257

THE ACADEMY CORNER

No. 34
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

Two items for your enjoyment this issue.

Memorial University Undergraduate
Mathematics Competition
March 2000
1. Find all roots of (b ; c)x + (c ; a)x + a ; b = 0 if a, b, c are in
2

arithmetic progression (in the order listed).

2. Evaluate x + y where x + y = 1 and x + y = 2.
3 3 2 2

3. In triangle ABC , we have \ABC = \ACB = 80 . P is chosen on

line segment AB such that \BPC = 30 . Prove that AP = BC .
4. Show that
;
= (1)(1998)+ (2)(1997)+ : : : + k(1999 ; k) + : : : +
2000
3

(1997)(2) + (1998)(1).
5. Let a , a , : : : , a be 6 consecutive integers. Show that the set
fa , a , : : : , a g cannot be divided into two disjoint subsets so that
1 2 6

1 2 6
the product of the members of one set is equal to the product of the
members of the other. (Hint: First consider the case where one of the
integers is divisible by 7.)
6. Let f (x) = x(x ; 1)(x ; 2) : : : (x ; n).
(a) Show that f 0(0) = (;1)nn!
(b) More generally, show that if 0  k  n,
then f 0(k) = (;1)n;k k!(n ; k)!
X
n
7. For each integer n  1; let n = 10; j .
( !)

j =1
(a) Show that nlim
!1 n exists.
(b) Show that nlim
!1 n is irrational.
Send me your nice solutions!
258

The Bernoulli Trials 2000

Christopher G. Small & Byung Kyu Chun
Since 1997, the Bernoulli Trials, an undergraduate mathematics competition,
has been held at the University of Waterloo. This is a double knockout com-
petition. At the start of each round, students are presented with a mathemat-
ics statement which can be true or false. They have 10 minutes to determine
the truth or falsehood of the proposition, and drop out after their second
incorrect answer.
In March of 2000, there were 36 student participants. The competi-
tion lasted for 3.5 hours and 13 rounds, after which the rst four places were
clearly determined. The winner was Scott Sitar, who was the sole contestant
not to be eliminated at the end of 12 rounds. Second place went to Megan
Davis, who won a 13th round tie-breaker with 3rd place going to Dennis The.
Adrian Tang came in 4th, having survived to round 11. In keeping with the
nature of the answers required, the prizes supplied by the Dean of Mathe-
matics were awarded in coins: 200 dollars (100 \toonies") for rst, 100 dollars
(\loonies") for second, 70 dollars for third, and 30 dollars for fourth.
1. A deck of 2000 cards has the numbers from 1 to 2000 labelled consec-
utively in order from top to bottom. The deck is shued as follows.
The second card from the top is placed on the top card, the third card
is placed below these two, the fourth above these three, the fth be-
low these four, and so on, until the 2000th card is placed above the
remaining 1999.
TRUE or FALSE? At the completion of this shue, every card is in a
di erent position in the deck than where it started; that is, for every
i = 1, : : : , 2000 the card labelled i is not in position i.
2. TRUE OR FALSE?
The equation
sin(sin(sin(x))) = x=3
has exactly one solution in real values x.
3. Let ABCD be a planar convex quadrilateral labelled clockwise as shown:
A B

C
D
 259

Suppose that
[ABC ]  [BCD]  [CDA]  [DAB ]
where [RST ] represents the area of triangle RST .
TRUE or FALSE? AD is parallel to BC .
(A quadrilateral is said to be convex if no vertex is within the triangle
formed by the other three vertices.)
4. We arrange dimes in rows on top of each other according to the following
rules:
 each coin must touch the next in its row;
 each coin except those in the bottom row touches two coins on the
row below.
Let A(n) be the number of distinct ways to arrange n coins. For exam-
ple, A(4) = 3 as shown.
u A(1) = 1
uu A(2) = 1
uuu uuu A(3) = 2
u uuuu
uuuu uuu A(4) = 3
TRUE or FALSE? A(n) is the nth Fibonacci number; that is, A(1) = 1,
A(2) = 1, : : : , A(n + 2) = A(n) + A(n + 1).
5. TRUE or FALSE? For every integer n  3, the equation
xn + y n = z n +1

has in nitely many solutions in positive integers x, y and z .

6. Consider a point P at random inside a circle of diameter 2. From P , a
ray is drawn in a random direction, and intersects the circumference of
the circle at Q.

Pq
QQ
QQ Q
QQs

TRUE OR FALSE? The average length of PQ is 1.

260

7. Let v , : : : , vn be any n vectors in Rn such that jjvi jj = 1, for all i = 1,

: : : , n.
1

TRUE OR FALSE? It is always possible to select  , : : : , n 2 f;1, +1g,

1
so that p
jj v +


+ n vnjj  n
1 1

and  , : : : , n 2 f;1, +1g, so that

jj v +


+ n vnjj  pn .
1

1 1

8. TRUE or FALSE? For all 0 < x < 1,

d
dx [ln(x) ln(1 ; x)] < 0 .
2000

2000

9. Two players, Arthur and Barbara, take turns selecting numbers from
the set
f1 , 2 , 3 , 4 , : : : , 9g .
A number, after selection, cannot be selected in a subsequent round.
The rst player to obtain a set of 3 numbers totalling 15 is the winner.
TRUE or FALSE? With best play by both sides, the rst player (Arthur)
can force a win.
10. Let f : R ! R be a continuous function. Suppose that for every rational
number q there exists a positive integer N such that f n(q ) = 0 for all
n  N , where f n denotes the n{fold iteration of f .
TRUE or FALSE? For every real number t
n
!1 f (t) = 0 .
nlim
11. TRUE or FALSE?
X1 8k + 4k + 1
3 X1 (2k + 1) + 4k + 3 3

(2k)! < (2k + 1)!

k
=0 k =0

12. Consider a sequence of positive integers

a ,a ,a , :::
0 1 2

with the property that an equals the number of positive divisors of

an; . (The number ai has both 1 and ai as divisors.) We set
a = 2000!.
1

TRUE or FALSE? For some positive integer n the number an is a perfect

square.
13. TRUE or FALSE? There exists a function f : [;1; +1] ! R with contin-
uous second derivative such that
X1 1 X1 1
f n converges and f n diverges.
n
=1 n
=1
 261

THE OLYMPIAD CORNER

No. 207
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.
I hope you spent the break working on problems and that I will soon
be receiving your nice solutions!
We rst give some Olympiad sets. We start with the Swedish Mathe-
matical Competition, Final Round, November, 1996. My thanks go to Richard
Nowakowski for collecting them when he was Canadian Team Leader at the
IMO in Argentina.
SWEDISH MATHEMATICAL COMPETITION
Final Round
November 23, 1996 | Time: 5 hours

1. Through an arbitrary interior point of a triangle, lines parallel to the

sides of the triangle are drawn dividing the triangle into six regions, three of
which are triangles. Let the areas of these three triangles be T , T , and T
and let the area of the original triangle be T . Prove that
1 2 3

p p p 
T = T + T + T .
2
1 2 3

2. In the country of Postonia, one wants to have only two values of

stamps. The two values should be integers greater than one and the di er-
ence between the two should be two. It should also be possible to combine,
in a precise way, stamps for each letter, the postage of which is greater than
or equal to the sum of the two values. What values can be chosen?
3. For all integers n  1 the functions pn are de ned for x  1 by
 p n  p n
pn(x) = 12 x + x ; 1 + x ; x ; 1 .
2 2

Show that pn (x)  1 and that pmn (x) = pm (pn(x)).

4. A pentagon ABCDE is inscribed in a circle. The angles at A, B,
C , D, E form an increasing sequence. Show that the angle at C is greater
than =2. Also prove that this lower bound is best possible.
262

5. Let n  1. Prove that it is possible to select some of the integers

1, 2, 3, : : : , 2n so that for all integers p = 0, 1, : : : , n ; 1, the sum of kp
over all selected integers k 2 f1, 2, 3, : : : , 2n g is the same as the sum of kp
over all non-selected integers k 2 f1, 2, 3, : : : , 2n g.
6. Tiles of dimension 6  1 are used to construct a rectangle. Prove
that one of the sides has a length divisible by 6.

Next we give the problems of the 48th Polish Mathematical Olympiad,

Final Round, written April 4{5, 1997. My thanks again go to Richard
Nowakowski, Canadian Team Leader to the IMO in Argentina, and to Marcin
E. Kuczma, Warszawa, Poland for sending me the problems.
48th POLISH MATHEMATICAL OLYMPIAD
Final Round { April 4{5, 1997
First Day | Time: 5 hours
1. The positive integers x , x , x , x , x , x , x
1 2 3 4 5 6 7 satisfy the condi-
tions:
x = 144 and xn
6 +3 = xn (xn + xn ) for n = 1, 2, 3, 4 .
+2 +1

Compute x . 7

2. Solve the following system of equations in real numbers x, y, z:

3(x + y + z ) = 1 ,
2 2 2

x y + y z + z x = xyz(x + y + z) .
2 2 2 2 2 2 3

3. In a triangular pyramid ABCD, the medians of the lateral faces

ABD, ACD, BCD, drawn from vertex D, form equal angles with the cor-
responding edges AB , AC , BC . Prove that the area of each lateral face is
less than the sum of the areas of the two other lateral faces.
Second Day | Time: 5 hours
4. The sequence a , a , a , : : : is de ned by
1 2 3

a = 0 , an = a n= + (;1)n n = for n > 1 .

1 [ 2]
( +1) 2

For every integer k  0 nd the number of all n such that

2k  n < 2k , an = 0 +1

([n=2] denotes the greatest integer not exceeding n=2).

 263

5. Given is a convex pentagon ABCDE with

DC = DE and \DCB = \DEA = 90 .
Let F be the point on AB such that AF : BF = AE : BC . Show that
\FCE = \ADE and \FEC = \BDC .
6. Consider n points (n  2) on the circumference of a circle of
radius 1. Let q be the number ofpsegments having those points as endpoints
and having length greater than 2. Prove that 3q  n . 2

Next we give the problems of the 18th Brazilian Mathematical Olympiad.

Thanks again go to Richard Nowakowski for collecting and forwarding them
to me when he was Canadian Team Leader to the IMO in Argentina.
18th BRAZILIAN MATHEMATICAL OLYMPIAD
1. Show that the equation
x + y + z = 3xyz
2 2 2

has in nitely many integer solutions with x > 0, y > 0 and z > 0.
2. Is there a set A of n points (n  3) in the plane such that:
(i) A does not contain three collinear points; and
(ii) given any three points in A, the centre of the circle which contains these
points also belongs to A?
3. Let f (n), n 2 Z , be the smallest number of ones that can be used
+

to represent n using ones and any number of the symbols +, , ( , ), (with

their usual meaning). For instance,
80 = (1 + 1 + 1 + 1 + 1)  (1 + 1 + 1 + 1)  (1 + 1 + 1 + 1)
and, therefore, f (80)  13. Show that
3 log n  f (n) < 5 log n ,
3 3

for all n > 1. (Note: 11, 111, 1111, etc. may not be used in the expressions;
only 1.)
4. Let D be a point of the side BC of the acute-angled triangle ABC
(D 6= B and D 6= C ), O be the circumcentre of 4ABD, O be the circum-
centre of 4ACD and O be the circumcentre of 4AO O . Determine the
1 2

locus described by the point O when D runs through the side BC (D 6= B

1 2

and D 6= C ).
264

5. A set of marriages is unstable if two persons who are not married

to each other prefer each other to their spouses. For instance, if Alessandra
and Daniel are married and if Julia and Robinson are married, but Daniel
prefers Julia to Alessandra, and Julia prefers Daniel to Robinson, then the
set of marriages Alessandra-Daniel and Julia-Robinson is unstable. If the
set of marriages is not unstable, we call it stable.
Consider now a group of people consisting of n boys and n girls. Each
boy makes his own list ordering the n girls according to his preferences and,
in the same way, each girl lists the n boys according to her preference. Show
that it is always possible to marry the n boys and the n girls obtaining a
stable marriage set.
6. Consider the polynomial T (x) = x + 14x ; 2x + 1. Show that
3 2

there exists a natural number n > 1 such that 101 divides T n (x) ; x for ( )

all integers x. (Note: T n (x) = T| (T (


{z(T (x))))} .)

( )

n times

As a fourth set to keep your solution skills nely honed, we give the
problems from the Selection Test for the Vietnamese Team 1997, written
May 16{17, 1997. Thanks again go to Richard Nowakowski who collected
them for us when he was Canadian Team Leader to the IMO in Argentina.

SELECTION TEST FOR

THE VIETNAMESE TEAM 1997
May 16{17, 1997
First Day | Time: 4 hours

1. Let ABCD be a tetrahedron with BC = a, CA = b, AB = c,

DA = a , DB = b , CD = c .
1 1 1

Prove that there exists one and only one point P satisfying the condi-
tions:
PA + a + b + c = PB + b + c + a
2 2
1
2 2 2 2
1
2 2

= PC + c + a + b = PD + a + b + c ,
2 2
1
2 2 2 2
1
2
1
2
1

and that for this point P , we have PA + PB + PC + PD  4R , where

2 2 2 2 2

R is the radius of the circumscribed sphere of the tetrahedron ABCD. Find

a necessary and sucient condition on the lengths of the edges so that the
preceding inequality becomes an equality.
2. In a country, there are 25 towns. Determine the least number k such
that one can set up ight routes connecting these towns (in both directions)
so that the following conditions are simultaneously satis ed:
 265

(i) from each town there are exactly k direct ight routes to k other towns,
(ii) if there is no direct ight route connecting two towns, then there exists
at least one town which has direct ight routes to these two towns.
3. Find the greatest real number such that there exists an in nite
sequence of whole numbers (an) (n = 1; 2; 3; : : : ) satisfying simultaneously
the following conditions:
(i) an > 1997n for every n 2 N ,
(ii) an  Un for every n  2, where Un is the greatest common divisor of
the set of numbers fai + aj j i + j = ng.
Second Day | Time: 4 hours
4. Let f : N ! Z be the function de ned by:
f (0) = 2; f (1) = 503; f (n+2) = 503f (n+1);1996f (n) for all n 2 N .
For every k 2 N, take k arbitrary integers s , s , : : : , sk such that si  k
for all i = 1, 2, : : : , k and for every si (i = 1, 2, : : : , k), take an arbitrary
1 2

prime divisor p(si) of f (2si ).

Prove that for positive integers t  k, we have:
X k
p(si) j 2t if and only if k j 2t .
i=1
5. Determine all pairs of positive real numbers a, b such that for every
n 2 N and for every real root xn of the equation
4n x = log (2n x + 1)
2
2
2

we have
axn + bxn  2 + 3xn .
6. Let three positive integers n, k, p satisfying k  2 and k(p +1)  n,
be given.
Let n distinct points on a circle be given. One colours these n points
blue and red (each point by a colour) so that there exist exactly k points
coloured blue, and on each arc, the extremities of which are two consecutive
(in clockwise direction) blue points, there exist at least p points coloured red.
What is the number of such colourings? (Two such colourings are dis-
tinct if there exists at least one point coloured with two di erent colours by
these colourings).
266

Next we give solutions by our readers to problems given in the Febru-

ary 1999 number of the Corner. We start with solutions to problems of the
Bundeswettbewerb Mathematik, Second Round 1995 [1999 : 4].
1. Starting in (1; 1), a stone is moved according to the following rules:
(i) From any point (a; b), it can be moved to (2a; b) or (a; 2b).
(ii) From any point (a; b), if a > b it can be moved to (a ; b; b), and if a < b
it can be moved to (a; b ; a).
Determine a necessary and sucient relation between x and y so that the
stone can reach (x; y ) after some moves.
Solutions by Mohammed Aassila, Strasbourg, France; and by Pierre
Bornsztein, Courdimanche, France. We give the solution by Aassila.
Since gcd(1; 1) = 1 and noting that gcd(p;q ) = gcd(p; q ; p), we
conclude that an odd common divisor can never be introduced. Hence
gcd(x; y) = 2r for some r 2 N if (x; y) is reachable.
Assume now that gcd(x; y ) = 2r and let us prove that (x; y ) is reach-
able. From the pairs (p;q ) from which (x; y ) can be reached, we choose the
pair which minimizes p + q . In fact p = q ; indeed if p > q then (p; q )
is reachable from ( p q ; q ), a contradiction; similarly for the case in which
+

p < q. Since gcd(p;q) = 2r and neither p nor q is even, otherwise p = q is

2

not minimal for (p; q ) from which (x; y ) can be reached. We conclude that
p = q = 1 and then (x; y) is reachable.
2. In a segment of unit length, a nite number of mutually disjoint
subsegments are coloured such that no two points with distance 0:1 are both
coloured. Prove that the total length of the coloured subsegments is not
greater than 0:5.
Comment by Mohammed Aassila, Strasbourg, France. [Ed: A solution
was received from Pierre Bornsztein, Courdimanche, France.]
This problem appeared as problem 6 of the Swedish Mathematical Com-
petition 1986 and a solution appeared in [1992 : 296].
3. Every diagonal of a given pentagon is parallel to one side of the pen-
tagon. Prove that the ratio of the lengths of a diagonal and its corresponding
side is the same for each of the ve pairs. Determine the value of this ratio.
Solutions by Mohammed Aassila, Strasbourg, France; and by Toshio
Seimiya, Kawasaki, Japan. We give the solution of Seimiya.
In the pentagon ABCDE , we assume that AB kCE , BC kAD,
CDkBE, DEkCA, and EAkDB. As shown in gure 1 on the next page,
we label the intersections of diagonals.
Since ATDE and SCDE are both parallelograms we get:
AT = ED = SC , so that AS + ST = ST + TC . Thus we have
AS = TC . Hence CE : AB = CS : SA = AT : TC = AD : BC .
 267
A

B S R E

T Q
P

C D
Figure 1.

Similarly we have
AD : BC = BE : CD = CA : DE = DB : EA .
We put
CE = AD = BE = CA = DB = k .
AB BC CD DE EA
Thus we have
CS = CE = k , DT = AD = k ,
SA AB BT BC
and
DR = BD = k , so that
DB = k + 1 and
TS = k ; 1 .
RA AE BT 1 SA 1
By Menelaus' Theorem for 4ATD we get
DB
TS
AR = 1 .
BT SA RD
Therefore we have k +1 k ;


k = 1. Thus k ; k ; 1 = 0, from which we
1 1 2
p 1 1

obtain k = 1+ 5
2
.
4. Prove that every integer k, (k > 1) has a multiple which is less
than k and which can be written in decimal representation with at most
4

four di erent digits.

Comment by Mohammed Aassila, Strasbourg, France; and solution by
Pierre Bornsztein, Courdimanche, France.
Aassila points out this problem was proposed by Poland but not used
by the jury at the 31st IMO in China. A solution appeared in [1993 : 10].
268

Next we turn to the First Round 1996 of the Bundeswettbewerb Math-

ematik given in [1999 : 4].
1. Is it possible to cover a square of length 5 completely with three
squares of length 4?
Solution by Sam Wong, student, Sir Winston Churchill High School,
Calgary, Alberta. E s
4 3

A s 5
sB
H s sF
1

J s
L

s
G

D
s s sC
1
K 4

This is a diagram showing the 5  5 square, ABCD, covered by two of

the three available 4  4 squares. One 4  4 square shown is JHCK , and
the other one is AEFG. The third 4  4 square is not shown, but occupies a
similar position to square AEFG, except that it is re ected over a line joining
AC . The reason that this square is not shown will be explained shortly.
The square JHCK covers a 4  4 area of the square ABCD, and
so leaves an L-shaped area (the polygon ABHJKD) left to be covered by
the other two squares. To cover this L-shaped area with two squares, each
square must cover, at the very least, half of this L-shaped area.
The square AEFG does cover over half of the area. This can be proven
as follows: Using Pythagoras' Theorem, the line EB can be calculated be-
cause it forms the triangle AEB and AB = 5, and AE = 4. Therefore, since
AB = EB + AE , EB = 5 ; 4 , or 9, so EB = 3.
2 2 2 2 2 2

On the left side of the diagram, \BAG must be greater than \BAC .
Since BAC is the diagonal of a square, \BAC = 45 . To nd \BAG, we
must rst nd \EAB (since \EAB and \BAG are complementary). The
sine of \EAB is , so when the inverse sine is taken for \EAB , the angle
3
5
 269

is 36:870 . Subtract this angle from 90 , and we get (90 ; 36:870 )
or 53:130 . Thus, \BAG is approximately 53 , and 53 > 45 . Thus \BAG
is greater than \BAC .
On the right side of the diagram, point L must extend past point H on
the line BC , or BL must be greater than BH (and BH is 1 unit long). Since
BL is the hypotenuse of 4BFL, and BF is 1, then BL must be greater
than 1. Thus BL is greater than BH .
When the third unit four square is placed as the re ection of AEFG
over the line AC , then this square will cover a similar area to AEFG. Both
squares together will completely cover the aforementioned L-shaped area,
and together with square JHCK , the unit ve square is completely covered
with three unit four squares.
2. The cells of an n  n{board are numbered according to the example
shown for n = 5. You may choose n cells, not more than one from each row
and each column, and add the numbers in the cells chosen. Which are the
possible values of this sum?

1 2 3 4 5

6 7 8 9 10

11 12 13 14 15

16 17 18 19 20

21 22 23 24 25

Solutions by Pierre Bornsztein, Courdimanche, France; by Edward T.H.

Wang, Wilfrid Laurier University, Waterloo, Ontario; and by Sam Wong,
student, Sir Winston Churchill High School, Calgary, Alberta. We give
Wang's solution.
Using the terminology of combinatorial matrix theory, a collection of
n entries from an n  n matrix A = (aij ) with no two entries lying in the
same row or the same column, is called a diagonal and the sum of the entries
from a diagonal is called a diagonal sum. Note that if  is a permutation of
f1, 2, : : : , ng then fai i j i = 1, 2, : : : , ng would be a diagonal of A, and
conversely, every diagonal gives rise to a permutation. Thus, A has exactly
( )

n! diagonals.
We show that for the matrix considered, all diagonal sums equal
n(n + 1)=2. To see this, note that by assumption, aij = (i ; 1)n + j
2

for all i, j = 1, 2, : : : , n. Hence

270

X
n X
n
ai i =
( ) ((i ; 1)n +  (i))
i=1 i=1
Xn X
n X
n
= ((i ; 1)n + i) = (n + 1) i; n
i=1 i=1 i=1
= n(n 2+ 1) ; n = n(n 2+ 1) .
2 2
2

Remark. The sum of all the n! diagonal sums is

XX n
ai i = n! n(n 2+ 1) .
2
S= ( )
 i =1

On the other hand, since each entry of A lies on exactly (n ; 1)! diagonals,
X
n X
n2
S = (n ; 1)! aij = (n ; 1)! k
i;j =1 k=1
= (n ; 1)! n (n2 + 1) = n! n(n 2+ 1) ,
2 2 2

and so we have a check!

3. There are four straight lines in the plane, each three of them deter-
mining a triangle. One of these straight lines is parallel to one of the medians
of the triangle formed by the other three lines. Prove that each of the other
straight lines has the same property.
Solutions by Michel Bataille, Rouen, France; Murray S. Klamkin, Uni-
versity of Alberta, Edmonton, Alberta; and by Toshio Seimiya, Kawasaki,
Japan. We rst give Seimiya's geometric solution.
Let the four lines be a, b, c, d, and let 4ABC be determined by a, b,
and c as shown in the gure. The line d intersects BC , CA, AB at P , Q, R
respectively.
We assume that the median AM of 4ABC is parallel to d. Let X be
a point on AC such that PX kAB . Since PX kAB and AM kPQ, we have
AC = BC = 2MC = 2
AC .
CX CP CP CQ
Thus CQ = 2CX ; that is, CX = XQ.
Hence median PX of 4PCQ is parallel to c. Let Y be a point on AB
such that PY kAC . Since PY kAC and AM kRP , we have
BA = BC = 2BM = 2
BA .
BY BP BP BR
 271

d
Y
A Z

P
B M C a
X
c Q
b

Thus BR = 2BY ; that is, BY = Y R.

Hence median PY of 4PBR is parallel to b. Let Z be a point on QR
such that AZ kBP . Since AZ kBP and AM kPQ we have
ZP AC MC BM BA ZP
ZQ = AQ = MP = MP = AR = ZR .
Thus ZQ = ZR.
Hence median AZ of 4AQR is parallel to a.
Next we give the solution of Bataille.
We denote the four straight lines by L , L , L , L , and we suppose
that the names are chosen so that L is parallel to one of the medians of the
1 2 3 4

triangle formed by L , L , L . More precisely, let L , L intersect at A,

4

L , L intersect at B, L , L intersect at C and let us suppose L kAM

1 2 3 2 3

where M is the mid-point of BC . We shall work in the system of axes with

3 1 1 2 4

origin A, and AB , AC as x{axis and y {axis respectively. Thus we have

1 1
A(0; 0) ; B(1; 0) ; C (0; 1) ; M 2 ; 2
and AM has equation y = x.
We readily nd:
L : x + y = 1 ; L : x = 0 ; L : y = 0 , and L : x ; y = k ,
1 2 3 4
272

where k 6= 0, 1, ;1 [this condition on k ensures that L determines a real

triangle with any two of the three lines L , L , L ].
4

1 2 3

It is now easy to compute the coordinates of the points A0 , B 0 , C 0 where

L intersects L , L , L respectively:
4 1 2 3

1 + k 1 ; k
A 2 ; 2 ; B0(0; ;k); C 0(k; 0) .
0

1+
;;!
The mid-point I of BC 0 has coordinates ( k ; 0) so that A0 I = (0; k; ). 1
2 2

Hence the median A0 I (of the triangle formed by L , L , L ) is parallel to L .

1 3 4 2

Similarly, the mid-point J of CB 0 has coordinates (0; ;k ) and 1

;A;0!
J = (; k ; 0) is parallel to L .
1+
2

2 3

the mid-point K of B 0 C 0 has coordinates ( k ; ; k ) and

;! =Lastly,
;AK ( k ; ; k ) is parallel to L .
2 2

2 2 1

4. Determine the set of all positive integers n for which n
2n; is a 1

perfect square.
Solutionsby Michel Bataille, Rouen, France; by Pierre Bornsztein, Cour-
dimanche, France; and by Edward T.H. Wang, Wilfrid Laurier University,
Waterloo, Ontario. We give Bornsztein's solution.
Let E be the set of such integers.
First Case. n is odd.
Write n = 2k + 1 with k  0. Then we have
n2n; = (2k + 1)(2k) with 2 and 2k + 1 coprime.
1 2

Then n 2 E if and only if 2k + 1 = n is a perfect square.

Second Case. n is even.
Write n = 2 k, k  1 an odd integer and  1. Then we have
n2n; = k2 k; with k and 2 coprime.
1 +2 1

Thus n 2 E if and only if + 2 k ; 1 is even and k is a perfect square.

Thus n 2 E if and only if is odd and k is a perfect square.
Further, n 2 E if and only if n = 2b where b 2 N .
2

In summary, n 2 E if and only if n is an odd square or n is twice a

square.

Now we turn to readers' comments and solutions to problems of the

XLV Lithuanian Mathematical Olympiad 1996 given in [1999 : 5].
 273

1. Solve the following equation in positive integers:

x ; y = xy + 61 .
3 3

Solutionsby Mohammed Aassila, Strasbourg, France; by Andrew Blinn,

student, Western Canada High School, Calgary, Alberta; by Michel Bataille,
Rouen, France; by Pierre Bornsztein, Courdimanche, France; by Murray S.
Klamkin, University of Alberta, Edmonton, Alberta; and by Edward T.H.
Wang, Wilfrid Laurier University, Waterloo, Ontario. We give Blinn's solu-
tion.
Manipulating the equation we obtain
(x ; y )(x + xy + yz) ; xy = 61
2

and
(x ; y)((x ; y) + 3yx) ; xy = 61 .
2

Set x ; y = a. Since the right-hand side is positive, a = x ; y > 0. Set

b = xy. Note that b > 0.
Rewriting the equation in terms of a and b yields
a + b(3a ; 1) = 61 ,
3

where a > 0 and b > 0.

Thus a < 61, so a = 1, 2, 3.
3

Also b = a;;a3 2 Z, so trying

61
3 1

a = 1 gives b = 30 ,
a = 2 gives b = 53=5 ,
a = 3 gives b = 17=4 .
Thus a = 1, x ; y = 1 and xy = 30, giving (y + 1)y = 30 with y = 5
(rejecting y = ;6) and x = 6.
The unique solution is x = 6, y = 5.
Comment. Aassila also points out that this problem was proposed at
the 15th All-Union Mathematical Olympiad held in Alma Ata, and a solution
in Russian appears in N.B. Vassiliev and A.A. Egorov, \The Problems of the
All-Union Mathematical Competitions", Moscow: Nauka, 1988.
2. Sequences a , : : : , an , : : : and b , : : : , bn, : : : are such that a > 0,
b > 0, and
1 1 1

an = an + b1 , bn = bn + a1 , n 2 N .
+1
n
+1
n
Prove that p
a + b > 10 2 .
25 25
274

Solutions by Mohammed Aassila, Strasbourg, France; by Pierre

Bornsztein, Courdimanche, France; by Murray S. Klamkin, University of Al-
berta, Edmonton, Alberta; and by Heinz-Jurgen Sei ert, Berlin, Germany.
We give the solution by Klamkin.
It follows that a + b  4 and 2 2

an + bn = (an + bn)(1 + 1=anbn)  (an + bn) + 4=(an + bn) .

+1 +1

Let xn = xn + 4=xn with x = 4. Then since x + 4=x is increasing for

x  2, we have an + bn  xn . Since (xn ) = (xn) +(4=xn) +8,
+1 2
2 2 2
+1 +1 +1 +1

(xn )  (xn ) + 8 . +1
2 2

Summing the latter set of inequalities for n = 2 to n ; 1, we obtain

(xn )  x + 8(n ; 2) = 8n .
2 2
2

It then follows that

(xn )  (xn) + 8 + 1=2n ,
+1
2 2

and summing this inequality for n = 2 to 24, we obtain

(x )  (x ) + 8(23) + (1=4 + 1=6 +


+ 1=48) .
25
2
2
2

p
Hence, (x ) > 16 + 184 or 10 2 < x  a + b .
2
25 25 25 25

Next we give the generalization by Sei ert.

More generally: Let u > 0, v > 0, and w > 0. If the sequences
fangn and fbngn satisfy a > 0, b > 0, and
1 1 1 1

an +1 = uan + bv , bn = bun + aw , n 2 N ,
n
+1
n
then
 
anbn > (n ; 1) uv + uw + 2pvw , n  3 , (1)
and
 s 
an + bn > 2 (n ; 1) uv + uw + 2pvw , n  3 . (2)

First, we note that an > 0 and bn > 0, n 2 N. We have

 v   bk w 
ak bk = uak + b
+1 +1
u + ak
k
= uv + uw + ak bk + avwb , k 2 N .
k k
 275

Summing as k ranges from 1 to n ; 1, where n  3, gives

v  nX;
anbn = (n ; 1) u + uw + a b + avwb
1

1 1
k k k
v  =1

> (n ; 1) u + uw + a b + avwb 1 1

v  p 1 1

 (n ; 1) u + uw + 2 vw ,
where we have used the AM-GM-Inequality.
p This proves (1). Then (2) fol-
lows from (1) and an + bn  2 an bn.
In the particular case u = v = w = 1, (1) and (2) give
p
anbn > 2n and an + bn > 2 2n , n  3 .
With n = 25, we then have a b > 50 and a + b > 10 2.
p
25 25 25 25

Remark. If u = 1 and v = w, then a bn = b an for all n 2 N, as is

easily veri ed by induction on n.
1 1

3. Two pupils are playing the following game. In the system

8 x + y + z = 0 ,
<
: xx + y + z = 0 ,
+ y + z = 0 ,
they alternately replace the asterisks by any numbers. The rst player wins
if the nal system has a non-zero solution. Can the rst player always win?
Solution by Pierre Bornsztein, Courdimanche, France.
Yes. Denote the system as follows:
a x + a y +a z = 0,
1 2 3 (1)
b x + b y + b z = 0,
1 2 3 (2)
c x + c y +c z = 0.
1 2 3 (3)
The rst player chooses any number for b . 2

Then form the pairs (a ; c ), (a ; c ), (a ; c ), (b ; b ).

1 1 2 2 3 3 1 3

Each time the second player chooses a number from one pair, then the
rst player gives the same number to the other member of the pair. Thus at
the end a = c , a = c , a = c , b = b . So (1) and (3) are equivalent.
1 1 2 2 3 3 1 3

And, since the system is homogeneous, it is consistent and must have

in nitely many solutions, in particular a non-(0; 0; 0) solution, and the rst
player wins.
4. How many sides has the polygon inscribed in a given circle and such
that the sum of the squares of its sides is the largest one?
276

Solutions by Mohammed Aassila, Strasbourg, France; by Pierre

Bornsztein, Courdimanche, France; and by Murray S. Klamkin, University
of Alberta, Edmonton, Alberta. We give Bornsztein's solution.
We will prove that the \maximal" polygon is the equilateral triangle.
Lemma. Let P = A A : : : An be a convex polygon with n  5 sides. Then
P has an obtuse angle.
1 2

P c = (n;2). Suppose, on the contrary,

Proof of the lemma. We have ni A i
Xn
ci   . Then, Ai  . Thus (n ; 2)  n . Then n  4,
n
=1

that each A 2
i 2 2
which is a contradiction.
Let P be a convex polygon with n  4 sides and with an obtuse angle.
Let c
B >  and A, C be the neighbouring vertices of B. Then, from the Law
of Cosines:
2

AC = AB + BC ; 2AB
BC cos c
2 2 2
B > AB + BC . 2 2

Thus, P is not maximal, because we have a larger sum by deleting B . Then,

using the above lemma: if a \maximal" polygon exists, it has n  4 sides,
and no obtuse angle.
Let Pn be a convex polygon with n  4 sides, no obtuse angle, inscribed
in the circle C with centre O and radius R.
Let n = 4. The proof of the lemma may be used to prove that all angles
of Pn are  (because they are   ). Then P is a rectangle.
2 2 4

D C

A B
Pythagoras' Theorem leads to:
AB + BC = 4R = CD + DA .
2 2 2 2 2

Thus the sum of the squares of the sides is S = 8R . Note that S is

4
2
4
independent of the rectangle.
Let n = 3. Suppose that 4ABC is a non-obtuse triangle inscribed
in C . Let G be the centre of gravity of 4ABC . For any point M , we have
AM + BM + CM = AG + BG + CG + 3GM .
2 2 2 2 2 2
(1) 2
 277

Then
AG + BG + CG + 3GO = 3R .
2 2 2
(2)
2 2

Moreover, for M = A, we have AB + AC = 4AG + BG + CG .

2 2 2 2 2

We also have the same relation for M = B , M = C .

Then, summing these three relations, we obtain
2(AB + BC + CA ) = 6(AG + BG + CG )
2 2 2 2 2 2

and, using (2) we get

AB + BC + CA = 9(R ; OG ) .
2 2 2 2 2

We deduce that the sum of the squares is S  9R with equality if and

2

only if O = G; that is, 4ABC is equilateral.

2

Then, the \maximal" polygon exists; it is the equilateral triangle and

the sum is S = 9R . 2

5. Given ten numbers 2, 3, 5, 6, 7, 8, 10, 11, 12, 13, one must cross
out several of them so that the total of any of the remaining numbers would
not be an exact square (that is, the sum of any two, three, four, : : : , and of
all the remaining numbers would not be an exact square). At most how many
numbers can remain?
Solutions by Mohammed Aassila, Strasbourg, France; and by Edward
T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario. We give Wang's
solution.
The maximum number is ve. First note that the numbers in each
of the following sets add up to a perfect square: f2; 7g, f3; 6g, f3; 13g,
f5; 11g, f6; 10g, f12; 13g, f3; 10; 12g. (And there are many more such sets,
of course.) Let S = f2, 3, 5, 6, 7, 8, 10, 11, 12, 13g and partition S as
S = f8g [ f2; 7g [ f5; 11g [ f3, 6, 10, 12, 13g. Let T  S be a set whose
elements never add up to a perfect square and suppose jT j  6. Then by
the strong Pigeonhole Principle we have jT \ f3, 6, 10, 12, 13gj  3 which
is impossible since if 3 62 T , then either f6; 10g  T or f12; 13g  T ,
a contradiction, and if 3 2 T , then 6 62 T and 13 62 T would imply that
f3, 10, 12g  T , again a contradiction. Therefore, jT j  5. On the
other hand, 5-element subsets of S whose elements do not add up to perfect
squares do exist. One such set is T = f3, 7, 8, 11, 12g since direct checkings
show that the values of the sums of any k elements of T , (1  k  5) are:
3, 7, 8, 10, 11, 12, 14, 15, 18, 19, 20, 21, 22, 23, 26, 27, 29, 30, 31, 33, 34,
38, and 41.

That completes the Corner for this issue. Send me your Olympiad
Contest materials and your nice solutions to problems from the Corner.
278

BOOK REVIEWS
ALAN LAW
Geometry from Africa | Mathematical and Educational Explorations
by Paulus Gerdes,
published by the Mathematical Association of America, 1999,
ISBN 0-88385-715-4, softcover, 244+ pages, $39.95 (U.S.)
Reviewed by Julia Johnson, University of Regina, Regina, Saskatchewan.
The thesis of this book is that the African peoples are actually doing
mathematics in their art. The author has a prestigious and proli c teaching
and research career. This book is a continuation of his inquiry into an emerg-
ing eld known as ethnomathematics. Particular formulations of mathemat-
ical ideas develop from cultural activities. The aim of ethnomathematics is
to uncover the common cognitive concepts underlying di erent cultures.
The book is divided into four parts which can be read independently
from each other. The rst part is entitled \On geometrical ideas in Africa
South of the Sahara". The peoples of this area have been particularly active
in geometrical thinking expressed in diverse cultural and social artifacts that
exhibit a high degree of symmetry. Examples include rock paintings found
in Northern Mozambique, petroglyphs from extreme East Angola, Adinkra
stamp motifs, archaeological nds of carved or wooden patterns from the
Republic of Mali and the geometrical structure of the Great Hall of King
Munza. This part of the book is lled with photographs and sketches to
illustrate the geometrical accents of African culture. The illustrations are
annotated with their area of origin and sometimes time period (e.g., design
from royal cloth from Northwest Cameroon, design from Nigeria embroidery,
body painting on wooden sculpture collected in Niangara in 1910). The sketch
of a roof structure of a Fulani house in Cameroon is mesmerizing, reminiscent
of today's test pattern for astigmatism. Beautifully illustrated decorative
designs on drums, facial tattoos from South Africa, semi-spherical basket sh
traps, and the hexagonally woven bottom of a conical quail basket leave the
reader with a graphic realization of what Cameroon mathematician George
Njock said: that \Black art is mathematics". The pictures reveal vividly the
geometry evident in every sphere of African life.
It is interesting to note Africa's connection with Egypt in terms of their
similarity in originating mathematical ideas. For example, the Pythagorean
proposition can be proved from designs derived from Mozambican decora-
tion. Part 2 of the book, like Part 1, is lled with graphics, but in Part 2
the graphics are used to illustrate the steps in the production of the proof
of the Pythagorean theorem. The theorem is proved again in a di erent
way, this time starting with Chokwe sand drawings with fourfold symmetry
which are easily transformed into Pythagorean designs. Part 2 is concluded
with coverage of other connections between African art and variations of the
 279

Theorem of Pythagoras. For example, the theorem can also be discovered in

mat weaving patterns.
Part 3 explores the uses of African art forms for teaching notions of sym-
metry. This part of the book provides a careful enumeration and organization
of many examples to further illustrate the appearance of geometrical ideas in
African artifacts. Alternative ways are developed for rectangle constructions
and for the determination of areas of circles and volumes of spatial gures.
Part 4 provides a summary of the author's previous books Sona
Geometry ( rst published 1993/94 with several later versions and transla-
tions), Luscona: Geometrical Recreations of Africa (published 1991, trans-
lated 1997), and Lunda Geometry (1995). As such, it provides an educational
exploration of geometry focusing on the sona sand drawing tradition among
the Chokwe people in Southern Central Africa. This part of the book, like
the others, is well illustrated with visual representations.
The pictorials that appear abundantly throughout all four parts convey
African art as mathematical symmetry. The educational uses of African art
forms teach geometrical notions in a very e ective way. A research method-
ology for advancing the eld of ethomathematics is evident from this text.
Paraphrasing Gerdes himself : If one tries to vary the geometrical forms and
patterns of traditional objects, a sub-optimal solution is reached which dis-
allows many practical advantages. This is to say that the traditional form is
never arbitrary but embodies mathematical knowledge that expresses itself
as accumulated knowledge and wisdom (1997; 1986).
I highly recommend this book both from the mathematical and artistic
points of view.
References
Gerdes, P. (1997). On culture, geometrical thinking and mathematics educa-
tion. In A.B. Powell and M. Frankenstein (Eds.), Ethnomathematics: Chal-
lenging Eurocentrism in Mathematics Education. New York: State University
of New York. pp. 223{248.
Gerdes, P. (1986). On culture, mathematics and curriculum development in
Mozambique. In S. Mellin-Olsen and M. Johnsen Hoines (eds.), Mathemat-
ics and Culture, A Seminar Report. Radal: Casper Forlag. pp. 15{42.
280

Some bounds for (n)(n)

Edward T.H. Wang
Let N denote the set of all natural numbers. In elementary number
theory, three multiplicative functions which are discussed most frequently
are  (n), the number of (positive) divisors of n 2 N;  (n), the sum of all
the divisors of n; and (n), Euler's totient function; that is, the number of
positive integers in f1, 2, : : : , ng which are coprime with n.
Though there are well-known formulae for computing the exact values
of these functions, these formulae all depend on the actual prime power fac-
torization of the natural number n. Hence it is of interest to nd upper and
lower bounds for these functions, preferably in terms of n only. Indeed, ex-
amples of such bounds abound in the literature. For example, the statements
in the following proposition are clearly true and trivial.
Proposition 1. Let n 2 N, n  2. Then
(a)  (n)  2,
(b) (n)  n ; 1,
(c)  (n)  n + 1.
In each of the three inequalities above, equality holds if and only if n is a
prime.
On the other hand, there are less trivial bounds for these functions. For
example, the inequalities in the next proposition can be found in ([1], p. 214
and p. 222).
Proposition 2. Let n 2 N. Then
(a) n is composite if and only if (n) < n ; pn,
(b) n is composite if and only if  (n) > n + pn.
In view of Proposition 1, (b) and (c), it is natural to ask whether there is
any inequality between (n) (n) and n ; 1. A quick numerical checking
2

for n  2 seems to suggest that (n) (n)  n ; 1. In our rst result

2

(Proposition 4 below), we will show that this is indeed the case, but rst, we
need a lemma.
Lemma 3. Let ai 2 N, i = 1, 2, : : : , k.
Then (a ; 1)(a ; 1)


(ak ; 1)  a a


ak ; 1. Equality holds if and
only if either k = 1 or ai = 1 for all i, 1  i  k.
1 2 1 2

Copyright c 2000 Canadian Mathematical Society

 281

Proof. We clearly have equality when k = 1. Hence we assume that

k  2. Let ai = 1 + bi where bi  0 for all i = 1, 2, : : : , k. Then
k !
Y Y
k Yk ! Yk kX
; 1

ai ; 1 ; (ai ; 1) = (1 + bi) ; 1 ; bi = Sm  0 ,
i=1 i=1 i=1 i=1 m=1

where; Sk m
is the mth elementary symmetric function de ned to be the sum

of all m possible products of the bi's taken m at a time.
If equality holds, then Sm = 0 for all m = 1, 2, : : : , k ; 1, from which
it follows that bi = 0 or ai = 1 for all i = 1, 2, : : : , k. This completes the
proof.
We now recall the familiar formulae for (n) and  (n).
If n = p 1 p 2


pk k is the prime power factorization of n 2 N, then
1 2

k 
Y  Y
k
(a) (n) = n 1 ; p1 = pi i ; (pi ; 1), 1

i=1 i i=1
Y
k ;
Yk pi i ; 1 +1

(b)  (n) = 1 + pi + pi +


+ pi =
2

pi ; 1 .
i
i=1 i =1

Now we are ready for our rst result.

Proposition 4. Let n 2 N, n  2. Then (n) (n)  n ; 1. Equality 2

holds if and only if n is a prime.

Proof. First suppose n has more than one prime factor. Thus,
let n = p 1 p 2


pk k be the prime power factorization of n where
k  2. Since  and  are multiplicative functions, we have by Lemma 3,
1 2

that
Y
k Y
k ;
(n)(n) = ((pi i )(pi i ))  pi i ; 1

2

i=1 ! i=1
Yk
< pi
2 i ; 1 = n ;1. 2

i=1
Next suppose n = p for some prime p where 2 N. Then we have
(n)(n) = p ; (p ; 1) p p ;;1 1 = p ; ;p ; 1

+1
1 1 +1

= p ;p ;  p ;1 = n ;1
2 1 2 2

with equality if and only if = 1; that is, if and only if n is a prime. This
completes the proof.
282

Similarly, the two bounds in Proposition 2 raise the following natural

question: Is there any inequality between (n) (n) and n ; n? Searching 2

for a possible answer to this question we rst do some computations and

compare the values of (n) (n) with n ; n for all n, 2  n  49, as shown
2

in the table below. Note that those integers n for which (n) (n) > n ; n 2

are circled.
n (n)(n) n2 ; n n (n)(n) n2 ; n n (n)(n) n2 ; n
j2 3 2 18 234 306 34 864 1122
j3 8 6 19 360 342 35 1152 1190
j4 14 12 20 336 380 36 1092 1260
j5 24 20 21 384 420 37 1368 1332
6 24 30 22 360 462 38 1080 1406
j7 48 42 23 528 506 39 1344 1482
j8 60 56 24 480 552 40 1440 1560
j9 78 72 25 620 600 41 1680 1640
10 72 90 26 504 650 42 1152 1722
11 120 110 27 720 702 43 1848 1806
12 112 132 28 672 756 44 1680 1892
13 168 156 29 840 812 45 1872 1980
14 144 182 30 576 870 46 1584 2070
15 192 240 31 960 930 47 2208 2162
16 248 240 32 1008 992 48 1984 2256
17 288 272 33 960 1056 49 2394 2352
The above table seems to indicate that (n) (n) is greater than n ; n 2

exactly when n is a prime power. Our second result below shows that this is
indeed the case.
Proposition 5. Let n 2 N, n  2. Then (n) (n) > n ; n if and only 2

if n = pk for some prime p where k 2 N.

Proof. The suciency is clear since if n = pk , then
(n)(n) = p k ; pk; > p k ; pk = n ; n .
2 1 2 2

To prove the necessity, we show that if n 6= pk , then in fact,

(n)(n) < n ; n. Assume rst that n has only two distinct prime factors,
2

so n = p q where , 2 N and p and q are distinct primes. Then using

the multiplicative property of  and  , we have
(n)(n) = (p )(p )(q )(q ) = ;p ; p ;
;q ; q ;
2 1 2 1

and so
n ; n ; (n)(n) = p q ; p q ; ;p ; p ;
;q ; q ;

2 2 2 2 1 2 1

= p q ; +p ; q ;p q ;p ; q ;
2 1 1 2 1 1

= p ; q ; (p + q ; pq ; 1)
1 1 +1 +1
 283

 p ; q ; (p + q ; pq ; 1)
1 1 2 2

= p ; q ; ((p ; q ) + pq ; 1)
1 1 2

> 0.
Therefore, (n) (n) < n ; n.
2

Now suppose (n) (n) < n ; n holds for all n 2 N with t distinct
2

prime factors for all t = 2, 3, : : : , m for some m  2, and suppose n 2 N

has m + 1 distinct prime factors. Then we can write n = pk s where p is a
prime and s 2 N has m distinct prime factors and (p; s) = 1. Hence

(n) (n) =  ;pk
;pk
(s)(s)

= p k ; pk; (s) (s)
2 1

 
< p k ; pk; ;s ; s
,
2 1 2

where the inequality is by the induction hypothesis. Therefore

n ; n ; (n)(n) > p ks ; pks ; ;p k ; pk;
;s ; s

2 2 2 2 1 2

= p ks +; pk; s ; pk s ; p
k; s
2 1 2 1

= pk; s pk ; p + s ; 1 1 +1

> 0,
from which (n) (n) < n ; n follows and our induction is complete.
2

The corollary below clearly follows from Propositions 4 and 5:

Corollary 6. lim sup
(n)(n) = 1.
n!1 n 2

Acknowledgement: The author would like to thank the referees for their
careful reading of the original manuscript and for making a number of valu-
able observations and suggestions which greatly improve the clarity of this
paper.
Reference
1. Kenneth H. Rosen, Elementary Number Theory and its Applications,
3rd ed., Addison-Wesley, 1993.
Wilfrid Laurier University
Waterloo, Ontario, Canada N2L 3C5
e-mail: [email protected]
284

Letter to the Editor

Regarding: Trevor Lipscombe & Arturo Sangalli; The Devil's dartboard
[2000 : 215].
The arrangement of the numbers on a dartboard has intrigued the many
players of the game and a number of mathematicians. I have outlined the
history of the dartboard in my paper: Arranging a dartboard [Bull. Inst.
Math. Appl. 16:4 (Apr 1980) 93-97; CMP 12:22 (1980) 1446; MR 81j:05005].
A simpler approach than used by Lipscombe & Sangalli is based on the idea
that large and small numbers tend to alternate, so for a cyclic arrangement
(ai ) of the rst n integers, one could consider the sums of two adjacent num-
bers, si = ai + ai , and seek to make these as equal as possible; that
+1
is, have the least standard deviation or variance. This approach is much
easier. It was used by Selkirk to determine the best and worst distribu-
tions, but the construction was a bit vague. I had also used the same ap-
proach and found an easy way to determine these optimal distributions. First
one notes
X that minimizing the variance corresponds to minimizing the sum
C = aiai . I considered a section : : : , a, b, : : : , c, d, : : : of the distrib-
+1

ution and noted that reversing the portion from b to c reduced (or preserved)
C unless (a ; d)(c ; b) > 0. But there is essentially only one distribu-
tion which satis es this necessary condition and it looks like: : : : , 4, n ; 2,
2, n, 1, n ; 1, 3, n ; 3, 5, : : : I extended the analysis, relating it to the
auto-correlation coecient of the cycle with itself shifted by one and deter-
mining the mean and standard deviation of this auto{correlation, from which
one can reasonably deduce that the designer of the standard dartboard must
have had something like the idea of putting big numbers next to little ones
in his mind.
I also considered making other sums as equal as possible, of which the
most natural next stage is si = ai; +X
1 ai + ai , as considered by Lipscombe
+1

& Sangalli, and, more generally, si = pj ai j , where pd is the probability

+

of hitting the value which is d values from the one aimed at. Then the average
of the sX i is the same as the average of the ai , which is a = (n +1)=2. Setting
Dd = (ai ; a)(ai d ; a), we have that Dd=nv is the auto{correlation
+

coecient of the cycle with itself shifted by d places | here v is the variance
of the rst n integers, namely (n ; 1)=12. StraightforwardX
2
manipulation
gives us an expression for the variance V of the si as nV = Dk;j pj pk .
X j ,k
We have D = nv ; Dd = D;d = Dn;d and Di = 0, which can be used
0

to simplify the expression for nV . One usually also assumes symmetry of the
pd, but even so, the problem generally involves at least two Dd and di erent
choices of the pd will give di erent optima and a given set of probabilities
may have several optima.
 285

For the version considered by Lipscombe & Sangalli, we take

p; = p = p = 1=3 and all other probabilities equal to 0, so we have
9nV = 3D + 4D + 2D . For n = 6, the unique best distribution is
1 0 1

0 1 2
1, 6, 3, 2, 5, 4, as also found by Lipscombe & Sangalli. However, for the
simpler version considered above, corresponding to p = p = 1=2 and
4nV = 2D +2D , the unique best distribution is 1, 6, 2, 4, 3, 5. Returning
0 1

to p; = p = p = 1=3, the case n = 7 has three best distributions: 1,

0 1

1 0 1
4, 7, 2, 3, 5, 6; 1, 4, 7, 3, 2, 5, 6; 1, 4, 7, 3, 2, 6, 5. These are rather better
than the distribution given by Lipscombe & Sangalli's algorithm, which I nd
is 7, 1, 4, 6, 2, 5, 3. The technique of reversing a part of the distribution can
be used here, but it leads to messy conditions which do not necessarily force
a global minimum, though a computer could easily use them to improve an
approximate minimum. The simplest case is reversing two adjacent terms in
the arrangement; changing : : : , a, b, c, d, e, f , : : : to : : : , a, b, d, c, e, f , : : :
decreases (preserves) the variance if (a + b ; e ; f )(c ; d) > 0(= 0). For
the result of Lipscombe & Sangalli, no such exchange reduces the variance,
but exchanging 2 and 5 preserves the variance and in that arrangement, ex-
changing 1 and 4 does reduce the variance and gives a minimal arrangement.
References.
Keith Selkirk. Re-designing the dartboard. Math. Gaz. 60 (No.413) (1976) 171-178.
Ian Cook. Unbiased dartboards and biased calculators. Math. Gaz. 61 (No.417)
(1977) 187-191. [He corrects and extends Selkirk, but considers di erent measures
than treated here.]
David Daykin, proposer; David Singmaster, solver. Problem 1059: Cyclic extrema.
X
Math. Mag. 52:1 (Jan 1979) 46 & 53:2 (Mar 1980) 115-116. [Asks for the extreme
values of ai ai+1 in a cycle of real values. See also Crux Math. 7:7 (Aug/Sep 1981)
210.]
Brian Bolt. The Amazing Mathematical Amusement Arcade. Cambridge Univ. Press,
X
1984. Prob. 116: Designing a new dartboard, pp. 67 & 123. [Asks to make
jai ; ai+1j as large as possible. In the solution he poses nding the smallest
value.]
Weixuan Li & Edward T.H. Wang, proposers; M.S. Klamkin & A. Meir, solvers. Prob-
lem E 3087 - Maximizing a cyclic sum of powers of di erences. Amer. Math. Monthly
X X
92 (1985) 287 & 94:4 (Apr 1987) 384-385. [Asks for the extreme values of
(ai ; ai+1 )2 where 0  ai  1. Solvers extend to jai ; ai+1 jp .]

David Singmaster
Song Bank University
London, England
[email protected]
286

THE SKOLIAD CORNER

No. 47
R.E. Woodrow
This issue we give the preliminary round of the Junior High School
Mathematics Contest of the British Columbia Colleges. This was written
in the schools by Grade 8 to 10 students on March 8, 2000. Thanks go to
Jim Totten, The University College of the Cariboo, one of the organizers, for
forwarding the contest materials to us.
BRITISH COLUMBIA COLLEGES
Junior High School Mathematics Contest
Preliminary Round | March 8, 2000

1. After 15 litres of gasoline was added to a partially lled fuel tank,

the tank was 75% full. If the tank's capacity is 28 litres, then the number of
litres in the tank before adding the gas was:
(a) 3 (b) 4 (c) 5 (d) 6 (e) 7
2. The following gures are made from matchsticks.
rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr
rr rr rr rr rr rr rr
rr rr rr
If you had 500 matchsticks, the number of squares in the largest such
gure you could build would be:
(a) 164 (b) 165 (c) 166 (d) 167 (e) none of these
3. The perimeter of a rectangle is 56 metres. The ratio of its length to
width is 4 : 3. The length, in metres, of a diagonal of the rectangle is:
(a) 17:5 (b) 20 (c) 25 (d) 40 (e) none of these
4. If April 23 falls on Tuesday, then March 23 of the same year was a:
(a) Saturday (b) Sunday (c) Monday (d) Wednesday (e) Thursday
5. Consider the dart board shown in the diagram. If a dart may hit
any point on the board with equal probability, the probability it will land in
the shaded area is:
 287

x
x r
x
x
x r
x
r x
x
x
x
r
(a) 0:07 (b) 0:24 (c) 0:25 (d) 0:28 (e) 0:32
6. The proper divisors of a number are those numbers that are factors
of the number other than the number itself. For example, the proper divisors
of 12 are 1, 2, 3, 4 and 6. An abundant number is de ned as a number for
which the sum of its proper divisors is greater than the number itself. For
example, 12 is an abundant number since 1 + 2 + 3 + 4 + 6 > 12. Another
example of an abundant number is:
(a) 13 (b) 16 (c) 30 (d) 44 (e) 50
7. The gure below is a right trapezoid with side lengths 4 cm, 4 cm,
and 6 cm as labelled. The circle has radius 2 cm. The area, in cm , of the 2

shaded region is:

4 cm
2 cm
4 cm r
2 cm

6 cm

(a) 20 ; 4 (b) 16 (c) 24 ; 2 (d) 20 ; 2 (e) 16 + 2

8. Three vertices of parallelogram PQRS were P (;3; ;2), Q(1; ;5),
and R(9; 1) with P and R diagonally opposite. The sum of the coordinates
of vertex S is:
(a) 13 (b) 12 (c) 11 (d) 10 (e) 9
288

9. Which shape cannot be lled, without any

overlapping, usingcopies of the tile shown on the right?

(a) (b) (c) (d) (e)

10. Arrange the following in ascending order:

2 5555
3 3333
62222

(a) 2 5555
3 3333
6 2222
(b) 2 6 5555
3 2222 3333

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

(e) 3 3333
2 5555
6 2222

11. 2000 days, 2000 hours, 2000 minutes, and 2000 seconds would be
equivalent to N million seconds. Of the choices o ered, the closest approx-
imation of N is:
(a) 1 (b) 15 (c) 45 (d) 180 (e) 2000
12. A three-digit decimal number abc may be expressed as
100a + 10b + c where each of the digits is multiplied by its respective place
value and subsequently summed. If a = b = c and a > 0, which of the
following numbers must be a factor of the three digit number abc?
(a) 7 (b) 11 (c) 13 (d) 19 (e) 37
13. If (x + y) ; (x ; y) > 0, then
2 2

(a) (x > 0 and y > 0) or (x < 0 and y < 0) (b) x > 0 and y < 0
(c) x < 0 and y > 0 (d) (x > 0 and y < 0) or (x < 0 and y > 0)
(e) x > y or x < y
14. Consider all non-congruent triangles with all sides having whole
number lengths and a perimeter of 12 units. The following statements cor-
respond to these triangles.
(i) There are only three such triangles.
(ii) The number of equilateral triangles equals the number of scalene
triangles.
(iii) None of these triangles are right angled.
(iv) None of these triangles have a side of length 1 unit.
Of the four statements made, the number of true statements is:
(a) 0 (b) 1 (c) 2 (d) 3 (e) 4
 289

15. An altitude, h, of a triangle is increased by a length m. How much

must be taken from the corresponding base, b, so that the area of the new
triangle is one-half that of the original?
(a) hbmm
+
(b) hbhm
2( + )
(c) b mm hh
(2
+
+ )

(d) b m
(
2 m h
+
h
+ )
(e) b hm mh
(2
2( +
+ )
)

Last issue we gave the rst round of two contests. Next we give short
\ocial" solutions to the rst of them. Thanks go to Richard Nowakowski,
Canadian Team Leader to the IMO in Buenos Aires, for collecting them.
BUNDESWETTBEWERB MATHEMATIK
Federal Contest in Mathematics (Germany) 1997
First Round

1. Can you always choose 15 from 100 arbitrary integers so that the
di erence of any two of the chosen integers is divisible by 7?
What is the answer if 15 is replaced by 16? (Proof!)
Solution. The answer to the rst question is yes. This is an elementary
application of the Pigeon-Hole Principle, as one remainder (mod 7) must
occur at least 15 times.
The answer to the second question is no. For a contradiction, take the
integers from 1 to 100.
2. Determine all primes p for which the system
p + 1 = 2x , 2

p + 1 = 2y ,
2 2

has a solution in integers x, y .

Solution. We can assume y > x  2 and p > y . Subtracting the given
equations, we have p(p ; 1) = 2(y ; x)(y + x). It follows p > y ; x and
2p > y + x, yielding p = x + y and p ; 1 = 2(y ; x). Eliminating y from
the given equations, we get p + 1 = 4x =) x = 2 =) p = 7. Indeed, 7
satis es the conditions.
3. A square Sa is inscribed in an acute-angled triangle ABC by placing
two corners on the side BC and one corner on AC and AB , respectively. In
a similar way, squares Sb and Sc are inscribed in ABC .
For which kind of triangle ABC do the sides of Sa, Sb and Sc have
equal length?
290

Solution. C

hc q
A c B
The inscribed squares are congruent if and only if triangle ABC is equi-
lateral. The if-direction is trivial, so let us assume the squares to be congru-
ent. Denoting the area of ABC by A and the length of the squares by q ,
we have A = q + q (hc ; q ) + (c ; q )q = q (hc + c). Cyclic permuta-
2 1 1 1

tion leads to A = q (hb + b) = q (ha + a), thus ha + a = hb + b. With

2 2 2
1 1

2A = aha = bhb we get aha + a = ahb + ab. Thus, bhb + a = ahb + ab,
2 2
2 2

and so, (a ; hb)(a ; b) = 0. But hb < a for an acute-angled triangle, so

a = b and cyclic permutation leads to a = b = c.
4. In a park there are 10 000 trees, placed in a square lattice of 100
rows and 100 columns. Determine the maximum number of trees that can
be cut down satisfying the condition:
sitting on a stump, you cannot see any other stump.
Solution. At most 2500 trees can be cut down and there is a way to
chop exactly 2500 trees satisfying the conditions.
Let the trees have integer coordinates from (0=0) to (99=99). We divide
the park into 2500 squares containing the trees (2i=2j ), (2i=2j + 1),
(2i + 1=2j ), (2i + 1=2j + 1). Each tree in one of these squares is
visible from the other lattice points in the square. Thus only one tree in
each square can be cut down.
Let the 2500 trees with coordinates (2i=2j ) be cut down
(i; j 2 f0, 1, : : : , 49g). For any two lattice points P (2a=2b) and Q(2c=2d)
(a, b, c, d 2 f0, 1, : : : , 49g, (a=b) 6= (c=d)), the mid-point R of segment
PQ has integer coordinates (a + c=b + d). If one of these coordinates is odd,
there is a tree left on R so that P and Q are invisible from each other. If both
coordinates are even we can replace Q by R and repeat the same argument.
After a nite number of steps we reach a tree. This completes the proof.

That completes the Skoliad Corner for this issue. We need good contest
materials as well as suggestions for future directions.
 291

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. M5S 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)

Shreds and Slices

On Logarithms
It's log! | The Ren & Stimpy Show
The logarithm is an important function in mathematics and has some
obvious uses, but it appears in some unexpected places as well. We give a
diverse list of such applications, but rst, we review some de nitions.
In the expression y = bx , b is called the base and x the exponent.
With base b xed, y is a function of x. The inverse function is called the
logarithm, denoted logb . Thus, logb y is the real number x such that y = bx .
For example, log 100 = 2 and log 16 = 4, since 10 = 100 and 2 = 16.
2 4

The only acceptable bases are 0 < b < 1 and b > 1.

10 2

Since the logarithm is based on the exponent function, the main prop-
erties of the logarithm are based on the main properties of exponents:

b0
= 1, (1)
b1
= b, (2)
bubv = bu v ,
+
(3)
bu=bv = bu;v , (4)
(bu)c = buc . (5)
The corresponding properties of logarithms are:
logb 1 = 0 , (6)
logb b = 1 , (7)
292

logb x + logb y = logb xy , (8)

logb x ; logb y = logb x=y , (9)
logb xc = c logb x . (10)
Let us prove these results.
Properties (6) and (7) follow directly from (1) and (2).
Let u = logb x and v = logb y , so x = bu and y = bv . Then xy = bu v , +

so u + v = logb xy . Also, x=y = bu;v , so u ; v = logb x=y , proving (8)

and (9). Also, xc = (bu)c = buc . Thus uc = c logb x = logb xc , proving (10).
There is an additional important property, sometimes called the change
of base formula:
logb x = log ax. (11)
log b a
Since bu = x, loga bu = loga x. But loga bu = u loga b, so that
u = logb x = loga x= loga b. This means that logarithms to di erent bases
are proportional to each other (see Problem 1).
 Base 10 is known as the common logarithm, but the most important
base is a constant known as e  2:71828. Why this constant? Let
Zx1
f (x) = t dt .
1

For those not familiar with calculus, you can think of f (x) as being the
area under the graph of y = 1=t from t = 1 to t = x. It turns out that
f (1) = 0 and f (xc) = cf (x) for all c and x > 0. This implies that f is
a logarithm to some base b. We de ne e to be this base. Hence,
Zx1
loge x = t dt .
1

n e,
Unless otherwise indicated, all following logarithms will nowbe to base
1
called the natural logarithm. One formula for e is e = nlim
!1 1 + .
n
log(1 + x) = x ; x2 + x3 ; x4 +


, ; 1 < x  1 .
2 3 4


What does the formula become for x = 1? What happens as x

approaches ;1?
 For all x  0, log(1 + x)  x.
 293

 For large n,
1 + 12 + 31 + 14 +


+ n1  log n .
In fact, the di erence approaches a constant:
1 1 1 1
!1 1 + 2 + 3 + 4 +


+ n ; log n  0:57722 ,
:= nlim
called the Euler-Mascheroni constant. It is currently not even known
whether is irrational or not.
 Stirling's Approximation:
log n!  n log n ; n .
 For a positive integer N , the number of digits in N expressed in decimal
notation is dlog (N + 1)e.
10

 One of the fundamental questions we can ask about prime numbers is

how many there are up to a certain number. Let  (n) be the number of
primes less than or equal to n, and graph  (n) in relation to n= log n:
n (n) n= log n (n)=(n= log n)
10 4 4:34 0:92
100 25 21:71 1:15
1; 000 168 144:76 1:16
10; 000 1; 229 1; 085:74 1:13
100; 000 9; 592 8; 685:89 1:10
1; 000; 000 78; 498 72; 382:41 1:08
The agreement, given the irregularity of primes, is good. In fact, the
Prime Number Theorem states that the two functions  (n) and n= log n
are asymptotic; that is, they will continue to approach each other:
 (n )
!1 n= log n = 1 .
nlim
 We give one nal application. In nance circles, there is a rule of thumb
called the Rule of 72: If an investment grows at a rate of r% annu-
ally, then the number of years it takes for the investment to double is
approximately 72=r.
For example, suppose that we have a dollar in a bank account, which
earns 3% annually, so r = 3. Assuming that there are no other trans-
actions, at the same time next year, there will be $1:03 in the account.
A year after that there will be $1:03  1:03 = $1:06 in the account,
and in general, after n years, there will be 1:03n in dollars. The time
in years n for the account to double to 2 dollars is given by 1:03n = 2.
294

Taking the logarithm of both sides, log 1:03n = n log 1:03 = log 2, so
n = (log 2)=(log1:03) = 23:44. The Rule of 72 gives n  72=3 = 24,
which is fairly close.
In general, we wish to solve for n in the equation
 r n
1+ = 2
 100 r  n  r 
==) log 1 + 100 = n log 1 + 100 = log 2
==) n = log(1log+2 r ) .
100

By the power series above,

log(1 + x) = x ; x2 + x3 ;


 x
2 3

for small values of x, so log(1 + r=100)  r=100, and

n  log 2 = 100 log 2 = 69:17:::  72 .
r r r r
100

We round 100 log 2 to 72, because 72 is divisible by many numbers, and

so provides for an easier calculation. The approximation is reasonable
for many r, as seen in the table below:
r n 72=r
1 69:66 72
2 35:00 36
3 23:45 24
4 17:67 18
6 11:90 12
8 9:01 9
9 8:04 8
12 6:11 6
Problems
1. In a table, list loge n and log n for various values of n. Show that
10
we can multiply all the values in one list by a constant to obtain all the
values in the other list.
2. Before calculators, there was a device called a slide rule which per-
formed numerical calculations. It consists of two, ruled pieces of ma-
terial which slide against each other. Find out how it works.
 295

Area of a Quadrilateral
The following is a useful, but not too well-known, formula for the area
of a quadrilateral:
Let ABCD be a quadrilateral, with sides a = AB , b = BC ,
c = CD, and d = DA. Let K denote the area, and s the semi-perimeter
(a + b + c + d)=2. Let A also denote the angle at vertex A, etc. Then
A + C 
K = (s ; a)(s ; b)(s ; c)(s ; d) ; abcd cos
2 2

2 .
We derive the formula as follows.
By the Cosine Law,
BD = a + d ; 2ad cos A = b + c ; 2bc cos C .
2 2 2 2 2

Therefore,
a ; b ; c + d = 2ad cos A ; 2bc cos C ,
2 2 2 2

implying that
(a ; b ; c + d ) = 4a d cos A ; 8abcd cos A cos C
2 2 2 2 2 2 2 2

+ 4b c cos C . 2 2 2

Now, the area of 4ABD is ad sin A, and the area of 4BCD is 1

bc sin C . Thus
2
1
2

K = 21 ad sin A + 21 bc sin C .
Therefore
16K = 4a d sin A + 8abcd sin A sin C + 4b c sin C
2 2 2 2 2 2 2

= 4a d sin A + 8abcd sin A sin C + 4b c sin C

2 2 2 2 2 2

+ 4a d cos A ; 8abcd cos A cos C + 4b c cos C

2 2 2 2 2 2

; (a ; b ; c + d ) 2 2 2 2 2

= 4a d + 4b c ; 8abcd cos(A + C ) ; (a ; b ; c + d )
2 2 2 2 2 2 2 2 2

= 4a d + 8abcd + 4b c ; 8abcd (cos(A + C ) + 1)

2 2 2 2

;(a ; b ; c + d ) 2 2 2 2 2

= (2ad + 2bc) ; (a; b ; c + d ) 2 2 2 2 2 2

;16abcd cos A +2 C 2

= ;a + 2ad + d ; b + 2bc ; c

2 2

 
2 2

 ;b + 2bc + c ; a + 2ad ; d
; 16abcd cos A +2 C

2 2 2 2 2
296

; 2 2

;
= (a + d) ; (b ; c) (b +c) ; (a ; d)
2 2


;16abcd cos A +2 C
2

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

(a + b + c ; d) ; 16abcd cos A +2 C 2

= (2s ; 2a)(2s ; 2b)(2

 As+;C2c)(2s ; 2d)
;16abcd cos 2

2 .

Therefore
A + C 
K = (s ; a)(s ; b)(s ; c)(s ; d) ; abcd cos
2 2

2 .

An immediate and important corollary of this result is Brahmagupta's

Formula: If ABCD is cyclic, then A + C = 180 , so
K = (s ; a)(s ; b)(s ; c)(s ; d) .
2

You should now be able to quickly solve Problem 10 on the 1993

Descartes Competition:
Suppose p, q , r, s are xed real numbers such that a quadrilateral can
be formed with sides p, q , r, s in clockwise order. Prove that the vertices of
the quadrilateral of maximum area lie on a circle.

Follow-up to \How to Solve the Cubic"

V.N. Murty writes to recommend the following textbooks as references
for the cubic, as well as classic textbooks on algebra:
S. Barnard and J.M. Child, Algebra
G. Chrystal, Algebra, two volumes
M. Abramowitz and I.A. Stegun, Handbook of Mathematical Functions
 297

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 problems | the
next issue will feature only solutions.
We warmly welcome proposals for problems and solutions. With the
schedule of eight issues per year, we request that solutions from this issue
be submitted in time for issue 6 of 2001.

High School Problems

Editor: Adrian Chan, 229 Old Yonge Street, Toronto, Ontario, Canada.
M2P 1R5 <[email protected]>
H273. Proposed by Jose Luis Daz, Universitat Politecnica de
Catalunya, Terrassa, Spain.
Let a, b, and c be complex numbers such that a + b + c = 0. Prove that
2ab ; c 2
b
2
a 2

b2
2bc ; a 2
c 2
= 0.
a2
c
2
2ac ; b 2

H274. Find a simpli ed expression for

X1 i
i k
i
=1

in terms of a real number k > 1.

H275. How many non-negative integers less than 10n are there
whose digits are in non-increasing order?
H276. Proposed by Ho-joo Lee, student, Kwangwoon University,
Seoul, South Korea.
Let ABCDE be a convex pentagon such that ACDE is a square, and
cot \BDE + cot \DEB + cot \EBD = 2 .
Show that 4ABC is a right triangle.
298

Advanced Problems
Editor: Donny Cheung, c/o Conrad Grebel College, University of
Waterloo, Waterloo, Ontario, Canada. N2L 3G6 <[email protected]>
A249. Proposed by Mohammed Aassila, Strasbourg, France.
A circle is circumscribed around 4ABC with sides a, b, c. Let A0 ,
B , C 0 denote the
0 mid-points of the arcs BC , CA, AB , respectively. The
straight lines A0 B 0 , B 0 C 0 , C 0 A0 intersect BC and AC , AC and AB , AB
and BC , in P , Q, R, S , T , U , respectively. Prove that
[PQRSTU ] = (a + b) + (b + c) + (c + a) ,
2 2 2

[ABC ] 2(a + b + c) 2

where [X ] denotes the area of the polygon X .

A250. Suppose polynomial P (x) has integer coecients such that for
any integer m, P (m) is a perfect square. Show that the degree of P is even.
A251. Proposed by Lee Ho-Joo, undergraduate, Kwangwoon Uni-
versity, Seoul, South Korea.
In a parallelogram ABCD, let P be the intersection of AC with BD.
Let M , N be the mid-points of PD, BC , respectively. Prove that the fol-
lowing two statements are equivalent:
(i) 4AMN is a (non-degenerate) right-angled triangle such that
AM = MN .
(ii) Quadrilateral ABCD is a square.
A252. Proposed by Mohammed Aassila, Strasbourg, France.
For every positive integer n, prove that there exists a polynomial of
degree n with integer coecients of absolute value at most n, which admits
1 as a root with multiplicity at least bpnc.

Challenge Board Problems

Editor: David Savitt, Department of Mathematics, Harvard University,
1 Oxford Street, Cambridge, MA, USA 02138 <[email protected]>
In Issue 3, C93 was accidentally printed as C98, and there was a typo
in the problem. It is corrected here.
C93. Let H be a subset of the positive integers with the property that
if x, y 2 H , then x + y 2 H . De ne the gap sequence GH of H to be the
set of positive integers not contained in H .
 299

(a) Prove that if GH is a nite set, then the arithmetic mean of the integers
in GH is less than or equal to the number of elements in GH .
(b) Determine all sets H for which equality holds in part (a).
C94. Proposed by Edward Crane and Russell Mann, graduate stu-
dents, Harvard University, Cambridge, MA, USA.
Suppose that V is a k-dimensional vector subspace of the Euclidean
space Rn which is de ned by linear equations with coecients in Q. Let 
be the lattice in V given by the intersection of V with the lattice Zn in Rn ,
and let ? be the lattice given by the intersection of the perpendicular vector
space V ? with Zn . Show that the (k-dimensional) volume of  is equal to
the ((n ; k)-dimensional) volume of ? . [2000 : 167]
C95. Prove that the curve x + y = 3xy has a horizontal tangent at
3 3

the origin. (This curve is known as the Folium of Descartes.)

C96. Recall that a bipartite graph is a graph whose vertices may be
divided into two nonempty disjoint sets (call them L and R, for left and right)
so that all of the edges of the graph connect a vertex in L to a vertex in R. In
other words, no two vertices in L are joined by an edge, and similarly for R.
Let G be a bipartite graph with 27 edges and in which L and R each contain
exactly 9 vertices. Show that we can nd three vertices l , l , l 2 L and
three vertices r , r , r 2 R such that at least six of the nine potential edges
0 1 2

l r , l r , l r , l r , l r , l r , l r , l r , l r are indeed edges of G.

0 1 2

0 0 0 1 0 2 1 0 1 1 1 2 2 0 2 1 2 2

1960 Bulgarian Mathematical Olympiad 1960

3rd Stage
1. Prove that the sum (di erence) of two irreducible fractions with
di erent denominators cannot be an integer.
2. Find the maximal and the minimal values of the function
y = xx ++2xx++12 ,
2

where x takes all real values.

3. Determine the tangents of x, y, z from the equations tan x : tan y :
tan z = a : b : c, if x + y + z = 180 and a, b, and c are positive numbers.
4. Two externally tangent circles of radii R and r are given.
(a) Prove that the quadrilateral whose sides are the two common tangents
and the chords connecting the points of contact is a trapezoid.
(b) Find the base and the altitude of the trapezoid.
300

5. The rays a, b, and c have a common origin and do not lie in a plane.
The angles = \(b;c), = \(c; a), and = \(a; b) are acute and are given
in a plane. Construct by ruler and compass the angle between the ray a and
the plane which passes through the rays b and c.
6. A sphere is inscribed in a cone. A second sphere tangent to the
rst is inscribed in the same cone. A third sphere tangent to the second is
inscribed in the same cone and so on. Find the sum of the areas of the spheres
if the altitude of the cone is equal to h and the angle at the vertex of its plane
section through the axes is equal to .

J.I.R. McKnight Problems Contest 1997

1. Given triangle ABC , with sides a, b, c, where a + b + c = 60 and
(b + c)=4 = (a + c)=5 = (a + b)=6.
(a) Prove that the sides a, b, c form an arithmetic sequence.
(b) Find sin A : sin B : sin C .
2. (a) Solve:
log(35 ; x ) = 3 .
3

log(5 ; x)
(b) Prove:
X
n
r log r x = n log x .
2 2
r=1
3. Consider three twin brothers: A, A, B , B , C , C . They are to be
arranged in a picture in such a way, that no pair of twins will be side by
side. Find the number of such arrangements.
4. The natural numbers are arranged in diamonds as shown below.
Conjecture and prove a formula for the sum of the numbers in the
nth diamond.
1
1 2 2
1 2 2 3 3 3
1 2 2 3 3 3 4 4 4 4
3 4 4 5 5 5
5 6 6
7
5. Find the limits N , M for which the inequality
N < (x + 1)(x x + 4) < M
is satis ed by no real values of x.
 301

6. Squares ABDE and BCFG are drawn outwards on the sides of 4ABC .
Prove that AC is parallel to DG if and only if 4ABC is isosceles.
G
D F
B
E

A C
7. Given f (x + y ) + f (x ; y ) = 2f (x) cos y , f (0) = a, and f (=2) = b.
Find f (t).
8. Two motorcycles are approaching each other at night on a straight, two-
lane highway. Each vehicle is travelling in the centre of its lane and the
centres of the 2 lanes are A metres apart. The eastbound cycle is trav-
elling at M metres per second. The westbound cycle is travelling at a
rate of N metres per second, and its headlight casts a shadow of the
eastbound cycle onto a fence, B metres from the centre of the east-
bound lane. How fast is the shadow of the eastbound cycle moving on
the fence? (Express your answer in terms of A, B , M , and N ).

6
B
?6

A
?
302

Problem of the Month

Jimmy Chui, student, University of Toronto
Problem. Prove that
1 1 3 5 1997 1
1999 < 2
4
6


1998 < 44 .
(1997 CMO, Problem 3)
Solution. For the left inequality,
1
3
5


1997 = 3
5
7


1997
1
2 4 6 1998 2 4 6 1996 1998
1
> 1
1
1


1
1998
1 > 1 .
= 1998 1999
For the right inequality, let
P = 21
34
56


1997
1998 .
Note that P is positive. Then,
P < 32
45
67


1998
1999
2 4 6 1998 1
= 1
3
5


1997
1999
1 ,
= 1999P
so that
P < 1999
2
1 < 1 = 1 ,
1936 44 2

which implies that P < 1=44. Thus,

1 < P < 1.
1999 44
 303

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 edi-
tor. When a submission 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 submitted 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 21 "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 March 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.

2551. Proposed by Panos E. Tsaoussoglou, Athens, Greece.

Suppose that ak (1  k  n) are positive real numbers. Let ej;k =
(n ; 1) if j = k and ej;k = (n ; 2) otherwise. Let dj;k = 0 if j = k and
dj;k = 1 otherwise.
Prove that
Y
n X
n Y
n X
n ! 2

ej;kak 
2
dj;kak .
j =1 k=1 j =1 k=1
2552. Proposed by Aram Tangboondouangjit, Carnegie Mellon
University, Pittsburgh, PA, USA.
Suppose that a, b, c > 0. If x  a pb c ; 1, prove that
+ +
3 3

(b + cx) + (c + ax) + (a + bx)  abc .

2 2 2

a b c
2553. Proposed by Aram Tangboondouangjit, Carnegie Mellon
University, Pittsburgh, PA, USA.
Find all real roots of the equation
p p 
2x ; 2x + 12 ; x ; 5
3
2

p = 2. 2

(5x ; 2x ; 3) 2x ; 2x + 12
2 2 9
304

2554. Proposed by Aram Tangboondouangjit, Carnegie Mellon

University, Pittsburgh, PA, USA.
In triangle ABC , prove that at least one of the quantities
A B 
(a + b ; c) tan 2 tan 2 ,
2

   
(;a + b + c) tan B2 tan C2 ,
2

C  A
(a ; b + c) tan 2

2 tan 2 ,
is greater than or equal to r , where r is the radius of the incircle of 4ABC .
2
3

2555. Proposed by Aram Tangboondouangjit, Carnegie Mellon

University, Pittsburgh, PA, USA.
In any triangle ABC , show that
X 1
p
A ; B C
< 433 .
cyclic tan + tan + tan
3
2 2 2
3

2556?. Proposed by Mohammed Aassila, CRM, Universite de

Montreal, Montreal, Quebec.
A lattice point is called visible (from the origin) if its coordinates are co-
prime numbers. Is there any lattice point whose distance from each visible
lattice point is at least 2000?
2557. Proposed by Gord Sinnamon, University of Western Ontario,
London, Ontario, and Hans Heinig, McMaster University, Hamilton, On-
tario.
(a) Show that for all positive sequences fxig and all integers n > 0,
0k 1
X
n X
k X
j X
n X 2

xi  2 @ xj A x;k . 1

k=1 j =1 i=1 k=1 j =1

(b)? Does the above inequality remain true without the factor 2?
(c)? [Proposed by the editors] What is the minimum constant c that can
replace the factor 2 in the above inequality?
2558. Proposed by Peter Y. Woo, Biola University, La Mirada, CA,
USA.
Let Z be a half-plane bounded by a line L. Let A, B and C be any three
points on L such that C lies between A and B . Denote the three semicircles
in Z on AB , AC and CB as diameters by K , K and K respectively. Let
F be the family of semicircles in Z with diameters on L (including all half-
0 1 2

lines in Z perpendicular to L). Denote by fX Y the unique semicircle passing

through the pair of distinct points X , Y in Z [ L. Let P , Q, R, be three
points on K , K , K , respectively.
2 1 0
 305

If fA P , fB Q and fC R concur at T , and the lines AP , BQ, CR con-

cur at S , prove that fA P , fB Q and fC R are orthogonal to K , K , K ,
respectively, and that the circle PQR is tangent to each semicircle Kj ,
2 1 0

(j = 0, 1, 2).
2559 Proposed by Ho-joo Lee, student, Kwangwoon University, Seoul,
South Korea.
Triangle ABC has incentre I . Show that CA + AI = CB if and only
if \CAB = 2\ABC .
2560? Proposed by Vaclav Konecny, Ferris State University, Big
Rapids, MI, USA.
Lines AB and AC are common tangents to the circles ; and ; with
distinct radii r and r respectively, as shown.
1 2

1 2

B
O1 E O2
;1 ;2
D r2 A
r1

C
B is a point of tangency on ; and C is a point of tangency on ; . The
intersection points of the circles, D and E , exist, CDB is a straight line, and
2 1

CD = DB.
Construct such a gure using straightedge and compass.
2561. Proposed by Hassan A. ShahAli, Tehran, Iran.
Let M disks from N di erent colours be placed in a row such that ki
disks are from the ith colour (i = 1, 2, : : : , N ) and k + k +


+ kN = M .
1 2

A move is an exchange of two adjacent disks.

Determine the smallest number of moves needed to rearrange the row
such that all disks of the same colour are adjacent to one another.
2562. Proposed by Bernardo Recaman Santos, Colegio Hacienda
Los Alcaparros, Bogota, Colombia.
(a) Show that for all suciently large n, it is possible to nd a set of n (not
necessarily distinct) positive integers whose sum is the square root of their
product.
(b)? Are there in nitely many n for which there is a unique set of n numbers
with property (a)?
306

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

The name of O. Ciaurri, Universidad de La Roija, Logro~no, Spain, a co-solver

with Manuel Benito and Emilio Fernandez, I.B. Praxedes Mateo Sagasta,
Logro~no, Spain, was unfortunately omitted in the solution to problem 2404.
Our apologies.

2436. [1999: 173, 191] Proposed by Vaclav Konecny, Ferris State

University, Big Rapids, MI, USA.
Find all real solutions of

2 cosh(xy) + 2y ; (2 cosh(x))y + 2 = 0 :


Solution by Nikolaos Dergiades, Thessaloniki, Greece; and by Heinz-

Jurgen Sei ert, Berlin, Germany (combined by the editor).
Correction: instead of
tp + t;p + 2p = t + t; + 2 .
1

read
tp + t;p + 2p = ;t + t;
p + 2 .
1

2451. [1999: 306] Proposed by Michael Lambrou, University of

Crete, Crete, Greece.
Construct an in nite sequence, fAn g, of in nite subsets of N with the
following properties:
(a) the intersection of any two distinct sets An and Am is a singleton;
(b) the singleton in (a) is a di erent one if at least one of the distinct sets
An, Am , is changed (so the new pair is again distinct);
(c) every natural number is the intersection of (exactly) one pair of distinct
sets as in (a).
Solutionby Jose H. Nieto, Universidad del Zulia, Maracaibo, Venezuela.
Arrange all the natural numbers in an in nite triangular array faij g,
1  i  j , lling column n with n consecutive integers, as follows:
 307

1 2 4 7 11 16 :::
3 5 8 12 17 :::
6 9 13 18 :::
10 14 19 :::
15 20 :::
21 :::
...

Then de ne:
A = f1, 3, 6, 10, 15, : : : g
0 (main diagonal)
A = f1, 2, 4, 7, 11, 16, : : : g ( rst row)
1

A = f2, 3, 5, 8, 12, 17, : : : g (union of column 2 and row 2)

2

An = union of column n and row n




Since A \ Ai = faiig for all 1  i and Ai \ Aj = faij g for all
1  i < j , it is clear that (a), (b) and (c) hold.
0

Also solved by NIKOLAOS DERGIADES, Thessaloniki, Greece; CHARLES DIMINNIE and

TREY SMITH, Angelo State University, San Angelo, TX, USA; RICHARD I. HESS, Rancho Palos
Verdes, CA, USA; GERRY LEVERSHA, St. Paul's School, London, England; KATHLEEN E. LEWIS,
SUNY Oswego, Oswego, NY, USA; ROBERT P. SEALY, Mount Allison University, Sackville, New
Brunswick; HARRY SEDINGER and DOUG CASHING, St. Bonaventure University, St. Bonaven-
ture, NY, USA; DEE SNELL, University of Wisconsin, La Crosse, WI, USA; JEREMY YOUNG,
student, University of Cambridge, Cambridge, UK; and the proposer.
Dergiades, Sealy, Sedinger and Cashing, Snell, and Young all gave solutions similar to
the above.

2452. [1999: 307, 428] Proposed by Antal E. Fekete, Memorial

University of Newfoundland, St. John's, Newfoundland.
Establish the following equalities:
X1 (2n + 1) 2 X1 (2n + 2) 2
(a) = .
n (2n + 1)!
=0 n (2n + 2)!
=0

X1 1) = X 1
(;1)n ((nn + ; n (n + 1) .
3 4
(b)
+ 1)! n ( 1) (n + 1)!
n =1 =1

X1 1) = X 1
(;1)n ((nn + (;1)n ((nn + 1) .
6 7
(c)
n =1
+ 1)! n =1
+ 1)!
308

Solution by Charles R. Diminnie, Angelo State University, San Angelo,

TX, USA. (modi ed slightly by the editor).
X1 (2n + 1) 2 X1 2n + 1 X
1 1 +X 1 1
(a) = =
(2n)! n (2n ; 1)! n (2n)!
n (2n + 1)!
=0 n =0 =1 =0

= sinh 1 + cosh 1 = e, while

X1 (2n + 2) X
21 2n + 2 X1 1 X1 1
(2 n + 2)! = (2n + 1)! = (2n )! + (2n + 1)!
n=0 n =0 n n =0 =0

= cosh 1 + sinh 1 = e.
(b) and (c). For k  1, let
X1 k X
1 k;
Ik = (;1)n ((nn ++ 1) ; n (n + 1) .
1
=
1)! n ( 1) (n)!
n =1 =1

It can be shown easily by the Ratio Test that Ik converges absolutely

for all k  1. 1
X
Note that I = 1 (;1)n 1 = 1 ; 1, while for k  2,
n! e
n=1
1 (;1)n kX
X ;1 k ; 1 kX
;1 k ; 1 X
1 (;1)n ni
Ik = ni =
n=1 n! i=0 i i=0 i n n!
=1

kX;1 k ; 1 X
1 (;1)n+1(n + 1)i
= i n=0 (n + 1)!
i=0
X1 (;1)n+1 kX ;1 k ; 1 X1 1)i
= (n + 1)! ; i n (;1)n ((nn +
+ 1)!
n=0 i=1 =0

X1 (;1)n ; k ; 1
kX 1

= n! i; i (1 + Ii)
n =1 =1

  kX ; k ; 1 kX ; k ; 1
= e1 ; 1 ;
1 1

i ; i Ii
i =1 i =1

  ;
kX ; k ; 1
= 1e ; 1 ; 2k; ; 1 ;
1

i Ii
1

i =1
 309

; k ; 1
kX
= 1e ; 2k; ;
1

i Ii.
1

i=1
;1 k ; 1
kX
De ne fak g by a = 1 and ak = 1 ; ai for k  2. We
1
i=1 i
a
show by induction that Ik = k ; 1 for all k  1.
e
Since I =
1 ; 1, the claim is true for k = 1.
1
e
a
Suppose Ik = k ; 1 for some k  1. Then using the recurrence
e
relation obtained above, we have

1 X k k 1 Xk k  ai 
k
= e ;2 ; k
Ik +1
i i Ii = e ; 2 ; i i e ; 1
=1 =1

k k
X k k
X
= 1e ; 2k ; 1e i ia + i
i=1 i=1
= 1e ; 2k ; 1e ;1 ; ak
+ ;2k ; 1
= ake ; 1, completing the
+1
+1

induction.
The rst ten terms of the sequence fak g are:
1, 0, ;1, ;1, 2, 9, 9, ;50, ;167, ;513,


.
Hence it follows immediately that I =
;1 ; 1 = I and
e 3 4

I = 9e ; 1 = I .
6 7

Also solved by MICHEL BATAILLE, Rouen, France; THE BOOKERY PROBLEM GROUP,
Walla Walla, WA, USA ; NIKOLAOS DERGIADES, Thessaloniki, Greece; RICHARD I. HESS,
Rancho Palos Verdes, CA, USA; WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria;
MICHAEL LAMBROU, University of Crete, Crete, Greece (three solutions); and the proposer.
Actually, as pointed out by Janous, all these results were contained in the pro-
poser's paper \Apropos Bell and Stirling Numbers" which appeared in CRUX with MAYHEM
[1999 : 274-281].
Diminnie commented that it would be interesting to know whether there are any val-
ues k  7, such that Ik = Ik+1 and in general, whether there are distinct values m, n,
m > n  7 such that Im = In . He conjectured that there are none, but was not able to come
up with a proof.
310

2453. [1999: 307, 428] Proposed by Antal E. Fekete, Memorial

University of Newfoundland, St. John's, Newfoundland.
Establish the following equalities:
X
1 n + 1) = ;3 X
1
(;1)n (2 ; n 1 .
3
(a)
(2n + 1)! ( 1) (2n + 1)!
n=0 n=0
X1 n) = ;3 X
1
(;1)n (2 (;1)n (21n)! .
3
(b)
n=0 (2n)! n=0
X
1 ! X
1 !
(;1)n (2n + 1) (;1)n (2n)
2 2
2 2
(c)
(2n + 1)! + (2n)! = 2.
n=0 n=0
Solution by Charles R. Diminnie, Angelo State University, San Angelo,
TX, USA.
X
1 n + 1) = X
1
n (2n + 1)
(;1)n (2 ;
3 2
(a)
(2n + 1)! ( 1) (2n)!
n=0 n=0
X
1 n ; 1) + 3 X
1 X
1
= (;1)n 2n(2
(2n)! (;1)n (22nn)! + (;1)n (21n)!
n=0 n=0 n=0
X1 X
1 X
1
= (;1)n (2n 1; 2)! + 3 (;1)n (2n 1; 1)! + (;1)n (21n)!
n=1 n=1 n=0
X1 X
1 X
1
= (;1)n (21n)! + 3 (;1)n (2n 1+ 1)! + (;1)n (21n)!
+1 +1

n =0 n n
=0 =0

X1
= ;3 (;1)n (2n 1+ 1)! = ;3 sin 1 .
n =0

X
1 n) = X
1
n (2n)
(;1)n (2 ;
3 2
(b)
(2n)! ( 1) (2n ; 1)!
n=0 n=1
X
1 ; 1)(2n ; 2) + 3 X
1 (2n ; 1)
= (;1)n (2n (2 n ; 1)! (;1)n (2 n ; 1)!
n=1 n=1
X
1 1
+ (;1)n (2n ; 1)!
n=1
X
1 X
1 X
1
= (;1)n (2n 1; 3)! +3 (;1)n (2n 1; 2)! + (;1)n (2n 1; 1)!
n=2 n=1 n=1
X1 X
1 X1
= (;1)n (2n 1; 1)! +3 (;1)n (21n)! + (;1)n (2n 1; 1)!
+1 +1

n=1 n n =0 =1
 311

X
1
= ;3 (;1)n (21n)! = ;3 cos 1 .
n=0
X
1 n + 1) = X
1
n 2n + 1
(;1)n (2 ;
2
(c)
(2n + 1)! ( 1) (2n)!
n=0 n=0
X
1 X
1
= (;1)n (22nn)! + (;1)n (21n)!
n=0 n=0
X1 X
1
= (;1)n (2n 1; 1)! + (;1)n (21n)!
n=1 n=0
X1 X1
= (;1)n (2n 1+ 1)! + (;1)n (21n)!
+1

n =0 n =0

= ; sin 1 + cos 1 ,
and
X
1 n) = X
1
n 2n
(;1)n (2 ;
2

(2n)! ( 1) (2n ; 1)!

n=0 n=1
X
1 1 +X1
= (;1)n (22nn ;; 1)! (;1)n (2n 1; 1)!
n=1 n=1
X1 X1
= (;1)n (2n 1; 2)! +
(;1)n (2n 1+ 1)! +1

n =1 n =0

X1 X1
= (;1)n (21n)! + (;1)n (2n 1+ 1)!
+1 +1

n =0 n =0

= ; cos 1 ; sin 1 .
Therefore,

X
1 ! X
1 !
(;1)n (2n + 1) (;1)n (2n)
2 2
2 2

(2n + 1)! + (2n)!

n=0 n=0
= (; sin1 + cos 1) + (; cos 1 ; sin1)
2 2

= 2.
Also solved by MICHEL BATAILLE, Rouen, France; PAUL BRACKEN, CRM, Universite de
Montre al, Montreal, Quebec; NIKOLAOS DERGIADES, Thessaloniki, Greece; KEITH EKBLAW,
Walla Walla, WA, USA; RICHARD I. HESS, Rancho Palos Verdes, CA, USA; WALTHER JANOUS,
Ursulinengymnasium, Innsbruck, Austria; MICHAEL LAMBROU, University of Crete, Crete,
Greece (three solutions); and the proposer.
312

Janous pointed out that, as in Problem 2452, the identities in this problem all follow
readily from results established in the proposer's paper Apropos Bell and Stirling Numbers.
[Ed: See comments at the end of the solution to Problem 2452.] More speci cally, (a) holds
if and only if c3 sin 1 + d3 cos 1 = ;3(c0 sin 1 + d0 cos 1) and (b) holds if and only if
c3 cos 1 ; d3 sin 1 = ;3(c0 cos 1 ; d0 sin 1). These are true since c0 = 1, c3 = ;3
and d0 = d3 = 0. [Ed: See Table 1 on p. 276 of the proposer's paper.]
As for (c), we have, more generally, that
P (2n + 1)k
2  P (2n)k
2
(;1)n (;1)n
1 1

n=0 (2 n + 1)!
+
n=0 (2n)!

2
= (ck sin 1 + dk cos 1) + (ck cos 1 ; dk sin 1)
2 = c2k + d2k .
[Ed: In particular, for k = 2 we get the answer 2 since c2 = ;1 and d2 = 1. For k = 3 and 4
we would get the answers 9 and 61 respectively. Hess commented that he determined these
two values \from the misprinted problem"].

2454. [1999: 307] Proposed by Gerry Leversha, St. Paul's School,

London, England.
Three circles intersect each other orthogonally at pairs of points A and
A0 , B0 and B0, and C and C 0 . Prove that the circumcircles of 4ABC and
4AB C 0 touch at A.
I. Solution by Michel Bataille, Rouen, France.
Among the three given circles, let ; be the one that does not pass
through A, and let I denote the inversion with centre A such that
I (;) = ;. Because of the mutual orthogonality of the three circles, I trans-
forms the two other circles into two perpendicular lines that are orthogonal
to I (;) = ;.
Hence these two lines are diameters of ;, meeting ; at I (B ), I (B 0 ),
and I (C ), I (C 0 ), respectively.
Clearly, the quadrilateral with vertices I (B ), I (C ), I (B 0 ) and I (C 0 ) is
a rectangle, so that the lines through I (B ), I (C ), and through I (B 0 ), I (C 0 ),
are parallel. Hence the images of these lines under I are circles, tangent at A.
But these circles are precisely the circumcircles of 4ABC and 4AB 0 C 0 , and
so we are done.
II. Solution by Toshio Seimiya, Kawasaki, Japan.
Let the three circles be ; , ; and ; , and let A and A0 be the inter-
sections of ; and ; , let B and B 0 be the intersections of ; and ; , and
1 2 3

let C and C 0 be the intersections of ; and ; . See Figure 1.

1 2 1 3

2 3

We assume that ; intersects ; and ; orthogonally.

3 1 2

Let O be the centre of ; . Then O B and O B 0 are tangent to ; at

B and B0 respectively, and O C and O C 0 are0 tangent to ; at C and C 0
3 3 3 3 1

respectively. Note that O lies on the line AA .

3 3 2

Hence we have
\CAA0 = \O CA0 and \B 0 AA0 = \O B 0 A0 .
3 3
 313

; 1
A ; 2

B0
C

C0
A0


B
;
O 3
3

Figure 1.
It follows that
\CAB 0 = \CAA0 + \B 0 AA0 = \O CA0 + \O B 0 A0 3 3

= \CA0 B 0 ; \CO B0 . 3 (1)

Since O is the centre of ; , we have
3 3

\CBB 0 = \CC 0 B = \CO B 0 . 1

2 3

Since \ABB 0 = \AA0 B and \AC 0 C = \AA0 C , we get

\ABC + \AC 0 B 0 = (\ABB 0 ; \CBB 0 ) + (\AC 0 C ; \CC 0 B 0 )
= (\ABB0 + \AC 0 C ) ; (\CBB 0 + \CC 0 B 0 )
= (\AA0 B 0 + \AA0 C ) ; \CO B 0 3

= \CA0 B 0 ; \CO B0 . 3

Thus we have, from (1),

\CAB 0 = \ABC + \AC 0 B 0 . (2)
A Let P be an interior point of \CAB 0
such that
\CAP = \ABC . (3)
C B0

C Then we obtain from (2)

0

B P \B 0 AP = \AC 0 B 0 . (4)
314

From (3) and (4), we have that the circumcircles of 4ABC and 4AB 0 C 0 are
tangent to AP at A. Thus, the circumcircles of 4ABC and 4AB 0 C 0 touch
at A.
Comment. As shown in the proof, the condition of the orthogonality of ; 1
and ; is not necessary.
2

Also solved by MIGUEL AMENGUAL COVAS, Cala Figuera, Mallorca, Spain;

CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; NIKOLAOS DERGIADES, Thessa-
loniki, Greece; PARAGIOU THEOKLITOS, Limassol, Cyprus, Greece; PETER Y. WOO, Biola Uni-
versity, La Mirada, CA, USA; JEREMY YOUNG, student, University of Cambridge, Cambridge,
UK; and the proposer.
All solvers other than Seimiya made use of inversion.

2455. [1999: 307] Proposed by Gerry Leversha, St. Paul's School,

London, England.
Three equal circles, centred at A, B and C intersect at a common
point P . The other intersection points are L (not on circle centre A), M
(not on circle centre B ), and N (not on circle centre C ). Suppose that Q is
the centroid of 4LMN , that R is the centroid of 4ABC , and that S is the
circumcentre of 4LMN .
(a) Show that P , Q, R and S are collinear.
(b) Establish how they are distributed on the line.
SolutionTby Ho-joo Lee, Seoul,
T South Korea (modiT ed by the editor).
Let AB PN = Z; BC PL = X , and CA PM = Y . Since
AP = AN = BP = BN; PANB is a rhombus whose diagonals AB and
PN are perpendicular and bisect one another (so that Z is the mid-point
of both AB and PN ). Similarly for BC ? PL and CA ? PM . Since
4XY Z is the mid-point triangle of 4ABC (with sides XY jjAB, etc.), we
have ZP ? XY; XP ? Y Z , and Y P ? ZX , so that P is the orthocentre of
4XY Z . Moreover, R (the centroid of 4ABC ) is the centroid of 4XY Z .
Let T denote the circumcentre of 4XY Z . Thus P , R, and T lie on the Euler
line of 4XY Z and satisfy
PR = 2RT . (1)
;

The dilatation P not only carries 4LMN to 4XY Z , but its centroid Q
1

to R | so that
2

PQ = 2PR , (2)
and its circumcentre S to T | so that
QS = 2RT . (3)
Since P , R, and T are collinear, so are P , Q, R, and S . Finally, we have
PR = RQ (from (2)) and PR = QS (from (1) and (3)), so that the points P ,
R, Q, and S are equally spaced in this order along the line.
 315


Also solved by SEFKET ARSLANAGI C,  University of Sarajevo, Sarajevo, Bosnia and
Herzegovina; MICHEL BATAILLE, Rouen, France; CHRISTOPHER J. BRADLEY, Clifton College,
Bristol, UK; NIKOLAOS DERGIADES, Thessaloniki, Greece; WALTHER JANOUS, Ursulinengym-
nasium, Innsbruck, Austria; MICHAEL LAMBROU, University of Crete, Crete, Greece; TOSHIO
SEIMIYA, Kawasaki, Japan; D.J. SMEENK, Zaltbommel, the Netherlands; CHOONGYUP SUNG,
Pusan, Korea; PARAGIOU THEOKLITOS, Limassol, Cyprus, Greece; PETER Y. WOO, Biola Uni-
versity, La Mirada, CA, USA; JEREMY YOUNG, student, University of Cambridge, Cambridge,
UK; and the proposer.
An immediate consequence of the featured proof is that P (the orthocentre of 4XY Z )
is also the orthocentre of 4LMN , and that the circle LMN is congruent to the three given
circles. Since PLMN forms an orthocentric quadrangle (meaning each point is the orthocentre
of the triangle formed by the other three), the con guration of the congruent circumcircles of
these four triangles has a long history. The converse | that the circle LMN is congruent to the
original 3 | can be traced back to R.A. Johnson [A circle theorem, American Math. Monthly
23 (1916), 161-162; see also Arnold Emch, Remarks on the foregoing circle theorem, 23 (1916),
162-164 ], and to G: Titeica (for whom the editor has never seen a speci c reference, so does
not know what he (or she) did and when it was done). This con guration was Dana McKenzie's
starting point; his work [Triquetras and porisms, College Math. J. 23 2 (March, 1992), 118-131]
would certainly be of interest to CRUX with MAYHEM readers.

2456. [1999: 307] Proposed by Gerry Leversha, St. Paul's School,

London, England.
Two circles intersect orthogonally at P . A third circle touches them at
Q and R. Let X be any point on this third circle. Prove that the circumcircles
of 4XPQ and 4XPR intersect at 45 .
Solution by Miguel Amengual Covas, Cala Figuera, Mallorca, Spain.
Let P be the second point of intersection of the circles which cut or-
1
thogonally.
Orthogonality invites inversion, so invert with respect to the circle with
centre P and radius PP . 1

The orthogonal circles invert into a pair of mutually perpendicular lines

through P , and the third circle becomes the circle tangent to these lines at
the maps, Q0 and R0 , of Q and P . That is to say, the inverse of the third
1

circle is a circle on the chord Q0 R0 which contains a right angle. Clearly, this
circle passes through X 0 , the inverse of the point X .
The circumcircles of 4XOQ and 4XPR become the straight lines
through X 0 Q0 and X 0 R0 respectively.
Since an inscribed angle equals, in degrees, half of its intercepted arc,
we clearly have
\Q0 X 0 R0 =  = 45
90
2
(Figure 1)
or \Q0 X 0 R0 = 270
2

= 135 = 180 ; 45 (Figure 2)
and the conclusion follows.
316

qX

q
Q q
Q
Qq P q Rq
qq q XRq
Qq P q X
0 0
1 1
q 0

qX R 0 R0

P
q 0

P
q

Figure 1 Figure 2

Also solved by SEFKET ARSLANAGI C,  University of Sarajevo, Sarajevo, Bosnia and
Herzegovina; MICHEL BATAILLE, Rouen, France; CHRISTOPHER J. BRADLEY, Clifton Col-
lege, Bristol, UK; NIKOLAOS DERGIADES, Thessaloniki, Greece; TOSHIO SEIMIYA, Kawasaki,
Japan; D.J. SMEENK, Zaltbommel, the Netherlands; PETER Y. WOO, Biola University, La Mi-
rada, CA, USA; JEREMY YOUNG, student, University of Cambridge, Cambridge, UK; and the
proposer.
All solvers used inversion, except for Arslanagic, Seimiya and Smeenk, who used direct
Euclidean methods.

2457. [1999: 308] Proposed by Gerry Leversha, St. Paul's School,

London, England.
In quadrilateral ABCD, we have \A + \B = 2 < 180, and BC =
AD. Construct isosceles triangles DCI , ACJ and DBK , where I , J and K
are on the other side of CD from A, such that \ICD = \IDC = \JAC =
\JCA = \KDB = \KBD = .
(a) Show that I , J and K are collinear.
(b) Establish how they are distributed on the line.
I. Solution by Peter Y. Woo, Biola University, La Mirada, CA, USA.
Since 4ICD  4JCA, we have 4JIC  4ADC [by side-angle-
side], so that \JIC = \ADC . Similarly, from 4ICD  4KBD, we have
4DIK  4DCB and \DIK = \DCB. Therefore, \JIK = \JIC +
\DIK ; \CID = (\ADC + \DCB ) ; \CID = (2 ; 2 ) ; ( ; 2 ) = 
[since in ABCD, \A + \B = 2 , while 4CID has base angles equal to ].
Consequently, JIK is a straight line.
IJ AD BC IK IK
Moreover,
IC = CD = CD = ID = IC . Therefore, IJ = IK ,
or I is the mid-point of JK . QED.
 317

II. Solution by Nikolaos Dergiades, Thessaloniki, Greece.

Let M be the matrix that represents the rotation about the origin through
angle 180 ; 2 6= 0. So
;! ;! ;! ;! ;! ;! ;! ;!
M
JA = JC , M
ID = IC M
KD = KB , M
DA = CB ,
and hence
;! ;! ;! ;!  ;! ;! ;!
M
IJ = M
ID + DA + AJ = M
ID +M
DA ;M
JA
;! ;! ;! ;! ;! ;!
= IC + CB ; JC =IB ; IC ; IJ
;! ;! ;!
= IB ; IC + IJ :
;! ;! ;! ;! ;! ;! ;!
M
IK = M
ID + DK = M
ID ;M
KD=IC ; KB
;! ;! ;!
= IC ; IB + IK :
;! ;!  ;! ;! ;! ;! !
By addition, M
IJ + IK = IJ + IK , which implies that IJ + IK = 0 ,
which means that the points I , J , K are collinear and that I is the mid-point
of the segment JK .

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 (a second solution); WALTHER
JANOUS, Ursulinengymnasium, Innsbruck, Austria; TOSHIO SEIMIYA, Kawasaki, Japan;
D.J. SMEENK, Zaltbommel, the Netherlands; and the proposer.

2458. [1999: 308] Proposed by Nikolaos Dergiades, Thessaloniki,

Greece.
Let ABCD be a quadrilateral inscribed in the circle centre O, radius
R, and let E be the point of intersection of the diagonals AC and BD. Let P
be any point on the line segment OE and let K , L, M , N be the projections
of P on AB , BC , CD, DA respectively.
Prove that the lines KL, MN , AC are either parallel or concurrent.
Solution by Toshio Seimiya, Kawasaki, Japan.
Let F and G be the feet of the perpendiculars from E to AD and BC ,
respectively. The triangles EAD and EBC are similar, because
\EAD = \CAD = \DBC = \EBC
and
\AED = \BEC .
318

D
A F NH

K O
E P M
B
G L
S
C
Since EF ? AD and EG ? BC , then 4EAF is similar to 4EBG and
4EDF is similar to 4ECG. Hence AF : BG = EF : EG = FD : GC
and therefore,
AF : FD = BG : GC . (1)
Let H and S be the feet of the perpendiculars from O to AD and BC , re-
spectively. Since O is the centre of the circle, AH = HD and BS = SC ,
giving
AH : HD = BS : SC . (2)
Since EF , PN and OH are perpendicular to AD, then EF k PN k OH ,
and therefore, FN : NH = EP : PO. Similarly, GL : LS = EP : PO, so
that
FN : NH = GL : LS . (3)
From (1), (2) and (3) we obtain
AN : ND = BL : LC ; (4)
denote these two ratios by p. Similarly,
AK : KB = DM : MC ; (5)
denote these two ratios by q .
If p = q , then, by Thales' Theorem, KL k AC and MN k AC .
If p 6= q , let T be the intersection of KL and AC . By Menelaus'
Theorem,
AT
CL
BK = 1 .
TC LB KA
Using (4) and (5), we obtain
AT
DN
CM = 1 .
TC NA MD
 319

T D
N
A
M
K

B L C
It follows (by the converse of Menelaus' Theorem) that T , M and N are
collinear. Therefore, KL, MN and AC are concurrent at T .
Also solved by MICHEL BATAILLE, Rouen, France; PETER Y. WOO, Biola University, La
Mirada, CA, USA; and the proposer.

2459. [1999: 308] Proposed by Vedula N. Murty, Visakhapatnam,

India, modi ed by the editors.
Let P be a point on the curve whose equation is y = x . Suppose that 2

the normal to the curve at P meets the curve again at Q. Determine the
minimal length of the line segment PQ.
Solution by Michel Bataille, Rouen, France.
Let P = (t; t ) be a point on the parabola with t 6= 0. Then
2

Q = (u; u ), where u 6= 0 and u 6= t. The slope of the2tangent

2
line at
P is 2t, so the slope of the normal at P is ; t ; that is, uu;;tt2 = ; t . 1 1

Hence, u = ;t ; t , and
2 2
1
2

PQ = (t ; u) + (t ; u ) = (t ; u) [1 + (t + u) ]
2 2 2 2 2 2 2

     
= 2t + 21t 1 + 41t = 4t + 41t + 2 1 + 41t .
2
2
2 2 2

Substituting z = t2 > 0, we obtain

1

 4
 
PQ = z + z1 + 2 1 + z1
2 2
2 2

1  1   4  27 2

= z + 3z + 3 + z = z ; 2 1 + z + 4 .
2

Hence PQ  with equality if pand only if z = . Therefore, the mini-

2 27
4 
1
2

mal length of the segment PQ is obtained when P is either p ; or

3 3 1 1

  2 2 2

;p ; .
1
2
1
2
320


Also solved by SEFKET ARSLANAGI C,  University of Sarajevo, Sarajevo, Bosnia and
Herzegovina; CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; JAMES T. BRUENING,
Southeast Missouri State University, Cape Girardeau, MO, USA; MIGUEL ANGEL CABEZ ON 
OCHOA, Logro~no, Spain; ELSIE CAMPBELL, Angelo State University, San Angelo, TX, USA;
CON AMORE PROBLEM GROUP, Royal Danish School of Educational Studies, Copenhagen,
Denmark (two solutions); NIKOLAOS DERGIADES, Thessaloniki, Greece; DAVID DOSTER,
Choate Rosemary Hall, Wallingford, CT, USA; KARTHIK GOPALRATNAM, student, Angelo State
University, San Angelo, TX, USA; RICHARD I. HESS, Rancho Palos Verdes, CA, USA; JOHN
G. HEUVER, Grande Prairie Composite High School, Grande Prairie, Alberta; JOE HOWARD,
New Mexico Highlands University, Las Vegas, NM, USA; WALTHER JANOUS, Ursulinengym-
nasium, Innsbruck, Austria; MURRAY S. KLAMKIN, University of Alberta, Edmonton, Alberta;

V ACLAV KONECN  Y,
 Ferris State University, Big Rapids, MI, USA; MICHAEL LAMBROU,
University of Crete, Crete, Greece; GERRY LEVERSHA, St. Paul's School, London, England;
ROBERT P. SEALY, Mount Allison University, Sackville, New Brunswick; HARRY SEDINGER,
St. Bonaventure University, St. Bonaventure, NY, USA; D.J. SMEENK, Zaltbommel, the Neth-
erlands; DIGBY SMITH, Mount Royal College, Calgary, Alberta; CHOONGYUP SUNG, Pusan
Science High School, Pusan, Korea; ARAM TANGBOONDOUANGJIT, Carnegie Mellon Uni-
versity, Pittsburgh, PA, USA; PARAGIOU THEOKLITOS, Limassol, Cyprus, Greece; PANOS E.
TSAOUSSOGLOU, Athens, Greece; DAVID VELLA, Skidmore College, Saratoga Springs, NY,
USA; KENNETH M. WILKE, Topeka, KS, USA; JEREMY YOUNG, student, University of Cam-
bridge, Cambridge, UK; and the proposer. There was also one incorrect solution submitted.
Klamkin noted that this problem has appeared as # 11 in the rst William Lowell
Putnam Mathematical Competition [1]. Most of the solvers used Calculus to nd the mini-
mum. Dergiades, Leversha and Sung gave \No Calculus" solutions.
[1] A.M. Gleason, R.E. Greenwood and L.M. Kelly, The William Lowell Putnam Mathematical
Competition, Problems and Solutions: 1938-1964, M.A.A., Washington, D.C., 1980, pp. 5, 91.

Congratulations, Paco Bellot!

The International Federation of Mathematics Competitions has awarded the Paul
Erdos} Prize to Francisco Bellot Rosado in recognition of his work, since 1988, with
Olympiad students. The prize will be presented in August 2000 during the Interna-
tional Conference of Mathematics Education in Japan. Francisco Bellot Rosado is the
rst Spaniard to receive this prize, and is the rst High School teacher in the world
to have been awarded it.
Congratulations, Paco Bellot (as he is known to his students and colleagues.)

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