CRUXv 24 N 7
CRUXv 24 N 7
CRUXv 24 N 7
line q cuts C1A1 at B2 ; line r cuts A1 B1 at C2 . Prove that the lines AA2 ,
BB2, CC2 concur at a point D lying on the circumcircle of triangle ABC .
6. The Alpine Club consisting of n members organizes four high-
mountain expeditions for its members. Let E1 , E2, E3, E4 be the four teams
participating in these expeditions. How many ways are there to compose
those teams, given the condition that E1 \E2 6= ;, E2 \E3 6= ;, E3 \E4 6= ;?
Problems of the Team Contest
June 30, 1995 (Time: 4 hours)
7. For every integer c consider the equation 3y4 + 4cy3 + 2xy +
48 = 0, with integer unknowns x and y. Determine all integers c for which
the number of solutions (x; y ) in pairs of integers satisfying the additional
conditions (A) and (B) is a maximum:
(A) the number jxj is the square of an integer;
(B) the number y is square-free (that is, there is no prime p with p2
dividing y ).
8. Consider the cube with vertices f1; 1; 1); that is, the set
f(x; y; z) : jxj 1; jyj 1; jzj 1g. Let V1; : : : ; V95 be points of that
cube. Denote by vi the vector from (0; 0; 0) to Vi. Consider the 295 vectors
of the form s1v1 + s2 v2 + + s95 v95, where si = 1 or si = ,1.
(a) Let d = 48. Show that among all such vectors one can nd a vector
w = (a; b; c) with a2 + b2 + c2 d.
(b) Find a number d < 48 with the same property.
Note: The smaller d, the better mark will be attracted by the solution.
9. Prove that the following inequality holds for all integers n;m 1
and all positive real numbers x; y :
(n , 1)(m , 1)(xn+m + yn+m ) + (n + m , 1)(xn ym + xm y n)
nm(xn+m,1y + xyn+m,1):
The next contest we give was also collected by Bill Sands while he
was assisting at the IMO in Toronto. These are the problems of the 9th
Iberoamerican Mathematical Olympiad held September 20, 21 in Fortaleza,
Brazil. Students were given 4 12 hours each day.
9th IBEROAMERICAN MATHEMATICAL OLYMPIAD
Fortaleza, Brazil, September 20{21, 1994
First Day | Time: 4.5 hours
1. (Mexico): A natural number n is called brazilian if there exists an
integer r, with 1 < r < n , 1, such that the representation of the number
387
n in base r has all the digits equal. For example, 62 and 15 are brazilian,
because 62 is written 222 in base 5 and 15 is 33 in base 4. Prove that 1993
is not brazilian, but 1994 is brazilian.
2. (Brazil): Let ABCD be a cyclic quadrilateral. We suppose that
there exists a circle with centre in AB , tangent to the other sides of the
quadrilateral.
(i) Show that AB = AD + BC .
(ii) Calculate, in terms of x = AB and y = CD, the maximal area that
such a quadrilateral can reach.
3. (Brazil): In each cell of an n n chessboard is a lamp. When a lamp
is touched, the state of this lamp, and also the state of all the lamps in its
row and in its column, is changed (switched from OFF to ON and vice versa).
At the beginning, all the lamps are OFF. Show that it is always possible, with
suitable sequence of touches, to turn ON all the lamps of the chessboard,
and nd, in terms of n, the minimal number of touches in order that all the
lamps of the chessboard are ON.
Second Day | Time: 4.5 hours
4. (Brazil): The triangle ABC is acute, with circumcircle k. Let P be
an internal point to k. The lines AP , BP , CP meet k again at X , Y , Z .
Determine the point P for which triangle XY Z is equilateral.
5. (Brazil): Let n and r be two positive integers. We wish to con-
struct r subsets of f0; 1; : : : ; n , 1g, called A1 ; : : : ; Ar , with card(Ai) = k
and such that, for each integer x, 0 x n , 1, there exist x1 2 A1 ;
x2 2 A2; : : : ; xr 2 Ar (an element in each subset), with
x = x1 + x2 + + xr :
Find, in terms of n and r, the minimal value of k.
6. (Brazil): Show that all natural numbers n 21000000 can be
obtained beginning at 1 with less than 1100000 sums; that is, there exists
a nite sequence of natural numbers x0 ; x1 ; : : : ; xk , with k < 1100000,
x0 = 1, xk = n, such that for each i = 1; 2; : : : ; k, there exists r, s, with
0 r < i, 0 s < i, and xi = xr + xs .
4. The incircle of a triangle divides one of its medians into three equal
parts. Find the ratio of the sides of the triangle.
5. The function f is given on the segment [0; 1]. It is known that
f (x) 0 and f (1) = 1. Besides that, for any two numbers x1 and x2, if
x1 0, x2 0 and x1 + x2 1, then f (x1 + x2 ) f (x1) + f (x2).
(a) Prove that f (x) 2x for any x.
(b) Does the inequality f (x) 1:9x hold for every x?
389
GRADE XI
1. (Same as IX.3)
2. (Same as IX.1)
3. How many solutions has the equation x = 1995 sin x + 199?
4. (Same as IX.4)
5. A natural number is written in each square of an m n rectangular
table. By one move, it is allowed to double all numbers of any row or subtract
1 from all numbers of any column. Prove that by repeating these moves
several times, all numbers in the table become zeros.
Next a bit of housekeeping. After the columns were set, and before they
appeared in print form, I received solutions from Pavlos Maragoudakis to the
six problems of the Swedish contest for which we published the solutions last
issue [1997: 196; 1998: 328{329]. He also submitted solutions to problems
1, 2, 3 and 5 of the Dutch Mathematical Olympiad, Second Round, 1993
[1997: 197; 1998: 329{332]. Last issue we gave solutions to all but the last
problem. It was rather unfortunate timing in terms of acknowledging his
contribution, but we are able to close out the le by having a complete set
of solutions from the readers.
5. P1; P2; : : : ; P11 are eleven distinct points on a line. PiPj 1 for
every pair Pi , Pj . Prove that the sum of all (55) distances Pi Pj , 1 i <
j 11 is smaller than 30.
Solution by Pavlos Maragoudakis, Pireas, Greece.
Without loss of generality we suppose that P1 ; P2 ; : : : ; P11 are adja-
cent.
P1 P2 qqq P11
Also setting the record straight, I found amongst the solutions for an-
other contest, the solution by Miguel Amengual Covas, Cala Figuera, Mal-
lorca, Spain to problem 2 of the Dutch Mathematical Olympiad, Second
Round, for which we published a solution last issue [1997: 197, 1998: 330{
331]. My apologies.
While we do not normally give solutions to problems of the USAMO,
I am giving two comments/solutions from our readers to problems of the
USAMO 1997 [1997: 261, 262].
2. Let ABC be a triangle, and draw isosceles triangles BCD, CAE,
ABF externally to ABC , with BC; CA; AB as their respective bases. ! !Prove !
that the lines through A; B; C perpendicular to the lines EF; FD; DE ,
respectively, are concurrent.
Comment by Mansur Boase, student, St. Paul's School, London, Eng-
land.
The result is immediate from Steiner's Theorem:
If the perpendiculars from the vertices A, B , C of a triangle ABC to the
sides B1 C1 , C1 A1 , and A1 B1 , respectively, of a second triangle A1 ; B1 ; C1
are concurrent, then the perpendiculars from the vertices A1 ; B1 ; C1 of the
triangle A1 B1C1 to the sides BC , CA, AB are also concurrent.
5. Prove that for positive real numbers a, b, c,
(a3 + b3 + abc),1 + (b3 + c3 + abc),1 + (c3 + a3 + abc),1 (abc),1:
Solutions by Mansur Boase, student, St. Paul's School, London, Eng-
land and by Murray S. Klamkin, University of Alberta, Edmonton, Alberta.
We give Klamkin's presentation.
Since the inequality is homogeneous, we can assume abc = 1. Then if
we let x = a3 , y = b3, z = c3 , the inequality becomes
1 1 + 1
1+x+y 1+y+z 1+z+x 1
+ (1)
where xyz = 1 and x, y , z are positive. On expanding, (1) is equivalent to
(x + y + z)(xy + yz + zx , 2) 3 :
391
Next we give two solutions by our readers to two problems of the 3rd
Ukrainian Mathematical Olympiad, March 26{27, 1994 given in [1997: 262].
2. (9{10) A convex polygon and point O inside it are given. Prove that
for any n ,,!
> 1 there ! exist points A1 ; A2 ;, : : : ; An on the sides of the polygon
such that OA1 + ,, OA2 + : : : + ,,! OAn = ! 0.
Solution by Murray S. Klamkin, University of Alberta, Edmonton, Al-
berta.
It follows by continuity that there always exists a chord A1 OA01 such
that A1 O = A01 O and hence ,, OA !1 + ,,!
OA 01 = , !
0 . Similarly, there exists
a chord A2 A03 which is bisected by the midpoint O1 of OA01 . It follows by
the parallelogram law that ,,
,,! OA!2 + ,,!OA03 = ,,! OA01 and hence ,, OA!1 + ,,OA!2 +
OA03 = ,! 0 . Again similarly there exists a chord A3 A04 which is bisected by
the midpoint of OA03 so that ,, OA!1 + ,,
OA!2 + ,, OA!3 + ,,!
OA04 = ,! 0 , and so on
for any number of vectors n > 1.
3. (10) A sequence of natural numbers ak , k 1, such that for each k,
ak < ak+1 < ak + 1993 is given. Let all prime divisors of ak be written for
every k. Prove that we receive an in nite number of di erent prime numbers.
Solution by Pavlos Maragoudakis, Pireas, Greece.
We suppose that there is a sequence of natural numbers such that
ak < ak+1 < ak + 1993, k 1, and the set of all prime divisors of all
ak is nite. Let p1a;1p2; : : a: ; pr list all the prime divisors of all ak . Now every
ak has the form p1 : : : pr r , ai = 0; 1; 2; : : : , i = 1; : : : ; r.
Let S = fpa1 1 pa2 2 : : : par r j ai = 0; 1; 2; : : : ; i = 1; 2; : : : ; rg:
De ne (xn) with x1 < x2 < : : : such that S = fxn = n 2 N g. We
have that ak < ak+1 and ak 2 S , k 1. Thus (ak ) is a subsequence of
392
(xk ). But ak < ak,1 + 1993 < ak,2 + 2 1993 < < a1 + (k , 1)1993 <
k(a1 + 1993); k 1.
Hence ak < k(a1 + 1993), k 1. Therefore
X1 1 X1 1 1 X1 1
> = a + 1993 n = +1
n=1 an n=1 n(a1 + 1993) 1 n=1
1
X 1 1
1 X X 1
and
n=1 na x =
n=1 n
1 pa2 : : : pa4
a1 ;:::;ar 0 p1 2 r
1 1 !
X 1
X 1
!
1
X 1
!
= a2 : : :
a1 =0 p1 a2 =0 p2 r =0 pr
1 ar
= p p,1 1 p p,2 1 p p,4 1 < +1 ;
1 2 4
which is a contradiction.
P
B C
393
The angles at the base of the isosceles triangle PAC are each 90 , . Also
4BPC is isosceles, having base angles
C , (90 , ) = 2B + , 90;
and so
\BPA = 180 , (\PBA + \BAP )
= 180 , [B , (2B + , 90 ) + 180 , 2 , 3B ]
= 4B + 3 , 90 :
As usual, let a, b and c denote the lengths of the sides BC , AC and AB .
By the Law of Cosines, applied to 4BPA, where PA = b and
PB = PC = 2b sin ,
c2 = b2 + (2b sin )2 , 2 b 2b sin cos(4B + 3 , 90) ;
so that
c2 = b2[1 + 4 sin2 , 4 sin sin(4B + 3 )] : (2)
We now use the fact that \C = 2\B is equivalent to the condition
c2 = b(b + a), which has appeared before in CRUX [1976: 74], [1984:
278] and [1996: 265{267]. Since a = 2 PC cos(2B + , 90 ) =
4b sin sin(2B + ), we have
c2 = b2[1 + 4 sin sin(2B + )] : (3)
Therefore, from (2) and (3), we get
b2[1 + 4 sin2 , 4 sin sin(4B + 3 )] = b2[1 + 4 sin sin(2B + )] ;
which simpli es to
sin , sin(4B + 3 ) = sin(2B + ) :
Since sin , sin(4B + 3 ) = ,2 cos(2B + 2 ) sin(2B + ), this equation
may be rewritten as
sin(2B + ) [1 + 2 cos(2B + 2 )] = 0:
Since, from (1), 2B + < 180 , we must have 1+2 cos(2B +2 ) = 0,
giving cos(2B + 2 ) = ,1=2 ; that is,
2B + 2 = 120 (4)
since, again from (1), 2B + 2 < 180 .
Finally, we may eliminate B between (1) and (4) to obtain = . The
result follows.
394
As known, the only rational roots of the latter are factors of n3, and
consequently x, y , z are integers.
The only triples of integers (x; y;z ), aside from permutations, which
satisfy x1 + y1 + 1z = n2 are
(1; 1; 1); (1; 2; 2); (2; 3; 6); (2; 4; 4); and (3; 3; 3):
5. Let A1; A2; : : : ; A8 be the vertices of a parallelepiped and let O be
its centre. Show that
4(OA21 + OA22 + + OA28 ) (OA1 + OA2 + + OA8 )2 :
Solution by Murray S. Klamkin, University of Alberta, Edmonton,
Alberta.
Let one of the vertices be the origin and let the vectors B + C ,
C + A, A + B denote the three coterminal edges emanating from this origin.
Then the vectors to the remaining four vertices are S + A, S + B , S + C ,
and 2S where S = A + B + C and which is also the vector to the centre.
The inequality now becomes
2(S2 + A2 + B 2 + C 2) (jS j + jAj + jB j + jC j)2 ;
or
S2 + A2 + B2 + C 2 2jSjfjAj + jBj + jC jg +2fjBj jC j + jC j jAj + jAj jBjg:
Since
S2 = A2 + B2 + C 2 + 2B C + 2C A + 2A B;
the inequality now becomes
S2 ,B C ,C A,AB jSjfjAj+jBj+jC jg+fjBj jC j+jC j jAj+jAj jBjg:
Clearly,
S2 jSjfjAj + jBj + jC jg
and
,B C , C A , A B jBj jC j jAj + jAj jBj :
There is equality if and only if the parellelepiped is degenerate,
for example, B = C = O.
BOOK REVIEWS
Edited by ANDY LIU
Dissections : Plane & Fancy
by Greg N. Frederickson,
published by Cambridge University Press, 1997,
ISBN# 0-521-57197-9, hardcover, 310+ pages, $34.95.
Reviewed by Andy Liu, University of Alberta.
This is a much awaited sequel to Harry Lindgren's 1964 classic work,
Geometric Dissections, which the author (G.N.F.) revised and augmented in
the 1972 Dover edition. Actually, the current volume is much more than just
a sequel. It is the most comprehensive treatise on the subject of geometric
dissections. It may be enjoyed on at least three levels.
First and foremost, this book is a collection of interesting dissection
puzzles, old and new. Only some background in high school geometry is
needed to fully enjoy these problems. Can you cut an octagon into ve pieces
and rearrange them into a square? How about turning a star into a pentagon?
The solutions, which are both appealing but for somewhat opposing reasons,
are shown below.
Octagon to square
Star to pentagon
This book is also an instructive manual on the art and science of geo-
metric dissections. While one may admire the ingenuity which produced
the spectacular solutions, the author probes into the underlying fabric which
might have led to such incisive insight. Many techniques are discussed, too
many to enumerate here. A favourite is that of tessellation. Below are two
tilings which might have suggested the dissections above.
397
p
(a) 25 2
p
(b) 252 2 (c) 20
p
(d) 252 (1 + 2)
p
(e) 25(1 + 2)
4. Last summer, I planted two trees in my yard. The rst tree came in
a fairly small pot, and the hole that I dug to plant it in lled one wheelbarrow
load of dirt. The second tree came in a pot, the same shape as that of the
rst tree, that was one-and-a-third times as deep as the rst pot and one-
and-a-half times as big around. Let us make the following assumptions:
i) The hole for the second tree was the same shape as for the rst tree.
399
ii) The ratios of the dimensions of the second hole to those of the rst
hole are the same as the ratios of the dimensions of the pots.
Based on these assumptions, the number of wheelbarrows of dirt that I lled
when I dug the hole for the second tree was:
(a) 2 (b) 2:5 (c) 3 (d) 3:5 (e) none of these
5. You have an unlimited supply of 5-gram and 8-gram weights that
may be used in a pan balance. If you use only these weights and place them
only in one pan, the largest number of grams that you cannot weigh is:
(a) 22 (b) 27 (c) 36 (d) 41 (e) there is no largest number of grams
6. If all the whole numbers from 1 to 1;000;000 are printed, the num-
ber of times that the digit 5 appears is:
(a) 100;000 (b) 500;000 (c) 600;000 (d) 1;000;000 (e) 2;000;000
7. The perimeter of a rectangle is x centimetres. If the ratio of two
adjacent sides is a : b, with a > b, then the length of the shorter side, in
centimetres, is:
(a) abx
+b
(b) x2 , b (c) a2+bxb (d) 2(aax+b) (e) 2(abx+b)
8. The sum of the positive solutions to the equation xxpx = (xpx)x
is:
(a) 1 (b) 1 12 (c) 2 41 (d) 2 21 (e) 3 41
9. Two circles, each with a radius of one unit, touch as shown. AB
and CD are tangent to each circle. The area, in square units, of the shaded
region is:
A B
C D
(a) (b) 4 (c) 2 , 2 (d) 4 , (e) none of these
10. A parabola with a vertical axis of symmetry has its vertex at (0; 8)
and an x{intercept of 2. If the parabola goes through (1; a), then a is:
(a) 5 (b) 5:5 (c) 6 (d) 6:5 (e) 7
11. A ve litre container is lled with pure orange juice. Two litres of
juice are removed and the container is lled up with pure water and mixed
400
thoroughly. Then two litres of the mixture are removed and again the con-
tainer is lled up with pure water. The percentage of the nal mixture that
is orange juice is:
(a) 27 (b) 25 (c) 30 (d) 36 (e) 24
12. The lengths of the sides of a triangle are b + 1, 7 , b and 4b , 2.
The number of values of b for which the triangle is isosceles is:
(a) 0 (b) 1 (c) 2 (d) 3 (e) none of these
13. The number of times in one day when the hands of a clock form a
right angle is:
(a) 46 (b) 22 (c) 24 (d) 44 (e) 48
14. In my town some of the animals are really strange. Ten percent of
the dogs think they are cats and ten percent of the cats think they are dogs.
All the other animals are perfectly normal. One day, I tested all the cats and
dogs in the town and found that 20% of them thought that they were cats.
The percentage of the dogs and cats in the town that really are cats is:
(a) 12:5 (b) 18 (c) 20 (d) 22 (e) 22:5
15. A short hallway in a junior high school contains a bank of lockers
numbered one to ten. On the last day of school the lockers are emptied and
the doors are left open. The next day, a malicious math student walks down
the hallway and closes the door of every locker that has an even number.
The following day, the same student again walks down the hallway and, for
every locker whose number is a multiple of three, closes the door if it is open
and opens it if it is closed. On the next day, the student does the same thing
with every locker whose number is divisible by four. If the student continues
this procedure for a total of nine days, the number of lockers that are closed
after the ninth day is:
(a) 4 (b) 5 (c) 6 (d) 7 (e) 8
Last issue we gave the Junior High School Mathematics Contest, Pre-
liminary Round 1998 of the British Columbia Colleges. Here are the \ocial
solutions", which come our way from the organizer, Jim Totten, The Univer-
sity College of the Cariboo.
BRITISH COLUMBIA COLLEGES
Junior High School Mathematics Contest
Preliminary Round 1998
Time: 45 minutes
1. A number is prime if it is greater than one and divisible only by one
and itself. The sum of the prime divisors of 1998 is: (c)
401
A E
If we choose AE , CE , and CD as bases of the triangles then the lengths
of the corresponding perpendicular heights are 5, 3, and 8 cm. Hence, the
area of the polygon is 12 4 5 + 12 8 3 + 12 4 8 = 38.
9. A point P is inside a square ABCD whose side length is 16. P is
equidistant from two adjacent vertices, A and B , and the side CD opposite
these vertices. The distance PA equals: (e)
403
x
16
P
x x
A E B
The Pythagorean Theorem applied to triangle PEB gives (16,x)2+82 = x2 ,
162 +82
so that 16 , 32x + x + 8 = x , and x = 32 = 10.
2 2 2 2
That completes the Skoliad Corner for this issue. We need suitable
contests and solutions. I welcome any comments, criticisms, or suggestions
for the future direction of this feature.
Advance Announcement
The 1999 Summer Meeting of the Canadian Mathematical Society will
take place at Memorial University in St. John's, Newfoundland, from Satur-
day, 29 May 1999 to Tuesday, 1 June 1999.
The Special Session on Mathematics Education will feature the topic
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 the
Mayhem Editor, Naoki Sato, Department of Mathematics, Yale University,
PO Box 208283 Yale Station, New Haven, CT 06520{8283 USA. The electronic
address is still
[email protected]
The Assistant Mayhem Editor is Cyrus Hsia (University of Toronto).
The rest of the sta consists of Adrian Chan (Upper Canada College), Jimmy
Chui (Earl Haig Secondary School), Richard Hoshino (University of Waterloo),
David Savitt (Harvard University) and Wai Ling Yee (University of Waterloo).
a b
c
The diagram shows the three line segments parallel to each other and
emanating from a common line. This gure and relationship between the
line segments appear a lot. The reader is encouraged to prove this.
407
A C B
a
q q b
qc
B C
P
The reader is probably already familiar with the famous case x = 2
known as the Pythagorean Theorem: A triangle with sides a, b, and c is a
right-angled triangle if and only if the lengths satisfy a2 + b2 = c2 .
b c
a
Of course, no discussion about algebraic relations in the above form is
complete without mentioning the notorious Fermat's Last Theorem and the
recent announcement that it has nally been laid to rest. If x is an integer
with x > 2, then the claim is that no solution in the natural numbers exists
for a, b, and c. However, in our general case, the values are real, so we are
not limited by the above result to nding wild and wacky geometric or other
interpretations for it.
409
If the reader is curious, as are we, then try the following exercises to
nd geometric interpretations for special cases of the above relation. The
exercises are explorational and may not have nice solutions, if any. Readers
are welcome to submit any interesting results they nd.
Exercises
1. Let a, b, and c be the lengths of three line segments. Determine how
these three line segments are related geometrically if they satisfy the
relation
(a) c3 = a3 + b3,
(b) pc = pa + b,
p
(c) c12 = a12 + b12 .
2. It is clear that algebraically, all the relations are similar. However,
the geometric interpretations do not appear to be related. Is there a
general geometric description where each of the above geometric gures
is a special case?
3. The algebraic relation clearly does not work for the case x = 0. Is
there a way to de ne the relation so that it would be consistent with
everything else mentioned so far?
Mayhem Problems
The Mayhem Problems editors are:
Richard Hoshino Mayhem High School Problems Editor,
Cyrus Hsia 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. We request
that solutions from this issue be submitted by 1 September 1999, for publi-
cation in issue 8 of 1999.
Advanced Problems
Editor: Cyrus Hsia, 21 Van Allan Road, Scarborough, Ontario, Canada.
M1G 1C3 <[email protected]>
A221. Construct, using straightedge and compass only, the common
tangents of two non-intersecting circles.
A222. Does there exist a set of n consecutive positive integers such
that for every positive integer k < n, it is possible to pick k of these numbers
whose mean is still in the set?
A223. Proposed by Mohammed Aassila, Universite Louis Pasteur,
Strasbourg, France.
Suppose p is a prime with p 3 (mod 4). Show that for any set of
p , 1 consecutive integers, the set cannot be divided into two subsets so
that the product of the members of the one set is equal to the product of the
members of the other set.
(Generalization of Question 4, IMO 1970)
A224. Proposed by Waldemar Pompe, student, University of War-
saw, Poland.
Let P be an interior point of triangle ABC such that \PBA =
\PCA = (\ABC + \ACB )=3. Prove that
AC = AB :
AB + PC AC + PB
IMO Report
Adrian Chan, student, UCC, Toronto
This year's Canadian IMO team began with a week of training at the Univer-
sity of Calgary with lavish meal tickets. Then they were o to beautiful and
rocky Kananaskis where we all had an \adventurous" time. Once the team
stepped o the air-conditioned plane and into hot and muggy Taipei, Taiwan,
it marked the team's ocial arrival to the 39th International Mathematical
Olympiad.
The team consisted of the following members: Adrian \Oops I dropped
my..." Birka, Adrian \If You Will" Chan, Jimmy \Nuclear Aerial Strike" Chui,
Mihaela \Baia" Enachescu, Jessie \So Cute" Lei, and Adrian \Nailing Radar"
Tang. Team leader Dr. Christopher \Focus" Small was driven to the edge,
while deputy leader J.P. \It's So Easy" Grossman calmly polished o old
competitions one by one. Special thanks to leader observer Arthur \Rubik's
Cube" Baragar and deputy observer Dorette \Dutch" Pronk for their coaching
and experience. Also, thanks must go to Dr. Bill Sands of the University of
Calgary for organizing such a fun training session.
The contest itself seemed to continue the trend of dicult IMO's and
low medal cuto s. With 76 countries competing, Canada fared extremely
well, bringing back 1 gold, 1 silver, 2 bronze and an honourable mention.
The scores were as follows:
CAN 1 Adrian Birka 10
CAN 2 Adrian Chan 31 Gold Medal
CAN 3 Jimmy Chui 14 Bronze Medal
CAN 4 Mihaela Enachescu 30 Silver Medal
CAN 5 Jessie Lei 13 Honourable Mention
CAN 6 Adrian Tang 15 Bronze Medal
In this year's contest, Canada placed 20th out of 76 countries, up from
last year's 29th ranking. Best of luck to CAN 1, 4, and 6 as they continue
university studies at MIT, Harvard, and Waterloo respectively. CAN 2, 3,
and 5 are all eligible for next year's team. Hopefully there won't be as much
moaning of \Where did I go wrong?" next year around!
Special thanks must also go to Dr. Graham Wright of the Canadian
Mathematical Society for again taking care of the tab, and Professor Ed
Barbeau for his hard work and dedication to the training of potential IMO
candidates through his year-long correspondence program.
Although sometimes things didn't make sense, and the IMO ag some-
how disappeared, the 39th International Mathematical Olympiad ran
smoothly and was de nitely a success. The new experience of a place half-
way around the world with a stimulating culture was new to most of us. Best
of luck to all IMO hopefuls for the 1999 team, as yet another Canadian IMO
journey begins next July in Bucharest, Romania.
413
In Euler's time, mathematics was faster and looser than today, and
niceties such as limits were blatantly ignored. Here is an argument of Euler's
that seems to have no right to work, but does nonetheless. We conclude with
some avenues for exploration and an open question.
If r1, : : : , rn are the zeros of a polynomial a0 + a1 x + + an xn , and
none of the ri are zero (or a0 6= 0), then the negative of the linear term over
the constant term is the sum of the reciprocals of the roots:
, aa10 = r11 + + r1n :
What about power series? For example, cos x = 1 , x2=2+ x4 =24 , ,
so that
cos px = 1 , x + x , :
2
2 24
The zeroes of this function are the squares of the odd multiples of =2:
((2n + 1)=2)2, n = 0, 1, 2, : : : . One might hope that the principle for
polynomials given above still holds:
1
1 = X 1
2 n=0 ((2n + 1)=2)2
which can be rewritten as
1
X 1 2 :
= (1)
n=0 (2n + 1) 8
2
Can you use the power series for (sin px)=x to \prove" the identity
X1 1 2 ?
= (2)
n=1 n 6
2
Can you relate (2) to (1) by arguing that (2) minus a quarter of (2) is (1)?
P
If you write a short computer program to compute 1000 n =1 n2 , and com-
1
pare it to 2 =6, you'll see that they are indeed very close. In fact, they di er
by almost exactly 1=1000. Can youPexplain why this might be? Can you
guesstimate the di erence between 1000 n=1 (2n+1)2 and 2 =8?
1
And nally, can you conjure up other examples of this sort of argument,
to \prove" other arcane identities? If so, please let us know!
Acknowledgements. This note was inspired by the example of cos px
given in [A].
References
[A] S. Abhyankar, Historical ramblings in algebraic geometry and re-
lated algebra, Amer. Math. Monthly 83 (1976), no. 6, 409{448.
[E] L. Euler, Introductio in Analysin In nitorum, Berlin Academy, 1748.
415
some n = k, k 1. For n = k + 1,
Fk2+1 , FkFk+2 = Fk2+1 , Fk(Fk+1 + Fk)
= (Fk+1 , Fk )Fk+1 , Fk2
= Fk,1Fk+1 , Fk2
= ,(,1)k,1 by the induction hypothesis
= (,1)k;
so the formula holds for n = k + 1. Therefore, by mathematical induction,
the formula holds for all n 1.
Theorem 2. For all n 1,
F1 + F3 + + F2n,1 = F2n :
Proof. Using the recurrence relation n times, we have
F2n = F2n,1 + F2n,2
= F2n,1 + F2n,3 + F2n,4
=
= F2n,1 + F2n,3 + + F3 + F1 + F0
= F2n,1 + F2n,3 + + F3 + F1 :
Copyright c 1998 Canadian Mathematical Society
416
Exercise 1. Prove n
X
Fj2 = FnFn+1 :
j =1
Divisibility
Theorem 3. For all m, n 1,
Fn+m = FnFm+1 + Fn,1Fm :
Proof. We will prove this by induction on m. When m = 1,
Fn+1 = Fn 1 + Fn,1 1 = FnF2 + Fn,1F1;
so the formula holds for all n when m = 1. Assume that the formula holds
for all n when m = M . For m = M + 1,
Fn+M +1 = Fn+M + F(n,1)+M
= Fn FM +1 + Fn,1FM + Fn,1FM +1 + Fn,2FM
by the induction hypothesis
= Fn FM +1 + (Fn,1 + Fn,2 )FM + Fn,1 FM +1
= Fn (FM + FM +1 ) + Fn,1 FM +1
= Fn FM +2 + Fn,1FM +1;
so the formula holds for all n for m = M + 1. By mathematical induction,
the formula holds for all m, n 1.
Corollary 4. For all m, n 1, FnjFnm .
Proof. We will prove this by induction on m. When m = 1, Fn certainly
divides itself for every positive integer n. Suppose the statement holds for
all n when m = M . For m = M + 1,
Fn(M +1) = FnM +n = FnM Fn+1 + FnM ,1Fn
by Theorem 3. Since FnM is divisible by Fn by the induction hypothe-
sis, FnM Fn+1 + FnM ,1Fn is also divisible by Fn. This is equivalent to
FnjFn(M +1), so the result holds for m = M + 1. By mathematical induc-
tion, the formula holds for all m, n 1.
Exercise 2. Prove that for every positive integer n, there exist n con-
secutive, composite Fibonacci numbers.
Number Theory 101
We will now de ne a few terms in the interests of formality. For in-
tegers a and b, we say that a divides b if there exists an integer q such that
b = aq, and a is called a divisor of b. Given two non-zero integers a and b,
the largest number which divides both of them, denoted gcd(a; b), is called
417
Fn = 2
p5 2
:
This is called Binet's Formula for the Fibonacci sequence.
Exercise 5 . Let = (1 +
p5)=2. Prove that F is the integer closest
n n
to p :
5
420
Problems
1. Prove that the product of every four consecutive Fibonacci numbers is
the area of a Pythagorean triangle.
2. Prove that every positive integer can be written as a sum of distinct
Fibonacci numbers.
3. Prove n+1
bX
2 c n , k
Fn = k,1 :
k=1
4. Prove that if Fn is prime and n 5, then n is prime.
5. Prove that Fn + 1 is always composite for n 4.
6. Show that for any positive integer n, among the rst n2 Fibonacci num-
bers, there exists at least one that is divisible by n.
7. De ne a Fibonacci prime to be a Fibonacci number that is prime. Prove
or disprove: There are in nitely many Fibonacci primes. (Note that this
is an open problem.)
421
r
5m -
6
r 6
10 m
6m
? ?
PROBLEMS
Problem proposals and solutions should be sent to Bruce Shawyer, De-
partment of Mathematics and Statistics, Memorial University of Newfound-
land, St. John's, Newfoundland, Canada. A1C 5S7. Proposals should be ac-
companied by a solution, together with references and other insights which
are likely to be of help to the editor. When a submission is submitted with-
out 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 inter-
esting problems may also be acceptable provided that they are not too well
known, and references are given as to their provenance. Ordinarily, if the
originator of a problem can be located, it should not be submitted without
the originator's permission.
To facilitate their consideration, please send your proposals and so-
lutions on signed and separate standard 8 12 "11" or A4 sheets of paper.
These may be typewritten or neatly hand-written, and should be mailed
to the Editor-in-Chief, to arrive no later than 1 April 1999. They may also
be sent by email to [email protected]. (It would be appreciated if
email proposals and solutions 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.
SOLUTIONS
No problem is ever permanently closed. The editor is always pleased to
consider for publication new solutions or new insights on past problems.
h ab sin v sin v b2 a2 b a
Similarly, r
1 = 1 1 + 1 , 2 1 1 cos v ;
h sin v d2 c2 d c
so that
r r
1 + 1 , 2 1 1 cos v = 1 + 1 , 2 1 1 cos v ; (1)
b2 a2 b a d2 c2 d c
and similarly,
r r
1 + 1 + 2 1 1 cos v = 1 + 1 + 2 1 1 cos v : (2)
a2 d2 a d b2 c2 b c
Now, consider another convex quadrilateral A1 B1 C1 D1 , with diagonals in-
tersecting in O1 , and such that \A1 O1 B1 = v , O1A1 = a1 , O1 B1 = 1b ,
O1C1 = 1c , and O1D1 = d1 . The equalities (1) and (2) imply that the oppo-
site sides of A1 B1 C1 D1 are equal in length, which means that A1 B1 C1 D1
is a parallelogram. So a1 = 1c , and 1b = 1d , implying a = c and b = d. This
proves that ABCD is a parallelogram.
II. Solution by the proposer (slightly edited).
If ABCD is a trapezoid with AB k CD, then
AB OK
CD = OM = 1 ;
which means that ABCD is a parallelogram. Hence, assume that ABCD is
not a trapezoid and set P = AB \ CD, Q = AD \ BC . Indeed, P 6= Q. The
assumption on the given quadrilateral says exactly that PO bisects \BPC
and QO bisects \AQB . Thus
AP = AO = AQ ;
PC OC QC
implying that P , O, and Q lie on the Apollonius circle with centre on the
line AC . Similarly, since
BP = BO = BQ ;
PD OD QD
429
P; O, and Q lie on the Apollonius circle with centre on the line BD. This im-
plies that O is the circumcentre of 4POQ; that is, points P and Q coincide,
a contradiction.
Also solved by CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK;
MICHAEL LAMBROU, University of Crete, Crete, Greece; TOSHIO
SEIMIYA, Kawasaki, Japan; and D.J. SMEENK, Zaltbommel, the Netherlands. There
were also ve incorrect solutions submitted.
Most of the submitted solutions are similar to the proposer's solution.
Janous also generalized the problem by showing that if a > 0, and > > 0
are given real numbers, then the largest constant C = C (a; ; )such that
(a + x) Cx holds for all x > 0 is given by C = a , .
,
The given problem is the special case when a = = 1 and = n + 1.
By applying the AM{GM Inequality to the n + 1 positive numbers:
x x x , Lord and the proposer obtained (1 + x)n+1 (n + 1)n+1 xn ,
1; n ; n ; : : : ; n
which is stronger than the proposed inequality for x > 1.
nn
Zarnowski et al. commented that when n is odd, the inequality is true for all
real x, while if n is even, there is a number xn ,2 such that the inequality holds
for all x xn .
Nlim
!1 1 + N + 2+1+::: ;
:::
where every fraction in this expression has the form
a + bc++::::::
b + ac++::::::
for some cyclic permutation a, b, c of 1, 2, N .
[Proposer's comment: this problem was suggested by Problem 4 of
Round 21 of the International Mathematical Talent Search, Mathematics and
Informatics Quarterly, Vol. 6, No. 2, p. 113.]
Solution by Keith Ekblaw, Walla Walla, Washington, USA.
It will be shown that
2 + N1++ 1 +
p5
1 + N + 1+ ,! 2 as N ! 1:
2+
First, consider
1+ 2+
JN = N + 2 + NN ++ :
1+
Note that
1 + N2++
2 + N1++ > 0:
Thus JN > N and hence JN ! 1 as N ! 1. Now let
2 + N +
2 + JN
KN = 1 + 2 + 1 :
KN = 1 + N + 1+
1+ = 1 +
JN JN KN
2+
431
The proof of part (b) is the same, except that it starts with the vectors
1 ; 1 and 1 ; 1 .
b c c0 b0
Also solved by CLAUDIO ARCONCHER, Jundia, Brazil; WALTHER JANOUS,
Ursulinengymnasium, Innsbruck, Austria; MICHAEL LAMBROU, University of Crete,
Crete, Greece; FLORIAN HERZIG, student, Cambridge, UK; RICHARD I. HESS, Ran-
cho Palos Verdes, California, USA; KEE-WAI LAU, Hong Kong; GERRY LEVERSHA,
St. Paul's School, London, England; VICTOR OXMAN, University of Haifa, Haifa,
Israel; HEINZ-JURGEN SEIFFERT, Berlin, Germany; TOSHIO SEIMIYA, Kawasaki,
Japan; GEORGE TSAPAKIDIS, Agrinio, Greece; JOHN VLACHAKIS, Athens, Greece;
and the proposer.
Most of the submitted solutions are similar to the one given above. Several
solvers pointed out that the restrictions on the sides are unnecessary, and that equal-
ity in (a) occurs if and only if \A = \A0 = 90 and b=c = b0 =c0 and in (b) if and
only if \A = \A0 = 90 and b=c = c0=b0 .
Janous proved more generally that for any real number p 1
1 (bb10)p + (cc10)p ; and 1 (bc10)p + (b01c)p .
(ha h0a)p (ha h0a)p
Xr Ar rD
,
rP rE r
B
r rC Y
Z
r
We denote the circumcircle of 4ABC by ,. Let BX and CY meet
at Z . Since \XBA + \Y CA = 180 , we get \XBA = 180 , \Y CA
= \ACZ , so that A; B; Z; C are concyclic, that is, Z lies on ,. Let D; E
be the second intersections of AX; AY respectively with ,. Since AX and
AY are xed lines, D and E are xed points. Let P be the intersection
of BE and CD. Since hexagon ADCZBE is inscribed in ,, by Pascal's
Theorem the intersections of AD and BZ , of DC and BE , and of CZ and
EA are collinear. Therefore variable line XY always passes through the
xed point P . [Editorial note: if the diagram di ers from the one shown,
for example if Z lies between X and B , the proof still works with minor
changes.]
Also solved by CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK;
FLORIAN HERZIG, student, Cambridge, UK; MICHAEL LAMBROU, University of
Crete, Crete, Greece; MAR I A ASCENSION
LOPEZ
CHAMORRO, I.B. Leopoldo Cano,
434
Valladolid, Spain; D.J. SMEENK, Zaltbommel, the Netherlands; and the proposer.
One incorrect solution and one comment were sent in.
Seimiya and the proposer had similar solutions.
Herzig notes that, if the xed angles are chosen to be AXB and AY C instead,
then the lines XY still pass through a xed point. Readers may like to show this
themselves.
N
K B
Y
D
X
P A
H
L C
Construct a square ABMN outwardly on 4ABC . Let X and Y be
the points of intersection of AK; BC and AB; CM respectively. The ro-
tation through a right angle about B maps 4MBC onto 4ABK . Hence
AK and CM are perpendicular and it follows that AP and CD are alti-
tudes in 4AY C . Therefore P is the orthocentre in that triangle, and as a
consequence Y P ? AC or Y P k BH . Thus
HP = BY = BM = AB :
PD Y D CD CD
435
L C
Let S be a point on AK such that DS ? BC and so DS k AH k BK .
Since AH k DS we get
HP = AH : (1)
PD DS
Since DS k BK and BK = BC , we have
AD = DS = DS : (2)
AB BK BC
Since AH ? BC and CD ? AB we get \HAD = \BCD. Moreover we
have \HDA = \BDC (= 90 ), so that 4HAD 4BCD. Thus
AH = AD : (3)
BC CD
From (2) and (3) we have
AH = AH BC = AD AB = AB ;
DS BC DS CD AD CD
so that we obtain from (1) that
HP = AB :
PD CD
Also solved by FRANCISCO BELLOT ROSADO, I.B. Emilio Ferrari, Valladolid,
Spain; CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; WALTHER JANOUS,
Ursulinengymnasium, Innsbruck, Austria; VACLAV KONECN Y,
Ferris State Univer-
sity, Big Rapids, Michigan, USA; MICHAEL LAMBROU, University of Crete, Crete,
Greece; GERRY LEVERSHA, St. Paul's School, London, England; D.J. SMEENK, Zalt-
bommel, the Netherlands; JOHN VLACHAKIS, Athens, Greece; and the proposer.
The proposer's solution was the same as Herzig's. Most other solvers used
either similar triangles (as in II), trigonometry, or coordinates.
436
A+B x+y
Therefore, we have either (x , y )2 = 0 (which implies that x = y ) or
A + B = x + y. Let us consider A + B = x + y:
A + B = x + y () (A + B)2 = 2(A2 + B2)
() (sA , B)2 = 0 () A = B
() x2 + y2 = pxy
2
() (x , y)2 = 0 () x = y
In conclusion, all solutions have x = y 6= 0, because x + y 6= 0.
Also solved by HAYO AHLBURG, Benidorm, Spain; PAUL BRACKEN,
CRM, Universite de Montreal; CHRISTOPHER J. BRADLEY, Clifton Col-
lege, Bristol, UK; CON AMORE PROBLEM GROUP, Royal Danish School
of Educational Studies, Copenhagen, Denmark (2 solutions); CHARLES R.
DIMINNIE, Angelo State University, San Angelo, TX, USA; DAVID DOSTER,
Choate Rosemary Hall, Wallingford, Connecticut, USA; RUSSELL EULER and
JAWAD SADEK, NW Missouri State University, Maryville, Missouri, USA;
WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria; MICHAEL
LAMBROU, University of Crete, Crete, Greece; KEE-WAI LAU, Hong
Kong; VICTOR OXMAN, University of Haifa, Haifa, Israel; HEINZ-JURGEN
SEIFFERT, Berlin, Germany; D.J. SMEENK, Zaltbommel, the Netherlands; DAVID
R. STONE, Georgia Southern University, Statesboro, Georgia, USA; ROGER
437
ZARNOWSKI, Angelo State University, San Angelo, Texas, USA; and the proposer.
There were 14 incorrect solutions submitted, 11 of which simply did NOT exclude
the origin from the solution set.
P to BC is x = 2 , 4a ;
2 2
P to CA is y = 2 , 4b ;
2 2
P to AB is z = 2 , 4c :
2 2
x2 , y2 b2 , a2
It follows that 2 2 = 2 2 , or
y ,z c ,b
x2(c2 , b2) + y2(a2 , c2) + z2(b2 , a2) = 0 : (1)
Considering x; y;z to be the triangular coordinates of P with respect to
4ABC , we conclude that (1) represents a conic K . Note that K passes
through the incentre I (1; 1; 1) and the excentres Ia (,1; 1; 1); Ib (1; ,1; 1),
and Ic (1; 1; ,1), [and also the circumcentre O(cos A; cos B; cos C )]. So K
is the conic through O of the pencil determined by I; Ia ; Ib ; Ic . Since the
degenerate conics of the pencil are degenerate orthogonal hyperbolas (that
is, pairs of perpendicular lines), K must be an orthogonal hyperbola.
Also solved by CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK.
Neither solver mentioned the obvious special cases:
The locus is a pair of perpendicular lines when 4ABC is isosceles, and just
the four points I; Ia ; Ib ; Ic when equilateral.
Bradley points out that it is clear from the statement of the problem (with no
need for coordinates) that the locus includes I; Ia ; Ib ; Ic (when = 0) and O (when
= 1).
438
On the other hand, note that the right inequality in (3) is equivalent to
a, n + x n + 1 2
<
n + x n + 1 2
, kn: (4)
2 2
Since kn >
n + xn 2, we have
2
kn + 1 n +2 xn + 1 n + x2n + 1 :
jpk
a u+1=
[Ed: The last inequality holds since n + xn is an integer.]
Hence it suces to establish (4) for a = n.
Substituting a = n into the right inequality of (3), we need to show that
kn < nxn + n. From n , 2pn , 1 < xn < n we get (n, xn )2 < 4(n , 1)
or (n + xn )2 + 4 < 4(nxn + n). Hence kn
n + xn 2 + 1 < nx + n.
2 n
Therefore (3) holds, and by setting a = u + 1; u + 2; ; n we get
kn < xn + 1 < kn < < xn + n , u , 1 < kn < xn + n , u < kn :
n n,1 u+1 u
440
Hence
kn - kn - kn - kn : (5)
n n,1 u+1 u
From (2) and (5) we conclude that kn satis es (1).
if k is any integer satisfying (1), then k kn . To
Now we show that
k
this end, let x =
n .We rst show that x xn. Since there exists an
k
integer z such that < z <
k and x = k < z, we have z , x 1
n n,1 n
k k k
and hence
n , 1 > x + 1. Similarly, n , 2 > x + 2, n , 3 > x + 3, ,
k > x + n , 1. That is, for all a = 1; 2; ; n we have k > (n , a)(x + a)
1
or k (n , a)(x + a) + 1 = ,a2 + a(n , x) + nx + 1: (6)
Hence
k n ,2 x + nx + 1 , a , n ,2 x :
2 2
(7)
k
Note that x 1. [Ed: If k n , 1, then 0 < <
k 1, contradicting
n n,1
k - k . Hence k n].
n n,1
On the other hand, it is clear that xn n , 1. Suppose, contrary to
what we claim, that x < xn . Then we have 1 x < xn n , 1 and so
2 n , x < n or 1 n ,2 x < n2 . (Here we must assume that n 3.
The case when n = 2 can be treated
separately, since it is easy to verify that
n , x
k2 = 3.) Hence we may let a = 2 in (6) and (7) and obtain
n
k 2 ,
x 2
+ nx + 1 ,
n , x
n , x 2
2 , 2 : (8)
Since the right side of (7) is an integer and since the last squared term in (8)
1
is either 0 or we get
4
$ % $ %
k n , x 2 + nx + 1 = n + x 2 + 1 : (9)
2 2
$ $ %% $ %
Thus x =
k 1
n + x 2 + 1 = 1 n + x 2 + 1 .
n n 2 n 2 n
441
bz c
z
[Ed: It is known and easy to show that
n = n for all real
numbers z and positive integers n.]
$ % $ %
Hence 0
1 n + x 2
1 1 n , x 2
1
n 2 !+ n , x = n 2 + n , which
implies that
1 n , x 2 + 1 1 or n , x 2 n , 1.
n 2 2
p p
Thus n , x < 2 n , 1 or x > n , 2 n , 1, from which we get x
bn +1 , 2pn , 1c = xn , a contradiction. Hence x xn . Therefore we may
$ x by
replace xn % in (6), (7), and (9) and conclude that
k n + x 2
n + 1 = k . This completes the proof.
2 n
Also solved by PETER TINGLEY, student, University of Waterloo, Waterloo,
Ontario. There was one incorrect solution. Tingley gave the answer
$ pn , 1 + 1c 2 %
p
kn = nbn , 2 n , 1 + 1c + n , bn , 2 +1
2
which is readily seen to be the same as the one obtained by Herzig. The proposer
had conjectured that
kn = 11 + (n , m)2 if m2 n , 2
+ (n , m)2 + (n , m) otherwise
p
where m = 1 + 24n , 7 and had veri ed it for 2 n 600 using a computer.
In a private communication Tingley has actually proved that this conjectured formula
is equivalent to the answer given by Herzig and himself. Interested readers may nd
the proof of this fact quite challenging.
n Y
m
(k + j ) = (m + 1)! m n+,n 1+ 1
X
k=1 j =0
= (m(m++2)(nn+,1)!1)! = n(n + 1) :m: :+(n2+ m + 1) :
B. The expressions reduce to sums of the form nk=1 km . With the help
P
of a calculator I got
!
n + 2 2(3n2 + 6n + 1) for m = 1;
3 5
!
n + 3 12(2n + 3)(5n2 + 15n + 1) for m = 2;
4 35
!
n + 4 8(35n4 + 280n3 + 685n2 + 500n + 12) for m = 3;
5 21
!
n + 5 40(126n5 + 1575n4 + 6860n3 + 12075n2 + 7024n + 60) for m = 4:
6 77
j =0 j =0 j =0
(easily veri ed by considering the common factor m j =0 (k + j ) of the two
Q
k=1 j =0 j =0 j =0 j =0
[Editorial note: Lambrou also solved parts B and C.]
All three parts also solved by G. P. HENDERSON, Garden Hill, Ontario. Parts A
and B only solved by THEODORE N. CHRONIS, Athens, Greece; WALTHER JANOUS,
Ursulinengymnasium, Innsbruck, Austria; and the proposer.
As Herzig mentions above, part A at least is a fairly familiar result. (For ex-
ample, see formula 2.50, page 50 of Concrete Mathematics by Graham, Knuth and
Patashnik.) In fact it was proposed for publication in CRUX in 1991 by an Edmonton
high school student, Jason Colwell, but was not accepted by the then editor.
Chronis notes that, in the solution for part B, when m = 4, the fth degree
polynomial has 2n + 5 as a factor.
445
where the square brackets [ ] and the parentheses ( ) denote the least com-
mon multiple and greatest common divisor respectively.
For what values of n does the identity un = (n , 1)un,1 hold?
Solution by Florian Herzig, student, Cambridge, UK.
We rst introduce some notation:
for any prime p, let (n)p = maxf jp ng, [n]p = maxf jp divides ng.
Thus p[n]p jjn. For each k = 1; 2; ; n we determine a = a(k) such that
pajjk. If a [n]p, then since pajn, we have pajj(k; n) and so p 6 j k . (k; n)
If a > [n]p , then since p[n]p jk we have p[n]p jj(k; n) and so pa,[n]p jj
k .
(k; n)
Thus the highest power of p in any
k
Y (k;n) arises when a = (n)p. This shows
that un = p(n)p ,[n]p , where the product is over all primes. Note that
Y
p
n,1= p[n,1]p and hence un = (n , 1)un,1 is equivalent to
p
(n)p , [n]p = (n , 1)p for all primes p (1)
We distinguish two cases:
Case (i) Suppose n is a prime power, say n = q b where q is a prime and
b > 0. For p 6= q, (1)b is satis ed since (n)p = (n , 1)p and [n]p = 0. For
p = q we have n = p and so (n)p = [n]p = b and (n , 1)p = b , 1. Hence
(1) holds if and only if b , 1 = 0; that is, b = 1.
Case (ii) If n is not a prime power, then (n)p = (n , 1)p for all primes
p. Hence (i) holds if and only if [n]p = 0 for all primes p, and so n = 1.
[Ed: Clearly n = 1 is not a solution, since u0 is unde ned.]
Therefore un = (n , 1)un,1 if and only if n is a prime.
447
2278. [1997: 431] Proposed by Joaqun Gomez Rey, IES Luis Bu~nuel,
Alcorcon, Madrid, Spain.
Determine the value of an , which is the number of ordered n{tuples
(k2 ; k3; : : : ; kn ; kn+1) of non-negative integers such that
2k2 + 3k3 + : : : + nkn + (n + 1)kn+1 = n + 1:
I. Solution by Michael Lambrou, University of Crete, Crete, Greece.
We show that an = p(n + 1) , p(n) for n 1 where p(m) denotes
the number of partitions of m into positive integral parts. Our argument is
based on the well-known observation that to a partition of m where lk k's
appear (k = 1; 2; : : : ; m), so that
1l1 + 2l2 + + mlm = m ; (1)
corresponds the ordered m-tuple (l1 ; l2 ; : : : ; lm ). Conversely, to any given
ordered m-tuple (l1 ; l2 ; : : : ; lm ) of positive integers satisfying (1), there cor-
responds a partition of m.
For xed n 1 consider the partitions of n + 1 as above. They are of
two types:
(a) those for which the number 1 is absent in the decomposition; or
(b) those for which the number 1 appears at least once.
The number of partitions of type (a) is clearly an . Moreover, for each
partition in case (b), if we delete one 1, we get a partition of n. Conversely,
every partition of n with an extra 1 added on gives a partition of n + 1 of
type (b). Clearly then p(n + 1) = an + p(n), as required.
II. Solution by Walther Janous, Ursulinengymnasium, Innsbruck, Aus-
tria.
We deal more generally with the equation:
k1 + 2k2 + + (j , 1)kj,1 + (j + 1)kj+1 + + (n + 1)kn+1 = n + 1
where j 2 f1; 2; : : : ; n + 1g is xed and determine the number an (j ) of its
non-negative solutions in Zn .
448
For this we recall that (since Euler's days) such problems are dealt with
best by generating functions, namely:
E(x) = (1 + x1 + x21 + : : : ) (1 + x2 + x22 + : : : ) : : :
1 1 1 1
X
= 1 , x 1 , x2 1 , x3 : : : = p(k)xk
k=0
where p(k) denotes the number of partitions of k; that is, the number of
unordered representations of k as k = s1 + s2 + + se with sj a positive
integer for j = 1; 2; : : : ; e, or equivalently k = 1n1 +2n2 + + knk , where
nj 0 is the number of appearances of summand j . Therefore, all partitions
with summand j forbidden are obtained via
1
X ,1
jX 1,
X
(1,xj )E (x) = (1,xj ) p(k)xk = p(k)xk+ p(k),p(k,j )xk :
k=0 k=0 k=j
Hence the desired amount an (j ) equals:
if n + 1 j , 1
an (j ) = p(n + 1);
p(n + 1) , p(n + 1 , j ); if n + 1 j
Also solved by HEINZ-JURGEN SEIFFERT, Berlin, Germany. There was one
incorrect solution submitted.
Janous remarks how his ideas above can be extended to include the case where
the excluded summand can be a subset of the values from 1 to n + 1.
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