcth vertex (k an integer mod 4) of the quadrilaterals B, C and V as follows:
(i) Bk is the midpoint of XAk\
(ii) Ck = AhBk+l n Ak+iBk]
(iii) Dk is the centroid of the quadrilateral BkAkAk+iBk+i-
Prove that B, C and V are squares whose centres lie on a line through the centre of A.
122

I. Comment by the editor (in this case Chris Fisher).

Before looking at the solution, note that the definition of the square V is ambiguous
since, as pointed out in the solution to Crux 1527 [1991: 150-151], there is no universally
approved meaning of centroid. Among the seven submitted solutions, one was independent
of the choice of centroid, three interpreted Dk to be the centroid of equal masses placed
at the vertices of the quadrangle BkAkAk+iBk+i, while the remaining three took Dk to be
the centroid of the quadrangular lamina. Of course, another interpretation seems possible
for V: Dk might be the centroid of the boundary of the quadrangle; no solver dealt with
this possibility, which would produce a counterexample to the claims made for V because
this centroid is not an affine invariant. A recent discussion of the three types of centroids
accompanies the solution to E3283 in the Amer. Math. Monthly 97 (1990) 849-850.
II. Combination of solutions by Jordi Dou, Barcelona, Spain; by Dan Pedoe,
Minneapolis, Minnesota; and by D.J. Smeenk, Zaltbommel, The Netherlands.
The solution is based on two observations:
(1) the image of a square under a dilatation having centre X is a square, with the
centres of the two squares collinear with X; and
(2) if M is the square formed by the midpoints of the sides of *4, then A and A4
have the same centre.
The dilatation D(X, 1/2) — with centre X and ratio of magnification 1/2 — takes
A to 5 , while D(X, 2/3) takes M to C. In the vertex case D{X, 3/4) takes M to £>, while
the dilatation D(X, 7/9) works for the lamina. [Details of the calculation can be deduced
from the cited Crux problem.] So, whichever of the two affine choices is made for D, the
centres of B and of A, of C and of M, and of V and of M are collinear with X , so that
the centres of .4, B, C, and V lie on a line through X as desired.
Also solved by CHRISTOPHER J. BRADLEY, Clifton College, Bristol, U.K.;
WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria; P. PENNING, Delft,
The Netherlands; CHRIS WILDHAGEN, Rotterdam, The Netherlands; and the proposer.
The proposer added various generalizations to his problem. In fact, one sees easily
that the featured solution invites a frenzied orgy of generalization:
Forn > 3 and A > 0, let A = AoAi .. .An-i be any convex n-gon, k be an integer
mod n, and the point Bk divide XAk in the ratio A : 1. If B, C, and V are defined as
in the original proposal, then they are similar to A and the centroids of their vertex
sets lie on a line through the centroid of the vertex set of A.
The dilatations required to verify the generalization have the ratios A for B,
2A/(1 -f A) for C, and either (A + l ) / 2 for D (when Dk is the centroid of the vertex set of
the quadrangle) or 2(A2 + A + 1)/(3(A + 1)) (when Dk is the centroid of the quadrangular
lamina).
 123

1 7 4 6 . [1992: 139] Proposed by Richard K. Guy, University of Calgary, and Richard

J. Nowakowski, Dalhousie University.
(i) Find infinitely many pairs of integers a, 6, with 1 < a < 6, so that ab exactly
divides a2 + b2 — 1.
(ii) With a and b as in (i), what are the possible values of (a 2 + b2 — l ) / a 6 ?
Solution by Richard Mcintosh, University of Regina.
Let a and b be integers such that 1 < a < b. We claim that ab divides a2 + b2 — 1
if and only if a and b are consecutive terms in the sequence

u2 - 1
Uo — 1 , U i — AC, ... , Un+1 — , ... ,

where k > 1 is an integer. Moreover, (a2 + b2 — l)/ab = fc and thus can take any integer
value greater than 1.
Proof. Let AC > 1 be any integer. Since

< + "n+l - 1 =ttnMn-l+ K - l ) 2 - * 4 - l = *4-l +1^-1

(u 2 - l)un_i Un-xlln

it follows by induction that if a and 6 are any two consecutive terms in the sequence then

a2 + b2 - 1 __ u2 + u\ - 1 _ 1 + k2 - 1

All pairs (a, 6) with 1 < a < b such that ab divides a2 + b2 — 1 arise in this fashion.
If a = 1 then put k = b and we are done. Now suppose that 1 < a < b and a6 divides
a2 •+ b2 — I. Let c = (a 2 — l ) / 6 so that c is an integer, 1 < c < a, and

c2 + a2 _ t ^ (q2-l)2 + (a2_1^2 ^ Q2 + &2 _ j

ca (a2 — l)ab ab

Hence the pair (a, b) gives rise to the smaller pair (c,a) satisfying the same divisibility
criterion. This process can be continued until we obtain t h e pair (l,fc), where
k = (a 2 + b2 — l)/ab. Therefore (a, 6) = ( u n , u n + i ) for some n. [Editor's comment Note
that the equation c = (a 2 — l)/fe says that the sequence 6, a, c , . . . ,fc,1 generated by this
process is just a sequence of u n 's in reverse order.]
Also solved by H.L. ABBOTT, University of Alberta; HAYO AHLBURG, Benidorm,
Alicante, Spain; CHARLES ASHBACHER, Cedar Rapids, Iowa; SAM BAETHGE,
Science Academy, Austin, Texas; MARGHERITA BARILE, student, Universitdt Essen,
Germany; CHRISTOPHER J. BRADLEY, Clifton College, Bristol, U.K.; DAVID
DOSTER, Choate Rosemary Hall, Wallingford, Connecticut; HANS ENGELHAUPT,
Franz-Ludwig-Gymnasium, Bamberg, Germany; RICHARD I. HESS, Rancho Polos
Verdes, California; RANDY HO, student, University of Arizona, Tucson; JUN-HUA
HUANG, The 4th Middle School of Nanxian, Hunan, China; WALTHER JANOUS,
124

Ursulinengymnasium, Innsbruck, Austria; FRIEND H. KIERSTEAD JR., Cuyahoga Falls,

Ohio; ANDY LIU, University of Alberta; DAVID E. MANES, State University of New
York, Oneonta; KAAREN MAY, student, St. John's, Newfoundland; J.A. MCCALLUM,
Medicine Hat, Alberta; P. PENNING, Delft, The Netherlands; R.P. SEALY, Mount
Allison University, Sackville, New Brunswick; DAVID R. STONE, Georgia Southern
University, Statesboro; P. TSAOUSSOGLOU, Athens, Greece; CHRIS WILDHAGEN,
Rotterdam, The Netherlands; and the proposers. Part (i) only (probably due to misunder-
standing part (ii)j solved by C FESTRAETS-HAMOIR, Brussels, Belgium; BEATRIZ
MARGOLIS, Paris, France; W.R. UTZ, University of Missouri, Columbia; and
KENNETH M. WILKE, Topeka, Kansas.

1 7 4 7 . [1992: 139] Proposed by K.R.S. Sastry, Addis Ababa, Ethiopia.

ABCDE is a convex pentagon in which each side is parallel to a diagonal. Two
of its angles are right angles. Find the sum of the squares of the sines of the other three
angles.

Solution by Jordi Dou, Barcelona, Spain.

The right angles can not be at consecutive
vertices, so we can suppose that LC = LE = 90°.
Put F = BC D AE, and let O be the midpoint of
DF. We may put OC = OD = OE = OF = 1
[since CDEF is cyclic with diameter DF]. We
have AD\\FC, AB\\CE, BE\\CD, BD\\AE. [Note
that BDAF is a parallelogram, so O is the mid-
point of AB.—Ed] Put

a = LDFC = LABE = LBEC = LECD.

[Editor's note. Here Dou seems to use symmetry

of the figure beyond what has been established.
All would be okay if we knew that DF JL AB, i.e.,
AD = BD. This can be proved as follows. From F
AE\\BD and AD\\BC we know LEAD = LDBC,
and since LE = LC = 90° we have AEAD - ACBD. If AE < BC then also DE < DC
and, since EC\\AB,

90° = LAEC + LDEC > LBCE + LDCE = 90°,

a contradiction. Similarly AE > BC is impossible, so AE = BC, thus AEAD = ACBD

and AD = BD. Or is it easier than this?]
It holds that
LA^LB = 90° + a, LD - 1 8 0 ° - 2a; (1)
 125

also /.COD = 2a, and so CE — 2 sin 2a. From this and

1 cos 2 a
CB = FC -FB = 2cosa-
cos a cos a
we have
CB cos 2a
Sma=
CE = cos a • 2 sin 2a '
or
cos 2a = sin 2 2a. (2)
From this,
- 1 = 4 sin 2 a cos 2 a = 4 cos 2 a -
4 cos 4 a - 2«cos2 a — 1 = o,

so
_ 1 + y/E
cos 2 a = (3)

Now using (1), (2) and (3),

sin 2 A + sin 2 B + sin 2 D = 2 cos 2 a + sin 2 2a = 2 cos 2 a + cos 2a

= 4 cos 2 a — 1 = Vo.

A/so so/?;ed 6y iL47<9 AHLBURG, Btnidorm, Spain; MIGUEL AMENGUAL

COVAS, Gala Figuera, Mallorca, Spain; CHRISTOPHER J. BRADLEY, Clifton College,
Bristol, U.K.; HANS ENGELHAUPT, Franz-Ludwig-Gymnasium, Bamberg, Germany;
C FESTRAETS-HAMOIR, Brussels, Belgium; RANDY HO, student, University of
Arizona, Tucson; L.J. HUT, Groningen, The Netherlands; WALTHER JANOUS,
Ursulinengymnasium, Innsbruck, Austria; MARCIN E. KUCZMA, Warszawa, Poland;
P. PENNING, Delft, The Netherlands; TOSHIO SEIMIYA, Kawasaki, Japan; D.J.
SMEENK, Zaltbommel, The Netherlands; and the proposer.
The proposer also wonders about the maximum number of right angles possible in a
convex n-gon A1A2 ... An in which A{Aj\\AkAi whenever i+ j = k + 1 mod n (an "affinely
regular" n-gon — the case n = 5 is the above situation).

1748. [1992: 140] Proposed by David Singmaster, South Bank Polytechnic,

London, England.
Several medieval arithmetic/algebra books give a problem with a fountain located
between two towers so that pigeons of equal speeds can get from the tops of the towers to
the fountain in equal times. The solution of this works out quite easily. However, in the
Trattato dAritmetica attributed to Paolo delTAbbaco, c. 1370, there are problems where
a rope is strung between the two tower tops and a weight is hung from a ring on the rope,
where the rope is just long enough for the weight to touch the ground. Solving this is a bit
trickier than the previous problem. To make it even trickier, suppose the rope isn't long
enough—suppose the towers are 1748 and 1992 uncia high, the rope is 2600 uncia long,
and the two towers are 2400 uncia apart. How far above the ground does the weight hang?
126

I. Solution by Dag Jonsson, Uppsala, Sweden

For notation see the figure. The rope forms
the same angle a at A and B (for the weight P does
not change its position if the rope is fixed at Q).
Thus, if the line BP is extended till A' we have
\AP\ = \A'P\ and \A'B\ = 2600. Since \A'D\ =
2400 we get \BD\ = V2600 2 - 24002 =" 1000. Now
\BC\ = 1 9 9 2 - 1748 = 244, \CD\ = 1 0 0 - 2 4 4 =
756, and \AP'\ = \P'A'\ = 756/2 = 378. Thus the
weight hangs 1748 — 378 = 1370 uncia above the
ground.

II. Solution by Hayo Ahlburg, Benidorm, Alicante, Spain.

This problem was inspired by work done hundreds of years ago. Its solution is
supplied by work done thousands of years ago. F2
The size of ring and weight are considered
to be negligible. Sliding along the tight rope which
is tied to the two tower tops (focal points F\ and
F2), the weight describes part of an ellipse. It
comes to rest at the lowest point P , where the
ellipse has a horizontal tangent t. The rope length
F\P + PF2 is equal to the large axis AB of the
ellipse,

F1P + PF2 = AB = 2600 uncia. (1)

Let the center of the ellipse (midpoint of F\F2 and of AB) be C. L and M are the
points where t intersects with the vertical towers F\G and F2H (which are normals to the
horizontal tangent t) and N is their midpoint. G, i?, and their midpoint J are assumed at
the same (horizontal) ground level, although this is not explicitly stated in the problem,
and you never know about those medieval towers. From GH = 2400 uncia, we have

GH
GJ = HJ = LN = MN = = 1200 uncia. (2)

With F\G = 1748 uncia and F2H = 1992 uncia we furthermore have

FiG + F2H
CJ = „ = 1870 uncia. (3)

Now let us cast our minds back 2200 years, to the great Apollonios. To him, the
figure was quite familiar, and he would have been happy to point out that

AB
CL = CM = CA=CB = = 1300 SaKTvXot} (4)

As he called uncia (inches) in his day.

 127

He had proved that A, L, M and B are points of the ellipse's circumcircle [1].
Using (1) to (4), we now simply have

2 2
CN = VCM* - MN* = W ( ^ ) - (^r) = 500 S&KTVXOI (5)

and
NJ = CJ-CN = 1370 SaKTvXot. (6)
This is also the elevation of P above ground.
If P were touching ground, the line LPNM would already signify ground level, and
the solution would stop at (5). Thanks to Apollonios — and to Pythagoras, of course —
solution (5) is not tricky, and (6) is not trickier, but "elementary, dear Watson".
By the way, since LLPFX = LMPF2 [2], we have

LP :PM = FXL : F2M = 378 : 622.

With LP + PM = 2400 SaKrvXoi, this gives

LP = 907.2 SaKTvXot and PM = 1492.8 SaKrvXoc.

References:
[1] It is a pleasure to honor this great mathematician whose works are so rarely quoted
today: A-KOXXUJVLOV Hspjalov KUJVLKUJV 7', 1/ (Apollonios of Pergae, c. 220 B.C.,
Conies III, 50).
[2] 'ATTOWUH/LOV Uepjalov KLOULKUJI/ 7', /i7/' (Conies III, 48).
Also solved by SAM BAETHGE, Science Academy, Austin, Texas; HANS
ENGELHAUPT, Franz-Ludwig-Gymnasium, Bamberg, Germany; RICHARD L HESS,
Rancho Palos Verdes, California; FRIEND H. KIERSTEAD JR., Cuyahoga Falls, Ohio;
KEE-WAI LAU, Hong Kong; J.A. MCCALLUM, Medicine Hat, Alberta; P . PENNING,
Delft, The Netherlands; P. TSAOUSSOGLOU, Athens, Greece; and the proposer. One
incorrect solution was sent in.

1 7 5 0 . [1992: 140] Proposed by Iliya Bluskov, Technical University, Gabrovo,

Bulgaria.
Pairs of numbers from the set { 1 1 , 1 2 , . . . , n} are adjoined to each of the 45 differ-
ent (unordered) pairs of numbers from the set { 1 , 2 , . . . , 10}, to obtain 45 4-element sets
Ai, A 2 , . . . . , A 45 . Suppose that |At- n Aj\ < 2 for all z ^ j . What is the smallest n possible?
Solution by Randy Ho, student, University of Arizona, Tucson.
The smallest n possible is n = 15.
n = 14 will not work. There are only 6 pairs of numbers from {11,12,13,14}.
Consider these 7 pairs from { 1 , 2 , . . . , 10}:

{1,2},{1,3},{1,4},{1,5},{1,6},{1,7},{1,8}.
128

By the Pigeonhole Principle, there are two pairs { l , i } and { l , j } that are assigned the
same pair from {11,12,13,14} and so the intersection of the resulting 4-element sets has
3 elements.
n = 15 will work. See the following chart for assigning pairs.

2 3 4 5 6 7 8 9 10 where a = 11,12
1 b c d e / 9 h i 3 b= 11,13
2 d e / 9 h i 3 a c = 12,13
3 / 9 h i 3 a b d= 11,14
4 h i 3 a b c c = 12,14
5 3 a b c d /=13,14
6 b c d e 0 = 11,15
7 d e / h = 12,15
8 / 9 i = 13,15
9 h j = 14,15

For example, the pair {5,9} gets assigned to c, or {12,13}, to form the 4-element set
{5,9,12,13}.
[Editor's note. Ho's rule here is simply to assign to each pair {x, y} from { 1 , 2 , . . . , 10} one
of the 10 pairs from {11,12,13,14,15} depending on what x + y is modulo 10. Thus it is
clear that no two pairs { x i , y i } and {22,2/2} w i t h an element in common can be assigned
the same pair, so the resulting 4-element sets must have at most two elements in common.]
Also solved by H.L. ABBOTT, University of Alberta; JUN-HUA HUANG, The 4th
Middle School of Nanxian, Hunan, China; WALTHER JANOUS, Ursulinengymnasium,
Innsbruck, Austria; STOYAN N. KAPRALOV, Technical University, Gabrovo, Bulgaria;
MARGIN E. KUCZMA, Warszawa, Poland; KEE-WAI LAU, Hong Kong; ANDY LIU,
University of Alberta; R.P. SEALY, Mount Allison University, Sackville, New Brunswick;
and the proposer.
Abbott generalized the result by replacing { 1 , 2 , . . . , 10} by { 1 , 2 , . . . , m } and
{ 1 1 , 1 2 , . . . , n } by {m + l , r a + 2 , . . . , m + s}. The minimum value of s (depending on
m) can be found as above.
As pointed out by both Kapralov and the proposer, the problem is related to the
construction of ^balanced tournament designs77.

*
 ^^^M^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^m
Dr. G.W. Sands, Editor-in-Chief Department of Mathematics and Statistics, University of Calgary,
Calgary, Alberta T2N 1N4

Dr. R.E. Woodrow, Editor-in-Chief Department of Mathematics and Statistics, University of Calgary,
i Calgary, Alberta T2N 1N4

Dr. G.P. Wright, Managing Editor Department of Mathematics, University of Ottawa,

Ottawa, Ontario KIN 6N5 |

Dr. R. Guy Department of Mathematics and Statistics, University of Calgary,

Calgary, Alberta T2N 1N4 |

Dr. C. Fisher Department of Mathematics, University of Regina,

Regina, Saskatchewan S4S 0A2

Dr. D. Hanson (Articles) Department of Mathematics, University of Regina,

Regina, Saskatchewan S4S 0A2 |

Dr. A. Liu (Book Reviews) Department of Mathematics, University of Alberta,

Edmonton, Alberta T6G 2G1 |

Dr. R. Nowiakowski Department of Math., Stats and Computing Science, Dalhousie

University, Halifax, Nova Scotia B3H 3J5

Dr. E. Wang Department of Mathematics, Wilfrid Laurier University

Waterloo, Ontario N2L 3C5
_J
Short articles intended for publication should be sent to Dr. Hanson, contest problem sets and solutions to Olympiad Comer
problems should be sent to Dr. Woodrow and other problems and solutions to Dr. Sands.

BACK ISSUES AND SUBSCRIPTION INFORMATION

INFORMATION SUR L'ABONNEMENT ET LES NUMEROS PRECEDENTS

1993 (Volume 19) SUBSCRIPTIONS / ABONNEMENTS 1993 (volume 19)

(10 issues per volume / 10 num6ros par volume)
Regular / R6gulier $42.00
CMS Members / membres de la SMC $21.00

These subscription rates include shipping and handling

Les frais de manutentions sont inc/us dans ces tarifs
CANADIAN ADDRESSES ADD 7% GST/ADRESSES CANADIENNES, AJOUTER 7% DE TPS)

BACK ISSUES / NUMEROS ANTERIEURS

Volume $45.00
Issue / Num6ro $ 4.50

BOUND VOLUMES / VOLUMES RELIES

(available / disponible: 1 &2, 3, 1, 8, 9)
Volume $10.00 each/chacun

FOR BACK ISSUES AND BOUND VOLUMES ONLY

POUR NUMEROS ANTERIEURS ET VOLUMES RELIES SEULEMENT
Canadian addresses, please add 7% GST (#R118 833 979).
Foreign addresses, please add 15% shipping and handling charges.

Les adresses canadiennes, veuillez a/outer 7% de TPS l#R118 833 979).

Les adresses 6trang§res, veuillez a/outer 15% pour les frais d'envois.
 CANADIAN MATHEMATICAL SOCIETY
SOCIETE MATHEMATIQUE DU CANADA
577 King Edward
P.O. Box 450, Station A
Ottawa, Ontario K1N 6N5 CANADA
Tel: (613) 564-2223, Fax: (613) 565-1539

ORDER FORM / BON DE COMMANDE

CRUX MATHEMATICORUM

Subscription / Abonnement X $42.00 = (1)

Back issues / NumGros anterieurs X $ 4.50 = (2)

. X $45.00 = _(3)

Bound volumes / Volumes relics _ _ _ X $10.00 = (4)

( 1 + 2 + 3 + 4)

Canadian addresses must add GST: 7% of total of lines 1 to 4 (# R118 833 979)
Les adresses canadiennes: veuillez ajouter 7% de TPS sur la somme
des lignes 1 a 4 (# R118 833 979) ______

Foreign addresses must add Shipping and Handling: 15% of total of lines 2, 3 and 4
Les adresses Strangles: veuillez ajouter 15% pour les frais de manutention
(15% de la somme des lignes 2, 3 et 4) ________

TOTAL $

• Cheque enclosed / cheque inclus • • Invoice / Facture

(micro-encoded cheques only
cheques encodes seulement)

• Please charge / Portez a mon compte • VISA • MasterCard

Account no. / Numero de compte Expiry date / date d'expiration

Signature

MAILING ADDRESS / ADRESSE POSTALE

Name / Norn:

Address / Adresse:

City, Province / Ville, Province:

Postal Code, Country / Code postal, pays: