Crux1991 All PDF
Crux1991 All PDF
Mathematicorum
Volume 17 , Number 1 January 1991
CONTENTS
1
1
are drawn. Determine, in terms of n, the total number of isosceles right-angled triangles
of all sizes.
1981 CELEBRATION OF CHINESE NEW YEAR CONTEST
January 26, 1981
1. What is the coecient of x when 2
( (((x , 2) , 2) , 2) , , 2)
2 2 2 2
f (x ) is divided by f (x).
100
'$'$ '$
'$
contact of C and C . If the area of either part is less than that of the other then move C
2 3 1
(or C ) so as to increase the area of that part. So we may assume the symmetry of either
C1
&%
&%
&%&%
S
S
S
C2
S
Z
Z
Z
Z
S
S
S
S
C3
C4
&%
&%
ZZ
C1
C2
&% &%
ZZ
S
S
S
S
C3
S
S
S
S
C4
Figure 1 Figure 2
3
In each gure, the area of the convex hull is made up of sectors of each of the 4
circles (total area ), 4 rectangles and either a parallelogram (Figure 1) or an equilateral
trapezium (Figure 2). So the area is either
+ 2 + 2 + 4 cos + 4 cos + (4 cos )(2 sin ) = + 4 + 8 cos (1 + sin )
or
p
So = =4 in Figure 2 yields the maximum area + 10 + 4 2:
Similarly, for the minimum we must look at + 4 + 8 cos (1 + sin p ); where 0
=3: Of course, = 0 gives + 12. But = =3 gives + 8 + 2 3 which is less.
After dierentiating, , sin , sinp + cos = 0, so 2 sin + sin = 1 and sin = 1=2:
2 2 2
For = =6, we have + 4 + 6 3 which is a local maximum. Thus the minimum area
occurs when = =3 in Figure 1.
4. [1984: 107] 1983 Chinese Mathematics Olympiad.
Determine the maximum volume of a tetrahedron whose six edges have lengths
2; 3; 3; 4; 5; and 5.
Solution by Richard K. Guy, University of Calgary.
A triangle of sides 2; 3; 5 is degenerate; moreover there is no point at distances 3; 4; 5
from the vertices of such a triangle. The only possible triangular faces containing edge 2
are thus 233; 234; 245; 255.
The 2 triangular faces with opposite and the other 2
which share edge 2 are edge triangular faces are
233 & 245 5 345 & 355
233 & 255 4 345 & 345
234 & 255 3 335 & 345
Therefore the only possible tetrahedra are those shown in the following gures.
4
can be unbounded (i.e., for every number M , there exists an xj such that xj > M ):
5
q
principle. Without loss of generality we can assume
that they are A and B . Then the p point E is covered 7
G is covered
p by the third disk since EG = 65=8 and K
GB > 65=8. This means that the point p F is covered E H F
by the second disk, since FA = FG = 65=8. Then the
point K , where FKp= 1=8; is covered by the third disk, A
pB
since KA > KE = 65=8. This means D is uncovered, since DB > DF > DK p > 65=8.
This contradicts our assumption that the three disks cover the square. So 65=8 is the
least possible radius.
2. [1984: 215] British Mathematical Olympiad.
For 0 r n, let an be the number of binomial coecients nr which leave
remainder 1 on division by 3, and let bn be the number which leave remainder 2. Prove
that an > bn for all positive integers n.
6
(1 + x3 )n (1 + x3 ,1 )n ,1 (1 + x3)n1 (1 + x)n0 :
k
k
k
k
For 0 i k, set
Fi (x) = (1 + x3 )n (1 + x3 ,1 )n ,1 (1 + x3)n1 (1 + x)n0 :
i
i
i
i
We claim that Fi (x) is satisfactory for 0 i k. This is+1certainly true for i = 0. Suppose
it holds for some i < k, and consider Fi+1(x) = (1 + x3 )n +1 Fi (x): i
i
If ni+1 = 0, we have Fi+1(x) = Fi (x), and the result follows from the induction
hypothesis.
If ni+1 = 1, then Fi+1+1(x) = Fi (x) + x3 +1 Fi (x). By the induction hypothesis Fi (x)
i
satisfactory.
If n = 2, then Fi+1(x) = Fi (x) + 2x3 +1 Fi (x) + +1 i
x23 +1 Fi (x). Again there are no i
like terms. Moreover, the numbers of coecients of 2x3 Fi (x) congruent to 1 and 2 are i
for all n 2 N.
(b) If b, ab = ak where k 2 N, then b, anbl = ank for all l; n 2 N.
1 1 l
= bn(bancn )cn n = ac = bn an c n n = :
( +1) 2 +1 +1 ( +1)( +2) 2
7
(b) This gave trouble, because it is misprinted (but true for l = 1 nevertheless).
Note
b, an b = (b, ab)n = akn :
1 1
However, assume inductively that b,lanbl = ank (true for l = 1 and all n). Then
l
M856. [1984: 283] Problems from KVANT.
(a) Construct a quadrilateral knowing the lengths of its sides and that of the segment
joining the midpoints of the diagonals.
(b) Under what conditions does the problem have a solution?
Solution by Richard K. Guy, University of Calgary.
Suppose rst that we are given the lengths of the sides 2a; 2b; 2c; 2d, in that cyclic
order, and that m is the length of the join of the midpoints of the diagonals. (Note that
m = 0 implies that the quadrilateral is a parallelogram and a = c; b = d are necessary;
this will be a special case of the general condition.) ,B
(a) The construction is easily discovered from a, BB b
,
the theorem that the midpoints of the sides of any , B
quadrilateral form a parallelogram. Draw a segment B, B
MN of length m. Construct two triangles with sides ,@B BA
a; c; m forming a parallelogram MANC , say. Similarly a , B@ ,B
This nishes the \archive" material we have for 1984. We now turn to problems
posed in the March 1989 number of the Corner.
3. [1989: 65] 1987 Annual Greek High School Competition.
Let A be an n n matrix such that A , 3A +2I = 0, where I is the identity matrix
2
and 0 the zero matrix. Prove that A k , (2k + 1)Ak + 2k I = 0 for every natural number
2
k 1.
Solution by Seung-Jin Bang, Seoul, Republic of Korea.
Since Am n = AmAn for all positive inters m; n, we see that if f (x) = p(x)q(x)
+
3. [1989: 66] 24th Spanish Olympiad-First Round.
Let C be the set of natural numbers
C = f1; 5; 9; 13; 17; 21; : : :g:
Say that a number is prime for C if it cannot be written as a product of smaller numbers
from C .
(a) Show that 4389 is a member of C which can be represented in at least two
distinct ways as a product of two numbers prime for C .
(b) Find another member of C with the same property.
Solution by Edward T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario.
Note rst that C is closed under multiplication since it consists of all positive
numbers of the form 4a + 1.
(a) Since 4389 = 3 7 11 19 is a product of four distinct primes of the form
4b + 3, the product of any two of these primes will have the form 4c + 1 which clearly
cannot be written as a product of smaller numbers from C . Hence 4389 can be written in
at least three dierent ways as a product of two members of C prime for C :
4389 = 21 209 = 33 133 = 57 77:
In fact, since it is obvious that the product of any three numbers from 3; 7; 11; 19 is prime
for C , there are four other ways of expressing 4389 in the described manner:
4389 = 3 1463 = 7 627 = 11 399 = 19 231:
(b) By the argument above, to obtain another member of C with the same property
it suces to replace 19 by 23, the next true prime of the form 4k + 3 and obtain 3 7
11 23 = 5313.
9
, ,
a circle. Thus the graph is as shown. @2 C,
(b) The area of 4AOB and p of 4 BCD is
2, the area of sector ABC is (2 2) = 2, and
1 2
(b) As n goes to innity the point of intersection of the line y = x and the graph
of y = tan x in the interval In goes o to innity in each coordinate. Thus the dierence
cn , n must go to =2.
[Editor's note. This solution was adapted from the rather more detailed one sub-
mitted by Seung-Jin Bang, Seoul, Republic of Korea.]
1. [1989: 67] 24th Spanish Olympiad.
Fifteen problems, numbered 1 through 15, are posed on a certain examination. No
student answers two consecutive problems correctly. If 1600 candidates sit the test, must
at least two of them answer each question in the same way?
Solutions by John Morvay, Springeld, Missouri, and by Edward T.H. Wang, Wil-
frid Laurier University, Waterloo, Ontario.
The answer is yes, assuming each question has only two possible answers, right
and wrong, and assuming no student leaves any question unanswered. First note that the
answer pattern of each student corresponds to a sequence of length 15, consisting of the
two symbols R and W in which there are no two consecutive R's. Clearly the maximum
number of R's is 8. For each 0 k 8, the number of those sequences with exactly k
R's is easily seen to be ,kk = k,k since each such sequence is equivalent to a way
15 +1 16
10
k k =0
Since 1; i are the roots we have that z ; z ; z are equal to 1; i; ,i in some order.
1 2 3
* * * * *
This completes the solutions submitted for problems from the March 1989 number
and this is all the space we have this month. Send me your contests!
11
MINI-REVIEWS
by ANDY LIU
BOOKS FROM DOVER PUBLICATIONS, INC.
The majority of Dover's publications are reprints of excellent (otherwise, why do
it?) books that are no longer available in other formats. The new editions are usually pa-
perbound and inexpensive (averaging about $5.00 U.S. each). Often, errors in the original
versions are corrected, and new material appended. While Dover has a large selection of
titles in main-stream mathematics (as well as in many other areas, academic or otherwise),
we will focus on the best of its line on popular mathematics. All are paperbacks.
Challenging Mathematical Problems with Elementary Solutions I, by A.M. Yaglom
and I.M. Yaglom, 1987. (231 pp.)
This is one of the nest collections of problems in elementary mathematics. The
100 problems in combinatorial analysis and probability theory are all easy to understand,
but some are not easy to solve, even though no advanced mathematics is required.
Challenging Mathematical Problems with Elementary Solutions II, by A.M. Yaglom
and I.M. Yaglom, 1987. (214 pp.)
In this second volume, 74 problems are selected from various branches of mathemat-
ics, in particular, number theory and combinatorial geometry. This book and the earlier
volume is a must for every school library.
Mathematical Baers, edited by A.F. Dunn, 1980. (337 pp.)
The baers in this book originally appeared as a most successful weekly corporate
advertisement in technical publications. They are contributed by the readers, with a
consequent diversity in their levels of sophistication. Some require almost no mathematics
while others are quite demanding. However, there is a beautiful idea behind each baer,
which is compactly stated and accompanied by a cartoon.
Second Book of Mathematical Baers, edited by A.F. Dunn, 1983. (186 pp.)
This second collection of baers, like the earlier volume, is organized by chapters,
each dealing with one area of mathematics. These include algebra, geometry, Diophantine
problems and other number theory problems, logic, probability and \insight".
Ingenious Mathematical Problems and Methods, by L.A. Graham, 1959. (237 pp.)
The 100 problems in this book originally appeared in the \Graham Dial", a pub-
lication circulated among engineers and production executives. They are selected from
areas not commonly included in school curricula, and have new and unusual twists that
call for ingenious solutions.
The Surprise Attack in Mathematical Problems, by L.A. Graham, 1968. (125 pp.)
These 52 problems are selected from the \Graham Dial" on the criterion that the
best solutions are not the ones the original contributors had in mind. The reader will
enjoy the elegance of the unexpected approach. Like the earlier volume, the book includes
a number of illustrated Mathematical Nursery Rhymes.
12
Mathematical Recreations and Essays, by W.R. Ball and H.S.M. Coxeter, 1988.
(418 pp.)
This is the foremost single-volume classic of popular mathematics. Written by two
distinguished mathematicians, it covers a variety of topics in great detail. After arithmeti-
cal and geometrical recreations, it moves on to polyhedra, chessboard recreations, magic
squares, map-colouring problems, unicursal problems, Kirkman's schoolgirls problem, the
three classical geometric construction problems, calculating prodigies, cryptography and
cryptanalysis.
Mathematical Recreations, by M. Kraitchik, 1953. (330 pp.)
This is a revision of the author's original work in French. It covers more or less the
same topics as \Mathematical Recreations and Essays". There is a chapter on ancient and
curious problems from various sources.
The Master Book of Mathematical Recreation, by F. Schuh, 1968. (430 pp.)
This is a translation of the author's original work in German. Four of the fteen
chapters are devoted to the analysis of mathematical games. The remaining ones deal with
puzzles of various kinds. General hints for solving puzzles are given in the introductory
chapter. The last chapter is on puzzles in mechanics.
Puzzles and Paradoxes, by T.H. O'Beirne, 1984. (238 pp.)
Like Martin Gardner's series, this book is an anthology of the author's column in
New Scientist. It consists of twelve largely independent articles.
* * * * *
PROBLEMS
Problem proposals and solutions should be sent to B. Sands, Department of Mathe-
matics and Statistics, University of Calgary, Calgary, Alberta, Canada T2N 1N4. Propos-
als should, whenever possible, be accompanied by a solution, references, and other insights
which are likely to be of help to the editor. An asterisk (*) after a number indicates a
problem submitted without a solution.
Original problems are particularly sought. But other interesting problems may also
be acceptable provided they are not too well known and references are given as to their
provenance. Ordinarily, if the originator of a problem can be located, it should not be
submitted by somebody else without permission.
To facilitate their consideration, your solutions, typewritten or neatly handwritten
on signed, separate sheets, should preferably be mailed to the editor before August 1,
1991, although solutions received after that date will also be considered until the time when
a solution is published.
1601. Proposed by Toshio Seimiya, Kawasaki, Japan.
ABC is a right-angled triangle with the right angle at A. Let D be the foot of the
perpendicular from A to BC , and let E and F be the intersections of the bisector of 6 B
with AD and AC respectively. Prove that DC > 2EF .
14
Xn
xi m + r : 2 2
i=1
1603. Proposed by Cliord Gardner, Austin, Texas, and Jack Garfunkel, Flushing,
N.Y.
Given is a sequence , ; , ; : : : of con- 1 2
centric circles of increasing and unbounded
radii and a ,! triangle A B C inscribed in
,! ,! 1 1 1
k j k j =0
15
is 1, 9, or prime.
1609. Proposed by John G. Heuver, Grande Prairie Composite H.S., Grande
Prairie, Alberta.
P is a point in the interior of a tetrahedron ABCD of volume V , and Fa ; Fb ; Fc ; Fd
are the areas of the faces opposite vertices A; B; C; D, respectively. Prove that
PA Fa + PB Fb + PC Fc + PD Fd 9V:
1610. Proposed by P. Penning, Delft, The Netherlands.
Consider the multiplication d dd ddd, where d < b , 1 is a nonzero digit in base
b, and the product (base b) has six digits, all less than b , 1 as well. Suppose that, when
d and the digits of the product are all increased by 1, the multiplication is still true. Find
the lowest base b in which this can happen.
* * * * *
SOLUTIONS
No problem is ever permanently closed. The editor will always be pleased to consider
for publication new solutions or new insights on past problems.
689. [1981: 276; 1982: 307; 1983: 144] Proposed by Jack Garfunkel, Flushing,
N.Y.
Let ma ; mb ; mc denote the lengths of the medians to sides a; b; c; respectively, of
triangle ABC , and let Ma ; Mb ; Mc denote the lengths of these medians extended to the
circumcircle of the triangle. Prove that
Ma + Mb + Mc 4:
ma mb mc
IV.Generalization by Dragoljub M. Milosevic, Pranjani, Yugoslavia.
In his solution [1982: 308{309], M.S. Klamkin showed that the problem is equivalent
to
X a
2
2(b + c ) , a 1
2 2 2
where the sum is cyclic over a; b; c: He also proved the related result
X a 3
k(b + c) , a 2k , 1
16
where k 1, and suggested the more general problem of nding all k 2 such that
X a 2
3 ;
k(b + c ) , a 2k , 1
2 2 2
for 2 and also for = 1, function f is convex, so for 0 < xi < p=q
X
3
xi = X f (x ) 3f x + x + x = 3 ,(x + x + x ) :
3
1 2 3
2
1 2 3
(2)
i
i=1 p , qxi i =1
3 3p , q(x + x + x ) 1 2 3
Putting in (2)
p = (an + bn + cn)k ; q = k + 1 ; x = an; x = bn; x = cn; 1 2 3
we obtain the desired inequality (1). Note that, for example, x < p=q is equivalent to
1
an < k(bn + cn );
so since k 2n, it is enough to prove that
1
1411. [1989: 47; 1990: 92] Proposed by D.J. Smeenk, Zaltbommel, The Nether-
lands.
ABC is acute angled with sides a; b; c and has circumcircle , with centre O. The
inner bisector of 6 A intersects , for the second time in A . D is the projection on AB of
1
A . L and M are the midpoints of CA and AB respectively. Show that
1
(i) AD = (b + c); 1
2
(ii) A D = OM + OL.
1
By using the relation (ii) (for acute triangles) we have an alternate proof of the well
known theorem: if ABC is an acute triangle with circumcentre O, circumradius R, and
inradius r, and L; M; N are the midpoints of the sides, then
OL + OM + ON = R + r:
(N.A. Court, College Geometry, p. 73, Thm. 114). [Editor's note: this theorem was also
used by Seimiya in his proof of Crux 1488, this issue.]
Let I be the incenter of ABC , and let
T; S be the feet of the perpendiculars from I to
AD; A D respectively. Then we get SD = IT = r.
1
It is well known that A I = A C , and because 1 1
6 A IS = 6 A AB = 6 A CB
1 1 1
and
ISA = 6 CNA = 90;
6 1 1
OL + OM + ON = A D + ON = A S + SD + ON 1 1
= A N + r + ON = R + r: 1
18
Another solution was received from K.R.S. SASTRY, Addis Ababa, Ethiopia, in
which he also shows that converses of (i) and (ii) need not hold, and nds similar formulae
in the case that AA1 is the external bisector of 6 A.
* * * * *
1432. [1989: 110; 1990: 180] Proposed by J.T. Groenman, Arnhem, The Nether-
lands.
If the Nagel point of a triangle lies on the incircle, prove that the sum of two of the
sides of the triangle equals three times the third side.
IV. Comment and solution by Dan Sokolowsky, Williamsburg, Virginia.
This is in response to L.J. Hut's claim [1990: 182] that, if D is the point at which
the incircle of ABC touches AB , then the incircles of triangles ABC and A0 B 0 C 0 touch
at D. The claim is true provided that some additional assumption (such as (2) below) is
made.
First, it is easy to see (referring to the gure on [1990: 181]) that this claim is
equivalent to
NI ? AB at D: (1)
We show, assuming that the Nagel point of ABC lies on its incircle, that (1) is equivalent
to: AB is the shortest side of ABC , that is, to
c a; b (2)
(which, incidentally, would justify Hut's selection of AB over the other edges of ABC ).
In the gure, let w denote the incircle of
ABC , touching AB at D, CA at Z , CB at Y .
Let P; Q be points on CA; CB respectively such
that PQ k AB and PQ touches w, say at F . Ob-
viously FI ? PQ, hence FI ? AB . Thus to show
(2) implies (1) it will suce to show that (2) im-
plies F is the Nagel point N of ABC .
Let CF meet w again at F 0 and AB at
V . Note that w is the excircle of CPQ on side
PQ, hence the Nagel point of CPQ lies on CF .
Then, since ABC CPQ, the Nagel point N
of ABC lies on CV . By hypothesis, it also lies
on w, hence it is either F or F 0 . We can assume
that F 0 lies on the arc DZ . Then BF 0 meets AZ
at a point X . If F 0 were the Nagel point of ABC
we would then have
s , c = AX < AZ = s , a
(s the semiperimeter), which implies a < c, contradicting (2). It follows that F 0 cannot be
the Nagel point of ABC , which must then be F , so (1) follows. Conversely, if (1) holds,
19
then by the preceding argument NI is perpendicular to the shortest side of ABC , which
must therefore be AB , so (2) holds.
A simple proof of the problem could now
go as follows. Parts marked in the adjoining gure
have the same meaning as before, but we assume
(2), so that by the above the point marked N is
the Nagel point of ABC . Let AN meet BC at
T , and let s0 be the semiperimeter of CPQ. We
then have
2s0 = (CP + PN ) + (NQ + QC )
= CZ + CY = 2(s , c):
Since CPQ CAB ,
CN = s0 = s , c
CV s s
and hence
CN = s , c :
NV c
Since N is the Nagel point of ABC ,
AV = s , b ; BT = s , c ; TC = s , b:
By Menelaus applied to CV B ,
1 = CN AV BT = s , c s , b s , c = (s , c) :
2
NV AB TC c c s,b 2
c
Thus s , c = c, which implies a + b = 3c.
Sokolowsky also pointed out a typo on [1990: 182], line 2: DG = 2GD0 should read
2DG = GD0 .
L.J. Hut submitted a further clarication of his solution, showing not only that
IN ? AB implies a + b = 3c, but also that IN k AB implies a + b = 2c. Interesting! Any
comments?
* * * * *
1481. Proposed by J.T. Groenman, Arnhem, The Netherlands, and D.J. Smeenk,
Zaltbommel, The Netherlands.
Let A; B; C be points on a xed circle with B; C xed and A variable. Points D
and E are on segments BA and CA, respectively, so that BD = m and CE = n where m
and n are constants. Points P and Q are on BC and DE , respectively, so that
BP : PC = DQ : QE = k;
20
also a constant. Prove that the length of PQ is a constant. (This is not a new problem.
A reference will be given when the solution is published.)
Solution by Hans Engelhaupt, Franz-Ludwig-Gymnasium, Bamberg, Germany.
Draw the lines XB and Y P paral-
lel to CE and of the same length, so that
X; Y; E are collinear and XE kBC . Also nd
Q on DE so that Y QkXD. Then
DQ = XY = BP = k; (1)
QE Y E PC
so Q is as dened in the problem. We get
the xed triangle BDX with sides BD = m
and BX = CE = n and included angle
(since A lies on a xed circle). Thus for every
position of A the side XD is xed. From (1)
the side QY is xed. Since also PY = CE =
n and 6 PY Q = 6 BXD is xed, side PQ is
constant.
Also solved by FRANCISCO BELLOT ROSADO, I.B. Emilio Ferrari, and MARIA
ASCENSION LOPEZ
CHAMORRO, I.B. Leopoldo Cano, Valladolid, Spain; JORDI DOU,
Barcelona, Spain; C. FESTRAETS-HAMOIR, Brussels, Belgium; RICHARD I. HESS,
Rancho Palos Verdes, California; L.J. HUT, Groningen, The Netherlands; WALTHER
JANOUS, Ursulinengymnasium, Innsbruck, Austria; and the proposers.
Although nobody mentioned it, it appears that point A must be restricted to one of
the arcs BC of the circle.
The problem was found by the proposers in Journal de Mathematiques E lementaires
(1912).
* * * * *
1482. Proposed by M.S. Klamkin, University of Alberta.
If A; B; C are vectors such that
jAj = jBj = jCj = jA + B + Cj;
prove that
jB Cj = jA (B + C)j:
21
= 3 + 2A (B + C) + 2B C;
so
A (B + C) = ,(B C + 1):
It follows from
(u v) = u v , (u v)
2 2 2 2
(twice) that
jA (B + C)j = A (B + C) , (A (B + C))
2 2 2 2
= (B + C) , (B C + 1) 2 2
= 1 , (B C) = jB Cj ; 2 2
where [ABC] = A (B C). Then it is known (e.g., Spiegel, Vector Analysis, Schaum,
1959, Chapter 2, exercises 53(c) and 103) that, reciprocally,
Z ; B = ZX ; C = XY ;
A = [YXYZ] [XYZ] [XYZ]
and
[XYZ] [ABC] = 1:
Substituting for A; B; C in the original problem, one gets the dual problem: if X; Y; Z are
vectors such that
jY Zj = jZ Xj = jX Yj = j(Y Z) + (Z X) + (X Y)j;
then (by symmetry)
jXj = jY , Zj ; jYj = jZ , Xj ; jZj = jX , Yj :
,! ,! ,!
Now consider a tetrahedron PXY Z where PX = X; PY = Y; PZ = Z. Then the above
shows that if the four faces of a tetrahedron have equal areas, the tetrahedron must be
isosceles, i.e., opposite pairs of edges are congruent. For a geometric proof, see N. Altshiller
Court, Modern Pure Solid Geometry, Macmillan, N.Y., 1935, Corollary 307.
* * * * *
14830. Proposed by George Tsintsifas, Thessaloniki, Greece.
Let A B C be a triangle inscribed in a triangle ABC , so that A0 2 BC; B 0 2
0 0
CA; C 0 2 AB , and so that A0B 0 C 0 and ABC are directly similar.
(a) Show that, if the centroids G; G0 of the triangles coincide, then either the trian-
gles are equilateral or A0 ; B 0 ; C 0 are the midpoints of the sides of ABC:
(b) Show that if either the circumcenters O; O0 or the incenters I; I 0 of the triangles
coincide, then the triangles are equilateral.
Solution by Jordi Dou, Barcelona, Spain.
When ABC is equilateral the triangles A0 B 0 C 0 are concentric with ABC and all of
G ; O ; I 0 coincide with O = G = I . Hereafter we suppose that ABC is not equilateral.
0 0
Let T 0 be the triangle whose vertices are the midpoints A0 ; B 0 ; C 0 of BC; CA; AB ,
0 0 0 0
respectively. The perpendicular bisectors of the sides of ABC (through A0 ; B 0 ; C 0 ) concur
0 0 0
at the circumcentre O of ABC . Any triangle T 0 = A0B 0 C 0 as described in the problem
can be obtained by applying a rotation of centre O and angle x, followed by a homothety
of centre O and ratio r = 1= cos x, to triangle T 0 . [Editor's note. Can someone supply a
0
reference?] The vertices A0x ; Bx0 ; Cx0 of the resulting triangle Tx0 will be on the sides of ABC .
Since T 0 is similar to ABC , Tx0 will also be similar to ABC for every x. Note that O is the
0
orthocentre H 0 of T 0 and thus also the orthocentre Hx0 of every Tx0 . Let G0x be the centroid
of Tx0 . From G0x O = (1= cos x)G0 O and 6 G0 OG0x = x it follows that G0x G0 ? G0 O. Thus
0 0
0 0 0 0
the locus of the centroids G0x is the line through G0 perpendicular to OG0 . Analogously
0 0
23
the locus of the circumcentres Ox0 and of the incentres Ix0 of the triangles Tx0 are the lines
through O0 and I 0 perpendicular to OO0 and OI 0 respectively.
Since G = G0 , G = G0x only for x = 0; this solves (a). Since O = H 0 6= O0 unless
0 0 0 0
0 0 0
T 0 (i.e., ABC ) is equilateral, O cannot coincide with Ox0 ; this solves the rst part of (b).
0
Finally, if I = Ix0 then, since G = G0 is on the line II 0 , 6 HIG = 6 H 0 I 0 G0 = 90.
0 0 0 0 0
Editor's note. Dou claimed not to have nished the proof of the impossibility of
I = Ix , but it now follows immediately from the known result that GH GI + IH , with
0 2 2 2
equality only for the equilateral triangle. See p. 288 of Mitrinovic, Pecaric, and Volenec,
Recent Advances in Geometric Inequalities, or the solution of Crux 260 [1978: 58].
Also solved by the proposer. Part (a) only was solved by JILL HOUGHTON, Syd-
ney, Australia.
For part (a), the proposer simply applied his two earlier problems Crux 1464 [1990:
282] and Crux 1455 [1990: 249].
* * * * *
1484. Proposed by Walther Janous, Ursulinengymnasium, Innsbruck, Austria.
Let 0 < r; s; t 1 be xed. Show that the relation
r cot rA = s cot sB = t cot tC
holds for exactly one triangle ABC , and that this triangle maximizes the expression
sin rA sin sB sin tC
over all triangles ABC .
Solution by Murray S. Klamkin, University of Alberta.
At most one of the angles rA; sB; tC can be greater than =2 so that, by the equality
conditions, none of them are greater than =2. Consequently the given cotangents are
monotonic in their angle argument. Now assume that A; B; C and A0; B 0 ; C 0 are dierent
solutions. Since A0 + B 0 + C 0 = A + B + C = , one pair of angles from A0; B 0 ; C 0 must
be bigger and smaller than the corresponding pair in A; B; C . This gives a contradiction
since the cotangents are monotonic. Consequently ABC is unique.
To maximize sin rA sin sB sin tC we take logs and use Lagrange multipliers with
Lagrangian
L = log sin rA + log sin sB + log sin tC , (A + B + C ):
Then
@L = @L = @L = 0
@A @B @C
yields the given cotangent relations for the maximum. On the boundary, i.e. for A or B
or C equal to 0, we obtain the minimum value 0.
More generally, a similar argument goes through to maximize
sinu rA sinv sB sinw tC
24
with the additional condition u; v; w > 0. Here the maximizing equations are
ur cot rA = vs cot sB = wt cot tC:
Also, a similar argument for maximizing
cosu rA cosv sB cosw tC
does not go through the same way. Here the extremal equations are
ur tan rA = vs tan sB = wt tan tC:
However, we now have to check the boundary. This entails setting one and then two of
A; B; C equal to 0. Then we have to decide the absolute minimum and maximum from
these seven possibilities. We leave this as an open problem.
Also solved by the proposer, who mentions that the problem contains as special cases
the items 2.10{2.13 of Bottema et al, Geometric Inequalities.
By \exactly one" triangle (the editor's wording) was meant of course \up to simi-
larity"!
* * * * *
1485. Proposed by Marcin E. Kuczma, Warszawa, Poland.
From a deck of 52 cards, 13 are chosen. Replace one of them by one of the remaining
39 cards. Continue
the process until the initial set of 13 cards reappears. Is it possible
that all the 52
13
combinations appear on the way, each exactly once?
Comment by Stanley Rabinowitz, Westford, Massachusetts.
It is known that the subsets of size k from a set of size n can be arranged in a
circular list such that adjacent sets in the list dier by the replacement of one element by
another. (The subsets are said to be in revolving door order.) A reference is [1] in which
an algorithm for forming such a list is given (not just an existence proof).
The idea behind the algorithm is as follows. If A(m; l) denotes a list of all the
l-subsets of f1; 2; : : : ; mg arranged in revolving door order beginning with f1; 2; : : : ; lg and
ending with f1; 2; : : : ; l , 1; mg, then it can be shown that
A(n; k) = A(n , 1; k) ; A(n , 1; k , 1) fng;
where the bar means that the order of the list is reversed and the cross means that the
element n is appended to each subset in the list. It is easy to check that if A(n , 1; k) and
A(n , 1; k , 1) are in revolving door order, then so is A(n; k). It follows by induction that
the list A(n; k) exists for each n and k.
Reference:
[1] Albert Nijenhuis and Herbert S. Wilf, Combinatorial Algorithms (second edition), Aca-
demic Press, New York, 1978, pp. 26{38.
Also solved by RICHARD I. HESS, Rancho Palos Verdes, California; and the pro-
poser.
25
* * * * *
1487. Proposed by Kee-Wai Lau, Hong Kong.
Prove the inequality
x + sin x 2 log(1 + x)
for x > ,1.
Combined solutions of Richard I. Hess, Rancho Palos Verdes, California, and the
proposer.
Let
f (x) = x + sin x , 2 log(1 + x) ; x > ,1:
We have
x + sin x = 2x , x3! + x5! , x7! + = 2x + (,(21)k +x 1)!
3 5 7 X1 k k 2 +1
k =1
and
2 log(1 + x) = 2x , 22x + 23x , 24x + = 2(,1)k x ;
2 3 4 X1 k k +1
k=1
and thus
f (x) = x (1 , 23x , 3!x ) + x2 (1 , 45x + 25!x ) + x3 (1 , 67x , 37!x ) +
4 6
2
X1 xk 2
2 k ( , 1)kk !
=
k =1
k 1 , 2k + 1 x + (2k + 1)! x :
Case (i): ,1 < x 1. Then for each k 1,
1 , 2k x + (,1) k x > 1 , 2k , k = (2k)! , k > 0;
k
2k + 1 (2k + 1)! 2k + 1 (2k + 1)! (2k + 1)!
so f (x) 0.
Case (ii): x 4:5. Since x , 2 log(1 + x) increases for x > 1, and
f (4:5) 4:5 , 1 , 2 log 5:5 0:0905038;
f (x) > 0 for x 4:5.
Case (iii): 1 < x < 4:5. The functions
g(x) = x + sin x and h(x) = 2 log(1 + x)
are both nondecreasing, and by means of a calculator we check that
g(a) , h(a + 0:05) > 0
for a = 1; 1:05; 1:1; 1:15; : : :; 4:5. [Editor's note. He's right! In fact the smallest value for
g(a) , h(a + 0:05) you get is g(4:05) , h(4:1) 0:0029937:] Hence for x 2 (a; a + 0:05]
where a is any of the above values,
f (x) = g(x) , h(x) g(a) , h(a + 0:05) > 0:
26
p
is an integer. Since (2 , 3) n < 1, it follows that
2
p p p
[(2 + 3) n] = (2 + 3) n + (2 , 3) n , 1:
2 2 2
x = ,a +2b + c ; y = a , b2 + c ; z = a + b2 , c ;
2 2 2 2 2 2 2 2 2
5 (a + pb + c ): 2 2
4
29
or f (a) 0, where
p
f (a) = 4a (c , b) , a bc + a(4b + 4b c + 4bc , 4c , 5bc b + c ) + 4bc(c , b ):
3 2 3 2 2 3 2 2 2 2
p
Since x 0, a b + c and we are to show that f (a) is negative for
2 2
p
bc a b +c : 2 2
(2)
If b = c, p
f (a) = ,ab [(a , b) + (5 2 , 7)b] < 0:
2
If b < c, f (a) is a cubic with rst and last coecients greater than 0. We have
f (,1) < 0 ; f (0) > 0 ; f (1) > 0;
and we nd p
f (c) = ,bc (5 b + c , 4b , 3c) < 0
2 2 2
because
25(b + c ) , (4b + 3c) = (3b , 4c) > 0;
2 2 2 2
further,
p p
f ( b + c ) = 2bc(4bp b + c , 5b , c )
2 2 2 2 2 2
= ,2bc( b + c , 2b) 0: 2 2 2
We see that f has three preal zeros. One is negative, one is between 0 and c, and
p one is
equal to or greater than b + c . Therefore f does not change
2 2
p sign in c a b +c 2 2
p The only solution of f (a) = 0 that satises (2) is a = pb + c , and then only 2 2
if b + c = 2b; that is, when (a; b; c) are proportional to (2; 1; 3) and so (x; y; z ) are
2 2
as Henderson to rewrite the inequality in the form (1). He then observes that since the
numerators on the left side of (1) are all positive, a; b; c are the sides of an acute triangle
ABC . Letting AP; BQ; CR be the altitudes, he obtains
a2 + b2 , c2 = 2ab cos C = a cos C = CQ ; etc:;
2b 2b
and thus the inequality takes the form
p
AR + BP + CQ p5 a2 + b2 + c2;
4 2
with equality for the 30{ 60{ 90 triangle. (Not exactly acute, as Kuczma notes!)
A lovely problem! It seems almost ungrateful to ask if there is a generalization to n
variables : : : .
* * * * *
1491. Proposed by J.T. Groenman, Arnhem, The Netherlands.
In triangle ABC , the internal bisector of 6 A meets BC at D, and the external
bisectors of 6 B and 6 C meet AC and AB (produced) at E and F respectively. Suppose
that the normals to BC; AC; AB at D; E; F , respectively, meet. Prove that AB = AC .
I. Solution by Toshio Seimiya, Kawasaki, Japan.
Because the normals to BC; AC; AB at D; E; F are
concurrent (at Q, say), we have
BD + DQ = BF + FQ ;
2 2 2 2
CE + EQ = CD + DQ ;
2 2 2 2
AF + FQ = AE + EQ ;
2 2 2 2
so, adding,
(BD , DC ) + (CE , EA ) + (AF , FB ) = 0: (1)
2 2 2 2 2 2
(b + c) 2
(c , a) (b , a)2 2
or
a (c , b) , b (a + c) + c (b + a) = 0:
2 2 2
(2)
b+c c,a b,a
In the case a > b; c, or b > a > c, or b < a < c, we have (2) similarly. [Because the above
formulas for CE , etc. will change only in sign.{Ed.] The left side of (2) becomes
(c , b)(a + b)(a + c)(b + c , a) : 2
(b + c)(c , a)(b , a)
[Editor's note. Seimiya gave an algebraic derivation. Can anyone nd a slick reason why
the left side of (2) factors so conveniently?] Because a; b; c > 0 and b + c > a, we obtain
c , b = 0. This implies AB = AC .
II. Solution by R.H. Eddy, Memorial University of Newfoundland.
More generally, let the lines AD; BE; CF intersect at a point P with trilinear
coordinates (x; y; z ) with respect to a given reference triangle ABC with sides a; b; c. If
we denote lines through D; E; F by d; e; f and assume that these pass through Q(u; v; w),
then it is easy to see that the coordinates of d; e; f are
[vz , wy; ,uz; uy] ; [vz; ,uz + wx; ,vx] ; [,wy; wx; uy , vx];
respectively. Since the condition that lines [l ; m ; n ] and [l ; m ; n ] are perpendicular is
1 1 1 2 2 2
One can check that the coordinates of the centroid (bc; ca; ab) and the Gergonne
point !
1 1 1
a(s , a) ; b(s , b) ; c(s , c)
(s the semiperimeter) also satisfy the above system, in which case Q is the circumcentre
and the incentre respectively. Are there any other interesting pairs?
Reference:
[1] D.M.Y. Sommerville, Analytical Conics, G. Bell and Sons, Ltd., London, 1961.
Also solved by C. FESTRAETS-HAMOIR, Brussels, Belgium; WALTHER
JANOUS, Ursulinengymnasium, Innsbruck, Austria; MARCIN E. KUCZMA, Warszawa,
Poland; MARIA ASCENSION LOPEZ CHAMORRO, I.B. Leopoldo Cano, Valladolid,
Spain; and the proposer.
The proposer mentioned that the analogous problem for three interior bisectors was
solved by Thebault. This case is contained in solution II above.
* * * * *
Please send notice of your desire to present a paper no later than May 31, 1991.
Crux
Mathematicorum
Volume 17 , Number 2 February 1991
CONTENTS
1
33
(c) Determine the limit of mp=n where p is xed and n tends to innity.
Question 2.
In this question we replace the two extra conditions of Question 1 by the two
following conditions:
i. p = n;
ii. for every ordered pair of positive integers (i; k) with i + k n, the integer aik
does not belong to the set Ei k .
+
(a) Show that for distinct i and j the sets Ei and Ej are dierent.
(b) Show that if n is at least 2q , where q is a positive integer, then N q + 1.
(c) Suppose that n = 2r , 1, where r is a xed positive integer. Show that N r.
Exercise II. Determine, for n a positive integer, the sign of n + 5n sin n + 1. For
6 5
n + 5n sin n + 1
6 5
34
Exercise III. Consider two spheres and and a straight line which does not
1 2
meet them. For i = 1 and i = 2, let Ci be the centre of i , Hi the orthogonal projection
of Ci on , ri the radius of i , and let di be the distance of Ci to . Let M be a point
on , and for i = 1 and i = 2, let Ti be the point of contact with i of a plane tangent to
i and passing through M ; set i(M ) = MTi . Situate M on so that (M ) + (M ) is 1 2
minimized.
Exercise IV. Consider ve points M , M , M , M , M situated on a circle C in
1 2 3 4
the plane. Show that the product of the distances of M from the lines M M and M M 1 2 3 4
equals the product of the distances from M to the lines M M and M M . What can one
1 3 2 4
deduce about 2n + 1 distinct points M ; : : : ; M n; M situated on C ?
1 2
* * *
Now we turn to solutions sent in response to the appeal to help \tidy up" the
archives.
K797. [1983: 270] Problems from Kvant.
It is well known that the last digit of the square of an integer is one of the following:
0,1,4,5,6,9. Is it true that any nite sequence of digits may appear before the last one, that
is, for any sequence of n digits fa ; a ; : : : ; ang there exists an integer whose square ends
1 2
with the digits a a : : : anb, where b is one of the digits listed above?
1 2
Alternatively, consider any sequence of the form 40k + 39. Squares cannot end as
: : : 90, : : : 91, : : : 94, : : : 95 or : : : 99, and 400k +396 = (10x 4) gives 20k +19 = 5x 4x,
2 2
*
The next block of solutions are for problems from the unused IMO proposals given
in the 1985 numbers of Crux.
17. [1985: 38] Proposed by Poland.
Given nonnegative real numbers x ; x ; : : : ; xk and positive integers k, m, n such
1 2
that km n, prove that
Y
k ! Xk
n xi , 1 m (xni , 1):
m
i=1 i=1
x ln P 0
Finally it is to be noted that m and n need not be positive integers, just being positive
reals with n km suces.
23. [1985: 39] Proposed by the USSR.
A tetrahedron is inscribed in a unit sphere. The tetrahedron is such that the center
of the sphere lies in its interior. Show that the sum of the edge lengths of the tetrahedron
exceeds 6.
36
where F = ab=2 is the area of T , F the area of T . Also R = c=2. The given inequality
1 1
now becomes 8mamb mc 5abc or, squaring,
(a + b )(a + 4b )(4a + b ) = 64(mamb mc ) 25a b c = 25a b (a + b ):
2 2 2 2 2 2 2 2 2 2 2 2 2 2
The last seven solutions are to problems from other Olympiads which appeared in
the 1985 issues.
2In. the
[1985: 168] 1984 Dutch Olympiad.
given diagram, B is a bat-
tery, L is a lamp, and S , S , S , S ,
@
@
,
,
S
S2 @@ @S3
1
S4 @@ @,
,@
S are switches. The probability that
5
1 2 3 4
@
L
switch S is on is 2/3, and it is 1/2 for ,
,
@
S5 @
3
the other four switches. These proba-
bilities are independent. Compute the
probability that the lamp is on. B
Solution by Murray S. Klamkin, University of Alberta.
More generally, let pi and qi denote switch Si being on or o, respectively, and also
their probabilities so 0 pi; qi 1, pi + qi = 1. The probability that the lamp is on is
given by
X X
p p p p p + q p p p p + q q p p p ,q q p p p ,q q p p p
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 2 5 1 3 4 1 4 2 3 5
+p p q q q + p p q q q :
1 2 3 4 5 4 5 1 2 3
The sums are symmetric over the p's and q's. Note that if at most one switch is o, the
lamp is on; if any two switches except q q or q q are o, the lamp is on, and nally if
2 5 1 4
only two switches are on they must be p p or p p . [Editor's note: this gives a probability
1 2 4 5
of 25=48 for the above problem.]
*
4. [1985: 237] 1981 Leningrad High School Olympiad.
A plane is partitioned into an innite set of unit squares by parallel lines. A triangle
ABC is constructed with vertices at line intersections. Show that if jAB j > jAC j, then
jABj , jAC j > 1=p, where p is the perimeter of the triangle. (Grades 8, 9, 10)
Solution by Murray S. Klamkin, University of Alberta.
Without loss of generality we can take the rectangular coordinates of A,B ,C to be
(0; 0), (x; y), (u; v), respectively, where x,y,u,v are integers. Then letting c = jAB j = 2 2
*
2. [1985: 239] 1983 Annual Greek High School Competition.
If a and b are given nonparallel vectors, solve for x in the equation
a + xa b = b + a b :
2 2
P a+b -
R
*
B@
B@
B @
B @
B a @
b
a,xb 0 B
B @
R
@
@
B Q
,x b
BBN 0
S
By the law of sines, a= sin( , ) = x0b= sin , and a= sin = b= sin( + ). Hence,
sin( + ) , sin( , ) = b sin , a sin0
a xb
or 2 cos = b=a , a=x0b. Finally,
x = ,x0 = b(b ,,2aa cos ) = b +,2aa b :
2 2
*
40
with equality just when = = = =3. (See O. Bottema et al, Geometric Inequalities,
1 2 3
item 2.3 on p. 18 or the equivalent inequality in item 2.23 on p. 25.) Thus f (3) = 9=4.
We show that f (n) = 9=4 for all n 3 by proving the following theorem.
Theorem: Let i 0 for all i = 1; 2; : : : ; n be such that Pni i = with n 3. Then
sin +sin + +sin n 9=4 with equality holding if and only if = = = =3
=1
2 2 2
1 2 1 2 3
and i = 0 for all i = 4; 5; : : : ; n, where we have, without loss of generality, renumbered
the indices so that : : : n .
1 2 3
= sin A cos B + 2 sinA sinB cosA cosB + cos A sin B , sin A , sin B
2 2 2 2 2 2
+1 +1
with n 3 would imply that ni i > , a contradiction. Hence by the lemma and the
=1
+1
=1
induction hypothesis we obtain
nX nX , X n
sin i = sin i + sin n + sin n sin i 94 :
+1 1
2 2 2 2 2
+1
i i =1 i
=1 =1
If equality holds it must hold in the lemma. This implies that either n = 0 or
n + n = =2. However, if n + n = =2, then + =2 and since Pni i = ,
+1
+1
+1 +1 1 2
we obtain + = =2 and n = 4. Since P , it follows immediately
=1
1 2 1 2 3 4
that i = =4, i = 1; 2; 3; 4, from which we obtain i sin i = 2, a contradiction. Thus 4
=1
2
41
x +x x x +x x2
1 2 xn, + xnx xn + x x
3
2
2 3 4
2
1 1
2
1 2
We prove by induction that Sn < n , 1 for all n 3. Consider the case n = 3. Since the
sum is cyclic and since the expression is unchanged when x and x are interchanged, we 2 3
may assume that x x x . Then 1 2 3
S = x +x x x + x +x x x + x +x x x < x +x x x + x +x x x + 1 = 2:
2 2 2 2 2
1 2 3 1 2
3 2 2 2 2 2
1 2 3 2 3 1 3 1 2 1 2 1 2 2 1
Now suppose Sn < n , 1 for some n 3 and consider Sn for n + 1 positive numbers +1
x ; x ; : : : ; xn . Without loss of generality, we may assume that xn = maxfxi : i =
1 2 +1 +1
1; 2; : : : ; n + 1g. Note that
Sn = Sn + x x+n,x x + x +xxn x + x xn+ x x , x xn+,x x , x +xnx x :
2 2 2 2 2
1 +1 1
+1 2 2 2
n, n n n n
1 n +1 n, n 2
n +1 1 +1 1 2 1 1
2
1 2
Since
xn, 2
xn, xn xn ; and xn
2 2 2 2
,
xn + x x < 1;
1 1 +1
xn, + xn xn
2
1 xn, + xn x xn + xn x xn + x x
+1
2
1 1
2
+1 1
2
1 2
2
+1 1 2
Sn = t t+ t + t +t + + t n,
2
+ t n,
4
+ tn 2 4 2 2 4
t2
t n, + t n, t n, + t n
5 4 7
t n+t 2 4 3 1 2 2 2 +1 4 3
,! n , 1 as t ,! 0 : +
That is, though the bound n , 1 can not be attained for n > 2, it is nonetheless sharp.
* * *
This completes the \archive" material for 1985 and the space available this number.
Contest season is upon us. Send me your contests and solutions!
42
BOOK REVIEW
30th International Mathematical Olympiad, Braunschweig, 1989 - Problems and
Results, edited by Hanns-Heinrich Langmann. K.H. Bock, 5340 Bad Honnef, Germany,
1990. ISBN 3-87066-213-1. 79 pp., softcover. Reviewed by Richard Nowakowski, Dalhousie
University.
This report on the 1989 IMO rst lists the rules and regulations governing the
competition. The program of events comes next followed by the competition problems
and solutions. All 109 proposed problems are listed next with solutions given to the 32
that were shortlisted by the jury. (These 32 problems have appeared in Crux in [1989:
196{197, 225{226, 260{262].) The contestants' scores and awards are given last. The book
is an excellent source of problems, but it would have been nice to see the solutions given
to more of the dicult problems which were not shortlisted. For example, I have yet to
see a good proof (one reasonably explained and under four pages long) of the following
proposed question:
HEL 2. In a triangle ABC for which 6(a + b + c)r2 = abc holds and where r denotes
the inradius of ABC , we consider a point M of its inscribed circle and projections D; E; F
of M on the sides BC; AC; AB , respectively. Let S and S1 denote the areas of the triangles
ABC and DEF , respectively. Find the maximum and minimum values of the quotient
S=S1.
* * * * *
CALL FOR PAPERS { ICME
There will be two 90{minute sessions at the ICME in Quebec City (August 1992)
on mathematical competitions. Papers are solicited on this topic which are of general
interest to the mathematics education community.
Enquiries and proposals for papers should be sent to
Please send notice of your desire to present a paper no later than May 31, 1991.
43
PROBLEMS
Problem proposals and solutions should be sent to B. Sands, Department of Mathe-
matics and Statistics, University of Calgary, Calgary, Alberta, Canada T2N 1N4. Propos-
als should, whenever possible, be accompanied by a solution, references, and other insights
which are likely to be of help to the editor. An asterisk (*) after a number indicates a
problem submitted without a solution.
Original problems are particularly sought. But other interesting problems may also
be acceptable provided they are not too well known and references are given as to their
provenance. Ordinarily, if the originator of a problem can be located, it should not be
submitted by somebody else without permission.
To facilitate their consideration, your solutions, typewritten or neatly handwritten
on signed, separate sheets, should preferably be mailed to the editor before September 1,
1991, although solutions received after that date will also be considered until the time when
a solution is published.
1611. Proposed by George Tsintsifas, Thessaloniki, Greece.
Let ABC be a triangle with angles A; B; C (measured in radians), sides a; b; c, and
semiperimeter s. Prove that
X X c,a 9
(i ) b + Ac , a 6s ; (ii ) b +aA :
1612. Proposed by Walther Janous, Ursulinengymnasium, Innsbruck, Austria.
Let x; y; z be positive real numbers. Show that
Xy ,x2 2
z + x 0;
where the sum is cyclic over x; y; z , and determine when equality holds.
1613. Proposed by Murray S. Klamkin, University of Alberta.
Prove that sin x p tan x p
2
x + x 2
for p 0 and 0 < x < =2. (The case p = 1 is problem E3306, American Math. Monthly,
solution in March 1991, pp. 264{267.)
1614. Proposed by Toshio Seimiya,Kawasaki, Japan.
Let D and E be points on side BC of a triangle ABC . Draw lines through D, E
parallel to AC; AB respectively, meeting AB and AC at F and G. Let P and Q be the
intersections of line FG with the circumcircle of ABC . Prove that D; E; P and Q are
concyclic.
44
sin ' 2
or
cot ' + 2R , R + (m , d) = 0:
2 2 2 2
(1)
Now
x = R ,d ;
cot ' = m 2
2 2 2
m 2 2
It is clear that R > m , d, so R , (m , d) > 0, and we conclude that (2) has two dierent
2 2
D = R m + m (R , d )[R , (m , d) ]
2 4 2 2 2 2 2
4
= m (R , 2R d + d + 2R dm , 2d m + d m)
2 4 2 2 4 2 3 2
= m (R , d + dm) :
2 2 2 2
So
= ,Rm + m (R , d + dm)
2 2 2
1
R ,d 2 2
= m ( R , d ) , m (R , d)
2 2 2
R ,d 2 2
and thus
1 = R+d = 1+ 1 = 1 + 1 ; (3)
m(R + d , m) m R + d , m PM PC
1
the \harmonic mean" property given on [1990: 27], with a maximum for when PM = 1
PC . Also
= ,Rm , m (R , d + dm)
2 2 2
2
R ,d 2 2
2 2
= ,m(R , d) , m = ,m(R , d + m) ;
2
R,d R,d
so
1 = 1 , 1 = 1 , 1 : (4)
R , d + m m PC 0 PM
2
sin ' 2
or (as above)
47
0 (R , d ) , 2Rm 0 , m [R , (m , d) ] = 0:
2 2 2 2 2 2 2
(5)
Comparing (2) and (5) we conclude: if (2) has the roots and , then (5) has the roots
1 2
, and , :
1 2
(3)
OP OE OP AP a 2
or
a ON = OP CM C 0M:
2
(4)
Also
ON = OI = R , IL = R , MN ;
OP OQ R , QL R,
48
= C 0M (a , OP CM ) = C 0M (m + (AM ) , OP CM )
2 2 2
= C 0M (m + CM (C 0M , OP )) = C 0M (R CM , (CM , m)m)
2
= C 0M (R CM , cm):
Thus CM (R , ) = R CM , cm, or
cm = cm :
= CM c+m
Therefore the diameter 2 of the incircle is the harmonic mean of CP and MP (and,
incidentally, also of C 0P and M 0 P ). The result then follows as before.
References:
[1] Marcel Berger, Geometry II (English translation), Springer, Berlin, 1987, Prop. 16.6.4
and Fig. 16.6.4(2), pp. 203 and 204.
[2] H.G. Forder, Higher Course Geometry, 1931/49, p. 224.
Editor's note. Walker also observes that, letting CQ and AMB meet at S , and CL and
PQ meet at T , one gets MS = PT = PA, and therefore OL; CS; PT concur (at Q). He
uses this approach to give a second solution, not using the Poncelet theorem.
A further solution has been received from TOSHIO SEIMIYA, Kawasaki, Japan.
Seimiya rst establishes the following lemma. ABC is a triangle with incenter I and
circumcircle , . Let D be a point on side BC . Let , be an inscribed circle tangent to the
1 2
segments AD and DC , at E and F respectively, and to the arc AC of , . Then E ,F and
1
I are collinear. Seimiya uses this lemma to prove the present problem, as well as to give
another proof of Crux 1260 [1988: 237; 1989: 51]. He also points out that the harmonic
mean relationship given above occurs as problem 1.2.7, p. 5 of H. Fukagawa and D. Pedoe,
Japanese Temple Geometry Problems (reviewed on [1990: 203]).
* * * * *
1430. [1989: 74; 1990: 158] Proposed by Mihaly Bencze, Brasov, Romania.
AD; BE; CF are (not necessarily concurrent) cevians in triangle ABC , intersecting
the circumcircle of ABC in the points P; Q; R. Prove that
AD + BE + CF 9:
DP EQ FR
When does equality hold?
49
DP EQ FR R
From the inequality (given in the published solution of Seimiya [1990: 159])
!
AD + BE + CF X (b + c) , 1 = X a X b + c , a ; 2
DP EQ FR a 2
a 2
where the sums are cyclic over a; b; c, and the fact that
2 r Y b + c , a P a(b + c , a) 2
4, R =4, a = Qa ;
it is enough to show that
X X X
a b c (b + c , a) a(b + c , a) :
2
2 2 2
Putting
x = b + c , a; y = c + a , b; z = a + b , c;
this becomes X X x + y x + z X y + z
2 2 2
x x 2 2
x 2 ; 2
or X X X
x x(x + y) (x + z ) , 4( x (y + z )) 0;
2 2
(1) 2 2
where the sums are cyclic over x; y; z . The left side of (1) becomes
X X X X
6
x + 3 (x y + x z ) , (x y + x z ) , 6 x y
5 5 4 2 4 2 3 3
X
+2 (x y z + x z y) , 9x y z
3 2 3 2 2 2 2
hX X i hX X i
= x , (x y + x z ) + 3x y z + 3 (x y + x z ) , 2 y z
6 4 2 4 2 2 2 2 5 5 3 3
hX i
+2 (x y z + x z y) , 6x y z 3 2 3 2 2 2 2
P(y + z , x ) (y , z ) + 2 P y z (y , z )
2 2 2 2 2 2 2 2 2 2 2 2
= 2XP x X
2
+3 yz (y , z ) + 2 x yz (y , z ) 2 2 2 2 2
0;
so (1) follows.
Editor's note. Murray Klamkin has obligingly supplied the editor with the following
simpler proof of inequality (1). By Cauchy's inequality,
X X X
x x(x + y) (x + z ) x(x + y)(x + z ) ;
2
2 2
50
which simplies to
X
x(x , y)(x , z ) 0: (2)
and similarly
B = b A + 1 , b C; C = c B + 1 , c A;
0 0
where a,b,c are the respective sides of ABC and is a nonzero constant.
For the case G = G0 , we choose the origin to be the centroid of ABC so that
A + B + C = 0 and also
0 = A + B + C = b , c + 1 A + c , a + 1 B + a , b + 1 C:
0 0 0
Thus
0 = 1b , 1c A + 1c , a1 B + a1 , 1b C = 2b , 1c , 1a A + 1b + 1c , a2 B;
and so
2 , 1 , 1 = 1 + 1 , 2 = 0;
b c a b c a
which implies a = b = c.
For the case O = O0, we choose the origin to be the circumcenter of ABC so that
jAj = jBj = jCj = R, the circumradius. Then also jA0j = jB0j = jC0j = R0 , or
2 2 2 2
2 2
R = a jCj + 1 , a jBj + 2 a 1 , a B C
2 0
2
2 2
= R + 1 , R + 2(a , ) R cos 2A
2 2
2 2 2
a
2
a! a 2
= 2 , 2 + 1 R + 2(a , ) R , 4(a , )R sin A
2 2 2
2 2
a2
a a 2
a 2
= R , (a , ); etc.;
2
or
(a , ) = R , R0 = (b , ) = (c , ):
2 2
Hence a = b = c.
II. Solution to part (a) by the proposer.
It is known (Crux 1464 [1990: 282]) that G = G0 if and only if
52
BA0 = CB 0 = AC 0 ;
A0C B 0 A C 0B A
B@
therefore according to our assumption we have CH
B@
B @
0 0 0
0
Let M , N , O be the midpoints of BC , E
Q
Q
0
PQ, EG, respectively. Let A be symmet-
h h P
h N Q
Q
Q
h h hhh Ohh Q Q
ric to A with respect to O. We have that
Q
Q
hh hQ
hG
Q
h
Q
AE is equal and perpendicular to BA, and
Q B
Q A
0
analogously EA (= AG) to AC . Therefore
L Q
h Q
h B
B
AA0 = BC , AA0 ? BC and OE = MA,
B
B B
OE ? MA. Let L on MN be such that QQ D
B
B
B
ALkOE . Then AO = LN , and ML > MA Q
Q
B
B F
B
because MA is perpendicular to AL. Hence Q B
Q
MN = LN + ML AO + MA, and Q
Q
B
B M C
BP + CQ = 2MN 2(AO + MA) = 2(BM + OE ) = BC + EG:
Equality occurs if and only if MN = MA + AO, i.e., MN coincides with MO, i.e.,
AB = AC .
53
ln( 5 , 1) + 2 , 5 : ,
, @
@
@x ,y
4 ,
,
@
=1
(b) If k > 2 is an integer, are there innitely many solutions in positive integers to
! !
a
ck = k , k ? b
55
(a + b , 1)(a , b) = 2c : 2
b = cD , D 2, 1 ; a = b + D = cD + D 2+ 1 :
2 2
6c = (a , 1) , (a , 1) , [(b , 1) , (b , 1)]
3 3 3
c = t ; b = t ; a = t + 2 (t arbitrary)
2 3 3
to produce an innite number of solutions. These are not all the solutions when k = 3.
Others include (a; b; c) = (9; 6; 4), (12; 4; 6) and (25; 19; 11).
Also solved (both parts) by H. L. ABBOTT, University of Alberta; HANS
ENGELHAUPT, Franz-Ludwig-Gymnasium, Bamberg, Germany; and RICHARD K. GUY,
56
University of Calgary. Part (a) only was solved by MATHEW ENGLANDER, Toronto,
Ontario; RICHARD A. GIBBS, Fort Lewis College, Durango, Colorado; RICHARD I.
HESS, Rancho Palos Verdes, California; FRIEND H. KIERSTEAD JR., Cuyahoga Falls,
Ohio; MURRAY S. KLAMKIN, University of Alberta; CHRIS WILDHAGEN, Breda, The
Netherlands; and the proposer.
The remarkable solution II was also found by Engelhaupt and by Guy. Abbott's
solution to part (b) was a recursively dened innite sequence of solutions, the rst being
the trivial ! !
1 =
3
3 2
, 3 ;
3
and the second being the already formidable
! !
16199 =3
30527 14328
, 3 !
3
These all have the additional property that a = b + c. Guy has since found other solutions
with this property, for instance
! !
26 =3
50 24
, 3 :
3
Example (1) of case (i) in Ahlburg's solution was also found by Engelhaupt, Eng-
lander and Gibbs, while example (2) of case (ii) was again given by Engelhaupt. Another
special case of Ahlburg's case (i), namely k = D(odd), was found by Hess, Kierstead and
Klamkin. This gives the nice solution
a = 3b , 1 ; c = 2b , 1 ; b arbitrary:
No solutions for k > 3 were sent in. However, Guy has come up with the single
(and singular!) example ! !
18
6 = 5 , 5
5 12
q
tor's note: see also Dou's comment at D D A
D A
M , @
,
the end of his solution to Crux 1493, DD D AA , @
@
this issue!] Analogously, DM ? CG. If D B
D ,@ ,
D, @,D @
@
F
AE = 2AD then AB = AN = AM = D D
@ O ,@
,
,
AD and 6 NAM = 6 BAD (= 90), so L
O1
@,
C HH @
H @ ,
BN is parallel to DM , and therefore
B
HH@
,
,
L, C , G are collinear. It is clear that if
0
H@ HH,G
@
AE 6= 2AD, then DM is not parallel to BN and L, C , G are not aligned.
II. Solution by Marcin,!E. Kuczma,,!
Warszawa, Poland.
Consider the vectors AD = u, AE = v. Let h denote rotation by 90 anticlockwise.
Since h is a linear map,
,! ,! ,! ,! ,! ,!
CL = CB + BL = ,u + h(BJ ) = ,u + h(AJ ) , h(AB ) = ,u , v , u = ,2u , v
and ,! ,! ,! ,! ,!
h(CG) = h(CD) + h(DG) = DA + DE = ,u , u + v = v , 2u:
Therefore
,! ,! ,! ,!
L; C; G are in line () CGkCL () CL ? h(CG)
() (2u + v)(2u , v) = 0
() jvj = 2juj:
Also solved by HANS ENGELHAUPT, Franz-Ludwig-Gymnasium, Bamberg,
Germany; C. FESTRAETS-HAMOIR, Brussels, Belgium; L. J. HUT, Groningen, The
Netherlands; WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria; MARCIN
E. KUCZMA, Warszawa, Poland (a second solution); KEE-WAI LAU, Hong Kong;
P. PENNING, Delft, The Netherlands; D.J. SMEENK, Zaltbommel, The Netherlands;
and the proposer. One incorrect solution was sent in.
The above solutions are related, but the dierent approaches are interesting.
The problem was taken from a lost sangaku dated 1826, and appears as problem
4.21, page 47, of Fukagawa and Pedoe's Japanese Temple Geometry Problems. Also given
there is the relationship between the sides of the four squares.
* * * * *
58
1497. [1989: 298] Proposed by Ray Killgrove and Robert Sternfeld, Indiana State
University, Terre Haute.
A translate g of a function f is a function g(x) = f (x + a) for some constant a.
Suppose that one translate of a function f : R ! R is odd and another translate is even.
Show that f is periodic. Is the converse true?
Solution by the St. Olaf Problem Solving Group, St. Olaf College, Northeld,
Minnesota.
We know that there exist a and b such that for every x, f (x + a) = f (,x + a) and
f (x + b) = ,f (,x + b). This implies that f (2a + x) = f (,x) and f (2b + x) = ,f (,x).
Thus,
f (x + 4(a , b)) = f (2a + (x + 2a , 4b)) = f (4b , x , 2a)
= ,f (2a + x , 2b) = ,f (2b , x) = f (x):
Thus f (x) has period 4(a , b). (If a = b, then f (x + a) = f (a , x) = ,f (a + x) which
implies that f (x) is the zero function.)
The converse is not true. The function f (x) = x , [x] provides a counterexample.
Also solved (in about the same way) by SEUNG-JIN BANG, Seoul, Republic
of Korea; WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria; MURRAY
S. KLAMKIN, University of Alberta; MARCIN E. KUCZMA, Warszawa, Poland; DAVID
E. MANES, SUNY at Oneonta, New York; CHRIS WILDHAGEN, Breda, The Nether-
lands; and the proposers.
* * * * *
1498 [1989: 298] Proposed by D.M. Milosevic, Pranjani, Yugoslavia.
:
Show that
Y ai 3
hi (3r) s ; 2
i=1
where a ; a ; a are the sides of a triangle, h ; h ; h its altitudes, r its inradius, and s its
1 2 3 1 2 3
semiperimeter.
I. Solution by Mark Kisin, student, Monash University, Clayton, Australia.
By the generalized A.M.-G.M. inequality,
Y ai ! = s Y ai ! a1 +a2 +a3 a h + a h + a h
1 2
1
3 3
hi = hi a +a +a : 1 1 2 2 3 3
i =1 i =1 1 2 3
But
a h = a h = a h = 2(Area) = 2rs;
1 1 2 2 3 3
so
Y ai ! = s 6rs
3 1 2
hi 2s = 3r;
i=1
and the result follows.
59
i=1
i.e. 2s s
Y3 2
3 :
ai ai (1)
i
=1
Y n Pn p x !Pni=1 pi xi
xi pixi i P i i =1
: (2)
=1 i n
Proof. The function f (x) = x log x, x > 0, is convex. Thus we get
X n Pn p x !
pi f (xi ) Pnf i P i i ; =1
=1i n
i.e., (2). 2
Since f (x) is strictly convex, equality holds in (1) if and only if a = a = a . 1 2 3
hi Fi [(n + 1)r]F ;
=1i
where hi are the altitudes, Fi the (n , 1)-dimensional areas of the faces, F = P Fi , and r
is the inradius.]
Also solved by L. J. HUT, Groningen, The Netherlands; MURRAY S. KLAMKIN,
University of Alberta; and MARCIN E. KUCZMA, Warszawa, Poland.
Klamkin's solution was very similar to solution II, obtaining a slightly weaker result.
* * * * *
1499. [1989: 298] Proposed by Herta T. Freitag, Roanoke, Virginia.
A second-order linear recursive sequence fAng is dened by An = An + An
1 +2 +1
for all n 1, with A and A any integers. Select a set S of any 2m consecutive elements
1
1 2
from this sequence, where m is an odd integer. Prove that the sum of the numbers in S is
always divisible by the (m + 2)nd element of S , and the multiplying factor is Lm, the mth
Lucas number.
60
= rn m + rn ( )m + s n m + s n ( )m ;
+2 +1 +1 +2 +1 +1
which simplies (by (1)) to showing that 2ab = d d . Now it is known that
1 2
d + d = 2(a + b )
2
1
2
2
2 2
ABCD is a parallelogram, E B
area ABD = area ACD:
Therefore we get
1 AB BD sin = 1 AD AC sin :
2 2
Hence from (2) we have sin = sin , and since + < we get = . Let F be the
point of intersection of AC with DE . As
6 AEF = + 6 EDB = + 6 ADE = 6 AFE;
62
we get AE = AF . Because
6 CDF = 6 AEF = 6 AFE = 6 CFD;
we have CF = CD. Hence
AE = AF = AC , CF = AC , CD = AC , AB:
Also solved by WILSON DA COSTA AREIAS, Rio de Janeiro, Brazil;
FRANCISCO BELLOT ROSADO, I.B. Emilio Ferrari, Valladolid, Spain; JORDI DOU,
Barcelona, Spain; HANS ENGELHAUPT, Franz{Ludwig{Gymnasium, Bamberg, Ger-
many; RICHARD A. GIBBS, Fort Lewis College, Durango, Colorado; RICHARD I. HESS,
Rancho Palos Verdes, California; L.J. HUT, Groningen, The Netherlands; WALTHER
JANOUS, Ursulinengymnasium, Innsbruck, Austria; ANTONIO LUIZ SANTOS, Rio de
Janeiro, Brazil; D.J. SMEENK, Zaltbommel, The Netherlands; and the proposer.
* * * * *
1501. [1990: 19] Proposed by J.T. Groenman, Arnhem, The Netherlands.
Two circles K and K touch each other externally. The equilateral triangle ABC
1
is inscribed in K , and points A ; B ; C lie on K such that AA ; BB ; CC are tangent
1 1 1 1 1 1 1
to K . Prove that one of the lengths AA ; BB ; CC equals the sum of the other two.
1 1 1 1
(The case when the circles are internally tangent was a problem of Florow in Praxis der
Mathematik 13, Heft 12, page 327.)
Solution by Marcin E. Kuczma, Warszawa, Poland.
Let O; O ; r; r be the centers and radii of K and K , and let T be the point of
1 1 1
tangency. Assume without loss of generality that T belongs to the shorter arc AB . Then
by Ptolemaeus, AT BC + BT CA = CT AB , whence
AT + BT = CT: (1)
[This is of course known. |Ed.] Produce AT to cut K again in D. The isosceles triangles
1
AOT and TO D are similar in ratio r : r , and so AD=AT = (r + r )=r. Thus
1 1 1
q s
AA = AT AD = r +r r AT;
1
1
DAG JONSSON, Uppsala, Sweden; TOSHIO SEIMIYA, Kawasaki, Japan; D.J. SMEENK,
Zaltbommel, The Netherlands; HUME SMITH, Chester, Nova Scotia; and the proposer.
Several solvers gave the above solution.
Janous located the problem in two Bulgarian sources: problem 38, page 58 of Davi-
dov, Petkov, Tonov, and Chukanov, Mathematical Contests, Soa, 1977; and problem
4.32, pages 55{56 of J. Tabov, Homothety in Problems, Soa, 1989. He also found similar
problems on pp. 172{175 of Honsberger's Mathematical Morsels, MAA, 1978.
* * * * *
1503. [1990: 19] Proposed by M.S. Klamkin, University of Alberta.
Prove that
1 + 2 cos(B + C ) cos(C + A) cos(A + B ) cos (B + C ) + cos (C + A) + cos (A + B );
2 2 2
cos =
2
q
= cos cos
(1 , cos )(1 , cos
)
2 2
R P R Q R R cos z cos x 1
On expanding out the determinant, we see that the given inequality corresponds to V 0.
2
There is equality if and only if x + y + z = 2 or one of x; y; z equals the sum of the other
two; correspondingly, if and only if A + B + C = or ABC = 0.
Also solved by SEUNG-JIN BANG, Seoul, Republic of Korea; WALTHER
JANOUS, Ursulinengymnasium, Innsbruck, Austria; MARCIN E. KUCZMA, Warszawa,
Poland; and KEE-WAI LAU, Hong Kong.
* * * * *
JACK GARFUNKEL
One of Crux's regular contributors since its earliest days, Jack Garfunkel, passed
away on New Year's Eve, 1990. His many proposals in geometry, sometimes very dicult,
but often beautiful, were well appreciated by Crux readers. He will be missed. The
following information on his life was kindly furnished by his son, Sol Garfunkel.
Jack was born in Poland in 1910 and came to the United States at the age of nine.
Although one of the top math majors at City College in New York, he left academic life
after graduation to help his family weather the depression. For the next 25 years, he
worked manufacturing candy. At the age of 45, he returned to his rst love and became a
high school mathematics teacher. Over the next 24 years, he taught at Forest Hills High
School, supervising over two dozen Westinghouse nalist and semi-nalist winners of the
talent search. When he \retired" from high school teaching, he immediately began work as
an adjunct professor at Queens College and later at Queensborough Community College,
where he taught until November of 1990.
Throughout his two careers as candyman and teacher, he continued to do mathemat-
ics, conjecturing and solving problems in synthetic geometry and geometric inequalities.
While his formal education ended with his undergraduate work, he never lost his curiosity
and love of his subject.
Crux
Mathematicorum
Volume 17 , Number 3 March 1991
CONTENTS
1
65
2. Each point of the plane (R2) is coloured by one of the two colours A and B.
Show that there exists an equilateral triangle with monochromatic vertices.
3. Determine all natural numbers N (in decimal representation) satisfying the
following properties:
(1) N = (aabb)10, where (aab)10 and (abb)10 are primes.
(2) N = P1 P2 P3, where Pk (1 k 3) is a prime consisting of k (decimal) digits.
2nd Day|June 29, 1989 (4 12 hours)
4. Let P be a convex polygon in the plane having A1, A2, : : :, An (n 3) as its
vertices. Show that there exists a circle containing the entire polygon P and having at
least three adjacent vertices of P on its boundary.
5. Let A be a vertex of a cube ! circumscribed about a sphere of radius 1. We
consider lines g through A containing at least one point of . Let P be the point of g \
having minimal distance from A. Furthermore, g \ ! is AQ. Determine the maximum
value of AP AQ and characterize the lines g yielding the maximum.
6. We consider sequences fan : n 1g of squares of natural numbers (> 0) such
that for each n the dierence an+1 , an is a prime or the square of a prime. Show that all
such sequences are nite and determine the longest sequence fan : n 1g .
66
Team Competition
June 30, 1989 (4 hours)
7. Functions f0, f1, f2, : : : are recursively dened by
(1) f0(x) = x, for x 2 R;
(2) f2k+1(x) = 3f2k (x) , where x 2 R, k = 0; 1; 2; : : : ;
(3) f2k (x) = 2f2k,1(x) , where x 2 R, k = 1; 2; 3; : : : .
Determine (with proof) the greater one of the numbers f10(1) and f9(2).
8. We are given an acute triangle ABC . For each point P of the interior or
boundary of ABC let Pa , Pb , Pc be the orthogonal projections of P to the sides a, b and
c, respectively. For such points we dene the function
f (P ) = AP c + BPa + CPb
PPa + PPb + PPc :
Show that f (P ) is constant if and only if ABC is an equilateral triangle.
9. Determine the smallest odd natural number N such that N 2 is the sum of an
odd number (> 1) of squares of adjacent natural numbers (> 0).
*
The second group of problems are from the other side of the globe, and we thank
S. C. Chan of Singapore for forwarding them to us.
SINGAPORE MATHEMATICAL SOCIETY
INTERSCHOOL MATHEMATICAL COMPETITION 1989
Part B, Saturday, 17 June 1989
1. Let n 5 be an integer. Show that n is a prime if and only if ninj 6= npnq for
every partition of n into 4 positive integers, n = n1 + n2 + n3 + n4, and for each permutation
(i; j; p; q) of (1; 2; 3; 4).
2. Given arbitrary positive numbers a, b and c, prove that at least one of the
following inequalities is false:
a(1 , b) > 14 , b(1 , c) > 14 , c(1 , a) > 14 :
3.(a) Show that v
u p
u
t 2 , 3:
tan( ) =
12
p
2+ 3
(b) Given any thirteen distinct real numbers, show that there exist at least two, say x and
y, which satisfy the inequality
v
u p
x , y u 2 ,
0 < 1 + xy < t p3 :
2+ 3
67
'$
Y
&%
A X
@
@
@
B
@
C bb
b @
b @
b
b @
b @
b
b@
b@
b
@Z
b
* * *
Before giving solutions, a remark. Seung-Jin Bang, Seoul, Republic of Korea, notes
that problem 2 [1989: 4] from the 1985{86 Flanders Mathematics Olympiad, for which
we discussed a solution on [1990: 233], had appeared in Loren C. Larson's book Problem-
Solving Through Problems, Springer-Verlag, 1983, p. 253, problem 7.2.9(b). Thanks for
pointing this out.
Having given over last month's number of the Corner to problems from the `archives',
we concentrate this month on solutions to problems given in the 1989 numbers of Crux.
First an acknowledgement. `Also solved' status should have been given solutions sent in
by Michael Selby, University of Windsor, for problems 1, 2 and 7 from the 24th Spanish
Olympiad [1989: 67{68], for which we gave solutions in January [1991: 9-10]. His solutions
arrived just as the issue was going to press.
*
When we gave the solutions, in the December 1990 number, to the 1986 Swedish
Mathematical Competition, one problem remained unanswered. An alert reader spotted
the gap and sent in a solution, which follows.
68
B0
DAA0 = DBB 0 . Thus 6 ADA0 = 6 BDB 0 and
6 A0 DB 0 = 6 ADB . Thus D is a point on the cir-
cumcircle of A0B 0 C .
6. N is a 4-digit perfect square all of whose decimal digits are less than seven.
Increasing each digit by three we obtain a perfect square again. Find N .
Solutions by Stewart Metchette, Culver City, California; John Morvay, Springeld,
Missouri; Bob Prielipp, University of Wisconsin{Oshkosh; D.J. Smeenk, Zaltbommel, The
Netherlands; and Edward T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario.
Let N = a 103 + b 102 + c 10 + d, where a, b, c and d are integers such that
1 a 6 and 0 b; c; d 6. By hypothesis N = n2 for some positive integer n. Hence
n2 = N 6666 so n 81. Also by hypothesis
(a + 3) 103 + (b + 3) 102 + (c + 3) 10 + (d + 3) = m2
for some positive integer m. Thus m2 , n2 = 3333, making (m + n)(m , n) = 3 11 101.
Because m + n > m , n and n 81, it follows that m + n = 101 and m , n = 33, so
n = 34. Therefore N = 1156.
7. Let a, b, c be the sides and , ,
be the opposite angles of a triangle. Show
that if
ab2 cos = bc2 cos = ca2 cos
then the triangle is equilateral.
71
Winning Ways I, by E.R. Berlekamp, J.H. Conway and R.K. Guy, Academic Press,
1982. (hardcover & paperback, 426 pp.)
This is the denitive treatise on mathematical games. As the subtitle \Games in
General" suggests, the general theory of mathematical games is presented in this rst
volume, but there are also plenty of specic games to be analysed, played and enjoyed.
The book is written with a great sense of humor, and is profusely illustrated, often in
bright colours.
Winning Ways II, by E.R. Berlekamp, J.H. Conway and R.K. Guy, Academic Press,
1982. (hardcover & paperback, 424 pp.)
The subtitle of this volume is \Games in Particular". Here, all kinds of mathemat-
ical games, classical as well as brand new, are presented attractively. Most of them are
two-player games. There are two chapters devoted to one-player games or solitaire puzzles,
and the book concludes with a chapter on a zero-player game!
Puzzles Old and New, by J. Slocum and J. Botermans, distributed by University of
Washington Press, 1986. (hardcover, 160 pp.)
Jerry Slocum has probably the largest puzzle collection in the world. This book
features a small subset of his mechanical puzzles, that is, puzzles made of solid pieces that
must be manipulated by hand to obtain a solution. They are classied into ten broad
categories, with enough information to make most of them and to solve some of them.
The book is full of striking full-colour plates.
The Mathematical Gardner, edited by D.A. Klarner, Wadsworth International,
1981. (hardcover, 382 pp.)
This book contains thirty articles dedicated to Martin Gardner for his sixty-fth
birthday. They re
ect part of his mathematical interest, and are classied under the
headings Games, Geometry, Two-Dimensional Tiling, Three-Dimensional Tiling, Fun and
Problems, and Numbers and Coding Theory.
Mathematical Snapshots, by H. Steinhaus, Oxford University Press, 1983. (paper-
back, 311 pp.)
This is an outstanding book on signicant mathematics presented in puzzle form.
Topics include dissection theory, the golden ratio, numeration systems, tessellations, geo-
desics, projective geometry, polyhedra, Platonic solids, mathematical cartography, spirals,
ruled surfaces, graph theory and statistics.
Mathematics Can Be Fun, by Y.I. Perelman, Mir Publishers, 1979. (hardcover, 400
pp.)
This is a translation of two books in Russian, \Figures for Fun" and \Algebra Can
Be Fun". The former is an excellent collection of simple puzzles. The latter is a general
discourse of algebra with quite a few digressions into number theory.
Fun with Maths and Physics, by Y.I. Perelman, Mir Publishers, 1984. (hardcover,
374 pp.)
The rst half of this beautiful book describes a large number of interesting experi-
ments in physics. The second half consists of a large collection of mathematical puzzles.
77
The Moscow Puzzles, by B.A. Kordemsky, Charles Scribner's Sons, 1972. (paper-
back, 309 pp.)
This is the translation of the outstanding single-volume puzzle collection in the
history of Soviet mathematics. Many of the three hundred and fty-nine problems are
presented in amusing and charming story form, often with illustrations.
The Tokyo Puzzles, by K. Fujimura, Charles Scribner's Sons, 1978. (paperback, 184
pp.)
This is the translation of one of many books from the leading puzzlist of modern-
day Japan. It contains ninety-eight problems, most of them previously unfamiliar to the
western world.
536 Puzzles and Curious Problems, by H.E. Dudeney, Charles Scribner's Sons, 1967.
(paperback, 428 pp.)
This book is a combination of two out-of-print works of the author, \Modern Puz-
zles" and \Puzzles and Curious Problems". Together with Dover's \Amusements in Mathe-
matics", they constitute a substantial portion of Dudeney's mathematical problems. Those
in this book are classied under three broad headings, arithmetic and algebra, geometry,
and combinatorics and topology.
Science Fiction Puzzle Tales, by Martin Gardner, Clarkson N. Potter, 1981. (pa-
perback, 148 pp.)
This is the rst of three collections of Martin Gardner's contribution to Isaac Asi-
mov's Science Fiction Magazine. The book contains thirty-six mathematical puzzles in
science ction settings. When solutions are presented, related questions are often raised.
Puzzles From Other Worlds, by Martin Gardner, Vintage Books, 1984. (paperback,
189 pp.)
This is the sequel to \Science Fiction Tales" and the predecessor of New Mathe-
matical Library's \Riddles of the Sphinx". It contains thirty-seven mathematical puzzles
plus further questions raised in the answer sections.
* * * * *
PROBLEMS
Problem proposals and solutions should be sent to B. Sands, Department of Mathe-
matics and Statistics, University of Calgary, Calgary, Alberta, Canada T2N 1N4. Propos-
als should, whenever possible, be accompanied by a solution, references, and other insights
which are likely to be of help to the editor. An asterisk (*) after a number indicates a
problem submitted without a solution.
Original problems are particularly sought. But other interesting problems may also
be acceptable provided they are not too well known and references are given as to their
provenance. Ordinarily, if the originator of a problem can be located, it should not be
submitted by somebody else without permission.
78
X
X X X X
X X X X
XX XX XX XX XX XX XX XX XX
1 2 3 4 5 6 7 8 9
X X X
X X X
X X X
XX XX XX X X X X X X X X X X X X
10 11 12 13 14 15 16 17 18
X X X X X X
X X
X
X X X X X X X X X X X X X X X X X X
19 20 21 22 23 24 25 26 27
X X
X X X X
XX X X X XX X X X X X
X X X X X X X X X
28 29 30 31 32 33 34 35 36
X X X X
X X XX X
X X X X X X X
X X X X X X X XX XX
37 38 39 40 41 42 43 44 45
X
X X X X
XX X X X X X X XX
XX XX XX X X X X X X
46 47 48 49 50 51 52 53 54
X
X X X X X X X
X X X X X X X X X X
X X X X X X X X X
55 56 57 58 59 60 61 62 63
X X X
X X XX X X X
X X X X X XX
X X X X X X X
64 65 66 67 68 69 70 71
81
For six other cases, X has a winning strategy, but he cannot dictate O's moves. The
positions in the following table are given by coordinates as follows:
11 12 13 14
21 22 23 24
31 32 33 34
41 42 43 44
Also given is X's rst move.
If O takes X takes
33,34,43 or 44 11,12 and 14
X1 11,12,21 or 22 14,34 and 44
XX 13,14,23 or 24 11,22 and 33
1 (16)
11,13,31,33,34,43 or 44 23 and 22
X X 12,21 or 22 14 and 44
X or X
14 44,22,21 and 11
X1 1X
30 (16) 58 (32) 23 44,34,33 and 22
12,13,31,42 or 43 14,23 and 22
X 14,32 or 34 22,23 and 43
X
X 1 11,21 or 22 14,23 and 43
34 (16) 23 34,31 and 11
13,14,23,24 or 34 42,22 and 11
32 43,11,33 and 13
33 11,32 and then either 14 or 42
X X 11 or 21 ( 42,32,34(and 14
1 or X 42 or 43 then
else
11,13 or 21 14 and 34 and 32
X
27 (16)
X 1
41 (16)
X 8 8 11
>
< then 21 >
< 23,13 and 43
22 > then 11 or 13 14 and > 32 and 34
: else : 11
This leaves eight classes:
X X X X X
X X X
X X X X
XX XX X X XX XX X X
10 (32) 11 (32) 39 (16) 40 (16) 44 (32) 48 (8) 53 (8) 66 (8)
Having been unable to nd a winning strategy for any of them and having convinced myself
that there were none, I put the problem to my Atari 1040ST computer. After about 5
hours, it agreed that O can always force a tie in these situations. Therefore X has a winning
strategy in 368 of the 520 starting positions.
* * * * *
83
and
2n,1
1,i n,1 3
d [x,k=m g (x)] = k k
m(m , 1) ( mk
, (n , 1)) d 4 X ( ,1) n ,
i xi,k=m5
n
dx (n , 1)! dx i=0 m ,i
k
k ( k , 1) ( k , (n , 1)) k +1) nX,1 n , 1!
= m m m (,1)x m
, ( (,1)n,1,ixi
(n , 1)! i=0 i
84
"
= x,( mk +1)
, mk ( mk , 1) ( mk , (n , 1)) (x , 1)n,1# :
(n , 1)!
Therefore
d hx,k=m (g (x) , f (x))i = 0;
n n
dx
so
x,k=m(gn (x) , fn(x)) = constant.
If either k=m is not an integer, or is an integer n, then since gn , fn is a polynomial (of
degree at most n , 1) the above constant must be 0, i.e. gn fn.
* * * * *
1502. [1990: 19] Proposed by D.J. Smeenk, Zaltbommel, The Netherlands.
AB is a chord, not a diameter, of a circle with centre O. The smaller arc AB is
divided into three equal arcs AC; CD; DB . Chord AB is also divided into three equal
segments AC 0; C 0D0; D0 B . Let CC 0 and DD0 intersect in P . Show that 6 APB = 31 6 AOB:
Solution by Mark Kisin, student, Monash University, Clayton, Australia.
Let P
A0 = PA \ CD, B 0 = PB \ CD: EA
EA
E A
Then E A
PAC 0 PA0 C
E A
E A
E A
and
E A
PC 0D0 PCD; O E
E
A
A
so that A
C
0 0E BA
AC 0 = PC 0 = C 0D0 :
D E A
E A
A0 C PC CD A0
EE AA
B0
But AC 0 = C 0D0, so A0C = CD = AC . Therefore C D
6 BAC = 6 ACA0 = , 6 CAA0 , 6 CA0A = , 26 CA0A
= , 6 CA0A , 6 BB 0 D = 6 APB:
But C and D trisect the arc AB , so
6 APB = 6 BAC = 21 6 BOC = 12 32 6 AOB = 13 6 AOB:
Also solved by WILSON DA COSTA AREIAS, Rio de Janeiro, Brazil;
FRANCISCO BELLOT ROSADO, I. B. Emilio Ferrari, and MARIA ASCENSION
LOPEZ CHAMORRO, I. B. Leopoldo Cano, Valladolid, Spain; JORDI DOU, Barcelona,
Spain; HANS ENGELHAUPT, Franz-Ludwig-Gymnasium, Bamberg, Germany;
C. FESTRAETS-HAMOIR, Brussels, Belgium; RICHARD I. HESS, Rancho Palos
Verdes, California; JEFF HIGHAM, student, University of Toronto; L. J. HUT,
85
* * * * *
1505. [1990: 19] Proposed by Marcin E. Kuczma, Warszawa, Poland.
Let x1 = 1 and q
xn+1 = x1 1 + xn , 1 :
2
n
Show that the sequence (2nxn) converges and nd its limit.
Solution by Je Higham, student, University of Toronto.
We rst show by induction on n that
xn = tan(=2n+1) (1)
for all n 2 N. Since x1 = 1 = tan(=4), (1) is true for n = 1. Suppose (1) is true for
n = k. Then
1 q
xk+1 = tan(=2k+1) 1 + tan (=2 ) , 1
2 k +1
Now, since p
cosh,1 u = ln(u u2 , 1);
88
we nd p p!
5 = ln 4 2 42 = 12 ln 2;
3
of the problem has the interesting property that its two local maxima have equal y-values
and its two local minima also have equal y-values. According to expert colleague Len Bos,
this is true because the above polynomial is actually
p p
4 8 T5(x= 8) , 12;
where T5 is the fth Chebyshev polynomial, which is well known to have this property.
* * * * *
1508. [1990: 20] Proposed by Edward T.H. Wang, Wilfrid Laurier University,
Waterloo, Ontario.
Let a b < c be the lengths of the sides of a right triangle. Find the largest
constant K such that
a2 (b + c) + b2(c + a) + c2(a + b) Kabc
holds for all right triangles and determine when equality holds. It is known that the
inequality holds when K = 6 (problem 351 of the College Math. Journal ; solution on p.
259 of Volume 20, 1989).
Solution by T. Leinster, Lansing College, England.
Let be the angle opposite the side with length a; then 0 < =4. Now
a2 (b + c) + b2(c + a) + c2(a + b) = a + a + b + b + c + c
abc c b a c b a
= sin + tan + cot + cos + sec + csc
= f (), say.
Then
f 0() = cos + sec2 , csc2 , sin + sec tan , csc cot ;
so putting s = sin , c = cos , f 0 () = 0 implies
0 = s2c3 + s2 , c2 , s3c2 + s3 , c3
= (s , c)(s + c + s2 + sc + c2 , s2c2 )
= (s , c)(s + c + s2 + sc + c4):
Since each component of the second term is 0 for 0 < =4, and they cannot all be 0
at once, the only turning point of f () in this range is where s , c = 0, i.e., = =4. f ()
is continuous over (0; =4], and f () ! +1 as ! 0 from above, therefore = =4 gives
the lowest possible value for f () in this range. So
p
Kmax = f (=4) = 2 + 3 2;
and equality holds for the isosceles right-angled triangle.
Also solved by HANS ENGELHAUPT, Franz-Ludwig-Gymnasium, Bamberg,
Germany; MORDECHAI FALKOWITZ, Tel-Aviv, Israel; C. FESTRAETS-HAMOIR,
90
Brussels, Belgium; RICHARD I. HESS, Rancho Palos Verdes, California; JEFF HIGHAM,
student, University of Toronto; WALTHER JANOUS, Ursulinengymnasium, Innsbruck,
Austria; FRIEND H. KIERSTEAD JR., Cuyahoga Falls, Ohio; MURRAY S. KLAMKIN,
University of Alberta; MARCIN E. KUCZMA, Warszawa, Poland; KEE-WAI LAU, Hong
Kong; J. A. MCCALLUM, Medicine Hat, Alberta; and the proposer.
For a related problem, see Crux 1227 [1988: 148].
* * * * *
1509. [1990: 20] Proposed by Carl Friedrich Sutter, Viking, Alberta.
Professor Chalkdust teaches two sections of a mathematics course, with the same
material taught in both sections. Section 1 runs on Mondays, Wednesdays, and Fridays
for 1 hour each day, and Section 2 runs on Tuesdays and Thursdays for 1.5 hours each
day. Normally Professor Chalkdust covers one unit of material per hour, but if she is
teaching some material for the second time she teaches twice as fast. The course began
on a Monday. In the long run (i.e. after N weeks as N ! 1) will one section be taught
more material than the other? If so, which one, and how much more?
Solution by Je Higham, student, University of Toronto.
Let x be the number of units of material covered by Section 1 (from the start of
the course) minus the number of units of material covered by Section 2. We will show by
induction that, in week n,
8
>
> (7 , 25,4n)=5 after Monday,
>
< (,4 ,32,44,n4n)=5 after Tuesday,
x = > (3 , 2 )=5 after Wednesday, (1)
>
> ( , 6 , 2 2,4n)=5 after Thursday,
: (2 , 21,4n)=5 after Friday.
(1) is easily veried when n = 1. Suppose (1) is true for n = k. Then on Monday
of week k + 1, Section 1 covers only \new material" (material not yet covered by the other
section), since x = (2 , 21,4k )=5 > 0 on the previous Friday by the induction hypothesis.
Thus, after this Monday,
x = 2 , 52 + 1 = 7 , 52 = 7 , 2 5
1,4k 1,4k 5,4(k+1)
:
On Tuesday of week k + 1, Section 2 covers the (7 , 25,4(k+1))=5 units of \old material"
(material already covered by the other section) in (7 , 25,4(k+1))=10 < 1:5 hours. Thus,
for the remaining
3 , 7 , 25,4(k+1) = 8 + 25,4(k+1) = 4 + 24,4(k+1)
2 10 10 10
hours, (4 + 24,4(k+1))=5 units of new material is covered, so x = (,4 , 24,4(k+1))=5 after
this Tuesday. On Wednesday of week k +1, old material is covered in (4+24,4(k+1))=10 < 1
hours and new material in
= 6 , 210 = 3,25
4,4(k+1) 4,4(k+1) 3,4(k+1)
1 , 4 + 210
91
= 4 , 210 = 2,25
2,4(k+1) 2,4(k+1) 1,4(k+1)
1 , 6 + 210
hours, so x = (2 , 21,4(k+1))=5 after this Friday. This completes the induction argument.
Clearly as n ! 1, x ! 2=5 after Friday; therefore in the long run Section 1 is
ahead of Section 2 by 2=5 units after Friday.
Also solved by DUANE M. BROLINE, Eastern Illinois University, Charleston;
CURTIS COOPER, Central Missouri State University, Warrensburg; RICHARD I. HESS,
Rancho Palos Verdes, California; MARCIN E. KUCZMA, Warszawa, Poland; HUME
SMITH, student, University of British Columbia; and the proposer. One incorrect solution
was received which was apparently due to the writer not understanding the problem.
* * * * *
1510. [1990: 20] Proposed by Jack Garfunkel, Flushing, New York.
P is any point inside a triangle ABC . Lines PA; PB; PC are drawn and angles
PAC; PBA; PCB are denoted by ; ;
respectively. Prove or disprove that
cot + cot + cot
cot A + cot B + cot C ;
2 2 2
with equality when P is the incenter of ABC:
Solution by Marcin E. Kuczma, Warszawa, Poland.
I choose to disprove.
Lines AP , BP , CP cut the boundary A
of triangle ABC in K , L, M . Let U , V be @
L
L @
the feet of perpendiculars from K to lines AB MQ L @
and CA, respectively. Then Q L @ L
Q L #@
QL# @
cot , cot A = sin( A , )
#P Q
L Q
# L Q
@
sin A sin UH ##
L Q
@
@V
= 1 KU
H
#HH L
Q
Q,@
Q
sin A KV ## HH L , Q@
= 1 BK sin B :
HH L , Q@
B
#
#
HHL ,
Q@@
QC
sin A KC sin C K
Hence (in cyclic notation)
92
X X BK sin B
(cot , cot A) = sin C sin A
Y KC
3 BK sin B 1=3
KC sin C sin A
= 3(sin A sin B sin C ),1=3; (1)
by the means inequality and Ceva's theorem. Equality holds if and only if each one of the
summands in (1) equals (sin A sin B sin C ),1=3, i.e. (writing a, b, c for the side lengths),
when ! !
BK = ca 2=3 ; CL = ab 2=3 ; AM = bc 2=3 : (2)
KC b2 LA c2 MB a2
In every triangle ABC there exists a unique point P for which (2) holds ( pick K ,
L, M to partition BC , CA, AB in ratios as in (2);Plines AK , BL, CM are concurrent by
the inverse Ceva theorem). For this point the sum cot attains minimum, equal by (1)
to X
cot A + 3(sin A sin B sin C ),1=3:
When P is the incenter, the ratios BK=KC , etc. are c=b, a=c, b=a, hence they dier from
those of (2), unless the triangle is regular.
For instance, in the isosceles right triangle of vertices A = (1; 0), B = (0; 1), C =
(0; 0), the optimal P = (x; y) has x = 21=3 , 1 = 0:259 : : :, y = 22=3 , 21=3 = 0:327 : : : and
gives the minimum value
cot = 2 + 3 21=3 = 5:779 : : : ;
compare with the incenter I = (u; v), u = v = 1 , 2,1=2 = 0:292 : : :, producing
cot(A=2) = 3 + 23=2 = 5:828 : : ::
Also solved by G. P. HENDERSON, Campbellcroft, Ontario.
* * * * *
1511. [1990: 43] Proposed by J.T. Groenman, Arnhem, The Netherlands.
Evaluate
Y
n
nlim
!1 (1 , tan4 2k ):
k=3
Solution by Beatriz Margolis, Paris, France.
Observe that
1 , tan4 (=2k ) = (1 , tan2(=2k ))(1 + tan2 (=2k ))
cos2 (=2k ) , sin2 (=2k ) 1
=
cos (=2 )
2 k cos (=2k )
2
k,1 ) k) !3
= cos( = 2 2 sin( = 2
cos(=2k ) 2 sin(=2k ) cos(=2k )
cos( = 2 k,1 ) 2 sin( = 2 k ) !3
= cos(=2k ) sin(=2k,1) :
93
Hence
Y
n Yn cos(=2k,1) Yn 2 sin(=2k ) !3
(1 , tan4(=2k )) = k
k=3 cos(=2 ) k=3 sin(=2 )
k,1
k=3
cos( = 2 2) sin( = 2 n) !3
=
cos(=2n)
2 sin(=22)
n, 2
= An Bn3 :
Now
!1 An
nlim = cos 4 = p1
2
and p
lim B = 1 lim sin(=2n) = 2 ;
n!1 n sin(=4) 4 n!1 =2n 4
so that 1 p2 !3 3
n
Y
!1 k=3 1 , tan 2k = p2
4 :
nlim 4
=2
4
Also solved by SEUNG-JIN BANG, Seoul, Republic of Korea; CURTIS COOPER,
Central Missouri State University, Warrensburg; C. FESTRAETS-HAMOIR, Brussels,
Belgium; WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria; MURRAY
S. KLAMKIN, University of Alberta; MARCIN E. KUCZMA, Warszawa, Poland;
T. LEINSTER, Lansing College, England; KEE-WAI LAU, Hong Kong; P. PENNING,
Delft, The Netherlands; EDWARD T. H. WANG, Wilfrid Laurier University, Waterloo,
Ontario; CHRIS WILDHAGEN, Breda, The Netherlands; KENNETH S. WILLIAMS,
Carleton University, Ottawa, Ontario; and the proposer. One other reader sent in an
approximation.
Janous, Klamkin, Penning, and the proposer in fact prove the more general result
Y
n
(1 , tan4 k ) = x cos x;
3
nlim
!1 k=1 2 sin x
3
which with x = =4 becomes the given problem, and which can be shown as above.
* * * * *
1512. [1990: 43] Proposed by Walther Janous, Ursulinengymnasium, Innsbruck,
Austria.
Given r > 0, determine a constant C = C (r) such that
(1 + z )r (1 + z r ) C (1 + z 2 )r
for all z > 0.
94
or
, xx2 +, 22xx ,, 11 = xr,1;
2
95
* * * * *
Crux
Mathematicorum
Volume 17 , Number 4 April 1991
CONTENTS
1
97
Pi Pj : Pj Pk = Qi Qj : Qj Qk
for all i 6= j 6= k, and suppose Q = Pn0 is the point of intersection of p and q, where
0
n 6= 0. Then our rst theorem was the subject of an earlier article [2] in Crux :
0
Proof. Let t be one position of the line t. Since a conic is uniquely determined if
4
we are given ve tangents, and the line at innity touches a parabola, there is a unique
parabola P which touches t ; t ; t and t . Let P ; P and P be the respective points
1 2 3 4 1 2 3
of contact of t ; t and t with the parabola P . Then if t is any tangent to P , and the
1 2 3
respective intersections of t with t ; t and t are R ; R and R , by [1], p. 330, Theorem
1 2 3 1 2 3
II,
fQ Q ; Q 1g = fP P ; P 1g = fR R ; R 1g;
1 3 2 1 3 2 1 3 2
where the rst and third cross-ratios are taken on their respective lines, but the sec-
ond cross-ratio is taken on the parabola P : that is, it is the cross-ratio of the pencil
98
Minneapolis, Minnesota
* * * * *
99
lation gives
! ! ! !
1000 (0:2) + 1000 (0:2) + 1000 (0:2) + + 1000 (0:2) = A + A + + A ;
0 1 2 1000
0 1 2 1000 0 1 1000
where Ak = k (0:2)k for k = 0; 1; 2; : : : ; 1000. For which k is Ak the largest?
1000
the product of the resulting numerator and denominator. For how many rational numbers
between 0 and 1 will 20! be the resulting product?
6. Suppose r is a real number for which
19 20 21 91
r + 100 + r + 100 + r + 100 + + r + 100 = 546:
100
Find b100rc. (For real x; bxc is the greatest integer less than or equal to x.)
7. Find A , where A is the sum of the absolute values of all roots of the following
2
equation:
p 91
x = 19 + p
19 + p p 91 91
19+
91
19+ p19+ 91
x
bbb is transmitted. Let p be the probability that Sa comes before Sb in alphabetical order.
a circle C of radius 1 in such a way that the twelve
disks cover C , no two of the disks overlap, and so
that each of the twelve disks is tangent to its two
neighbors. The resulting arrangement of disks is
shown in the gure to the right. The sum of the
areas of theptwelve disks can be written in the
form (a , b c), where a; b; c are positive integers
and c is not divisible by the square of any prime.
Find a + b + c.
12. Rhombus PQRS is inscribed in rectangle ABCD so that vertices P; Q; R;
and S are interior points on sides AB; BC; CD; and DA, respectively. It is given that
PB = 15, BQ = 20, PR = 30, and QS = 40. Let m=n, in lowest terms, denote the
perimeter of ABCD. Find m + n.
13. A drawer contains a mixture of red socks and blue socks, at most 1991 in all.
It so happens that, when two socks are selected randomly without replacement, there is a
probability of exactly 1/2 that both are red or both are blue. What is the largest possible
number of red socks in the drawer that is consistent with this data?
14. A hexagon is inscribed in a circle. Five of the sides have length 81 and the
sixth, denoted by AB , has length 31. Find the sum of the lengths of the three diagonals
that can be drawn from A.
101
k=1
where a ; a ; : : : ; an are positive real numbers whose sum is 17. There is a unique positive
1 2
integer n for which Sn is also an integer. Find this n.
*
This month we also give the 1990 Australian Olympiad. I particularly want to
thank Andy Liu, University of Alberta, for having collected these, and other problem sets
we shall use, while he was at the IMO last summer.
3. Let ABC be a triangle and k be a circle through the points A and C such that
1
k intersects AB and BC a second time in the points K and N respectively, K and N
1
being dierent. Let O be the centre of k . Let k be the circumcircle of the triangle KBN ,
1 2
and let the circumcircle of the triangle ABC intersect k also in M , a point dierent from
2
B . Prove that OM and MB are perpendicular.
4. A solitaire game is played with an even number of discs, each coloured red on
one side and green on the other side. Each disc is also numbered, and there are two of
each number; i.e. f1; 1; 2; 2; 3; 3; : : : ; N; N g are the labels. The discs are laid out in rows
with each row having at least three discs. A move in this game consists of
ipping over
simultaneously two discs with the same label. Prove that for every initial deal or layout
there is a sequence of moves that ends with a position in which no row has only red or
only green sides showing.
102
R 2
6. Up until now the National Library of the small city state of Sepharia has had n
shelves, each shelf carrying at least one book. The library recently bought k new shelves,
k being positive. The books will be rearranged, and the librarian has announced that each
of the now n + k shelves will contain at least one book. Call a book privileged if the
shelf on which it will stand in the new arrangement is to carry fewer books than the shelf
which has carried it so far. Prove: there are at least k + 1 privileged books in the National
Library of Sepharia.
7. For each positive integer n, let d(n) be the number of distinct positive integers
that divide n. Determine all positive integers for which d(n) = n=3 holds.
8. Let n be a positive integer. Prove that
1n , 1n + 1n , + (,1)n + + 1n = 1 :
k, 1
2
1
2
2
2
3
2
k
n+1 2
2
n,1
* * * * *
Before turning to solutions submitted by the readers, I want to give two comments
received about recent numbers of the Corner.
4. [1985: 304; 1991: 41] Proposed by Canada.
Prove that
x + x
2
1
+ + xn n , 1
2
2
2
x +x x x +x x
2
1 2 3
2
2xn + x x
3 4
2
1 2
Editor's note. In the October 1990 issue of Crux I discussed a solution to this
problem, saying it had not been considered in the Corner. Edward T.H. Wang, Wilfrid
Laurier University, Waterloo, Ontario, takes me to task for not remembering what I have
published in earlier numbers! He points out that this problem was also given as number
1 of the 1984 Brazilian Mathematical Olympiad [1987: 70] and that in [1988: 231] I had
given George Evagelopoulos's observation that it was also problem 4 of the 1983 Australian
Olympiad, and its solution is in [1986: 23]. So the same problem was used at least three
times in three dierent Mathematical Olympiads, and in three consecutive years! Wang
asks if this is a record.
* * * * *
We next give some solutions to \archive problems". I hope they haven't been
discussed before!
2. [1981: 47] 1978 Romanian Mathematical Olympiad, Final Round (12th class).
Let P and Q be two polynomials (neither identically zero) with complex coecients.
Show that P and Q have the same roots (with the same multiplicities) if and only if the
function f : C ! R dened by f (z ) = jP (z )j , jQ(z )j has a constant sign for all z 2 C if
it is not identically zero.
Solution by Murray S. Klamkin, University of Alberta.
The only if part is easy.
For the if part, we can assume without loss of generality that f (z ) 0. If r is any
root of P , it immediately follows that it must also be a root of Q (note if P is a constant,
then so also Q is a constant). Also the multiplicity of any root of P must be at most the
corresponding multiplicity in Q. For if a root r had greater multiplicity in P than in Q,
by setting z = r + ", where " is arbitrarily small, we would have f (z ) < 0. Next the degree
of P must be at least the degree of Q. Otherwise by taking jz j arbitrarily large, we would
have f (z ) < 0. It follows that P and Q have the same roots with the same multiplicities.
*
4. [1982: 301] U.S. Olympiad Student Proposals.
Find all solutions (x; y; z ) of the Diophantine equation
x + y + z + 6xyz = 0:
3 3 3
References:
[1] L.J. Mordell, Diophantine Equations, Academic Press, New York, 1969, p. 78.
[2] L.E. Dickson, History of the Theory of Numbers, Vol. II, Stechert, New York, 1939,
p. 589.
7. [1982: 301] U.S. Olympiad Student Proposals.
In R let X = (x ; x ; : : : ; xn ), Y = (y ; y ; : : : ; yn), and, for p 2 (0; 1), dene
n 1 2 1 2
0 p ,p p ,p p ,p1
x 1 y 1
1
x 2 y 2
1
x y
Fp(X; Y) @ p 1 , p ; p 1 , p ; : : : ; pn 1 ,n p A :
1
Prove that
kXkm + kYkm kFp(X; Y)km ;
where
kXkm = (jx jm + jx jm + + jxnjm) =m:
1 2
1
, p) ,p 1
m m , p) ,p 1
*
3. [1983: 107] 1983 Bulgarian Winter Competition.
Determine all values of the real parameter p for which the system of equations
x+y+z =2
yz + zx + xy = 1
xyz = p
has a real solution.
105
3 3
AG = 2GH , we have B C
AG = DE : H I
AD DO
Obviously, DE is perpendicular to AH , so that 6 ODE = 90 , 6 ADE = 6 DAG. From
this, ADG DOE . Since the angle between AD and DO is 90, the angle between
DG and EO must be 90, too. Thus OE is perpendicular to CD.
106
Sn is the sum and Pn is the product of the rst n terms of the sequence y ; y ; y ; : : :, where1 2 3
yn = 1 +1 x :
n
Prove that aSn + Pn = 1 for all n.
Solution by Murray S. Klamkin, University of Alberta.
It follows easily that
1 = 1 , 1 + 1; (1)
2
yn +1 yn yn
where y = 1=(1 + a). Now let
1
yn , 1 'n, , ayn 1
yn , 1 yn
Hence 'n , ayn = yn +1 +1
'n, , ayn yn 1
or
'nyn = 'n, yn : 1 +1 (3)
An easy calculation shows that y = 1=(1 + a + a ), so that
2
2
1
' = P (y , 1) + ay = 1 + a 1 + a + a , 1 + 1 + aa+ a
1 1 2
1
2 2 2
= 1 + a(1 + a) , 1 = 0:
(1 + a)(1 + a + a ) 1 + a 2
Then since yi 6= 0 for all i, from (3) 'i = 0 for all i. Since aS + P = 1, by (2) aSn + Pn = 1
1 1
for all n.
*
107
holds for all n (where n is the number of logarithm functions in the denominator)?
Solution by Murray S. Klamkin, University of Alberta
The answer to the given problem is in the armative and it is a special case of a
result of du Bois-Reymond [1].
First, if f (x)=g(x) ! 1 as x ! 1, we say that the order of f is greater than the
order of g and we write it as f g. The theorem of du Bois-Reymond is that given a scale
of increasing functions 'n such that
' ' ' : : : 1;
1 2 3
then there exists an increasing function f such that 'n f 1 for all values of n. Here
we choose ' = ln x; ' = ln ln x; ' = ln ln ln x, etc.
1 2 3
More generally we have the following: given a descending sequence f'ng : ' 1
' ' 'n ' and an ascending sequence f n g :
2 3 1 2 3
p such that p 'n for all n and p then there is f such that p f 'n for
all n and p.
References:
[1] P. du Bois-Reymond, U ber asymptotische Werthe, innitare Approximationen und
innitare Au
osung von Gleichungen, Math. Annalen 8 (1875), p. 365.
*
1. [1984: 74] West Point Proposals.
Given six segments S ; S ; : : : ; S congruent to the edges AB; AC; AD; CD; DB; BC;
1 2 6
respectively, of a tetrahedron ABCD, show how to construct with straightedge and com-
pass a segment whose length equals that of the bialtitude of the tetrahedron relative to
opposite edges AB and CD (i.e., the distance between the lines AB and CD).
Solution by Murray S. Klamkin, University of Alberta.
We use the known construction for an altitude of a tetrahedron [1] and the known
theorem [2] that the volume of a tetrahedron equals one-sixth the product of two opposite
edges times the sine of the angle between those edges and times the shortest distance
between those edges. The volume also equals one-third the product of an altitude and the
area of the corresponding face.
Let s = jS j, etc. Let be the angle between edges AB and CD and let d be the
1 1
distance between AB and CD. Also let hD be the altitude of the tetrahedron from D and
h0 be the altitude of triangle ABC from C . Then six times the volume of the tetrahedron
equals
s s d sin = 2hD [ABC ] = hD s s sin 6 CAB = hD s h0 :
4 1 1 2 1 (1)
108
Here [ABC ] denotes the area of triangle ABC . To express sin as a ratio of two con-
structible segments, we have
s s cos = j(A , B) Cj = jA C , B Cj = js s cos 6 ADC , s s cos 6 BDC j; (2)
4 1 3 4 5 4
We nish this month's Corner with solutions to four of the ve problems of the
1989 Asian Pacic Mathematical Olympiad [1989: 131]. We're still waiting for a correct
proof of number 2.
1. Let x ; x ; : : : ; xn be positive real numbers, and let S = x + x + + xn. Prove
1 2 1 2
that
(1 + x )(1 + x ) (1 + xn) 1 + S + 2! S + S + + S :
1 2
3! n!
Correction and solution by George Evagelopoulos, Athens, Greece, and by Edward
T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario.
The correct inequality should be
n
(1 + x )(1 + x ) (1 + xn) 1 + S + S2! + + Sn! :
2
1 2
This is well known and can be found, for example, in Analytic Inequalities by D.S. Mitri-
novic (xx3.2.42). Now by the AM-GM inequality
n + x + : : : + xn n S n
(1 + x ) (1 + xn) 1
n = 1+ n 1
S n(n , 1) S S n 2
= 1+n n + 2 n + + n
using the binomial theorem. Since (n , m)!nm n!, the coecient of S m is
!
n 1 = n! n! = 1 ;
m n m m!(n , m)!n m m!n! m!
from which the result is immediate.
3. Let A ; A ; A be three points in the plane, and for convenience, let A = A ,
1 2 3 4 1
A = A . For n = 1; 2 and 3 suppose that Bn is the midpoint of AnAn , and suppose
5 2 +1
that Cn is the midpoint of AnBn . Suppose that An Cn and BnAn meet at Dn, and that +1 +2
AnBn and CnAn meet at En. Calculate the ratio of the area of triangle D D D to
+1 +2 1 2 3
the area of triangle E E E . 1 2 3
1 1 1 L @ C2
B X = 3A B ; B X = 3A B ; B X = 3A B :
1 3 1 2 1 2
LL@@
3 2 3
E2 L ,@B2
Using perspectives from A the cross-ratios
(A ; A ; B ; C ) and (B ; A ; X; D ) are equal. There- D1 X ,L,DL 2E3 @@
1
2 3 2 2 1 3
B1 , L
1
@A3
fore , L
LC3
AB CA =BXDA 2 2 2 3
1
, , D 1
3
,E,1 B3
3
B A A C XA B D 2 3 2 2 3 1 1
from which
1 D A
C1
,
13= 2 B D : , 1 3
1 1
A1
,
110
D X = B X , B D = , = 4:
1 1 1 1
1
3
1
7
BX 1 BX 7 1
1
3
[D D X ] = 4 = 16 ; 1 2
2
[B B X ] 7 49 1 2
[B B X ] [B B X ] 49
2 3 1 3
It follows that
[D D D ] = 16 [B B B ] = 4 [A A A ];
49 49
1 2 3 1 2 3 1 2 3
(A ; A ; B ; C ) = (B ; A ; X; E )
1 3 3 3 1 3 3
and so
1 1 = A B C A = B X E A = 1 E A : 1 3 3 3 1 3 3 3 3
3 B A A C XA B E 2 B E 3 3 1 3 3 1 3 1 3
2 2 4BA:
XE = 3 , 5 B A = 15 3 1 3 1 3
[D D D ] = 25 [E E E ]:
49
1 2 3 1 2 3
4. Let S be a set consisting of m pairs (a; b) of positive integers with the property
that 1 a < b n. Show that there are at least
4m m , n =4
2
3n
triples (a; b; c) such that (a; b), (a; c) and (b; c) belong to S .
Solution by George Evagelopoulos, Athens, Greece.
Draw a graph whose vertices are the integers 1; 2; : : : ; n with an edge between x and
y if and only if either (x; y) or (y; x) belongs to S . Those 3-element subsets of S in which
111
any two elements are connected by an edge will be called \triangles" and the number of
edges to which a vertex x belongs will be denoted by d(x). 1
For example, if S = f(1; 2); (1; 3); (2; 3); (2; 4); (3; 4)g where @
n = 4, the subsets f(1; 2); (1; 3); (2; 3)g and f(2; 3); (2; 4); (3; 4)g form 4 @@
triangles as illustrated. Also d(2) = 3, for example. The problem is @ ,2
to determine a lower bound for the number of triangles in the general @@ ,
case. @3,
Let x be joined to y by an edge. Then d(x) + d(y) , 2 edges are attached to the
remaining n , 2 vertices and therefore at least d(x)+ d(y) , 2 , (n , 2) vertices are attached
to both x and y. Hence at least d(x) + d(y) , n triangles contain both x and y. It follows
that the total number of triangles is at least
X d(x) + d(y) , n
3
;
( x;y 2S
)
(x;y)2S x=1
because each d(x) occurs exactly d(x) times in the sum over the elements of S . Therefore,
by Chebyshev's inequality, we get
X d(x) + d(y) , n 1 X n !
3
=
3 x
d(x) , nm
2
x;y 2S
( )
0 n
=1
! 1
1 1 X 2
3 @n d(x) , nmA
x
=1
= 4m(m , n =4)
2
3n
P
because nx d(x) = 2m.
=1
5. Determine all functions f from the reals to the reals for which
(i) f (x) is strictly increasing,
(ii) f (x) + g(x) = 2x for all real x where g(x) is the composition inverse function to
f (x). (Note: f and g are said to be composition inverses if f (g(x)) = x and g(f (x)) = x
for all real x.)
Solution by George Evagelopoulos, Athens, Greece (with an assist by the editors).
We will prove that f (x) = x + d for some constant d, i.e., f (x) , x is a constant
function.
For each real d, denote by Sd the set of all x for which f (x) , x = d. Then we must
show that exactly one Sd is nonempty. First we prove two lemmas.
112
Lemma 1. If x 2 Sd then x + d 2 Sd .
Proof. Suppose x 2 Sd . Then f (x) = x + d, so g(x + d) = x, and f (x + d)+ g(x + d) =
2x + 2d. Therefore f (x + d) = x + 2d and x + d 2 Sd : 2
Lemma 2. If x 2 Sd and y x then y 62 Sd0 for any d0 < d.
Proof. First let y satisfy x y < x + (d , d0). Note that by monotonicity f (y)
f (x) = x + d. Hence y 2 Sd0 would imply y + d0 x + d or y x +(d , d0), a contradiction.
Thus y 62 Sd0 in this case. Now by induction it follows that for all x 2 Sd ,
if x + (k , 1)(d , d0 ) y < x + k(d , d0 ) then y 62 Sd0 :
The base case k = 1 is proved above. Assume the statement holds for some k and let
x + k(d , d0) y < x + (k + 1)(d , d0):
Then
x + d + (k , 1)(d , d0) y + d0 < x + d + k(d , d0):
But x + d 2 Sd , and so by the induction hypothesis y + d0 62 Sd0 . The lemma follows. 2
Now suppose that two Sd 's are nonempty, say Sd and Sd0 where d0 < d. If 0 < d0 ,
then Sd0 must contain arbitrarily large numbers by Lemma 1. But this is impossible by
Lemma 2.
Editor's note. The above, slightly rewritten, is Evagelopoulos's argument, except
he hadn't noted that his argument required 0 < d0. We now complete the proof.
Lemma 3. If Sd and Sd0 are nonempty and d0 < d00 < d then Sd00 is also nonempty.
Proof. Since Sd and Sd0 are nonempty, there are x and x0 so that f (x) , x = d and
f (x0 ) , x0 = d0. Since f is increasing and has an inverse, it is continuous, so the function
f (x) , x is also continuous. Thus by the Intermediate Value Theorem there is x00 so that
f (x00 ) , x00 = d00, so Sd00 6= ;: 2
Now by Lemma 3 we need only consider two cases: 0 < d0 < d, which was handled
by Evagelopoulos, and d0 < d < 0. We do the second case. Choose some y 2 Sd0 . By
Lemma 1, Sd contains arbitrarily large negative numbers, so we can nd x 2 Sd such that
x < y. But then y 62 Sd0 by Lemma 2. This contradiction completes the proof.
* * *
That's all the room we have this issue. Send me your contests and nice solutions!
* * * * *
113
PROBLEMS
Problem proposals and solutions should be sent to B. Sands, Department of Mathe-
matics and Statistics, University of Calgary, Calgary, Alberta, Canada T2N 1N4. Propos-
als should, whenever possible, be accompanied by a solution, references, and other insights
which are likely to be of help to the editor. An asterisk (*) after a number indicates a
problem submitted without a solution.
Original problems are particularly sought. But other interesting problems may also
be acceptable provided they are not too well known and references are given as to their
provenance. Ordinarily, if the originator of a problem can be located, it should not be
submitted by somebody else without permission.
To facilitate their consideration, your solutions, typewritten or neatly handwritten
on signed, separate sheets, should preferably be mailed to the editor before November 1,
1991, although solutions received after that date will also be considered until the time when
a solution is published.
1631. Proposed by Murray S. Klamkin, University of Alberta. (Dedicated to
Jack Garfunkel.)
Let P be a point within or on an equilateral triangle and let c1; c2; c3 be the lengths
of the three concurrent cevians through P . Determine the largest constant such that
c1 ; c2 ; c3 are the sides of a triangle for any P .
1632. Proposed by Stanley Rabinowitz, Westford, Massachusetts.
Find all x and y which are rational multiples of (with 0 < x < y < =2) such
that tan x + tan y = 2.
1633. Proposed by Toshio Seimiya, Kawasaki, Japan.
In triangle ABC , the internal bisectors of 6 B and 6 C meet AC and AB at D
and E , respectively. We put 6 BDE = x; 6 CED = y. Prove that if 6 A > 60 then
cos 2x + cos 2y > 1.
1634. Proposed by F.F. Nab, Tunnel Mountain, Alberta.
A cafeteria at a university has round tables (of various sizes) and rectangular trays
(all the same size). Diners place their trays of food on the table in one of two ways,
depending on whether the short or long sides of the trays point toward the centre of the
table:
or Z .
S
SS SS ZZ
ZZ S SS ZZZZ
S S Z
Moreover, at the same table everybody aligns their trays the same way. Suppose n math-
ematics students come in to eat together. How should they align their trays so that the
table needed is as small as possible?
114
* * * * *
115
SOLUTIONS
No problem is ever permanently closed. The editor will always be pleased to consider
for publication new solutions or new insights on past problems.
1513. [1990: 43] Proposed by M.S. Klamkin, University of Alberta.
(a) A planar centrosymmetric polygon is inscribed in a strictly convex planar cen-
trosymmetric region R. Prove that the two centers coincide.
(b) Do part (a) if the polygon is circumscribed about R.
(c) Do (a) and (b) still hold if the polygon and region are n-dimensional for n > 2?
I. Solution to (a) and (b) by Jordi Dou, Barcelona, Spain.
(a) Let M be the centre of the polygon P and O the centre of R. Let AB be a side
of P and ST the symmetrically opposite side with respect to M . Let A0B 0 be symmetric
to AB with respect to O. If O and M are distinct, we have three chords AB; TS and
B 0 A0 of R which are of equal length and are parallel, which is incompatible with the strict
convexity of R.
(b) Let a be a side of P and s the opposite side. Let a0 be symmetric to a with
respect to O. If O and M are distinct [and a is chosen not parallel to OM ], we have three
tangents to R which are mutually parallel, which is again incompatible with the strict
convexity of R.
II. Solution to (b) and (c) by Marcin E. Kuczma, Warszawa, Poland.
[Kuczma also solved part (a). | Ed.]
(b) Suppose the polygon is circumscribed about R. Choose a pair of parallel sides.
They are contained in two parallel supporting lines of R. To every direction there exists
exactly one pair of supporting lines, situated symmetrically relative to O, the center of
symmetry of R. Thus O lies midway between (the lines containing) the chosen sides.
Repeat the argument taking another pair of parallel sides; point O is also equidistant from
these sides. The only point with these properties is the center of symmetry of the polygon.
Note that strict convexity is not needed in this part.
(c) The analogue of part(b) is true in any dimension n 2. The proof is the same as
above; just write hyperplanes for lines; and instead of two pairs of opposite sides consider
n pairs of opposite faces of the polyhedron.
The analogue of part (a) is false in every dimension n 3. Here is a counter-
example.
Consider the points (in Rn)
ith place th place
#
i
#
Ai = (q; : : : ; q; ,1 ; q; : : : ; q) ; Bi = (,q; : : : ; ,q; 1 ; ,q; : : : ; ,q)
for i = 1; : : : ; n, where q is the root of the equation
3
q + 3q = n ,4 1 ; (1)
note that
1 <q<1 (2)
n,1
116
since n 3.
Let d be the metric in Rn induced by the 3-norm
Xn
kX k = ( jxij ) = 3
3 1 3
i=1
for X = (x ; : : : ; xn ). [Editor's note: that is, the \distance" from X to Y is dened to be
1
kX , Y k .] By (1),
3
So if we take
k=R R R ; 1 2 (2) 3
properties of inversion (see e.g. N. Altshiller Court, College Geometry, New York, 1962,
xx526 and 413),
R0 = k ; (3)
R jR , d j 2 2
conclusion is: p
if t > R, the locus of P is the circle of center O and radius r + t ; p 2 2
P 0 Pn
1
DD
A P 0 DD
, A
3
,T @,
02
,
r
An
DD ,,
2
DD , ,, @
P
A0
, , @ O , 3
, D V A P ,
0
@ ,
1
,
1
,, DD
3
, DA0n @ ,T,
A0 , ,P
2
, DD 2
A 0
DD
3
Figure 1 Pn0 P1
Figure 2
Choose a pair of opposite sides whose distance apart (D) is minimal. Consider the
circle ! of centre O and diameter D; it is tangent to the two opposite sides at interior
points T and T 0, and thus is interior to P . Then
Xn
D < perimeter of P = si = 1;
i=1
so D < 1=. Therefore the orthogonal projection of all 2n sides of P onto TT 0 has total
length 2D < 2=, as was to be proved.
Note that TT 0 is the line for which the sum of projections is minimal. It is also
clear that the largest diagonal of P gives the direction of maximal sum of projections of
the n segments.
Also solved (almost the same way!) by MURRAY S. KLAMKIN, University of
Alberta; T. LEINSTER, Lancing College, England; P. PENNING, Delft, The Netherlands;
and the proposer.
The proposer did give a second proof using integration, and suggests the analogous
problem for segments of total length 1 in three-dimensional space. With integration he
obtains that there is a line so that the sum of the lengths of the projections of the segments
to this line is less than 1=2 (and this is best possible, as was 2= in the two-dimensional
case). He would like a simpler proof. Might there also be a generalization to planar regions
in three-space?
The problem was suggested by the proposer for the 1989 IMO, but not used (see
#73, p. 45 of the book 30th International Mathematical Olympiad, Braunschweig 1989|
Problems and Results, which was reviewed in February).
* * * * *
119
JJ
Draw AF parallel to BC so that
JJ
AF = CD. Draw AM = MN where
M
ss s s s s s s s
n , 2k, i.e., the number of n-sequences of n , k + 1 R's and k , 1 L's; there are k,n of
1
,
,
q s s s
,@ ,@ ,
q
on that part of the walk from 0 to 0, @ , @,
qs q q sq q qs s
the rst time ,1 is reached. Draw- @ ,@ ,
ing walks as \graphs" (as shown), ,1 @, @ ,
this amounts to \re
ecting" an ini- ,2
@ @, ,
tial piece of the graph about the step
2
step
,2
horizontal line y = ,1, giving the step
n k
n= " n!
X
[ 2]
n
!#
n
!
k =0
k , k , 1 = [n=2] (2)
(follows by telescoping).
Finally we prove that walks of n unit steps on Z Z (i.e., the integer points of
+
the upper half-plane) starting at (0; 0) are equinumerous with walks of 2n + 1 steps on Z +
starting at 0.
To see this, consider a walk of n steps on Z Z . It can be written as a sequence
+
of moves L, R, U(up), D(down). [At each stage the number of U's must equal or exceed
the number of D's.] Replace each L by LR, R by RL, U by RR, D by LL, and prex this
sequence with an R. This process gives a walk of length 2n + 1 on Z , and the process is
+
reversible.
Therefore the number of walks of n steps on Z Z starting at (0; 0) is, by (2),
+
!
2n + 1 :
n
[Editor's note. Hirschhorn sent in a second proof, based on a formula for the number of
walks of n steps on Z Z from (0; 0) to (x; y).]
+
121
X k 2n + 1! X n 2n + 1! 1 X n 2 n + 1
!
= 1 2 n = 4n;
2 +1
= = 2 +1
j
=1
n,j i
=0
i 2i
=0
i 2
so (5) holds. For (4) we have
! ! !
2 n , 1 2 n
g (n) = 2 n , 1 + n , 1 + n , 2 , 1 2 n , 1
0
! ! !
= 2 n , 1 2 n , 1 2 n , 1
n! + 2 !n , 1 + !n , 2
= 2nn + n2,n 1 = 2nn+ 1 ;
k 2n , 1! !
k 2 n , 1 k 2 n , 1 2 n , 1 2n , 1
=2
j n,1,j +j n,2,j + n,1 +j n,j , n
k " 2n , 1 ! !# X k " 2n , 1 ! !#
=0 =0 =0
X 2 n , 1 2 n , 1
=
j n,1,j + n,2,j +j n,1,j + n,j
X k
=0
! Xk ! X k 2n + 1!
=0
2 n 2n
=
j =0
n,1,j +j n,j =j n,j :
=0 =0
122
,! ,!
, 2R + 2 OB OC = R , 2 cos 2A = ,1
2 2
, 6 A = 60 or 120:
123
Case (ii): OH = OA
,! ,! ,! ,! X ,! ,!
, (OA + OB + OC )2 = (OA)2 , 3R2 + 2 OB OC = R2
X
, cos 2a = ,1 , cos A cos B cos C = 0
X
[using 4 cos A cos B cos C = 1 + cos 2A]
, 6 B = 90 or 6 C = 90:
Case (iii): HO = HA
,! ,! ,! ,! ,! ,! ,! ,!
, j OH j = j OH , OA j , (OA + OB + OC ) = (OB + OC ) 2 2
X ,! ,! ,! ,!
, 3R + 2 OB OC = 2R + 2 OB OC
2 2
, 2(cos 2B + cos 2C ) = ,1
, 4 cos A cos(B , C ) = 1
[using cos X + cos Y = 2 cos X Y cos X ,Y with X = 2B , Y = 2C ]. This solves the
+
2 2
rst part of the problem.
To cope with the last question, we must restate the resultant equalities in terms of
side lengths. In case (i) we are led to the alternatives
a = b , bc + c
2 2 2
or a = b + bc + c ;
2 2 2
(1)
in case (ii) to
b =a +c or 2
c =a +b :
2 2 2 2 2
(2)
If a = b , bc + c in (1), without loss of generality let b > c. Then a , c = b
2 2 2 2 2 2
, bc
so that
a,c = b = m
b,c a+c n
in lowest terms. Thus
m m , n m
a , c = n (b , c) or a+ n c= n b
and n
a + c = m b:
Solving for a and c, we have
! !
m ,
a = 2mn , m mn + n 2
n ,
b ; c = 2mn , m b : m 2 2 2
2 2
Thus
a = k(m , mn + n ) ; b = k(2mn , m ) ; c = k(n , m ):
2 2 2 2 2
(3)
Similarly from a = b + bc + c we obtain
2 2 2
As m, n range over all pairs of relatively prime integers with n > m > 0, the sets of
formulas in (3) and (4) for k = 1 produce triples of positive integers a; b; c which are either
mutually coprime or divisible by 3; after possible reduction by 3 we obtain the complete
solution to the equations in (1) in irreducible triples of positive integers.
For (2) the irreducible solutions are the Pythagorean triples
a = 2xy ; b = x + y ; c = x , y
2 2 2 2
or
a = 2xy ; b = x , y ; c = x + y ;
2 2 2 2
for x; y; z integral and not all even, which, by reducing modulo 4, clearly has no solution.
Thus there are no solutions in case (iii).
There were two partial solutions submitted.
The proposer's original problem actually contained the additional condition that the
area of the triangle also be integral; as the proposer points out, case (i) above then has no
solution, since sin 60 is irrational, and so the only solutions are the right-angled triangles
of case (ii). The editor, however, inadvertently omitted the area condition from the problem.
* * * * *
1519. [1990: 44] Proposed by P. Penning, Delft, The Netherlands.
Find all prime numbers which, written in the number system with base b, contain
each digit 0; 1; : : : ; b , 1 exactly once (a leading zero is allowed).
Solution by Richard I. Hess, Rancho Palos Verdes, California.
If p is such a prime number, with base b representation (ab, ab, a a )b , then 1 2 1 0
p = a + a b + a b + + ab, bb,
0 1 2
2
1
1
Thus for b = 2k p will be divisible by b , 1, so the only such primes for even bases are for
b = 2; in fact, only 2 = (10) . For b = 2k +1 p will be divisible by k, so the only such primes
2
for odd bases are for b = 3, and we nd the solutions (012) = 5; (021) = 7; (102) = 11,
3 3 3
and (201) = 19.
3
Also proved (usually by one or the other of the above methods) by H.L. ABBOTT,
University of Alberta; RICHARD I. HESS, Rancho Palos Verdes, California;
MURRAY S. KLAMKIN, University of Alberta; MARCIN E. KUCZMA, Warszawa,
Poland; T. LEINSTER, Lancing College, England; R.C. LYNESS, Southwold, England;
P. PENNING, Delft, The Netherlands; DAVID SINGMASTER, South Bank Polytechnic,
London, England; ST. OLAF PROBLEM SOLVING GROUP, St. Olaf College, Northeld,
Minnesota; and the proposer. A comment was also received from a reader who evidently
didn't understand the problem.
One of the above solvers (whose identity will be kept secret) claims that the same
result is true for a countable number of parabolas. Is it?
126
This is the third solution of Jordi Dou to be featured in this issue. It seems tting
to disclose here that Dou recently wrote to the editor, and (while conveying 80th birthday
greetings to Dan Pedoe) mentioned that he had himself just turned 80 on June 7! Crux
readers, and certainly this admiring and appreciative editor, would wish Professor Dou all
the best.
* * * * *
1521. [1990:74] Proposed by J.T. Groenman, Arnhem, The Netherlands.
Triangle ABC has angles ; ;
, circumcenter O, incenter I , and orthocenter H .
Suppose that the points A; H; I; O; B are concyclic.
(a) Find
:
(b) Prove HI = IO:
(c) If AH = HI , nd and .
Solution by D.J. Smeenk, Zaltbommel, The Netherlands.
(a) If A; H; I; O; B are concyclic, then
C
6 AHB = 6 AIB = 6 AOB; ll
l
p p
ll
so %% lOl
180 ,
= 90 +
= 2
; %
2 hh%% I ll
%Hhhhhhhhhhlhlhl
so
= 60.
Remark: if two of the three points H; I; O are con-
cyclic with A and B , then the third one is as well. A%
hl B
[Editor's note. This assumes ABC is acute. If
6 C = 120 then A; H; O; B are concyclic but I
does not lie on the same circle.]
(b) From 6 ABH = 90 , = 6 OBC we obtain
6 OBI = 1 ( ,
) = 6 IBH;
2
so HI = IO.
(c) If AH = HI then
90 , = 6 ABH = 6 HBI = 2 ,
2 = 2 , 30;
so = 80 and thus = 40.
Also solved by WILSON DA COSTA AREIAS, Rio de Janeiro, Brazil;
FRANCISCO BELLOT ROSADO, I.B. Emilio Ferrari, Valladolid, and JOSE YUSTY
PITA, Madrid, Spain; JORDI DOU, Barcelona, Spain; HANS ENGELHAUPT, Franz-
Ludwig-Gymnasium, Bamberg, Germany; HIDETOSI FUKAGAWA, Aichi, Japan; L.J.
HUT, Groningen, The Netherlands; WALTHER JANOUS, Ursulinengymnasium, Inns-
bruck, Austria; MARCIN E. KUCZMA, Warszawa, Poland; KEE-WAI LAU, Hong Kong;
127
2s 1 2s 2s
3 2s 2 4
or
r + r = Fs + Fs = Fs + Fs = r + r ;
1 3
1 3 2 4
2 4
1 3 2 4
where r ; r ; r ; r are respectively the inradii of the above triangles. The relation r + r =
1 2 3 4 1 3
r + r is true and has been shown by H. Forder in \An ancient Chinese theorem", Math
2 4
Note 2128, p. 68 of Mathematical Gazette 34, no. 307 (1950). It was also part (b) of Crux
1226 [1988: 147].
Also solved by DUANE M. BROLINE, Eastern Illinois University, Charleston;
HANS ENGELHAUPT, Franz-Ludwig-Gymnasium, Bamberg, Germany; C. FESTRAETS-
HAMOIR, Brussels, Belgium; GEORGE P. HENDERSON, Campbellcroft, Ontario;
RICHARD I. HESS, Rancho Palos Verdes, California; WALTHER JANOUS, Ursulinen-
gymnasium, Innsbruck, Austria; MARCIN E. KUCZMA, Warszawa, Poland; KEE-WAI
LAU, Hong Kong; TOM LEINSTER, Lancing College, England; J.A. MCCALLUM,
Medicine Hat, Alberta; VEDULA N. MURTY, Penn State University at Harrisburg; P.
PENNING, Delft, The Netherlands; TOSHIO SEIMIYA, Kawasaki, Japan; D.J. SMEENK,
Zaltbommel, The Netherlands; and the proposer.
Murty points out that the result holds for all real a; b; c; d; x; y provided only that the
denominators in the problem are nonzero (as is clear from solution I).
The geometrical interpretation given in (partial) solution II was also noted by
Festraels-Hamoir, Kuczma, Smeenk and the proposer. Most of them, however, gave al-
gebraic proofs, thus avoiding the cases (e.g., a > b + c + d) missed by Areias and Bellot.
Kuczma gave a separate and rather involved argument to cover these cases. Is there a
\nice" way to complete solution II?
* * * * *
Crux
Mathematicorum
Volume 17 , Number 5 May 1991
CONTENTS
1
129
Determine the radius of a fourth circle contained in the same sphere and tangent to the
three given ones.
6. Prove that the inequality
ab + ac + ad + bc + bd + cd = abc + abd + acd + bcd =
!1 2 !1 3
6 4
holds for any positive numbers a; b; c; d.
*
A second Olympiad for this issue comes from Sweden via Andy Liu.
130
can be written in base n + 1 with the same digits but in opposite order.
2
2. Determine all continuous functions f such that f (x) + f (x ) = 0 for all real 2
numbers x.
3. Find all positive integers n such that n , 18n + 115n , 391 is the cube of a
3 2
positive integer.
4. Let ABCD be a regular tetrahedron. Where on the edge BD is the point P
situated if the edge CD is tangent to the sphere with diameter AP ?
5. Assume that x ; : : : ; x are positive real numbers such that x < x and assume
1 5 1 2
that x ; x ; x are all greater than x . Prove that if > 0, then
3 4 5 2
1 + 1 + 1 < 1 + 1 + 1 :
(x + x ) (x + x ) (x + x ) (x + x ) (x + x ) (x + x )
1
3 2
4
2 5
1
2 2 3 4 5
2. Suppose there are 997 points given on a plane. If every two points are joined
by a line segment with its mid-point coloured in red, show that there are at least 1991 red
points on the plane. Can you nd a special case with exactly 1991 red points?
3. Let a ; a ; : : : ; an; b ; b ; : : : ; bn be positive real numbers such that
1 2 1 2
X n n
ak =
X
bk :
k=1 k=1
Show that n n
X (ak ) 1
2 X
ak :
k=1 ak + bk 2 k=1
4. During a break n children at school sit in a circle around their teacher to play a
game. The teacher walks clockwise close to the children and hands out candies to some of
them according to the following rule: he selects one child and gives him a candy, then he
skips the next child and gives a candy to the next one, then he skips 2 and gives a candy
to the next one, then he skips 3, and so on. Determine the values of n for which eventually
(perhaps after many rounds) all children will have at least one candy each.
5. Given are two tangent circles, C ; C , and a point P on their radical axis, i.e.
1 2
on the common tangent of C and C that is perpendicular to the line joining the centres
1 2
of C and C . Construct with compass and ruler all the circles C that are tangent to C
1 2 1
and C and pass through the point P .
2
*
I am giving more problems than usual this issue to give readers some sources of
pleasure for the summer break. Also the June issue is normally taken up by two contests
for which we do not usually publish readers' solutions.
* * *
Last issue we gave the problems of the A.I.M.E. for 1991. As promised, we next give
the numerical solutions. The problems and their ocial solutions are copyrighted by the
Committee of the American Mathematics Competitions of the Mathematical Association
of America, and may not be reproduced without permission. Detailed solutions, and
additional copies of the problems, may be obtained for a nominal fee from Professor Walter
E. Mientka, C.A.M.C. Executive Director, 917 Oldfather Hall, University of Nebraska,
Lincoln, Nebraska, U.S.A., 68588-0322.
1: 146 2: 840 3: 166 4: 159 5: 128
6: 743 7: 383 8: 010 9: 044 10: 532
11: 135 12: 677 13: 990 14: 384 15: 012
* * *
We now turn to some further solutions submitted by readers for problems from the
`Archives'. First a problem from April 1984.
132
:
Y Y
=
i=1 3 3 ,1 i=1 3 3 ,1
Solution by Murray S. Klamkin, University of Alberta.
Let (
3 i !) tan 3 + tan( 33ni,,11 )
Ai = tan 3 1 + 3n , 1 =
1 , tan 3 tan( 33ni,,11 )
and
i tan , tan( ni,,1 )
Bi = tan 3 1 , 3n3, 1
( !)
3
= 3
: 3 1
Now since
tan 3 = tan 3 , tan
" #
2
1 , 3 tan 2
we get
3 , tan ( ni,,1 ) tan( n,i )
2 3 3
AiBi = = 3 1
: 3 1
Ai Bi = tan( ) = = 1:
Y
3 1 3 1
i =1 n, tan n,
3 1 3 1
y + x , yx , 9y=8 = 0:
4 2 3
= (x , y)(2xy , 1)(x + xy + y ) : 2 2
candidates (0; 0) and (9=8; 9=8). If 2xy = 1, then substituting y = (2x), in (1) and 1
simplifying we obtain
8x + x1 , 9x = 0
4
2
2 2 2 2 2 2
Finally, suppose
x + xy + y = 0 :
2 2
(3)
From (1) minus (2) we obtain
8(x , y ) , 8(x , y ) + 8xy(x , y ) , 9(x , y) = 0 ;
4 4 2 2 2 2
and disregarding the possibility that x = y, a case already considered above, we have
(x + y)[8(x + y ) , 8 + 8xy] = 9 :
2 2
(4)
Substitution of (3) into (4) now yields
x + y = , 98 : (5)
From (5) and (3) we get
xy = (x + y) , (x + xy + y ) = 81
2
64
: 2 2
(6)
From (5) and (6) we see that x and y are the two roots of the equation
u + 98 u + 81
2
64
=0;
p
or 64u + 72u + 81 = 0. Solving, we get u = (,9 9 3i)=16. This gives two more
2
(complex) candidates
( 9! ; 9! ); ( 9! ; 9! ):
2 2
8 8 8 8
Straightforward substitution into (1) and (2) shows that all of these ten candidates are
solutions.
134
Editor's note. Solutions were also received from Nicos Diamantis, student, Univer-
sity of Patras, Greece, and Hans Engelhaupt, Gundelsheim, Germany.
8. [1986: 20] 1985 Austria-Poland Mathematical Competition.
The consecutive vertices of a given convex n-gon are A0; A1; : : : ; An,1. The n-gon
is partitioned into n , 2 triangles by diagonals which are non-intersecting (except possibly
at the vertices). Show that there exists an enumeration 1 ; 2; : : : ; n,2 of these triangles
such that Ai is a vertex of i for 1 i n , 2. How many enumerations of this kind
exist?
qqq
Solution by Hans Engelhaupt, Gundelsheim, Germany.
There is always just one such enumeration.
We argue by induction on n. The cases n = 3 and n = 4 are trivial. Suppose n > 4.
qq
An,2 H Ak
H
A
A
An,1
C
C
C
C
C
C
C (,
(A
C (((( 2
A0 A1
two convex polygons [A ; : : : ; Ak ] and [Ak ; : : : ; An, ]. Also the original triangulation in-
0 1
duces a triangulation of each of these, since the diagonals do not intersect except at end-
points. Existence of a numbering is now immediate. Uniqueness follows since the triangles
; : : : ; k, must be in [A ; : : : ; Ak ] and k ; : : : ; n, must be in [Ak ; : : : ; An, ].
1 1 0 +1 2 1
*
We now turn to the March 1986 numbers, and solutions to some of the problems of
the 1982 Leningrad High School Olympiad (Third Round) [1986: 39-40].
1. P ; P
1 2and P are quadratic trinomials with positive leading coecients and real
3
roots. Show that if each pair of them has a common root, then the trinomial P + P + P 1 2 3
also has real roots. (Grade 8)
135
and
P : gx + hx + i = 0 with g > 0 and the real roots w ; w :
3
2
1 2
Case 1: u = w or u = w .
1 1 1 2
Then P + P + P has the real root u so the other root is real as well.
1 2 3 1
Case 2: u = w and v = w .
2 1 2 2
Then
P + P + P = a(x , u )(x , u ) + d(x , u )(x , v ) + g(x , u )(x , v ) (= f (x)):
1 2 3 1 2 1 2 2 2
j =0 j j =0 j
j even j odd
p
[Thus A is the sum of the even-numbered terms in the expansion of ( 1982 + 1)k and B
is the sum of the odd-numbered terms in that expansion.] Note also that
p k k (p1982)k,j (,1)j = B , A if k is odd
! (
1982 , 1)k
X
( = A , B if k is even .
j =0 j
Case 1: k is odd. Let n = A2. Then
q
p q
p p
n + 1981k + n = A + ( 1982 , 1)k ( 1982 + 1)k + A
2
q
= A + (B , A)(A + B ) + A 2
p
= A + B p, A + A 2 2 2
= B + A = ( 1982 + 1)k :
Case 2: k is even. Let n = B . Then 2
q
p q p
n + 1981k + n = B + (A , B )(A + B ) + B 2 2
p p p
= A + B = ( 1982 + 1)k : 2 2
11. Show that it is possible to place non-zero numbers at the vertices of a given
regular n-gon P so that for any set of vertices of P which are vertices of a regular k-gon
(k n), the sum of the corresponding numbers equals zero. (Grade 10)
Solution by Nicos Diamantis, student, University of Patras, Greece.
Consider a coordinate system with O(0; 0) the centre of the n-gon (and therefore
the centre of all regular k-gons (k n) which have vertices some of those of the n-gon).
At the vertices of the n-gon, place the x-coordinates. [The,!n-gon can be rotated so that
none of these x-coordinates are zero.] But we have ki OAi = 0, where the Ai are the
P
=1
vertices of a regular k-gon. From this, looking at the x-coordinates we have ki xi = 0, P
=1
where x ; x ; : : : ; xk are the x-coordinates of A ; : : : ; Ak . This solves the problem.
1 2 1
*
Before turning to more recent problems, I want to apologize for leaving S.R. Cavior
of the University of Bualo o the list of solvers when I discussed problem 1 of the 24th
Spanish Olympiad [1989: 67] in the January number [1991: 9]. His solution somehow had
found its way into the collection for a later month.
*
For the remainder of this column, we turn to problems given in the June 1989
number of the Corner. We give solutions to all but numbers 3 and 6 of the 3rd Ibero-
American Olympiad [1989: 163{164].
1. The angles of a triangle are in arithmetical progression. The altitudes of the
triangle are also in arithmetical progression. Show that the triangle is equilateral.
Solution by Bob Prielipp, University of Wisconsin{Oshkosh.
Let A, B , C be the angles of the given triangle and let ha ; hb ; hc be the corresponding
altitudes. Without loss of generality, we may assume A B C . Since the angles are
in arithmetic progression A + C = 2B , and since A + B + C = 180, B = 60. Now also
hc hb ha and a b c where a; b; c are the side lengths opposite A; B; C , respectively.
From the law of cosines b = a + c , 2ac cos60 = a + c , ac. Now 2hb = ha + hc
2 2 2 2 2
implies that 4F=b = 2F=a + 2F=c, where F is the area of triangle ABC , so
2 = 1 + 1 = a + c ; or b = 2ac :
b a c ac a+c
Since b = a + c , ac, 4a c = (a + c) (a + c , ac). From this we get
2 2 2 2 2 2 2 2
0 = (a + c) (a + c , ac) , 4a c
2 2 2 2 2
= (a , c) + 5ac(a , c)
4 2
= (a , c) [(a , c) + 5ac];
2 2
is rational.
Solution by Michael Selby, University of Windsor, and by Edward T.H. Wang,
Wilfrid Laurier University, Waterloo, Ontario. A
Consider the triangle shown with sides represented by ,AK
vectors. Let Bj be the vector joining the vertex B to the , A
corresponding point of subdivision on side b, and dene Cj ,
, A
A
and Aj analogously. Then ,c bA
, A
j
Bj = a + n b;
,
, A
-A
B a C
so
jBj j = jaj + nj jbj + 2nj (a b):
2
2 2 2
2
Therefore
n, n, n,
jBj j = (n , 1)a + b nj + n2 j (a b)
X1 X1 2 X1
2 2 2
2
j
=1 j =1 j =1
= (n , 1)a + b (n , 1)(2
2 2
6n
n , 1) + (n , 1)(a b):
Similarly
jCj j = (n , 1)b + c (n , 1)(2 n , 1) + (n , 1)(b c);
X1n,
2 2 2
j =1 6n
139
j =1 6n
Therefore
,1
nX
S= (jAj j + jBj j + jCj j ) = (n , 1)(a + b + c ) +
2 2 2 2 2 2
j =1
(a + b + c )(n , 1)(2n , 1) + (n , 1)(a b + b c + a c)
2 2 2
(1)
6n
Since a + b + c = 0,
0 = ja + b + cj = a + b + c + 2(a b + a c + b c):
2 2 2 2
S = (a + b + c ) n , 1 + (n , 1)(2 n , 1) , n , 1 :
!
2 2 2
6n 2
Hence
S = (n , 1)(5n , 1) :
a +b +c 2
6n 2 2
such that
(x + yt + zt )(u + vt + wt ) = 1 : 2 2
Editor's note. The result is still true if t = 2 replaces t = 2 in the problem.
3 2
* * *
Send me your nice solutions, and also your national and regional contests.
* * * * *
140
PROBLEMS
Problem proposals and solutions should be sent to B. Sands, Department of
Mathematics and Statistics, University of Calgary, Calgary, Alberta, Canada T2N 1N4.
Proposals should, whenever possible, be accompanied by a solution, references, and other
insights which are likely to be of help to the editor. An asterisk (*) after a number indicates
a problem submitted without a solution.
Original problems are particularly sought. But other interesting problems may also
be acceptable provided they are not too well known and references are given as to their
provenance. Ordinarily, if the originator of a problem can be located, it should not be
submitted by somebody else without permission.
To facilitate their consideration, your solutions, typewritten or neatly handwritten
on signed, separate sheets, should preferably be mailed to the editor before December 1,
1991, although solutions received after that date will also be considered until the time when
a solution is published.
1641. Proposed by D.J. Smeenk, Zaltbommel, The Netherlands.
Quadrilateral ABCD is inscribed in circle ,, with AD < CD. Diagonals AC and
BD intersect in E , and M lies on EC so that 6 CBM = 6 ACD. Show that the circumcircle
of BME is tangent to , at B .
1642. Proposed by Murray S. Klamkin, University of Alberta.
Determine the maximum value of
x(1 , y )(1 , z ) + y(1 , z )(1 , x ) + z (1 , x )(1 , y )
2 2 2 2 2 2
subject to yz + zx + xy = 1 and x; y; z 0.
1643. Proposed by Toshio Seimiya, Kawasaki, Japan.
Characterize all triangles ABC such that
AIa : BIb : CIc = BC : CA : AB;
where Ia; Ib ; Ic are the excenters of ABC corresponding to A; B; C , respectively.
1644. Proposed by Walther Janous, Ursulinengymnasium, Innsbruck, Austria.
Let f : R ! R be continuous such that it attains both positive and negative values,
and let n 2 be an integer. Show that there exists a strictly increasing arithmetic sequence
a < < an such that f (a ) + + f (an ) = 0.
1 1
has a + b + c , ab , bc , ca as a factor.
2 2 2
141
un = un + 2nu+n 6 ; n = 0; 1; : : : :
+3 +2
2
where the sums are cyclic over the angles ; ;
of a triangle.
1650. Proposed by Iliya Bluskov, Technical University, Gabrovo, Bulgaria.
Find all real numbers for which the equality
p p p
[ n + + n] = [ 4n + 1]
holds for every positive integer n. Here [ ] denotes the greatest integer function. (This
problem was inspired by problem 5 of the 1987 Canadian Mathematics Olympiad [1987:
214].)
* * * * *
SOLUTIONS
No problem is ever permanently closed. The editor will always be pleased to consider
for publication new solutions or new insights on past problems.
875. [1983: 241; 1984: 338] Proposed by Stanley Rabinowitz, Digital Equipment
Corp., Nashua, New Hampshire.
Can a square be dissected into three congruent nonrectangular pieces?
II. Solution by Sam Maltby, student, University of Calgary.
We show that the answer is no.
Let the unit square ABCD be cut into three congruent pieces P1 ; P2 and P3 , and
suppose these pieces are not rectangles. Here we will assume that our \pieces" contain
their boundaries, and therefore that two pieces are allowed to overlap on their boundaries
but not elsewhere. We must also assume that the pieces are connected, and moreover that
they contain no \isthmuses" or \tails"; otherwise it is easy to nd counterexamples, as
suggested by the following gures.
142
The eect of these assumptions is that we take the pieces to have the following
three properties:
(1) If a piece contains a vertex of the square but does not contain at least part of
both edges at that vertex, then some other piece must also contain that vertex;
(2) A segment of the boundary of the square cannot belong to more than one piece;
(3) Between any two points X and Y of a piece P there is a curve (called an incurve
in P between X and Y ) with X and Y as endpoints such that all points of the curve other
than X and Y are interior points of P . An incurve of one piece cannot intersect another
piece except possibly at its endpoints.
Note that since the pieces are congruent it must be possible through some combi-
nation of translation, rotation or re
ection to place one piece on another. This will take
each curve in the rst piece to some curve in the second; we shall say that these two curves
correspond.
We now give a series of lemmas and eventually arrive at the desired conclusion.
q&%
Lemma 1. The intersection of a piece with a side of the square is connected.
Proof. Suppose for a contradiction that the intersection of one of the pieces, say P ,
1
with one of the sides of the square, say AB , is not connected. Then there are two points
E and F on AB in P such that EF is not entirely in P ; say A
1 1
E G F B
that point G between E and F is in P . Let c be an incurve
2
in P between E and F . If there is some point H in P on the
1 2
c
other side of c from G, then an incurve in P between G and
2
H must intersect c, but it cannot have a common endpoint
with c. This contradicts (3), so H cannot exist. Thus P is 2
enclosed by c and EF . But then the convex hull of P is larger
1
than the convex hull of P , which contradicts P and P being D
2 1 2
C
congruent. 2
Lemma 2. No piece contains two opposite corners of the square.
Proof. Suppose that, for instance, P contains A and C . Then clearly P and P
1 2 3
must each also contain opposite corners of the square. If P contains B and D, then there is
2
an incurve in P between A and C and one in P between B and D. These must intersect,
1 2
but they do not have a common endpoint, which contradicts (3). Thus P and likewise P
2 3
must contain A and C .
Now some piece, say P , contains B . Then P and P must each contain a point
1 2 3
at distance 1 from both A and C . But B and D are the only such points, so two of the
143
pieces (say P and P ) contain, say, B . By Lemma 1, both P and P contain BC , which
1 2 1 2
is impossible by (2). 2
By the Pigeonhole Principle, one of the three pieces,
say P , contains at least two corners of the square, and by A
1
B
Lemma 2 these corners must be adjacent, say A and B . By S
Lemma 1 P contains the entire side AB . Then each of P S w
S
1 2
and P contains a segment which corresponds to AB ; call
3
2
these segments S and S respectively. It follows that
2 3
If P is a rotation of P then P cannot contain any line segment in AD, since by the
2 1 1
rotation P would contain some line segment in AB which would also be in P , contrary
2 1
144
1 2 ,3
P does not contain B . Then S must bisect the right angle ,
3
2
ABC , and P must be a 45 rotation of P around B . Since P G S 2 ,
2 1 , 2
is bounded by S , BC , and the perpendicular to S through
2 @ , 2
E , Pp must contain the segment DF along DC with length
3
@
@,
,
2 , 2. Since P is contained in BEFC , by rotation P is
2 E@ 1
contained inp ABEG, so P also contains the segment DG of D
3
@
@F C
length 2 , 2, where EF meets AD at G.
Now S is not parallel to BC by Lemma 3, and it cannot be parallel to CD because
3
if it were it would have to be CD itself, but then we could replace AB by CD and P by P 1 3
in case (b) or (c) to get the result. So S makes an acute angle with CD. By applying (4)
3
to P we see that G; D and F must all be on the same side of S , but S cannot intersect
3 3 3
the interiorpof EB since EB lies on the boundary of P and P . Thus S must be GF ; but 1 2 3
jGF j = 2( 2 , 1) < 1, a contradiction. This nishes the proof that P (and P ) cannot 2 3
contain BC . Similarly, P and P cannot contain AD.
2 3
Now suppose that P contains CD but S is not parallel to CD. By (4) S must
2 2 2
have either C or D as an endpoint, say C . S cannot be perpendicular to CD since it 2
would then be BC contrary to Lemma 3. But since P does not contain C , P contains 1 3
C by (1). Now C in P cannot correspond to A in P , since then P would contain no
2 1 1
point of AD other than A, and A cannot be in P , so P would contain A and C contrary 2 3
to Lemma 2. So C in P corresponds to B in P . But then P contains B and thus BC ,
2 1 3
145
pp p p p p p p
2 3
So P must be a rotation of P through 90 about the centre of the square. Then
2 3
S is perpendicular to S , and since neither S nor S are edges of the square it follows
2 3 2 3
that S and S must cross at a point interior to both. (Otherwise we can rotate both lines
p pp p p ppp
2 3
by 180 and extend the resulting four lines until they meet,
as in the diagram. Let x be the length of the segment past x
the point of intersection.
p The convex hull of this gure is a
hhhh
hhh
square of side at least 1 + x , and so has area at least 1+ x ,
2
2
where the sums are cyclic over the angles A; B; C of a triangle. [This generalizes Murray
Klamkin's problem E3180 in the Amer. Math. Monthly (solution p. 771, October 1988).]
I. Solution by Seung-Jin Bang, Seoul, Republic of Korea.
Our solution is essentially the same as O.P. Lossers in the October 1988 Monthly.
Let
V (A; B; C; t) = cos tA , sin tA:
X X
2
= 2 cos t(A , B ) cos t( , C ) , sin t( , C ) + cos tC , sin tC:
!
2 2 2
Note that
p
V (A; B; 0; t) = 2 2 cos t(A , B ) cos t + + 1
2 2 4
p t t
2 2 cos 2 cos 2 + 4 + 1
p
= 2 cos(t + 4 ) + 2:
147
2 4 2 2
f xi f (xi );
X X
(2)
i=1 i=1
for x ; : : : ; xn 0,
1
P
=2. (This is a yet more natural form of the statement.)
n x
i=1 i
It suces to prove (2) for n = 2; obvious induction then does the rest. And for
n = 2, writing x and y for x and x , we have
1 2
2 2 2 2 2
x y
= 4 sin sin cos x
+ y x + y
, sin 2 0
2 2 2
because (x + y)=2 2 [0; =4].
Also solved by C. FESTRAETS-HAMOIR, Brussels, Belgium; MURRAY S.
KLAMKIN, University of Alberta; KEE-WAI LAU, Hong Kong; and the proposer. There
was one incorrect solution submitted.
The problem was also proposed independently (without solution) by Robert E. Shafer,
Berkeley, California.
148
* * * * *
1524. [1990:74] Proposed by George Tsintsifas, Thessaloniki, Greece.
ABC is a triangle with sides a; b; c and area F , and P is an interior point. Put
R = AP , R = BP , R =p CP . Prove that the triangle with sides aR ; bR ; cR has
1 2 3 1 2 3
circumradius at least 4F=(3 3).
Solution by Walther Janous, Ursulinengymnasium, Innsbruck, Austria.
In any triangle,
R p1 (a + b + c) (1)
3 3
([1], item 5.3). Thus, using item 12.19 of [1], we get for the circumradius R~ under consid-
eration:
R~ p1 (aR1 + bR2 + cR3) p2 (ar1 + br2 + cr3)
3 3 3 3
= p2 (2F ) = 4pF :
3 3 3 3
Reference:
[1] Bottema et al, Geometric Inequalities.
Also solved by C. FESTRAETS-HAMOIR, Brussels, Belgium; MURRAY S.
KLAMKIN, University of Alberta; MARCIN E. KUCZMA, Warszawa, Poland; and the
proposer.
Klamkin and Kuczma note that inequality (1) is equivalent to the fact that the
perimeter of a triangle inscribed in a circle is maximized when the triangle is equilateral.
* * * * *
1525. [1990: 75] Proposed by Marcin E. Kuczma, Warszawa, Poland.
Let m; n be given positive integers and d be their greatest common divisor. Let
x = 2m , 1, y = 2n + 1:
(a) If m=d is odd, prove that x and y are coprime.
(b) Determine the greatest common divisor of x and y when m=d is even.
Solution by Kenneth M. Wilke, Topeka, Kansas.
Putting m = m=d and n = n=d, we have (m ; n ) = 1. Let = (x; y) and put
1 1 1 1
k = x=, ` = y=. We want to nd .
(a) If m = m=d is odd, then
1
= 2 d , 1 = (2d + 1)(2d , 1) :
2
k=1
can be proved easily from the Binomial Theorem. Establish this identity by a combinatorial
argument.
Solution by H.L. Abbott, University of Alberta.
The Mathematics Department at a certain university has n members. The admin-
istration of the department is handled by an executive committee whose chairman also
serves as chairman of the department. There are no restrictions, except the obvious ones,
on the size of the executive committee. For example, during those years when many oner-
ous problems are expected to arise it may be a committee of one, while in times when little
of any consequence needs attention it may consist of the whole department. Each member
of the department who is not a member of the executive committee is required to serve on
exactly one of q committees. There is no restriction on the size of these committees either.
Indeed, some of them need not have any members at all. This, for example, will be the
case when the size k of the executive committee is such that q + k > n.
150
Late one evening as the chairman was leaving the department he remarked to his
secretary that there is a simple expression for the number of possible administrative struc-
tures for the department. \Observe," he said, \that the number of ways of choosing an
executive committee of size k is nk . The chairman of this committee, and thus of the
department, may then be chosen in k ways and the remaining n , k members of the de-
partment may then be assigned their tasks in qn,k ways. Thus the number of possible
bureaucracies is nk k nk qn,k ."
P
=1
His secretary, almost without hesitation, replied, \Surely there is a much simpler
expression for this number. The chairman of the executive committee may be chosen in n
ways and after this choice has been made the remaining n , 1 members of the department
may be assigned their administrative chores in (q + 1)n, ways. Thus the number is
1
A few moments later the chairman related this conversation to the caretaker as they
rode the elevator to the rst
oor. \It shows," said the caretaker, \that your secretary is
the one who counts."
A student who was in the elevator was overjoyed upon hearing this discussion.
She realized that she could now solve the one remaining question on her combinatorics
assignment which was due at the next class. The problem called for a combinatorial proof
that n
k(k , 1) nk qn,k = n(n , 1)(q + 1)n, :
!
X
2
k=2
As soon as she arrived home she wrote out the solution: The Mathematics Department
at a certain university has n members. The administration of the department is handled
by an executive committee whose chairman and associate chairman also serve as chairman
and associate chairman of the department.... .
Also solved by JACQUES CHONE, Clermont-Ferrand, France; C. FESTRAETS-
HAMOIR, Brussels, Belgium; WALTHER JANOUS, Ursulinengymnasium, Innsbruck,
Austria; TOM LEINSTER, Lancing College, England; and the proposers.
* * * * *
1527. [1990: 75] Proposed by D.J. Smeenk, Zaltbommel, The Netherlands.
In quadrilateral ABCD the midpoints of AB , BC , CD and DA are P , Q, R
and S respectively. T is the intersection point of AC and BD, M that of PR and QS .
G is the centre of gravity of ABCD. Show that T , M and G are collinear, and that
TM : MG = 3 : 1.
Solution by John Rausen, New York.
The \center of gravity" of a quadrilateral is ambiguous. For a triangle, the centroid
G (point of intersection of the medians) is a \center of gravity" in two dierent ways: (1)
G is the center of mass of a system of three equal point-masses at the vertices; (2) G is
also the center of mass of a uniform mass distribution on a thin plate (\lamina") covering
the triangle. In the case of a (plane) quadrilateral, we get two dierent points: (1) the
center of mass of a set of four equal masses, say unit masses, at the vertices. This is point
151
,! ,! ,! ,! 3 3
goes into the corresponding point associated to A0 B 0 C 0D0. Thus, since M goes into itself,
3
Reference:
[1] Nathan Altshiller-Court, College Geometry, 2nd edition, 1952, New York.
Also solved by WILSON DA COSTA AREIAS, Rio de Janeiro, Brazil; JORDI
DOU, Barcelona, Spain; C. FESTRAETS-HAMOIR, Brussels, Belgium; RICHARD I.
HESS, Rancho Palos Verdes, California; L.J. HUT, Groningen, The Netherlands;
WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria; MARCIN E. KUCZMA,
Warszawa, Poland; KEE-WAI LAU, Hong Kong; TOM LEINSTER, Lancing College,
England; P. PENNING, Delft, The Netherlands; TOSHIO SEIMIYA, Kawasaki, Japan;
DAN SOKOLOWSKY, Williamsburg, Virginia; and the proposer.
Lau located the problem as a worked example in x7:12 of Humphrey's Intermediate
Mechanics Vol. 2 (Statics), Second Edition, Longmans, 1964. (On p. 237-238 of x151 in
152
the 1961 edition.) The proposer took the problem from Journal de Math. E lementaires,
April 1917, no. 8477.
* * * * *
1528 . [1990: 75] Proposed by Ji Chen, Ningbo University, China.
If a; b; c; d are positive real numbers such that a + b + c + d = 2, prove or disprove
that
a + b + c + d
2
16 :
2 2 2
(a + 1) (b + 1) (c + 1) (d + 1) 25
2 2 2 2 2 2 2 2
2 3 2 4
125 2
Since f 00(1=2) < 0, the curve is below the tangent near x = 1=2. At x = 0, it is above the
tangent. They intersect at x , the real root of
1
If a; b; c; d x ,
1
F
X
4 + 48 a , 1 = 1 (48a , 4) = 48 2 , 4 4 = 16 ;
X
At x = t, the curve is below the tangent because t is greater pthan the abscissa of the point
of in
ection.pIt is easily veried that the tangent at x = 1= 3 passes through the origin.
Since t > 1= 3, (1) is still above the curve at x = 0. Therefore
f (x) f (t) + m(x , t) :
Using this for x = b; c; d,
F f (a) + 3f (t) + m(b + c + d , 3t) = f (a) + 3f (t) :
It remains to show that for t = (2 , a)=3,
G(a) = f (a) + 3f (t) < 16=25 :
We have
G0 (a) = f 0(a) , f 0 (t) max1 f 0 (a) , 5min2 f 0 (t)
a 8
0
8 t 3
= f 0(1=8) , f 0(2=3) = 63 64 16 , 20 27 < 0 :
65 3
13 3
Therefore
G(a) G(0) = 108 <
169 25
16 :
1530. [1990: 75] Proposed by D.S. Mitrinovic, University of Belgrade, and J.E.
Pecaric, University of Zagreb.
Let =2 sin2k x dx
R
Ik = R =2
0
0 sin2k+1 x dx
where k is a natural number. Prove that
1 Ik 1 + 21k :
Ik = 2 12 32 34 45 : : : 2k2,k 1 2k2+k 1 :
!
(3)
This is just a piece of (1). If we however truncate (1) one step earlier, we obtain another
partial product of (1), which we denote Jk . Thus
Jk = 2 21 32 : : : 2k2,k 1 ; Jk = Jk 2k2+k 1 22kk +
+1
+2
1: (4)
Relations (2) and (4) show that < Ik > is a decreasing sequence, < Jk > is increas-
ing, and both converge to 1, in agreement with (1). So we have for each k
Ik > 1 > Jk = 2k2+k 1 Ik ;
155
Ik = 2j , 1 2j + 1 :
Y
(5)
j =k+1
It follows from the Lagrange (intermediate value) theorem that
1 < ln x , ln(x , 1) < 1
x x , 1 for x > 1;
which with x = 4j gives 2
1 < ln 4j < 1 : 2
4j 4j , 1 4j , 1
2 2 2
Thus, by (5),
!8
,
1 < 1 = 1 P1
= 4k1+2
ln 4j4j, 1
P
2 < 1 1 1
ln Ik =
X
j =k+1 4j 2,1 2 j =k+1 2j ,1 2j +1
and nally
< exp k = (, 1 ) < , 4k1+2 = 1 + k ;
8
<
1 1 1 1
Ik (6)
4 +2 4 +1
4k+2 exp 1
2
> exp k > 1 + k +
: 5
20 +12 k >1+ k :5
20 +12
1
2
5
20 +12
1
4 +2
Note. Equality (3) can be rewritten as Ik = (k + )((2k)!) (2k k!), ; consequently, the
1
2
2 4
estimates (6) can be also derived by brute force (with much more calculation, though)
from the Stirling formula, taken for instance in the form
p n + 1 < < 12n :
n! = 2nnne,nn ; 1212 n n 12n , 1
The problem is an old one and, as Kuczma and others point out, comes from the
usual proof of the Wallis product. Falkowitz found the given inequality, with solution, in
R. Courant, Dierential and Integral Calculus Vol I, pp. 223-224. Leinster spotted it in
Spivak's Calculus, page 328, problem 26.
* * * * *
1531. [1990: 108] Proposed by J.T. Groenman, Arnhem, The Netherlands.
Prove that
v + w bc + w + u ca + u + v ab 4(a + b + c);
u s,a v s,b w s,c
where a; b; c; s are the sides and semiperimeter of a triangle, and u; v; w are positive real
numbers. (Compare with Crux 1212 [1988: 115].)
I. Solution by Niels Bejlegaard, Stavanger, Norway.
By the A.M.-G.M. inequality the left hand side of the given inequality is greater
than or equal to
v s
u
uY
33
t
u+v
abc 2 2 2
= 33 u + v 16R r s Y
2 2 2
w (s , a)(s , b)(s , c) w rs 2
= 33 u + v 16R s ; Y
2
w
where R is the circumradius and r the inradius, and the products are cyclic over u; v; w.
So if I can show that s
33 u + v 16R s 8s ;
Y
2
w
the problem is solved. Cubing both sides gives
Y u + v 32 s = 32 sin A ;
2 X
(1)
2
w 27 R 27
where the sum is cyclic over the angles A; B; C of the triangle. Now it is known that
p
sin A 3 2 3 ;
X
and also
Y
u+v = u + v v + w w + u
w w w u u v v
= 2+ u + w + v + u + w + v
w u u v v w
2+32=8 :
157
Therefore p
32 32 3 3 = 8 u+v ;
!2
X 2
A
Y
27 sin 27 2 w
which is (1). Equality holds when a = b = c and u = v = w.
II. Generalization by Murray S. Klamkin, University of Alberta.
First we prove that
v + w (bc) p 6 p 4F p ;
!2
X
2
(2)
u 3
where F is the area of the triangle, the sums here and subsequently are cyclic over u; v; w
and a; b; c, and for now p 1. Regrouping the left side of (2) and applying the A.M.-G.M.
inequality to each resulting summand, we get
X v + w (bc) p =
2
v (bc) p + u (ca) p 2(abc)p(ap + bp + cp ) :
X
2 2
u u v
We now use abc = 4RF (R the circumradius) and the following known inequalities:
p
ap + bp + cp (a +3pb,+ c) ; p 1 (by the power mean inequality);
1
R 4pF
2
(the largest triangle inscribed in a circle is equilateral);
3 3
p
(a + b + c) 12F 3 (the largest triangle with given perimeter is equilateral):
2
u 3p, 1
6(4 F ) p 4F p= !
p p= 2
4 F p !2
3p 3p3 (12F 3) = 6 p3 ; 2
i.e., (2).
We now extend the range of (2) by showing that it is also valid for 0 p < 1.
The rest of the proof is similar to Janous's solution of Crux 1212 [1988: 115-116] and
uses results
p mentioned there. For 0 p < 1, ap ; bp; cp are the sides of a triangle of area
Fp F p ( 3=4) ,p. From this and the case p = 1 of (2), we get
1
p
v + w (bc) p 32F 6 p 4F
!2
:
X
2 2
u p 3
Now if a; b; c are the sides of a triangle, then so are
q q q
a(s , a) ; b(s , b) ; c(s , c) ;
158
u 3
Dividing by F p = (s(s , a)(s , b)(s , c))p , we obtain
2
v + w bc p 6 4s p :
!
X
(3)
u s,a 3
The proposed inequality corresponds to the special case p = 1.
As a companion inequality, we obtain
v + wa p 6 p 4F p
!
X
2
(4)
u 3
for p 0. We get as before (via regrouping and the A.M.-G.M. inequality) that
X v + wa p =
2
X v a p + u b p 2 (ab)p :
2 2
X
u u v
For p 1, the rest follows from the known inequalities
p
ab
P !
(ab) 3
p ; p 1 (power mean)
X
3
and p
ab 4F 3 :
X
The extension of (4) to the range 0 p < 1 is carried out the same way as (2) was
extended. q
By letting a = a(s , a), etc. as before, we obtain a dual inequality to (4), i.e.,
v + w ap(s , a)p 6 p 2F p :
!
X
(5)
u 3
Finally, since we always have the representation
a = y + z ; b = z + x ; c = x + y ; s = x + y + z ; F = xyz (x + y + z ) ; 2
u 3
v + w (x + y) (x + z ) 6 4(x + y + z ) p ;
p p !
X
u xp 3
v + w (y + z ) p 6 16xyz (x + y + z ) p ;
!
X
4
u 3
v + w (y + z ) p x p 6 4xyz (x + y + z ) p ;
!
X
2 2
u 3
159
X n R , ri x 2 x x :2
X
i j
i=1 ri i i<j
Proof. Indeed,
n R , ri x = R n xi , n x n n
!2
2
xi , xi = 2 xi xj ;
X X X X X X
2 2 2
i=1 ri i i ri i
i
=1 =1 i=1 i=1 i<j
where we have used the Cauchy-Schwarz inequality. Equality holds if and only if
x =r = : : : = xn=rn: 2
1 1
Putting n = 3 and
r = u ; r = v ; r = w ; x = bc ; x = ca ; x = ab
1 2 3 1 2 3
we get
v + w (bc) 2
X
2
X
caab = 2abc
X
a = 4abcs ; (6)
u
where the sums are cyclic.
[Editor's note. Janous now uses
4abcs = 16FRs 32Frs = 32F 2
to obtain (2) for the case p = 1, then mimics his proof of Crux 1212 (exactly as Klamkin
does) to obtain inequalities (2) and (3) for 0 p 1. Then he uses his lemma with
n = 3; r = u; r = v; r = w; x = a ; x = b ; x = c , and
1 2 3 1
2
2
2
3
2
x + y + z xy + yz + zx
2 2 2
u
and also extends this to obtain inequality (4) for 0 p 2 only. He points out that (4)
can be considered a \dual" of the published generalization of Crux 1051 (inequality (2) on
[1986: 252], due independently to him and Klamkin), in the same way that (3) is a \dual"
of his generalization of Crux 1212 (see [1988: 116]) and (2) is a \dual" of Crux 1221 (see
(3) on [1988: 116]). Janous now continues... ]
160
Obviously all the special cases stated in the solution of Crux 1051 can be translated
literally into their \dual" versions, e.g., (v) on [1986: 254] becomes
X 2s + a a 32F : 4 2
2s , a
Using the stronger inequality in (7), putting u = a , etc. and dividing by 2F = aha =
3
ha F
Etc., etc., etc. q
Applying the transformation a ! a(s , a), etc. to (6), we get
X v + w bc(s , b)(s , c) 2 r
Y
a(s , a)
Xq
a(s , a) ;
u
i.e., s
v + w bc 2 abc
s
(s , a) = 4 R
Xq Xq
a a(s , a) :
X
u s,a (s , a)
Q
r
It seems that the right-hand quantity is greater than or equal to 8s (which if true would
strengthen the proposed problem), but I can't prove or disprove this. Therefore I leave to
the readers the following
Problem. Prove or disprove that
a 2 s ;
s r
X
(8)
ra R
where ra ; rb ; rc are the exradii of the triangle.
Since s , a = rs=ra, etc., this inequality is equivalent to the one I can't prove.
Furthermore, (8) should be compared to item 5.47, p. 59 of Bottema et al, Geometric
Inequalities, namely
a 3rs !
s
X
ra 2 r
Also solved by C. FESTRAETS-HAMOIR, Brussels, Belgium; STEPHEN D.
HNIDEI, student, University of British Columbia; MARCIN E. KUCZMA, Warszawa,
Poland; D.M. MILOSEVIC, Pranjani, Yugoslavia; and the proposer.
Milosevic also proved inequality (7), with right-hand side 4abcs.
* * * * *
Crux
Mathematicorum
Volume 17 , Number 6 June 1991
CONTENTS
1
161
All communications about this column should be sent to Professor R.E. Woodrow,
Department of Mathematics and Statistics, The University of Calgary, Calgary, Alberta,
Canada, T2N 1N4.
This month we begin with the Canadian Mathematics Olympiad for 1991, which
we reproduce with the permission of the Canadian Olympiad Committee of the Canadian
Mathematical Society. My thanks to Ed Barbeau for sending me the contest along with
the \ocial" solutions which will be given in the next issue.
* * *
162
The next set of problems are from the twentieth annual United States of America
Mathematical Olympiad, written in April. These problems are copyrighted by the Com-
mittee on the American Mathematics Competitions of the Mathematical Association of
America and may not be reproduced without permission. Solutions, and additional copies
of the problems, may be obtained for a nominal fee from Professor Walter E. Mientka,
C.A.M.C. Executive Director, 917 Oldfather Hall, University of Nebraska, Lincoln, NE,
U.S.A. 68588-0322. As always we welcome your original \nice" solutions and generaliza-
tions.
1. In triangle ABC , angle A is twice angle B, angle C is obtuse, and the three side
lengths a, b, c are integers. Determine, with proof, the minimum possible perimeter.
2. For any nonempty set S of numbers, let (S) and (S) denote the sum and
product, respectively, of the elements of S . Prove that
X (S ) 1 1 1
(S ) = (n + 2n) , 1 + 2 + 3 + : : : + n (n + 1) ;
2
where the \P" denotes a sum involving all nonempty subsets S of f1; 2; 3; : : : ; ng.
3. Show that, for any xed integer n 1, the sequence
2
2; 22; 222 ; 222 ; : : : mod n
is eventually constant. [The tower of exponents is dened by a1 = 2, ai+1 = 2ai . Also,
ai mod n means the remainder which results from dividing ai by n.]
4. Let a = mmm+1m++nnnn+1 , where m and n areN positive integers. Prove that am + an
mm + nn. [You may wish to analyze the ratio a a,,NN N for real a 0 and integer N 1.]
5. Let D be an arbitrary point on C
Z
B
side AB of a given triangle ABC , and let E B ZZ
be the interior point where CD intersects the B
B
Z
Z
external, common tangent to the incircles of BE Z
Z
triangles ACD and BCD. As D assumes all
B Z
Z
positions between A and B , prove that point
B
B Z
Z
E traces the arc of a circle.
B
B
Z
Z
Z
A D B
* * *
163
Next is a selection of problems from the 15th All Union Mathematical Olympiad {
10th Grade. These appeared in Kvant and were translated by Hillel Gauchman and Duane
Broline of Eastern Illinois University, Charleston. My thanks to them for sending these in.
15TH ALL UNION MATHEMATICAL OLYMPIAD{TENTH GRADE
First Day
1. Find natural numbers a1 < a2 < : : : < a2n+1 which form an arithmetic sequence
such that the product of all terms is the square of a natural number. (5 points)
2. The numbers 1; 2; : : : ; n are written in some order around the circumference
of a circle. Adjacent numbers may be interchanged provided the absolute value of their
dierence is larger than one. Prove in a nite number of such interchanges it is possible
to rearrange the numbers in their natural order. (5 points)
3. Let ABC be a right triangle with right angle at C and select points D and E
on sides AC and BC , respectively. Construct perpendiculars from C to each of DE , EA,
AB , and BD. Prove that the feet of these perpendiculars are on a single circle. (10 points)
4. Let a 0, b 0, c 0, and a + b + c 3. Prove
a + b + c 3 1 + 1 + 1 :
1 + a2 1 + b2 1 + c2 2 1 + a 1 + b 1 + c
(rst inequality 3 points, second inequality 7 points)
Second Day
5. Prove, for any real number c, that the equation
x(x2 , 1)(x2 , 10) = c
cannot have ve integer solutions. (10 points)
6. A rook is placed on the lower left square of a chessboard. On each move, the
rook is permitted to jump one square either vertically or horizontally. Prove the rook may
be moved so that it is on one square once, one square twice, : : : , one square 64 times, and
(a) so that the last move is to the lower left square,
(b) so that the last move is to a square adjacent (along an edge) to the lower left
square.
(part (a) 4 points, part (b) 4 points)
7. Three chords, AA1, BB1, C
B1
CC1, to a circle meet at a point K TT
where the angles B1 KA and AKC1 are A1 T A
60, as shown. Prove that KT
T
T
KA + KB + KC = KA1 + KB1 + KC1 :
T
B T
T
(10 points) T
C1
164
8. Is it possible to put three regular tetrahedra, each having sides of length one,
inside a unit cube so that the interiors of the tetrahedra do not intersect (the boundaries
are allowed to touch)? (12 points)
* * *
We now turn to solutions of \archive" problems received from the readership. The
problems are from the 1985 Spanish Olympiad that appeared in the May 1986 number of
the Corner.
4. [1986: 97] 1985 Spanish Mathematical Olympiad, 1st Round.
L and M are points on the sides AB and AC , respectively, of triangle ABC such
that AL = 2AB=5 and AM = 3AC=4. If BM and CL intersect at P , and AP and BC
intersect at N , determine BN=BC .
Solution by Hans Engelhaupt, Gundelsheim, Germany. C
From Ceva's theorem one has S
M #SN
S
AL BN CM = 1 QQ#
#Q
P
S
S
Q
LB NC MA ## Q S
Q S
# Q
This gives BN=NC = 9=2 and so BN=BC = 9=11. A #
L
Q
SB
where 3m is the constant value of the sums. [Editor's note. To see this, rst observe that
we can assume that the row sum is zero, hence m = 0, since subtracting the same amount
from all entries gives another \sum-magic" matrix. We may then suppose that the matrix
has the form 2 3
,
64 ,a
x x+y , y
a+b ,b 75 :
x + a ,(x + y + a + b) y + b
Now from the diagonal sums,
a + 2b + y , x = 0 and 2a + b + x , y = 0:
Adding these two equations gives a = ,b, so a = y , x and b = x , y, from which the
claim is immediate.]
If the square is \product-magic" then from the diagonals,
(m , x)m(m + x) = (m , y)m(m + y) : (1)
Case 1. m = 0. From the row and column products,
xy(x + y) = 0 = xy(x , y) :
Solving, we obtain x = 0 or y = 0 (or both). This gives the two forms
2 3 2 3
r ,r 0 0 s ,s
64 ,r 0 r 75 and 64 ,s 0 s 75 :
0 r ,r s ,s 0
Case 2. m 6= 0. Then from (1), x2 = y2 . If x = y, then from the product for the second
row and column,
m(m2 , 4x2) = m3
so m2 , 4x2 = m2 and x = 0, giving all entries equal. If x = ,y we similarly obtain
m3 = m(m2 , 4x2) and x = 0 giving all entries equal.
In summary, a \sum-magic" square that is also \product-magic" has all entries
equal unless one of the diagonals contains only zeroes, and then the matrix is uniquely
determined by an o-diagonal entry.
* * *
We now turn to solutions to the problems from the October 1989 number of the
Corner. These were problems proposed to the jury, but not used, at the 30th IMO at
Braunschweig, (then) West Germany [1989: 225{226]. We give the solutions we've re-
ceived which dier from the \ocial" solutions published in the booklet 30th International
Mathematical Olympiad (see the review on [1991: 42]).
167
6. Proposed by Greece.
Let g : C ! C, ! 2 C, a 2 C, with !3 = 1 and ! 6= 1. Show that there is one and
only one function f : C ! C such that
f (z ) + f (!z + a) = g(z ) ; z 2 C:
Find the function f .
Solutions by George Evagelopoulos, Athens, Greece, and by Edward T.H. Wang,
Wilfrid Laurier University, Waterloo, Ontario.
If f : C ! C satises
g(z ) = f (z ) + f (!z + a) for all z 2 C; (1)
then
g(!z + a) = f (!z + a) + f (!2 z + !a + a) for all z 2 C: (2)
Since !3 = 1 and 1 + ! + !2 = 0, we have
!2 z + !a + a = !2 z + a(1 + !) = !2 (z , a) = z ,! a ;
and so
z , a z , a z , a
g ! = f ! + f w ! + a = f (!2 z + !a + a) + f (z ): (3)
Therefore (1) minus (2) plus (3) yields
g(z ) , g(!z + a) + g z ,! a = 2f (z );
and so z , a
1
f (z ) = 2 g(z ) , g(!z + a) + g ! :
On the other hand, it is easily checked that this choice of f satises (1) and hence it is the
unique function with this property.
7. Proposed by Hungary.1
Dene the sequence fan gn=1 of integers by
X
ad = 2n :
djn
Show that njan . [Editor's note. Of course xjy means that x divides y.]
168
3S 2(ab + ac + ad + bc + bd + cd)
169
whence
4S (a + b + c + d)2 :
Thus 4 1989 m4 and m 9.
But (a + b + c + d)2 > a2 + b2 + c2 + d2 so m4 > 1989 and m > 6. Since a2 + b2 + c2 + d2
is odd, so is m2 = a + b + c + d, thus m = 7 or m = 9. Suppose for a contradiction that
m = 7. Now (49 , d)2 = (a + b + c)2 > a2 + b2 + c2 = 1989 , d2. Thus d2 , 49d + 206 > 0.
It follows that d > 44 or d 4. However, d < 45, since 452 = 2025. Also, d 4 implies
a2 + b2 + c2 + d2 64 < 1989.
It follows that m = 9. Now n2 = d > 16, since d 16 implies a + b + c + d 64 < 81.
As before d 44 so n2 = 25 or 36. If d = n2 = 25 then let a = 25 , p, b = 25 , q,
c = 25 , r, with p; q; r 0. Furthermore a + b + c = 56 implies p + q + r = 19.
(25,p)2+(25,q)2 +(25,r)2 = 1364 implies p2 +q2 +r2 = 439. Now (p+q +r)2 > p2 +q2 +r2
gives a contradiction.
Thus the only possibility is that n = 6, and there is a solution with a = 12, b = 15,
c = 18, d = 36. Thus m = 9 and n = 6.
11. Proposed by Mongolia.
Seven points are given in the plane. They are to be joined by a minimal number of
segments such that at least two of any three points are joined. How many segments has
such a gure? Give an example.
Generalization and solution by Edward T.H. Wang, Wilfrid Laurier University,
Waterloo, Ontario.
We use the terminology of graph theory. For each positive integer n, let `n denote
the minimum number of edges that a graph on n vertices must have so that among any
three vertices at least two are joined by an edge. Then we claim
$ 2%
`n = ( n , 1) :
4
To see this it is useful to consider mn, the maximum number of edges thatagraph
on n vertices can have, but contain no triangle. We rst show that `n + mn = n2 . Let
G be a graph on n vertices with `n edges so that there are no \empty triangles". Then
the complementary graph G, obtained by interchanging edges and non-edges,
is a graph
without triangles, so n2 , `n mn. Similarly one argues that `n n2 , mn .
Now by Turan's theorem mn = bn2=4c (c.f.
j n2 Ex.
k 30, p. 68 of Combinatorial Problems
and Exercises by L. Lovasz). Thus `n = 2 , 4 . By checking separately n even and n
n
odd, one obtains `n = b(n , 1)2=4c in all cases. This situation is realized by the disjoint
union of two complete graphs on (n , 1)=2 and (n + 1)=2 points respectively if n is odd,
and by two copies of the complete graph Kn=2 if n is even.
* * *
This completes the correct solutions received for problems from the October 1989
number and the space we have this month. Send me your problems and nice solutions!
170
* * * * *
BOOK REVIEW
Mathematical Olympiad in China, edited by W. Ouyang, published by Hunan Ed-
ucation Publishing House, Changsha, 1990, ISBN 7-5355-1152-X, hardcover, 312+ pages.
Reviewed by Andy Liu, University of Alberta.
This book was put together by the Chinese Mathematical Olympiad Committee as a
gift to the Leaders and Deputy Leaders of the teams participating in the 31st International
Mathematical Olympiad in Beijing. It consists of four parts.
The rst part consists of an essay sketching the history and development of the
Olympiad movement in China, climaxing in the 1990 IMO. The second part consists of
eight articles on mathematics that arose from the Olympiads. Some are translations of
articles of exceptional merit that were published previously in Chinese, while others are
contributed especially for this volume.
Here is a problem considered in the article \A conjecture concerning six points in a
square" by L. Yang and J. Zhang, translated from the Chinese version published in 1980.
Everyone knows that 9 points in a square of area 1 are sucient to guarantee that 3 of
them will determine a triangle of area at most 1/4. Actually, 9 points are not necessary.
What is the minimum?
The answer is 6. A key lemma states that among 4 points determining a convex
quadrilateral in a triangle of area 1, 3 will determine a triangle of area at most 1/4. The
simple proof is an elegant combination of Euclidean geometry and the Arithmetic-Mean
Geometric-Mean Inequality.
Another easy corollary of this lemma is that 5 points in a triangle of area 1 are
sucient to guarantee that 3 of them will determine a triangle of area at most 1/4. Here,
it is easy to see that 5 points are also necessary.
The third part of the book consists of three articles on the art and science of
problem-proposing and problem-solving. The fourth part consists of the problems and
solutions of the rst ve annual Mathematical Winter Camps in China.
A very limited number of extra copies of this book have been printed. The price at
US $5.80 (plus 25% overseas postage) is a true bargain. It may be ordered with prepayment
from:
Hunan Education Publishing House,
1 Dongfen Road, Changsha, Hunan,
410005, People's Republic of China.
* * * * *
171
PROBLEMS
Problem proposals and solutions should be sent to B. Sands, Department of
Mathematics and Statistics, University of Calgary, Calgary, Alberta, Canada T2N 1N4.
Proposals should, whenever possible, be accompanied by a solution, references, and other
insights which are likely to be of help to the editor. An asterisk (*) after a number indicates
a problem submitted without a solution.
Original problems are particularly sought. But other interesting problems may also
be acceptable provided they are not too well known and references are given as to their
provenance. Ordinarily, if the originator of a problem can be located, it should not be
submitted by somebody else without permission.
To facilitate their consideration, your solutions, typewritten or neatly handwritten
on signed, separate sheets, should preferably be mailed to the editor before January 1,
1992, although solutions received after that date will also be considered until the time when
a solution is published.
1651. Proposed by George Tsintsifas, Thessaloniki, Greece.
Let ABC be a triangle and A1; B1; C1 the common points of the inscribed circle with
the sides BC; CA; AB , respectively. We denote the length of the arc B1 C1 (not containing
A1) of the incircle by Sa , and similarly dene Sb and Sc . Prove that
p
a + b + c 9 3:
Sa Sb Sc
1652. Proposed by Murray S. Klamkin, University of Alberta.
Given xed constants a; b; c > 0 and m > 1, nd all positive values of x; y; z which
minimize !
xm + ym + z m + am + bm + cm , x + y + z + a + b + c m :
6 6
1653. Proposed by Toshio Seimiya, Kawasaki, Japan.
Let P be the intersection of the diagonals AC; BD of a quadrangle ABCD, and let
M; N be the midpoints of AB; CD, respectively. Let l; m; n be the lines through P; M; N
perpendicular to AD; BD; AC , respectively. Prove that if l; m; n are concurrent, then
A; B; C; D are concyclic.
1654. Proposed by Walther Janous, Ursulinengymnasium, Innsbruck, Austria.
Let x; y; z be positive real numbers. Show that
X
q x 1;
x + (x + y)(x + z )
where the sum is cyclic over x; y; z , and determine when equality holds.
172
SOLUTIONS
No problem is ever permanently closed. The editor will always be pleased to consider
for publication new solutions or new insights on past problems.
93. [1975: 97; 1976: 45, 111; 1984: 293] Proposed by H.G. Dworshak, Algonquin
College.
Is there a convex polyhedron having exactly seven edges?
V. Solution by William Pippin and James Underwood, students, The Ohio State
University.
Since the vertices and edges of a polyhedron form the vertices and edges of a simple
graph in which each vertex has degree at least 3, the following result provides an immediate
negative answer.
A simple graph in which each vertex has degree at least 3 cannot have exactly 7
edges.
Proof. Suppose that the graph has V vertices. Then the number of edges must be
at least 3V=2, since each vertex is incident to at least 3 edges and each edge has 2 ends.
But if V 5, then
43V=2 15=2 > 7. On the other hand, if V 4 then the graph cannot
have more than 2 = 6 edges.
* * * * *
964. [1984: 217; 1985: 296] Proposed by R.B. Killgrove, Alhambra, California.
Let T be the image of the Euler -function, that is,
T = f(n) : n = 1; 2; 3; : : :g :
Prove or disprove that T is a Dirichlet set, as dened in the proposer's article \Elementary
Dirichlet sets" [1984: 206{209, esp. p. 206 and last paragraph p. 209].
Solution by Sam Maltby, student, University of Calgary.
We need to show that if S is an arithmetic sequence,
S = fa; a + d; a + 2d; a + 3d; : : :g
with a and d positive integers, then jS \ T j 2 f0; 1; 1g. We in fact prove: if S \ T contains
an element greater than or equal to d, then it contains innitely many elements.
Assume t 2 S \ T with t d. Since t 2 T ,
t = (m) = p1 1 (p1 , 1)p2 2 (p2 , 1) : : : pr r (pr , 1) (1)
where m = p1 1 +1p2 2 +1 : : : pr r +1, the pi's are distinct primes and each i 0.
Now if we can nd innitely many elements of T congruent to a modulo d, we are
done. Since t a mod d, it is sucient to nd these elements congruent to t mod d.
Let
d = kp11 p22 : : : pr r ; i 0 for all i ; (k; pi) = 1 for all i :
174
The xed lines of X are the tangents from X to !. In order that these be perpen-
dicular it is necessary and sucient that X be on the circle (of Monge) that is the locus
of points from which the tangents to ! form a right angle.
Therefore the required points P are the points of intersection of S 0 T with .
Construction of P . Since S 0T is the axis of it must contain D = AC 0 \ A0 C and
E = AB 0 \ A0B (as well as BC 0 \ B 0 C ). But S 0T is also the polar of S with respect to !;
thus if M is the midpoint of S 0 T , then SM passes through the centre O of ! (which is also
the centre of ). Analogously, letting F = AA0 \ BB 0 , G = A0B \ FS and R = GT \ AA0,
then RT is the polar of A. Letting N be
E
the midpoint of RT , AN passes through ,
,
O. Thus O = SM \ AN . It remains to ,
,
nd a point on the circle . Let A ; B ; T 00 00 00 ,
be the orthogonal projections of A0 ; B 0 ; T 0 ,
,
on l (note T 00 = T 0). Let U and U 0 be the ,
xed points of the projectivity A; B; T ! ,
A00; B 00 ; T 00 . Then the perpendiculars to l ,
,
through U; U 0 are tangent to ! and there- B,0 ,
0
fore U and U lie on . (The xed points of
,@
a projectivity are easily found; for exam- , @
ple, project A; B; T; A00 ; B 00 ; T 00 onto any , ""@ A
B0 P
, R"
@ S 0
convenient circle from a point V of , "
F "E
B @
G D C
B
0
. Let a; b; t; a00 ; b00; t00 be these projections , "
, "
"
E B @
E Oa
@
onto . The line through points at \ a t 00 00
,""
N E Maa
a B
B aa @
l0
and bt \ b t cuts in u and u , which
00 00 0
"
,
"
"
q q
E
E q 00 CB q 0
B aa@
project back to give us U and U .) 0 A ,
l B U B 00 T A U
@ 0
a
S = T 00
=T
P
* * * * *
1532. [1990: 109] Proposed by Murray S. Klamkin, University of Alberta.
Determine all (possibly degenerate) triangles ABC such that
(1 + cos B )(1 + cos C )(1 , cos A) = 2 cos A cos B cos C:
Solution by Kee-Wai Lau, Hong Kong.
We show that the equality of the problem holds if and only if A = 0 and B = C =
=2.
Since
2 cos A cos B cos C = (1 + cos B )(1 + cos C )(1 , cos A) 0 ;
so 0 A; B; C =2. Clearly A 6= =2, and if A = 0 then B = C = =2. In what
follows we assume that 0 < A < =2 and 0 B; C < =2. For 0 x < =2, let
f (x) = ln(sec x + 1). We have
f 00(x) = 1(1++cos x , cos2 x > 0
cos x) cos2 x
177
p p
,
d01 = b2 , s2 , r and d03 = b2 , s2 + r; d3
0
6 d1
- 0
so
d012 + d032 = 2b2 , 2s2 + 2r2: b x
0
d4
0
Similarly
p p
d02 = b2 , r2 , s and d04 = b2 , r2 + s;
?
so
d022 + d042 = 2b2 , 2r2 + 2s2:
Thus
d012 + d022 + d032 + d042 = 4b2; (1)
independent of the placement of P .
In going from (x0; y)-coordinates to (x; y)-coordinates as shown on the next page,
we have
a 2! a 2! 1
d2i = 1 + b2 x02 = 1 + b2 2 d0i2;
and so by (1),
!
d21 + d22 + d23 + d24 = 1 1 + a2 4b2 = 2(a2 + b2):
2 b2
180
y y
@ HH
0
x0 @d2 x0 d2
HH
@ d01 HH
Hd
1
@,
x0 , a x0 H d4
04 b HH
, d@ @ H
x0 -
, x
d03
,
, d3
,
,
,
II. Solution par Jacques Chone, Lycee Blaise Pascal, Clermont{Ferrand, France.
Soit (; ) les coordonnees de P , avec 2 =a2 + 2=b2 , 1 0. Un vecteur unitaire
des cordes considerees a pour coordonnees:
!
a "b
p 2 2 ; p 2 2 ; " 2 f,1; 1g:
a +b a +b
Les points des cordes considerees ont pour coordonnees:
!
a "b
+ p 2 2 ; + p 2 2 ; reel.
a +b a +b
Les reels di (i = 1; : : : ; 4) sont les valeurs absolues des solutions de l'equation
!2 !2
1 + p a 1 b"
+ b2 + p 2 2 = 1;
a2 a2 + b2 a +b
c'est a dire de
!
2 2 + p 2 + " + 2 + 2 , 1 = 0:
a2 + b2 a2 + b2 a b a2 b2
La somme des carres des solutions de l'equation Ax2 + Bx + C = 0 etant (,B=A)2 , 2C=A,
on obtient
2 !2 !23 !
2 2 2 2 4 5 2 2
d1 + d2 + d3 + d4 = a + b + a , b (a + b ) , 2 a2 + b2 , 1 (a2 + b2)
2 2
= 2(a2 + b2);
ce qui est le resultat desire.
Also solved by SEUNG-JIN BANG, Seoul, Republic of Korea; FRANCISCO
BELLOT ROSADO, I.B. Emilio Ferrari, Valladolid, Spain; JORDI DOU, Barcelona,
181
Hence p p
v 100 , 6400
18
+ 36p or v 100 + 6400 + 36p :
18
Taking p = 0 provides the useful bounds v 10=9 or v 10.
We rst take v 10. Then by (2) q = 90+ k for some integer k such that 0 k 9.
But since q is the last nonzero remainder and k is the units digit of q, we must have k = q=v.
Hence 90 + k = q = kv or
90 = k : (3)
v,1
Since 0 k 9, (3) yields (v; k) = (11; 9); (15; 6); (19; 5); (31; 3); (46; 2) and (91; 1). Sub-
stituting q = 90 + k into (2) and rearranging yields
p = 100 , (10 , k)v : (4)
Using each of the above possibilities for (v; k), we discard the last three since they yield
p < 0. Hence we have the following solutions:
v = 11; k = 9; p = 89; P = 1089;
v = 15; k = 6; p = 36; P = 1536;
v = 19; k = 5; p = 5; P = 1805:
It remains to consider v 10=9 or v = 1. Here we have nine trivial solutions
(v; k; p) = (1; k; k) where 1 k 9. They correspond to the rst nine problems published
in Eureka (Crux's predecessor).
Hence the next and last problem number satisfying the conditions of the problem
will be Crux 1805 appearing in volume 19.
Also solved by EMILIO FERNANDEZ MORAL, I.B. Sagasta, Logro~no, Spain;
RICHARD I. HESS, Rancho Palos Verdes, California; L.J. HUT, Groningen, The Nether-
lands; FRIEND H. KIERSTEAD JR., Cuyahoga Falls, Ohio; KEE-WAI LAU, Hong Kong;
J.A. MCCALLUM, Medicine Hat, Alberta; P. PENNING, Delft, The Netherlands; and
K.R.S. SASTRY, Addis Ababa, Ethiopia. One incorrect solution was received.
Readers may like to contribute an appropriate problem to be numbered 1805, to carry
on the \tradition" newly established by Janous! The editor suggests that such a problem (i)
involve the number 1805 and possibly the volume number (19) or even the year (which will
be 1993), and (ii) have exactly one integer solution larger than 1805. (Not too large!)
* * * * *
1537. [1990: 110] Proposed by Isao Ashiba, Tokyo, Japan.
ABC is a right triangle with right angle at A. Construct the squares ABDE and
ACFG exterior to ABC , and let P and Q be the points of intersection of CD and AB ,
and of BF and AC , respectively. Show that AP = AQ.
183
Heuver found the problem with solution on pp. 83{84 of Coxeter and Greitzer,
Geometry Revisited, L.W. Singer, New York.
McCallum notes that another problem, rather familiar, has the same answer
pq=(p + q) as was obtained for AP and AQ above: namely, given two vertical posts of
heights p and q on a
at plain, and wires connecting the top of each post with the bottom
of the other, nd the height of the intersection of the wires. He wonders if there is a nice
way to turn each problem into the other.
* * * * *
1538. [1990: 110] Proposed by K.R.S. Sastry, Addis Ababa, Ethiopia.
Find all functions y = f (x) with the property that the line through any two points
(p; f (p)); (q; f (q)) on the curve intersects the y-axis at the point (0; ,pq).
Solution by Ka-Ping Yee, student, St. John's{Ravenscourt School, Winnipeg,
Manitoba.
All functions y = f (x) = x(c + x), where c is some constant, have this property,
and no other functions do. We can show this as follows.
If a line passes through the points (p; f (p)) and (q; f (q)), then
f (p) = mp + b ; f (q) = mq + b :
Eliminating the slope m and solving for the y-intercept b we get
, qf (p) = ,pq (as given) ,
b = pf (qp) , q
which implies
pf (q) , qf (p) = pq(q , p)
and so
f (q) , q = f (p) , p :
q p
Since this is true for all [nonzero] real p and q, f (x)=x , x has a constant value for all
x[6= 0]. If we let this constant value be c, then f (x) = x(c + x).
[Editor's note. To this solution should be added two points, missed by many solvers:
(i) from the statement of the problem, f (0) must be 0, and the functions f (x) = x(c + x)
all have this property, thus f (x) = x(c + x) for all x (not just x 6= 0); (ii) all functions
f (x) = x(c + x) do in fact satisfy the problem.]
Also solved by SEUNG-JIN BANG, Seoul, Republic of Korea; J. CHONE, Lycee
Blaise Pascal, Clermont{Ferrand, France; JORDI DOU, Barcelona, Spain; HANS
ENGELHAUPT, Franz-Ludwig-Gymnasium, Bamberg, Germany; EMILIO FERNANDEZ
MORAL, I.B. Sagasta, Logro~no, Spain; C. FESTRAETS-HAMOIR, Brussels, Belgium;
RICHARD I. HESS, Rancho Palos Verdes, California; STEPHEN D. HNIDEI and
ROBERT PIDGEON, students, University of British Columbia; WALTHER JANOUS,
Ursulinengymnasium, Innsbruck, Austria; MARCIN E. KUCZMA, Warszawa, Poland;
185
KEE-WAI LAU, Hong Kong; BEATRIZ MARGOLIS, Paris, France; P. PENNING, Delft,
The Netherlands; R.P. SEALY, Mount Allison University, Sackville, New Brunswick;
DAVID C. VAUGHAN, Wilfrid Laurier University, Waterloo, Ontario; KENNETH M.
WILKE, Topeka, Kansas; and the proposer. (Three of these solutions assumed that the
required functions were dierentiable, but we'll let it go this time.)
The problem is the converse of a result proved in Herb Holden's article \Chords
of the parabola", Two Year College Math. Journal, June 1982, pp. 186{190, and was
suggested by that article.
* * * * *
1539. [1990: 110] Proposed by D.M. Milosevic, Pranjani, Yugoslavia.
If ; ;
are the angles, s the semiperimeter, R the circumradius and r the inradius
of a triangle, prove or disprove that
X 2 2 2R , r 2
tan tan ;
2 2 s
where the sum is cyclic.
Solution by Stephen D. Hnidei, student, University of British Columbia.
Starting with
v
u v
u v
u
u
t ( s , b )( s , c ) u
t ( s , a )( s , c )
u
t (s , a)(s , b) ; (1)
tan 2 = s(s , a) ; tan 2 = s(s , b) ; tan 2 = s(s , c)
where a; b; c are the sides opposite the angles ; ;
respectively, the given inequality
becomes
(s , a)2 + (s , b)2 + (s , c)2 (2R , r)2 ;
or equivalently
a2 + b2 + c2 (2R , r)2 + s2 : (2)
Using the fact that
a2 + b2 + c2 + r(4R + r) = s2 ; (3)
2
(2) becomes
a2 + b2 + c2 8R2 + 4r2 :
This is inequality 5.14, page 53 of Bottema et al, Geometric Inequalities.
[Editor's note. (1) can be proved from the formulas
tan = r ; etc. and r2 s = (s , a)(s , b)(s , c) :
2 s,a
(3) is proved on page 52 of Mitrinovic et al, Recent Advances in Geometric Inequalities.]
Also solved by NIELS BEJLEGAARD, Stavanger, Norway; JACK GARFUNKEL,
Flushing, N.Y.; WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria;
186
In general, once we prove that g(k) is nite for every positive odd integer k, then
we shall have that
f (k) = g(k) + cardf odd ti : 0 i g(k)g: (1)
We shall show that the coincidence between the sequence < tn > (reversed) and
the sequence of residues of the 2n's modulo k, for example,
< 2n modulo 15 > : 1; 2; 4; 8; 1; ;
< 2n modulo 13 > : 1; 2; 4; 8; 3; 6; 12; 11; 9; 5; 10; 7; 1; ;
< 2n modulo 9 > : 1; 2; 4; 8; 7; 5; 1;
is not accidental. In fact we can prove that, for every odd k > 1 (already excluding the
case k = 1, for which f (1) = 2), we have:
(a) tn < k for all n;
(b) 2ntn 1 mod k for all n; and so
(c) if h is a positive integer such that 2h 1 mod k, then tn 2h,n mod k for
n h, and so the period of the sequence < tn > is
g(k) = minfh > 0 : 2h 1 mod kg(:= the order of 2 mod k):
Proof of (a). By induction. Note that
t0 = 1 < k ; t1 = 1 +2 k < k:
If we suppose tn,1 < k and we put tn,1 " mod 2 (" = 0 or 1), then
tn = tn2,1 + " k2 < k2 + k2 = k:
Proof of (b). By induction. Note that t0 = 1 and 2t1 = 1 + k 1 mod k. If we
suppose that 2n,1tn,1 1 mod k, then we have (as above)
!
n
2 tn = 2 n t n ,1 k
+ " = 2n,1tn,1 + k2n,1"
2 2
2 tn,1 mod k 1 mod k:
n , 1
Proof of (c). According to (b), 2ntn 1 mod k, and (as k is odd) by division tn
2 mod k for every n h. By this and (a), the terms t1 ; t2 ; : : : ; th of the sequence < tn >
h , n
are, in opposite order, just the same as the terms 20 mod k, 21 mod k; ; 2h,1 mod k of
the sequence < 2n mod k >, for every positive h such that 2h 1 mod k. The periods of
< tn > and < 2n mod k > are therefore equal.
Remembering (1) above, we have now that sequence < an > is pure periodic for all
odd k, with period f (k) = ` + s with
` = the order of 2 mod k;
s = card fodd residues of 2i mod k ; 0 < i `g:
188
f (k) = ` + s < k2 + k2 = k;
a contradiction. Thus ` = '(k). In this case the sequence < 2i mod k : 1 i ` >
contains all the positive numbers less than k and prime to k. If m < k is prime to k, then
k , m is prime to k too and has opposite parity to m since k is odd; therefore
s = '(2k) = cardfall positive odd integers < k and prime to kg:
So we have by the hypothesis
which is seen, by adding the rows together, to have the value zero. Thus the lines are
concurrent.
Also solved by FRANCISCO BELLOT ROSADO, I.B. Emilio Ferrari, Valladolid,
Spain; JORDI DOU, Barcelona, Spain; C. FESTRAETS-HAMOIR, Brussels, Belgium;
HIDETOSI FUKAGAWA, Aichi, Japan; P. PENNING, Delft, The Netherlands; JOHN
RAUSEN, New York; D.J. SMEENK, Zaltbommel, The Netherlands; D. SOKOLOWSKY,
Williamsburg, Virginia; E. SZEKERES, Turramurra, Australia; and the proposer.
The generalization given by Eddy was also found by Rausen, who later located it
in Aubert et Papelier, Exercices de Geometrie Analytique, Vol. 1, 10th ed., Paris, 1957,
problem 52, p. 35. Rausen also points out that the lines I1P1, I2P2, I3P3 may be parallel
rather than concurrent: just pick P1 on line A2A3 and P2 on A3A1 such that lines I1P1
and I2P2 are parallel, then dene P to be A1P1 \ A2P2.
* * * * *
1543. [1990: 143] Proposed by George Tsintsifas, Thessaloniki, Greece.
Show that the circumradius of a triangle is at least four times the inradius of the
pedal triangle of any interior point.
Solution by Walther Janous, Ursulinengymnasium, Innsbruck, Austria.
Let R1; R2; R3 and r1 ; r2; r3 be the distances of the interior point P from the vertices
and sides, respectively, of triangle A1A2A3 having sides a1; a2 ; a3 . The sides of the pedal
triangle are given by
Ri sin Ai = a2i R
Ri ; i = 1; 2; 3
(e.g. [1], p. 296), and its area is given by
X X
Fped = 12 r2r3 sin A1 = 41R a1 r2r3
where the sums are cyclic. (This can be seen by the trigonometric area formula and the
fact that r2 and r3 make the angle , A1, etc.) Therefore the claimed inequality R 4rped
reads Rsped 4Fped or X X
R a1 R1 4 a1r2 r3: (1)
Now it is known ([2], item 12.19) that
X X
a1R1 2 a1r1 = 4F:
Therefore (1) would follow from RF P a1 r2r3 , i.e.,
X
a1a2 a3 4 a1 r2r3 :
But this inequality can either be found in [1] (p. 333, item 7.1 or p. 339, item 7.28) or in
Crux ([1990: 64] or [1987: 260]).
191
References:
[1] Mitrinovic et al, Recent Advances in Geometric Inequalities, Kluwer, Dordrecht, 1989.
[2] Bottema et al, Geometric Inequalities, Wolters{Noordho, Groningen, 1968.
Also solved by JOHN G. HEUVER, Grande Prairie Composite H.S., Grande Prairie,
Alberta; MURRAY S. KLAMKIN, University of Alberta; D.J. SMEENK, Zaltbommel, The
Netherlands; and the proposer.
* * * * *
1544. [1990:3 143] Proposed by Stanley Rabinowitz, Westford, Massachusetts.
One root of x + ax + b = 0 is times the dierence of the other two roots (jj 6= 1).
Find this root as a simple rational function of a, b and :
Solution by Jean-Marie Monier, Lyon, France.
Letting the roots be x1; x2 ; x3, we have
!
x21 = [(x2 , x3)]2 = 2[(x2 + x3)2 , 4x2x3] = 2 x21 + x4 b
1
Thus
ax1 + b = ,x31 = 42,b1 ;
2
so !
x1 = a 2 , 1 , b = ba(3(2 ,+1)1) :
1 4 2b 2
[Editor's note. Monier's solution actually contained a small error, but in other respects
was the nicest one received, so the error has been corrected in the above write-up.]
Also solved by HANS ENGELHAUPT, Franz-Ludwig-Gymnasium, Bamberg,
Germany; MATHEW ENGLANDER, student, University of Waterloo; M. MERCEDES
SANCHEZ BENITO, Madrid, and EMILIO FERNANDEZ MORAL, I.B. Sagasta,
Logro~no, Spain; C. FESTRAETS-HAMOIR, Brussels, Belgium; JACK GARFUNKEL,
Flushing, N.Y.; RICHARD I. HESS, Rancho Palos Verdes, California; STEPHEN D.
HNIDEI, student, University of British Columbia; L.J. HUT, Groningen, The Netherlands;
WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria; O. JOHNSON, student,
King Edward's School, Birmingham, England; FRIEND H. KIERSTEAD JR., Cuyahoga
Falls, Ohio; KEE-WAI LAU, Hong Kong; MARIA ASCENSION LOPEZ
CHAMORRO,
I.B. Leopoldo Cano, Valladolid, Spain; BEATRIZ MARGOLIS, Paris, France; J.A.
MCCALLUM, Medicine Hat, Alberta; P. PENNING, Delft, The Netherlands; KENNETH
M. WILKE, Topeka, Kansas; and the proposer.
As was pointed out by several solvers, the solution is valid only for a 6= 0.
192
* * * * *
1545. [1990: 144] Proposed by Marcin E. Kuczma, Warszawa, Poland.
A sphere is said to be inscribed into the skeleton of a convex polyhedron if it is
tangent to all the edges of the polyhedron. Given a convex polyhedron P and a point O
inside it, suppose a sphere can be inscribed into the skeleton of each pyramid spanned by
O and a face of P:
(a) Prove that if every vertex of P is the endpoint of exactly three edges then there
exists a sphere inscribed into the skeleton of P:
(b) Is this true without the assumption stated in (a)?
Solution by the proposer.
(a) Label the faces F1; : : : ; Fn in such a way that F1 and F2 be adjacent and every
successive Fk (3 k n) be adjacent to some two previous faces. [For instance, F3 would
be the third face meeting at one of the endpoints of the common edge of F1 and F2 .|Ed.]
Let Pj be the pyramid of base Fj and vertex O, let Dj be the disc inscribed in Fj (it does
exist) and let `j be the line through the center of Dj perpendicular to the plane of Fj . The
two spheres inscribed into the skeletons of P1 and P2 touch the common edge of F1 and
F2 in the same point (they both contain the incircle of triangle P1 \ P2). The two planes
passing through that point and through line `1 , resp. `2 , are perpendicular to the edge,
hence they are identical. Thus `1 and `2 intersect.
Face F3 is adjacent to F1 and F2. The same argument shows that `3 cuts `1 and
`2. Since `1; `2 ; `3 do not lie in the same plane, we infer that they have exactly one point
in common.
By induction, each successive line `4; : : : ; `n passes through this point, which is
therefore equally distant from all edges of P [because it is equidistant from all the edges
of any one face]; the claim results.
(b) The assumption given in (a) is indeed essential. As an example consider the
following construction. ABC is an equilateral triangle of side a, O is its center. Points
D and E lie symmetrically with respect to plane ABC , at a distance h from it; O is the
midpoint of DE . Polyhedron P is dened as the union of pyramids ABCD and ABCE .
A sphere can be inscribed into the skeleton p of any one of the six pyramids spanned by O
and the faces of P if and only if h = ( 3 , 1)a=4 (verication is routine and therefore
omitted). A sphere inscribed into the skeleton of P exists if and only if h = a=3.
There was one incorrect solution sent in, which failed to take into account the con-
dition in part (a).
* * * * *
Crux
Mathematicorum
Volume 17 , Number 7 September 1991
CONTENTS
1
193
ON AN IDEA OF GROENMAN
R.H. Eddy
In Crux [1987: 256] and [1987: 321], the late J.T. Groenman, a prolic and much
appreciated contributor, proposed the following two related problems.
1272. (solution [1988: 256]) Let A A A be a triangle. Let the incircle have center
1 2 3
I and radius , and meet the sides of the triangle at points P ; P ; P . Let I ; I ; I be the
1 2 3 1 2 3
excenters and ; ; the exradii. Prove that
1 2 3
(a) the lines I P ; I P ; I P concur at a point S ;
1 1 2 2 3 3
d :d :d = : : :
1 2 3 1 2 3
1295. (solution [1989: 17{19]) Let A A A be a triangle with I1; I2; I3 the excen-
1 2 3
ters and B1; B2 ; B3 the feet of the altitudes. Show that the lines I1B1 ; I2B2; I3 B3 concur at
a point collinear with the incenter and circumcenter of the triangle.
A third point, with a similar construction, may be found in Nagel [9], where in
Groenman's notation we may write
Let S1; S2 ; S3 denote respectively the midpoints of the sides A2A3; A3 A1; A1A2 of the
given triangle; then the lines I1S1 ; I2S2; I3S3 concur at the MITTENPUNKT of the given
triangle.
We found out about this interesting point, while on sabbatical leave in Europe in
the fall of 1986, from Peter Baptist, a faculty member at Bayreuth University who has
done considerable work in the geometry of the triangle, particularly as it relates to special
points. (For an example of his work, see [3].) A direct translation would seem to yield the
term middlespoint in English, and this is a good description since the point is constructed
using middles, i.e., centres of circles and midpoints of line segments. This term is used in
[5] and [6], but since it is rather clumsy we shall use mittenpunkt throughout this note.
Let us consider the point from a dierent angle. Consider the pair of triangles
S1S2S3 and I1I2I3 which are in perspective with the triangle A1A2A3 from the centroid
and incentre respectively. Then the triangles S1 S2S3 and I1I2I3 are in perspective from the
mittenpunkt of A1A2A3. While Nagel's proof is synthetic|in fact he seems to dislike other
types, especially trigonometric|the problem is easily solved analytically using trilinear
coordinates. Since these have appeared in Crux several times, further details will not be
given here. It suces to remark that the trilinear coordinates of the mittenpunkt are
(s , a1 ; s , a2 ; s , a3) where, as usual, s is the semiperimeter of the given triangle. Since
the construction seems to be something of an unnatural \buddying-up", so to speak, of the
incentre and the centroid, one immediately gets the feeling that some sort of generalization
is lurking around. Further justication for such a possibility is immediately obtained by
replacing the centroid by the symmedian point, see Crux [11], which has the rather nice
194
The restriction to P being in the interior of the triangle is unnecessary: see the solution
of Crux 1541 [1991: 189]. For alternative formulations, see [1] or [2]. We are told that the
theorem also appears in [8] and [10]. (We thank the referees for supplying references [1],
[2], [8] and [10] which were previously unknown to us.)
For Groenman's problem 1272, P is the Gergonne point and Q is the incentre,
while for 1295, P is the orthocentre and Q is again the incentre. Also noted in [6] is the
fact that the point in 1272 is the isogonal conjugate of the mittenpunkt. This problem
was discovered at the proofreading stage of [6] and thus was able to be included.
It is interesting that Groenman seemed to be approaching the same generalization.
In his solution to 1295 he has P in general position while Q is still the incentre. A related
class of points referred to by Nagel as interior mittenpunkts (dened by replacing one of
the excentres by the incentre and interchanging the other two) is also given in [9]. This
class is also generalized in [6]. A dual notion for T , the mittenlinie (middlesline), is given
in [5].
For those readers who have access to a symbolic manipulation program, the follow-
ing is an elementary exercise.
Under what circumstances are P; Q, and T collinear?
If the trilinear coordinates of P and Q are (zi ) and (yi ) respectively, i = 1; 2; 3, then it is
not too dicult to show that the coordinates of T (xi ) are
y y y y y y y y y
y ,z + z + z ; y z , z + z ; y z + z , z :
1
1 2 3
2
1 2 3
3
1 2 3
1 2 3 1 2 3 1 2 3
Hence yi=yi = zi =zi for some i, and thus P; Q; T must lie on a line through a vertex of
+1 +1
the given triangle.
An exhaustive search for some other references to this point has met with only
one partial success. In Gallatly [7], it is mentioned only as the Lemoine point of the
triangle I I I . The coordinates of are given as (s , a ; s , a ; s , a ), thus it is Nagel's
1 2 3 1 2 3
195
mittenpunkt but under another guise. I am surprised that a point with such a pretty
and ready construction seems not to have found its way into the modern geometry of the
triangle. Perhaps the reader can shed some new light on the problem.
References :
[1] N. Altshiller-Court, College Geometry, supplementary exercise 7, page 165.
[2] Aubert et Papelier, Exercices de Geometrie Analytique, Vol. 1, 10th ed., Paris, 1957,
problem 52, p. 35.
[3] P. Baptist, U ber Nagelsche Punktepaare, Mathematische Semesterberichte 35, 1988,
118{126.
[4] H.S.M. Coxeter, The Real Projective Plane, 2nd Ed., Cambridge University Press, 1961.
[5] R.H. Eddy, A Desarguesian dual for Nagel's middlespoint, Elem. Math. 44/3 (1989)
79,80.
[6] R.H. Eddy, A generalization of Nagel's middlespoint, Elem. Math. 45/1 (1990) 14{18.
[7] W. Gallatly, The Modern Geometry of the Triangle, 2nd Ed., Francis Hodgson, London,
circa 1920.
[8] S. Iwata, Encyclopedia of Geometry, Vol. 3, 1976, problem 676. (Japanese)
[9] C.H. von Nagel, Untersuchugen uber die wichtigsten zum Dreiecke gehorenden Kreise,
Leipzig, 1836.
[10] Y. Sawayama, Tokyo Buturi Gakko, 1903.
[11] Problem 1359, proposed by G.R. Veldkamp, solution (II) by R.H. Eddy, Crux Math.
15 (1989) 244{246.
Memorial University of Newfoundland
St. John's, Newfoundland
* * * * *
THE OLYMPIAD CORNER
No. 127
R.E. WOODROW
All communications about this column should be sent to Professor R.E. Woodrow,
Department of Mathematics and Statistics, The University of Calgary, Calgary, Alberta,
Canada, T2N 1N4.
In recent years the September number of the Corner has been the issue in which the
year's IMO is discussed. I rely on our readers and my colleagues who may attend the IMO
for information, and there is frequently a panic and some delay in getting the September
number out since mathematicians are notorious for forgetting to carry out routine tasks
like sending o copies of questions and so on. We have also been experiencing some delays
with the switch over to LaTEX and so we have decided to hold over to a later issue the 1991
IMO results. We begin this number with problems submitted to the jury but not used for
196
the 31st IMO in China. I am indebted to Professor Andy Liu of the University of Alberta,
who was involved with training the Chinese team, for taking the time to send me these
problems.
1. Proposed by Australia.
The integer 9 can be written as a sum of two consecutive integers: 9 = 4 + 5;
moreover, it can be written as a sum of more than one consecutive integer in exactly two
ways, namely 9 = 4 + 5 = 2 + 3 + 4. Is there an integer which can be written as a sum of
1990 consecutive integers and which can be written as a sum of more than one consecutive
integer in exactly 1990 ways?
2. Proposed by Canada.
Given n countries with 3 representatives each, a list of m committees
A(1); A(2); : : : ; A(m) is called a cycle if
(1) each committee has n members, one from each country;
(2) no two committees have the same membership;
(3) for 1 i m, committee A(i) and committee A(i + 1) have no member in
common, where A(m + 1) denotes A(1);
(4) if 1 < ji , j j < m , 1, then committees A(i) and A(j ) have at least one member
in common.
Is it possible to have a cycle of 1990 committees with 11 countries?
3. Proposed by Czechoslovakia.
Assume that the set of all positive integers is decomposed into r disjoint subsets
N= A [ [ Ar . Prove that one of them, say Ai, has the following property: there exists
1
a positive integer m such that for any k, one can nd numbers a ; a ; : : : ; ak in Ai with
1 2
0 < aj , aj m, 1 j k , 1.
+1
4. Proposed by France.
Given ABC with no side equal to another side, let G, K and H be its centroid,
incentre and orthocentre, respectively. Prove that 6 GKH > 90.
5. Proposed by Greece.
Let f (0) = f (1) = 0 and
f (n + 2) = 4n f (n + 1) , 16n f (n) + n2n ;
+2 +1 2
n = 0; 1; 2; : : : . Show that the numbers f (1989), f (1990) and f (1991) are divisible by 13.
6. Proposed by Hungary.
For a given positive integer k, denote the square of the sum of its digits by f (k)1
and let fn (k) = f (fn (k)). Determine the value of f (2 ).
+1 1 1991
1990
7. Proposed by Iceland.
A plane cuts a right circular cone into two parts. The plane is tangent to the
circumference of the base of the cone and passes through the midpoint of the altitude.
Find the ratio of the volume of the smaller part to the volume of the whole cone.
197
8. Proposed by Ireland.
Let ABC be a triangle and ` the line through C parallel to the side AB . Let the
internal bisector of the angle at A meet the side BC at D and the line ` at E . Let the
internal bisector of the angle at B meet the side AC at F and the line ` at G. If GF = DE
prove that AC = BC .
9. Proposed by Japan.
On the coordinate plane a rectangle with vertices (0; 0), (m; 0), (0; n) and (m; n) is
given where both m and n are odd integers. The rectangle is partitioned into triangles in
such a way that
(1) each triangle in the partition has at least one side, to be called a \good" side,
which lies on a line of the form x = j or y = k where j and k are integers, and the altitude
on this side has length 1;
(2) each \bad" side is a common side of two triangles in the partition.
Prove that there exist at least two triangles in the partition each of which has two \good"
sides.
10. Proposed by Mexico.
Determine for which positive integers k the set X = f1990; 1991; 1992; : : :; 1990+ kg
can be partitioned into two disjoint subsets A and B such that the sum of the elements of
A is equal to the sum of the elements of B .
11. Proposed by the Netherlands.
Unit cubes are made into beads by drilling a hole through them along a diagonal.
The beads are put on a string in such a way that they can move freely in space under
the restriction that the vertices of two neighbouring cubes are touching. Let A be the
beginning vertex and B the end vertex. Let there be p q r cubes on the string where
p; q; r 1.
(a) Determine for which values of p, q and r it is possible to build a block with
dimensions p, q and r. Give reasons for your answer.
(b) The same as (a) with the extra condition that A = B .
* * *
The Canadian Mathematics Olympiad for 1991 saw two students tie for rst prize.
One of them went on to place in the top eight for the USAMO. The top prize winners are
listed below: (
First Prize Ian Goldberg
J.P. Grossman
Second Prize Janos Csirik
Third Prize 8 Adam Logan
>
< Jie
> Lou
Fourth Prize > Kevin
Peter
Kwan
Milley
>
: Mark Van Raamsdonk
198
The \ocial" solutions to the 1991 CMO are reproduced below with the permission
of the CMO Committee of the Canadian Mathematical Society.
1991 CANADIAN MATHEMATICS OLYMPIAD
April 1991
Time: 3 hours
1. Show that the equation x 2
+ y = z has innitely many solutions in integers
5 3
2. Let n be a xed positive integer. Find the sum of all positive integers with the
following property: in base 2, it has exactly 2n digits consisting of n 1's and n 0's. (The
rst digit cannot be 0.)
When n = 1, the sum is clearly 2. Let n 2. The left digit is 1 and
Solution.
there are nn, possibilities for arranging (n , 1) 1's and n 0's in the other 2n , 1 digital
2 1
positions. Considerany digital position but the rst. The digit in it is 1 for nn, of the
2 2
numbers and 0 for n, of the numbers, so that the sum of the digits in this position is
2 2
n, n , 1
! ! ! !
2n , 2 (1 + 2 + + 2 n, ) + 2n , 1 2 n, = 2n , 2 (2 n, , 1) + 2n , 1 2 n, :
2 2 2 1 2 1 2 1
n n n n
3. Let C be a circle and P a given point in the plane. Each line through P which
intersects C determines a chord of C . Show that the midpoints of these chords lie on a
circle.
Solution. Let O be the centre of circle C and X be the midpoint of a chord through
P . Then OX is perpendicular to XP , so that X lies on that portion of the circle with
diameter OP which lies within C . (When P = O, the locus degenerates to the single point
O.)
4. Ten distinct numbers from the i i
set f0; 1; 2; : : : ; 13; 14g are to be chosen to , A
, A@
ll in the ten circles in the diagram. The , A A@
absolute values of the dierences of the two i i i i A i@@ i
, A
1; 2; : : : ; 13; 14, so that their sum is 105, an odd number. Denoting the numbers entered
in the circles by xi (1 i 10), we have that
X X X
105 = jxi , xj j (xi , xj ) (xi + xj ) mod 2:
i<j i<j i<j
In the nal sum, each xi is counted as often as a segment emanates from its circle, an even
number of times. This yields a contradiction. Hence the task is not possible.
5. In the gure, the side length of the TT
large equilateral triangle is 3 and f (3), the number
of parallelograms bounded by sides in the grid, is T
T
T
15. For the general analogous situation, nd a T T
formula for f (n), the number of parallelograms, T
T
T T T
T
for a triangle of side length n. T T T
* * *
The prize winners for the 1991 USA Olympiad follow. A total of 139 students from
118 schools were invited to participate.
J.P. Grossman Toronto
Ruby Breydo New York
Kiran Kedlaya Washington
Joel Rosenberg West Hartford
Robert Kleinberg Elma
Lenhard Ng Chapel Hill
Michail Sunitsky New York
Dean Chung Mountain Lakes
Notice that J.P. Grossman was also a winner of the CMO. Apparently the eligibility
rules for the USAMO are changing next year to exclude Canadian entries.
* * *
In the April number of Crux, E.T.H. Wang asked if three was a record for the
number of times one problem has reappeared in dierent Olympiads. Andy Liu responded
by pointing me to the note \The art of borrowing problems" by Rene Laumen in the World
Federation Newsletter, No. 6, August 1987. There Laumen discusses eight problems that
200
have been used more than once, one repeated three times, but three still seems to be the
record.
Andy also noticed that we didn't publish the solution to Question 2 of the 1987
Asian Pacic Mathematical Olympiad. To complete the picture he sends in the following
solution. Interestingly, the published \ocial solution" contains an error.
2. [1989: 131] 1989 Asian Pacic Mathematical Olympiad.
Prove that the equation
6(6a + 3b + c ) = 5n
2 2 2 2
and c = 3d. If the original equation has a non-trivial solution, then this equation has
one with gcd(a; b; d; m) = 1. Clearly b d mod 2. Now the quadratic residues modulo
16 are 0, 1, 4 and 9. The following table gives the possibilities for 2a + b + 3d mod 16, 2 2 2
Hence m is even so that 10m 0 or 8 mod 16. From the above table, we must have a; b; d
2
all even. However this contradicts that gcd(a; b; d; m) = 1. It follows that 6(6a +3b +c ) = 2 2 2
* * *
We turn now to solutions to problems from the \Archives".
2. [1986: 202] 1986 USAMO Training Session.
Determine the maximum value of
x +y +z ,x y,y z,z x
3 3 3 2 2 2
for 0 x; y; z 1.
Solution by Edward T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario.
Let f (x; y; z ) denote the given expression. We show that f (x; y; z ) 1 with equality
if and only if (x; y; z ) equals one of the six triples: (1; 0; 0), (0; 1; 0), (0; 0; 1), (1; 1; 0), (1; 0; 1)
and (0; 1; 1).
Suppose rst that 0 x y z 1. Then
1 , f (x; y; z ) = (1 , z ) + x (y , x) + y (z , y) + z x 0
3 2 2 2
201
4 4
3. [1986: 230; 1988: 165] 1986 Austrian Mathematical Olympiad.
Determine the set of all values of x ; x 2 R such that the sequence dened by
0 1
xn = 3x xn,,xn2x ; n 1
+1
1
n, 1 n
contains innitely many natural numbers.
Alternate solution by G.R. Veldkamp, De Bilt, The Netherlands.
From the recurrence relation, xn = 0 gives xn = 0 and the sequence breaks
+1
down. So we assume xn 6= 0. Now put un = 1=xn. The recurrence relation now gives
un , 3un + 2un = 0, with characteristic equation t , 3t + 2 = 0, having 1 and 2
+2 +1
2
*
5. [1986: 231] 1986 Austrian{Polish Mathematical Competition.
Determine all quadruples (x; y; u; v) of real numbers satisfying the simultaneous
equations
x + y + u + v = 4;
2 2 2 2
xu + yv = ,xv , yu;
xyu + yuv + uvx + vxy = ,2;
xyuv = ,1:
Solution by Hans Engelhaupt, Franz-Ludwig-Gymnasium, Bamberg, Germany.
The second equation can be rewritten
(x + y)(u + v) = 0
and the third is xy(u + v) + (x + y)uv = ,2.
Case 1. x + y = 0. Then the last two equations give x (u + v) = 2 and x uv = 1.
2 2
From this u + v = 2uv and x = y = 2=(u + v). Using the rst equation,
2 2
4 + (u + v) , (u + v) = 4:
2
u+v
Setting z = u + v we obtain
(z , 1)(z , 2)(z + 2) = z , z , 4z + 4 = 0:
3 2
* * *
Now we turn to the November 1989 number of Crux and the rest of the problems
proposed but not used at the 1989 IMO. The \ocial" solutions can be found in the booklet
30th International Mathematical Olympiad, Braunschweig, 1989, edited by Hanns-Heinrich
Langmann.
12. [1989: 260] Proposed by Australia.
Ali Baba the carpet merchant has a rectangular piece of carpet whose dimensions
are unknown. Unfortunately, his tape measure is broken and he has no other measuring
instruments. However, he nds that if he lays it
at on the
oor of either of his storerooms,
then each corner of the carpet touches a dierent wall of that room. He knows that the
sides of the rooms are integral numbers of feet, and that his two storerooms have the
same (unknown) length, but widths of 38 and 55 feet respectively. What are the carpet's
dimensions?
Correction by Hans Engelhaupt, Franz-Ludwig-Gymnasium, Bamberg, Germany.
Let the carpet have length y, width x. Let the length
of the storerooms be L, and set y=x = k. The diagram shows
the larger storeroom. By similar triangles, BE=y = b=x or A a
EQ B
6
BE = kb. Similarly DF = ka. Now a + kb = 55 and ka + b =
Q
Q
L, so b
x yQQ
a = kL , 55 ; b = 55k , L :
QQ
L
k ,1
2
k ,1 2
F
Q
This gives Q
Q
Q
! ! Q
kL , 55
x = k ,1 + k ,1
2 55
2
k , L 2
D
QQ
C
?
2 2 55 -
so
(k , 1) x = (kL , 55) + (55k , L) :
2 2 2 2 2
Equating we have
(38k , L) + (Lk , 38) = (55k , L) + (Lk , 55) :
2 2 2 2
Editor's Note. The \ocial" solution in the IMO booklet starts out with the assumption
that the widths are 38 and 50, even though the problem is stated as above, with widths
of 38 and 55!
14. [1989: 260] Proposed by Hungary.
At n distinct points of a circle-shaped race course there are n cars ready to start.
They cover the circle in an hour. Hearing the signal each of them selects a direction and
starts immediately. If two cars meet both of them change directions and go on without
loss of speed. Show that at a certain moment each car will be at its starting point.
Solution by Curtis Cooper, Central Missouri State University, Warrensburg.
We will show that after n hours, each car will be back at its starting point. Consider
another car race such that if two cars meet they pass through each other instead of chang-
ing directions. Although the races are dierent, the \passing through" car race and the
\changing direction" car race are similar in that at any given time, the points on the course
occupied by cars are the same and the directions the cars at these points are moving are
the same. After one hour of the \passing through" car race, each car is back at its starting
point and going in its starting direction. Thus, although the cars may not be the same,
we see that after one hour of the \changing direction" car race, there will be a car at each
starting point and that this car will be going in the same direction that the starting car
at that point went. In addition, although we may not know which car is at each starting
point in one hour, we do know that the cars in the \changing direction" car race maintain
the same relationship with the other cars. That is, in the \changing direction" car race,
the cars in the clockwise direction and counterclockwise direction with respect to a given
car are always the same. Therefore, after one hour of the \changing direction" race, the
cars at the starting points on the course are some rotation (possibly 0) of the original cars
at the starting points on the course and are going in the same directions the original cars
were going. Thus after n hours, the cars in the \changing direction" car race will be back
at the starting point.
* * *
In the remaining space for this month's number of the Corner, we given solutions
sent in by readers to problems of the 11th Austrian{Polish Mathematics Competition that
was given in the December 1989 number of Crux [1989: 289{291].
1. Let P (x) be a polynomial having integer coecients. Show that if Q(x) =
P (x) + 12 has at least six distinct integer roots, then P (x) has no integer roots.
Generalization and solution by David Singmaster, South Bank Polytechnic, London,
England.
Dene M (2k) = (k!) and M (2k + 1) = k!(k + 1)!, where k is a positive integer.
2
We show:
THEOREM. If Q(x) has integer coecients and k distinct integer roots and if
P (x) = Q(x) , y has an integral root with y 6= 0, then jyj M (k).
The given problem has k = 6, y = 12, so that M (k) = 36 6 jyj, and the problem
follows.
205
To prove the theorem, let the given roots of Q(x) be x ; x ; : : : ; xk and let
1 2
B (x) = (x , x )(x , x ) : : : (x , xk ):
1 2
1 1 1 + 1 1 1
B (x ) = (x , x ) : : : (x , xk )jy:
0 0 1 0 (1)
The values x , xi are k distinct integers so the least absolute value of such a product is
0
1 1 2 2 3 3 (with k terms) = M (k); hence jyj M (k). This proves the theorem.
In place of 12, the original problem could have had any integer in the range [1; 35].
Editor's Note. The problem was also solved by Seung-Jin Bang, Seoul, Republic of
Korea; by Curtis Cooper, Central Missouri State University, Warrensburg; and by Edward
T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario.
4. Determine all strictly monotone increasing functions f : R ! R satisfying the
functional equation
f (f (x) + y) = f (x + y) + f (0)
for all x; y 2 R.
Solutions by Seung-Jin Bang, Seoul, Republic of Korea; by George Evagelopoulos,
Athens, Greece; and by David C. Vaughan, Wilfrid Laurier University, Waterloo, Ontario.
Let f be a strictly increasing function on R satisfying the functional equation. In
particular, if we set y = ,x, we must have f (f (x) , x) = f (0) + f (0) = 2f (0). Unless
f (x) , x is a constant we have a contradiction to the fact that f is 1 , 1 since f (f (x) , x)
is constant. Thus f (x) = x + c where c = f (0). Any such f is strictly increasing, and we
have
f (f (x) + y) = f (x + c + y) = (x + y + c) + c = f (x + y) + f (0);
that is, any such f is a solution.
7. In a regular octagon each side is coloured blue or yellow. From such a colouring
another colouring will be obtained \in one step" as follows: if the two neighbours of a
side have dierent colours, the \new" colour of the side will be blue, otherwise the colour
will be yellow. [Editor's note: the colours are modied simultaneously.] Show that after a
206
nite number, say N , of moves all sides will be coloured yellow. What is the least value of
N that works for all possible colourings?
Solution by Curtis Cooper, Central Missouri State University.
Consider a colouring of the sides of a regular octagon where each side is coloured
blue or yellow. Label the sides of the octagon S ; : : : ; S and code the colours of the sides
of the regular octagon by the column vector x = (x ; : : : ; x ), where
1 8
1 8
(
xi = 10 ifif SSi isis coloured blue,
i coloured yellow,
for i = 1; 2; : : : ; 8. Let 00 1 0 0 0 0 0 11
BB 1 0 1 0 0 0 0 0 CC
BB 0 1 0 1 0 0 0 0 CC
BB CC
A=B B 0 0 1 0 1 0 0 0 C
BB 0 0 0 1 0 1 0 0 CCC :
BB 0 0 0 0 1 0 1 0 CC
B@ 0 0 0 0 0 1 0 1 CA
1 0 0 0 0 0 1 0
Then Anx is the coded colouring of the sides of the regular octagon after n steps, where the
addition and multiplication in the matrix operations are performed mod 2. Since A = 0 4
it follows that after 4 steps all sides will be coloured yellow. This is the least possible value
of N that works for all possible colourings, since for x = (1; 0; 0; 0; 0; 0; 0; 0)T ,
Ax = (0; 1; 0; 0; 0; 0; 0; 1)T ; A x = (0; 0; 1; 0; 0; 0; 1; 0)T ; A x = (0; 1; 0; 1; 0; 1; 0; 1)T :
2 3
* * *
This completes the space for this month. Send in your nice solutions and your
Olympiads.
* * * * *
PROBLEMS
Problem proposals and solutions should be sent to B. Sands, Department of
Mathematics and Statistics, University of Calgary, Calgary, Alberta, Canada T2N 1N4.
Proposals should, whenever possible, be accompanied by a solution, references, and other
insights which are likely to be of help to the editor. An asterisk (*) after a number indicates
a problem submitted without a solution.
Original problems are particularly sought. But other interesting problems may also
be acceptable provided they are not too well known and references are given as to their
provenance. Ordinarily, if the originator of a problem can be located, it should not be
submitted by somebody else without permission.
207
R R 2
1
2
where F and F are the areas of triangles A B C and ABC respectively, with equality
1 1 1 1
when ABC is equilateral.
1665n . Proposed by P. Penning, Delft, The Netherlands.
1665 ends in 5 for n 1, and in 25 for n 2. Find the longest string of digits
which ends 1665n for all suciently large n.
1666. Proposed by Marcin E. Kuczma, Warszawa, Poland.
(a) How many ways are there to select and draw a triangle T and a quadrilateral Q
around a common incircle of unit radius so that the area of T \ Q is as small as possible?
(Rotations and re
ections of the gure are not considered dierent.)
(b) The same question, with the triangle and quadrilateral replaced by an m-gon
and an n-gon, where m; n 3.
208
n=1
1668. Proposed by Stanley Rabinowitz, Westford, Massachusetts.
What is the envelope of the ellipses
x +y =1 2 2
a b 2 2
2
is an nth root of unity for some n. Show that q must be an integer.
1670. Proposed by Juan Bosco Romero Marquez, Universidad de Valladolid,
Spain.
Let B ; B ; C ; C be points in the plane and let lines B B and C C intersect in
1 2 1 2 1 2 1 2
A. Prove that the four points G ; G ; G ; G form the vertices of a parallelogram when
11 12 21 22
Gij is determined in any of the following ways: (i)Gij is the centroid of ABi Cj ; (ii) Gij
is the orthocenter of ABi Cj ; (iii) Gij is the circumcenter of ABi Cj .
* * * * *
SOLUTIONS
No problem is ever permanently closed. The editor will always be pleased to consider
for publication new solutions or new insights on past problems.
1542. [1990: 143] Proposed by Murray S. Klamkin, University of Alberta.
For xed n, determine the minimum value of
Cn = j cos j + j cos 2j + + j cos nj:
It is conjectured that min Cn = [n=2] for n > 2.
Solution by Manuel Benito Mu~noz and Emilio Fernandez Moral, I.B. Sagasta,
Logro~no, and Maria Mercedes Sanchez Benito, I.B. Luis Bunuel, Alcorcon, Madrid, Spain.
We will prove that the conjecture is true for all n > 2 except for n = 4 and n = 6.
Consider the function
X n
Fn (x) = cos jx 2 :
j =1
209
As we shall see later, this lemma will reduce our proofs to searching the values of
Fn(x) for x = 1=b with b n.
Till now we had in mind that n was a xed number and b was a variable number
less than n. Now we must change this point of view, and we consider b 3 xed and n
variable.
PROPOSITION 1. Let b 3 be an integer. If b n < 2b with n even, then
Fn(1=b) > n=2 except for the cases
F (1=3) 1:87 < 2 and F (1=5) 2:97 < 3:
4 6
Proof. Fn (1=b) > n=2 is equivalent to: the average of the terms
j
cos ; j = 1; 2; : : : ; n;
2b
is greater than 1=2. Note that
8
< > 1=2 for 1 j < 2b=3;
j >
cos < 1=2 for 2b=3 < j < 4b=3;
2b >
: > 1=2 for 4b=3 < j < n:
Thus it suces to show that Fn (1=b) > n=2 for n = whichever of [4b=3] , 1, [4b=3], [4b=3]+1
are even, that is,
(i) for b = 3c, F c (1=b) > 2c;
4
211
Since the function j cos(x=2)j is concave, we can bound it on [0; 2] from be-
low by the polygonal path obtained by connecting the points (x; j cos(x=2)j) for x =
0; 1=3; 2=3; 1; 4=3; 5=3; 2 by straight lines. This gives a function H (x) dened by
8 p,
>
> 1p+ x , p = u + v x for x 2 [0; 1=3];
3 3 6
>
>
2 0 0
> 3, + 1
x = u + v x for x 2 [1=3; 2=3];
3 3 3
>
< 3=2 + (,3=2)x 2
=u +v x
2
for x 2 [2=3; 1];
1 1
>
> ,p p, 3 3
>
> 5 4
+ p x = u + v x
3 3 3 3
for x 2 [4=3; 5=3];
> p2
: (3 3 , 5) +
2 4 4
, x = u + v x for x 2 [5=3; 2]:
6 3 3
2 5 5
For (i),
1 Xc j Xc j X iX c j ( +1)
4 4 3
F c 3c = cos 6c H 3c =
4 ui + vi 3c
j =1 j i j ic =1 =0 = +1
X X
= c ui + 31c vi c[(2i + 21)c + 1]
3 3
i i
p 1 c + 1 X
=0 =0
= c 3+ + v c X iv 3 3
i+
2 6 i 3i i
p 1 c + 1 ,3 c 6 , 3p3 !
=0 =0
= c 3+ + +
2 6 2
p p 3 2
= c(5 + 24 3) , 1 = 2c + (2 3 ,43)c , 1
> 2c for c 3;
while for c = 2, by direct calculation,
X j 8
F (1=6) = cos 12 4:05 > 4:
8
j =1
For c = 1 appears the rst special case: direct computation
p shows that
F (1=3) = 1 + 23 < 2: 4
For (ii),
X c Xc X c
F c 3c 1+ 1 u + v 3c j+ 1 + u + v 3c j+ 1 + u + v 3c j+ 1
2 3 +1
4 0 0 1 1 2 2
j =1 j c j c = +1 =2 +1
Xc
u + v 3c j+ 1
4
+ 3 3
j c =3 +2
212
X
ui + u , u + 6c 1+ 2 [c(c + 1)v + (3c + 1)cv + (5c + 2)(c + 1)v
3
=c 2 3 0 1 2
i =0
12c + 4 12c + 4
> 2c for c 2;
while for c = 1, F (1=4) > 2. Also, from the above calculation,
4
1 p p 3
F c 3c + 1 2c + (6 3 , 9) c , (9 , 4 3) c 2
+ 2b (4c + 1) , 2 3
4 +2
12cp+ 4 p!
+ 3 3 , 3 (4c + 2) + 5 , 4 3
2b 2
p p p
= 2c + 1 + (6 3 , 9)c , (9 , 4 3)c + 4 3 , 6 2
12c + 4
> 2c + 1 for all c 1:
For (iii),
1 X c j X c j 2 +1
F c 3c + 2
4 +2 u + v 3c + 2 + 0 0u + v 3c + 2 1 1
j =1 j c = +1
c
X j X
3 +2 c j 4 +2
+ u + v 3c + 2 + u + v 3c + 2
2 2 3 3
j c =2 +2 j c =3 +3
X
= c ui + u + u + 6c 1+ 4 [c(c + 1)v + (3c + 2)(c + 1)v
3
1 2 0 1
i =0
= 2c + 1 + (6 3 , 9)c + (8 3 , 15)c , (6 , 2 3)
2
12c + 8
> 2c + 1 for c 2:
When c = 1 we have the other exceptional case, because
F (1=5) 2:97 < 3 :
6
2
j 2b j b 2b
a
=1
1
= +1
1
= 2Fb
b + 1 = 2Fb b + 1 = F b b : 2 (1)
Case (i): b = 3c. Then
1 X iX c X X X
ui + vi 3c = c ui + c +6 1 vi + 3c ivi
j
5 ( +1) 5 5 5
F c 3c
6
i j ic i i i
p p p
=0 = +1 =0 =0 =0
c
= c(2 3 , 2) + 0 + (12 , 3 3) = 3c + ( 3 , 1)c; (2)
3
and since 6c < n < 9c,
a a X n ja
Fn 3c = F c 3c + cos : (3)
6
j c
=6 +1
6c
214
But all terms j cos(ja=6c)j of the above sum are greater than or equal to 1=2, except for
at most c of them. [Editor's note. In general, j cos(ja=2b)j < 1=2 for 2b j n if and
only if j belongs to an interval of the form
!
2b(t + a) + 2b ; 2b(t + a) + 4b
a 3a a 3a
for some nonnegative integer t, where n < 3b means t < a=2; thus there are less than a=2
such intervals, each containing at most 2b=3a such j 's, so at most b=3 such j 's altogether.]
Hence from (1), (2) and (3),
a p 1 n p 3 n
Fn 3c 3c + ( 3 , 1)c + (n , 7c) 2 = 2 + 3 , 2 c > 2 :
Case (ii): b = 3c + 1. As in the case above, we have
a 1
F c 3c + 1 = F c 3c + 1
6 +2 6 +2
X c Xc
u + v 3c j+ 1 + u + v 3c j+ 1
2
0 0 1 1
j
=1 j c = +1
cX
3 +1
j X c 4 +1
j
+ u + v 3c + 1 +
2 2 u + v 3c + 1
3 3
j c =2 +1 j c =3 +2
Xc
5 +1
j X c 6 +2
j
+ u + v 3c + 1 +
4 4 u + v 3c + 1
5 5
j c =4 +2 j c =5 +2
X
= c ui + u + u + 6c 1+ 2 [c(c + 1)v + c(3c + 1)v + (c + 1)(5c + 2)v
5
2 5 0 1 2
=0 i
p +c(7c + 3) pv + c(9c + 3)v + (cp+ 1)(11c+ 4)v ]
3 4 5
p 3c + 1 3c + 1
> 3c + 1 + ( 3 , 1)c:
Now since 6c + 2 < n < 9c + 3,
a a X n ja
Fn 3c + 1 = F c 3c + 1 +6 +2
cos 6c + 2
j c =6 +3
p
> 3c + 1 + ( 3 , 1)c + n , 72c , 2 > n2 ;
because, as before [see the editor's note in case (i)], of the n , 6c , 2 terms in the summation,
at most c will be less than 1=2.
Case (iii): b = 3c + 2. In the same way,
F c 3c a+ 2 = F c 3c 1+ 2
6 +4 6 +4
215
c
X X c
u + v 3c j+ 2 + u + v 3c j+ 2
2 +1
0 0 1 1
j =1 j c = +1
cX
3 +2
j X c 4 +2
j
+ u + v 3c + 2 +
2 2 u + v 3c + 2 3 3
j c
=2 +2 j c =3 +3
Xc
5 +3
j X c 6 +4
j
+ u + v 3c + 2 +
4 4 u + v 3c + 2 5 5
j c
=4 +3 j c =5 +4
X
= (c + 1) ui , u , u + 6c 1+ 4 [c(c + 1)v + (c + 1)(3c + 2)v
5
0 3 0 1
i =0
p 2c + 1 p
= 3c + 2 + ( 3 , 1) c + > 3c + 2 + ( 3 , 1)c;
3c + 1
and then we have, since 6c + 4 < n < 9c + 6,
X n
Fn 3c a+ 2 = F c 3c a+ 2 + cos ja
6 +4
j c 6c + 4
=6 +5
p
> 3c + 2 + ( 3 , 1)c + n , 72c , 4 > n2 :
Thus Proposition 3 is proved. 2
To establish (), it only remains to consider the case n 3b, proceeding inductively.
PROPOSITION 4. Let a and b be positive integers with 1 a < b, a odd and
gcd(a; b) = 1. Then for all even integers n with n 3b, Fn(a=b) > n=2.
Proof. Note
a a nX , b (2b + i)a
Fn b = F b b + cos 2b
2
2
i
a nX , b a nX , b ia
=1
=Fb
2
ia
2
2
b + i cos a + 2b = F b b + i cos 2b 2
a =1
a =1
=Fb 2
b + Fn, b b : 2 (4)
If we suppose the proposition is true when (2k , 1)b n < (2k +1)b, then when (2k +1)b
n < (2k + 3)b we obtain by (4) and Proposition 3 that
a
Fn b > b + n ,2 2b = n2 ;
so the proposition follows provided we can establish the base case. If b 6= 2, 3 or 5 the
proposition is true for k = 1, i.e., in the case b n < 3b, by Propositions 1, 2 and 3.
216
Finally, for the remaining cases b = 2, 3 and 5 we could begin the induction on the interval
3b n < 5b, since by direct computation
F (1=2) 3:1 > 3 ; F (1=3) 5:5 > 5 ; F (1=5) 9:2 > 8:
6 10 16
[Editor's note. Here are a few more details. For b = 2 and a odd, one nds as in the proof
of Proposition 3 that
F a2 = 3F a2 + 1 = 3F 12 + 1 = F 12 ;
6 2 2 6
1
> 3 + (n , 6) 2 (as is easily checked)
= n2 :
Similarly for b = 3, a odd and not divisible by 3, we use
F a3 = 3F a3 + 32 = 3F 31 + 32 = F 13
10 3 3 10
To nish, we need to verify that F (x) 1, F (x) 2 and F (x) 3 for all x. It
3 5 7
is possible to do this quite easily with the aid of a computer. By Lemma 1 it suces to
compute the values on the terms of the Farey sequence with odd numerator. The following
tables give the results.
x 1=3 1=2 1
F3(x) 1:36+ 1:41+ 1
x 1=5 1=4 1=3 1=2 3=4 3=5 1
F5(x) 2:65+ 2:39+ 2:73+ 3:12+ 2:93+ 2:65+ 2
217
x 1=7 1=6 1=5 1=4 1=3 3=7 1=2 3=5 5=7 3=4 5=6 1
F7(x) 3:93+ 3:55+ 3:55+ 4:02+ 4:59+ 3:93+ 3:82+ 4:41+ 3:93+ 4:02+ 4:26+ 3
No other solutions for this problem were received. The examples refuting the
proposer's conjecture for n = 4 and 6 were also found by RICHARD I. HESS, Ran-
cho Palos Verdes, California; WALTHER JANOUS, Ursulinengymnasium, Innsbruck,
Austria (n = 4 only); and CHARLTON WANG, student, Waterloo Collegiate Institute,
and EDWARD T.H. WANG, Wilfrid Laurier University, Waterloo, Ontario. Hess gave
the correct answer to the problem without proof.
The proposer also asks for the maximum value of
j sin j + j sin 2j + + j sin nj
for xed n. The editor is grateful he didn't make this Crux 1542, part (b)! However, readers
may like to work on it.
Janous recalls a somewhat similar problem in a recent U.S.S.R. Olympiad, but has
no more details. Maybe a reader can add more information?
* * * * *
1546. [1990: 144] Proposed by Graham Denham, student, University of Alberta.
Prove that for every positive integer n and every positive real x,
Xn xk 2
k xn n = : ( +1) 2
k =1
1+
n+1 n+2 , n+2 = 1+ n+1 n+2 n + 2 , (n + 1)
!n
n 2( n + 1)
2
= 1, n+2 n+2 :
This minimum value is positive because for n 2,
n + 2 n= !n=
2
= 1 + 1 2 >
2
2(n + 1) :
n n=2 n+2
218
We now prove the inequality of the problem by induction. For n = 1 the inequality
reduces to an identity. Suppose that the inequality holds for n = m 1. Then
mX xk
+1 2
x m ( +1)2
k , x m m ( = x
+1)(m m =
+2) 2
+ m+1 , x m m =
( +1) 2 ( +1)( +2) 2
k =1
!
x m m = ( +1)( +2) 2
m+1 ,x
=x m m = 1+ ( m +1) 2 +1
Pk n k x k k :
1
=1 k =1
3 =1
k =1
Since ! n !
Xn n ( n + 1) X 2 2
k = 2
3
= k ;
k=1 k=1
we get
X
n
kxk n(n2+ 1) xn n
2
( +1) 2 = :
k=1
Finally,
X n xk n Zx
X Z
k , dx n(n + 1) x xn n
2
= kx 2 1 ( +1) 2 = , 1 dx = xn(n+1)=2:
k=1
k k =1
02 0
k k = 2
k k ;
k=1 k=1
then Pn
X
n
k xk x k=1 k k
k=1
for all x 0.
219
* * * * *
1547. [1990: 144] Proposed by Toshio Seimiya, Kawasaki, Japan.
Let P be an interior point of a parallelogram ABCD, such that 6 ABP = 26 ADP
and 6 DCP = 26 DAP . Prove that AB = PB = PC .
Solution by Dan Sokolowsky, Williamsburg, Virginia. D
Let 6 ADP = and 6 DAP = , so that B C 2
6 ABP = 2 and 6 DCP = 2 . Draw POjjAB , PO cut-
O
B
ting AD, with PO = AB = CD. Then ABPO, CDOP Q
H BP
H S S
HH
are parallelograms, so 6 AOP = 2 and 6 DOP = 2 . S
HH S S
S
Extend PO to Q with OQ = OA. Then HH S S
B 2
A
6 OQA = 6 OAQ = = 6 ADP;
so A; P; D; Q lie on a circle . Hence 6 DQP = 6 DAP = , so
6 ODQ = 6 DOP , 6 DQO = 2 , = ;
therefore wait to see what the Monthly publishes before deciding which solution, if any,
to feature in Crux. The list of solvers of Crux 1548 will also be given at that time. It
shouldn't be too long a wait, although the Monthly seems to be even further behind in
their solutions than we are!
* * * * *
1549. [1990: 144] Proposed by D.J. Smeenk, Zaltbommel, The Netherlands.
In quadrilateral ABCD, E and F are the midpoints of AC and BD respectively. S
is the intersection point of AC and BD. H; K; L; M are the midpoints of AB; BC; CD; DA
respectively. Point G is such that FSEG is a parallelogram. Show that lines GH; GK; GL,
GM divide ABCD into four regions of equal area.
Solution by C. Festraets-Hamoir, Brussels, Belgium.
Il s'agit d'un theoreme connu sous le nom de THEOREME DE BRUNE (2000
theoremes et problemes de geometrie, A. Dalle, ed. La Procure-Namur). En voici une
demonstration. B
Demontrons que l'aire de ,
,@
GKCL vaut le quart de l'aire de ,
, @@
ABCD. @
Joignons EK , EL, KL. ,
Hq
,
,
@K
"@
,
"
L'homothetie de centre C et ,
,
G""
", @
,
de rapport 2 applique CKEL , F , @
, D, @
sur CBAD (puisque K; E; L A, D
, @C
S EJ D
sont les milieux respectifs de J
J J D
BC; AC; DC ), d'ou Jq
JD
MJ
L
1
aire(CKEL) = aire(CBAD):
J
J
4 D
GE k KL (k BD), d'ou aire(GKL) = aire(EKL),
aire(GKCL) = aire(GKL) + aire(KLC ) = aire(EKL) + aire(KLC )
= aire(EKCL) = 1 aire(ABCD):
4
On demontre de m^eme que
aire(GHBK ) = aire(GMAH ) = aire(GLDM ) = 41 aire(ABCD):
( (
f (x) = ,11 for x 6= 0; f (x) = ,11 for x 6= 0;
3
for x = 0; 4
for x = 0:
That these four functions satisfy the condition is easily checked. We show that there are
no others.
Letting x = 0 and y = 1, we see that jf (0)j = 1. Letting x = ,y 6= 0, we
get jf (,x) + f (x)j 2, which implies (since jf (x)j 1) that f (,x) = f (x) = 1.
Hence jf (x)j = 1 for all x 2 A. Suppose there exists x; y 2 A satisfying xy 6= 0 and
f (x)f (y) = ,1. Then (since f (,x) = f (x)) there exists x; y 2 A satisfying xy < 0 and
f (x)f (y) = ,1. Then jx + yj jx , yj, which is impossible for xy < 0.
Also solved by SEUNG-JIN BANG, Seoul, Republic of Korea; NICOS D.
DIAMANTIS, student, University of Patras, Greece; MATHEW ENGLANDER, student,
University of Waterloo; MORDECHAI FALKOWITZ, Tel-Aviv, Israel; C. FESTRAETS-
HAMOIR, Brussels, Belgium; WALTHER JANOUS, Ursulinengymnasium, Innsbruck,
Austria; JEAN-MARIE MONIER, Lyon, France; M. PARMENTER, Memorial University
of Newfoundland; and the proposer.
* * * * *
1551. [1990: 171] Proposed by J.T. Groenman, Arnhem, The Netherlands.
Find a triangle ABC with a point D on AB such that the lengths of AB; BC; CA
and CD are all integers and AD : DB = 9 : 7, or prove that no such triangle exists.
I. Solution by Hayo Ahlburg, Benidorm, Spain.
Let AD = kx, DB = `x, AD : DB = k : ` (given), AB = (k + `)x, BC = y,
CA = z , CD = t. These lengths are related by Stewart's Theorem [1]
(k + `)(t + k`x ) = ky + `z ;
2 2 2 2
in our case (k = 9, ` = 7)
16t + 1008x = 9y + 7z :
2 2 2 2
(1)
222
more solutions.
References:
[1] H.S.M. Coxeter and S.L. Greitzer, Geometry Revisited, New Math. Library, Vol. 19,
M.A.A., 1967, p. 6.
[2] L.J. Mordell, Diophantine Equations, Academic Press, 1969, pp. 49{51.
II. Solution by P. Penning, Delft, The Netherlands.
Consider the situation that ABC is similar to ACD. Then
AB = BC = AC :
AC CD AD
Write AD = 9f , DB = 7f , CD = x, BC = a, AC = b, where f is an integer determining
the scale. Then
16f = a = b ;
b x 9f
or
b = 12f and xa = 43 :
Points C lie on a circle with A as centre g
qZ q g( = 5)
g q
and radius b = 12f . With g an integer in ( = 6) ( = 4)
!1 n = e
1
!
j , 1 j < e, for all j: 1
j
Hence n j , 1 !n X
X n
xn = j < e,n=j for all n;
j =1 j =1
224
and thus Z
xn 1 X n
sup n lim sup n e = e, =x dx:
, 1
L = lim n=j 1
n!1 n!1 j =1
0
!
j , 1 j > e, , for all j N:
1
j
Then for n large enough so that N=n < , we have
1 X n 1 Xn j , 1 !n x
(e , )
, n=j 1
= n:
n j n=[ ]
nj j n =1
Taking the lim inf gives (ii). [Since , 1=n < [n]=n , the lim inf of the left side is, via
Riemann sums, the integral in (ii).]
The method given above can be used to prove the following slightly more general
result. Let 0 aj 1 for j = 1; 2; : : : ; and suppose (aj )j ! a. Put
Xn
xn = (aj )n
j =1
for each n. Then Z
lim xn = 1
a =x dx:
1
n!1 n 0
CONTENTS
1
225
All communications about this column should be sent to Professor R.E. Woodrow,
Department of Mathematics and Statistics, The University of Calgary, Calgary, Alberta,
Canada, T2N 1N4.
Since we still haven't heard from our IMO team representatives about the events in
Sweden, I am putting o until the next issue a discussion of this year's contest. Anyway,
we still have some of the problems from last year's IMO in China that were proposed to
the jury, but not used. Once again I would like to thank Andy Liu, of the University of
Alberta, who sent them to me. I also must correct a mistake I made in the last issue. Andy
was an observer at the marking sessions in China but not a trainer of the team. I had
remembered that he has helped to train the Hong Kong team in the past, and confused
the rest. My apologies.
UNUSED PROBLEMS FROM THE 31ST IMO
1. Proposed by Hungary.
The incentre of the triangle ABC is K . The midpoint of AB is C and that of AC
1
the triangles AB C and ABC are equal, what is the measure of 6 CAB ?
2 2
2. Proposed by Ireland.
An eccentric mathematician has a ladder with n rungs which he always ascends
and descends in the following ways: when he ascends each step, he covers a rungs, and
when he descends each step, he covers b rungs, where a and b are xed positive integers.
By a sequence of ascending and descending steps, he can climb from ground level to the
top rung of the ladder and come back down to ground level again. Find, with proof, the
minimum value of n expressed in terms of a and b.
3. Proposed by Norway.
Let a and b be integers with 1 a b, and M = [(a + b)=2]. Dene the function
f : Z ,! Z by (
f (n) = nn + a if n < M;
, b if n M:
Let f (n) = f (n), and f i (n) = f (f i (n)), for i = 1; 2; : : : . Find the smallest positive
1 +1
5. Proposed by Poland.
Prove that every integer k > 1 has a multiple which is less than k that can be
4
+ c
3
+ d 3
1:
b+c+d c+d+a d+a+b a+b+c 3
9. Proposed by the U.S.A.
Let P be a cubic polynomial with rational coecients, and let q ; q ; q ; : : : be a
sequence of rational numbers such that qn = P (qn ) for all n 1. Prove that there exists
1 2 3
Find all positive integers n for which every positive integer whose decimal repre-
sentation has n , 1 digits 1 and one digit 7 is prime.
11. Proposed by the U.S.S.R.
Prove that on a coordinate plane it is impossible to draw a closed broken line such
that
(1) the coordinates of each vertex are rational;
(2) the length of each edge is 1; and
(3) the line has an odd number of vertices.
* * *
Repeating ourselves? L.J. Upton, of Mississauga, Ontario writes pointing out that
problem 2 [1991: 68{9] was previously discussed in Eureka (the original name of Crux
Mathematicorum), in an article by T.M. Apostol [1977; 242{44].
* * *
Now we turn to the \archive problems".
227
f (a , a + 1) = 0. Thus a 6= 0. If a + 1 6= 0, then
2 2
a, a + a + 1, a , a + 1, ,a are the vertices of a
2 2
parallelogram, and ja + a + 1j or ja , a + 1j is
2 2
,a a2
+ +1
a
greater than jaj, contradicting its choice. So a = @
@
are roots of f (x) and f (x) = (x + 1)mg(x) where
2
0
Now
(x + 1)(x + 2x + 2) = x + 2x + 3x + 2x + 2:
2 2 4 3 2
(3)
This gives that
g(x) g(x + 1) = g(x + x + 1);
2
i.e. g(x) satises the same functional equation as f (x). By the argument at the beginning
of this case, g(x) (not being divisible by x +1) must be a constant polynomial, and hence,
2
by Case 1, g(x) 1. Thus if f (x) is non-constant and satises (1) we must have that
f (x) = (x + 1)m:
2
On the other hand (3) shows that (1) is then satised.
*
1. [1987: 71] Second Balkan Mathematical Olympiad.
Let O be the centre of the circle through the points A, B , C , and let D be the
midpoint of AB . Let E be the centroid of the triangle ACD. Prove that the line CD is
perpendicular to the line OE if and only if AB = AC .
Solution
,!
by G.R.
,!
Veldkamp,
,!
De Bilt, The Netherlands.
Set OA= a, OB = b, OC = c. Then
OE =OA + OC + OD= 23 OA + 12 OB + OC = 16 (3a + b + 2c)
,! ,! ,! ,! ,! ,! ,!
and
CD= 12 (CA + CB ) = 21 (OA , OC + OB , OC ) = 12 (a + b , 2c):
,! ,! ,! ,! ,! ,! ,!
language. This pair forms 1983 triads with the remaining participants, each of which must
229
have a common language with P or P (or both). It follows that one of the pair, say P ,
1 2 1
has a common language with each of at least 992 participants. Since P only speaks at 1
most ve languages, some one of them is spoken by at least 199 of the 992 people. Then
that language is spoken by at least 199 + 1 = 200 people, including P . 1
* * *
We now give solutions for some of the problems proposed but not used on the 1987
IMO in Havana, Cuba. These were given in the October and November 1987 numbers of
the Corner.
Finland 1. [1987: 246]
In a Cartesian coordinate system, the circle C has center O = (,2; 0) and radius
1 1
3. Denote the point (1; 0) by A and the origin by O. Prove that there is a positive constant
c such that for any point X which is exterior to C , 1
OX , 1 c minfAX; AX g: 2
X 2D .
2
St , (OX , 1)=AX = t , 1=AX is minimized when AX is minimal; this is clearly the case
1 2
calculates that q
1 + 24 sin (!=2) , 1 2
t , 1=AX = 6 sin(!=2)
;
where ! is the angle XO A. This is an increasing function of !. So even here t , 1=AX
1
is minimized at X .
If X 2 D nD , we again consider St such that X 2 St . On St , (OX , 1)=AX =
0
2
2 1
t=AX , 1=AX . This function of AX assumes its minimum either when AX takes its
2
largest value or when it takes its smallest value, i.e., either on the boundary C or the
x-axis, where it reduces in either case to t , 1 and is minimized at X , or on the boundary
2
[6 sin(!=2)] : 2
q
This decreases with ! and is minimized at X . So one can choose c = 5=3 , 1. It is also
0
that there exist real numbers p, q, r and a polynomial S such that P = pS , Q = qS and
R = rS . 2
holds for
P + Q = R , with 2 f,1; +1g:
4 4 2
(1)
The case h(P; Q; R) = 0 is obvious. Let us assume h(P; Q; R) > 0.
Suppose that an irreducible polynomial F divides two of the polynomials P , Q , 4 4
and R . Then F divides the third one, and the uniqueness of factorization implies F jP ,
2
p
A + B = C ; where C = 2Q:
1
4
2
4 2
and if Q and R were constant, so would be P . Now, the induction hypothesis implies that
A, B , C are constant. Hence, by (1) and (2), also P , Q and R are constant.
Poland 1. [1987: 278]
Let F be a one-to-one mapping of the plane into itself which maps closed rectangles
into closed rectangles. Show that F maps squares into squares. Continuity of F is not
assumed.
Solution by George Evagelopoulos, Athens, Greece.
We consider an arbitrary rectangle ABCD. Let O be the center of the rectangle,
and X , Y , Z , T the midpoints of the sides AB , BC , CD, DA respectively. Let P , PAB ,
PBC , PCD , PDA denote the rectangles ABCD, ABY T , BCZX , CDTY , DAXZ , and a, b
the segments Y T , XZ , respectively. Thus we have
P = PAB [ PCD = PBC [ PDA ; a = PAB \ PCD ; b = PBC \ PDA :
231
1 2 +1
Editor's comment. Edward T.H. Wang, Wilfrid Laurier University, Waterloo, On-
tario, points out that this problem is identical to problem 3 of the 1987 Arany Daniel
Competition [1989: 5]. A solution was submitted by Curtis Cooper, Central Missouri
State University, Warrensburg, but we have already given a solution for the earlier occur-
rence. See [1990: 264{265].
3. Show that 25x + 3y and 3x + 7y are multiples of 41 for the same integer values
of x and y.
Editor's comment. Several readers pointed out that the problem as stated can not be
correct. These included Richard A. Gibbs, Fort Lewis College; Richard K. Guy, University
of Calgary; Stewart Metchette, Culver City, CA; Bob Prielipp, University of Wisconsin{
Oshkosh; and Edward T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario. Guy
points out that 25x + 3y is a multiple of 41 just if x 13y mod 41 (or y 19x mod 41),
while 3x + 7y is a multiple of 41 just if x 25y (or y 23x) mod 41. The only solutions
are x y 0 mod 41. Gibbs suggested that perhaps the 7 should be replaced by 2, since
2(25x + 3y) = 41x + 3(3x + 2y). This gives 25x + 3y is a multiple of 41 i 3x + 2y is,
since 41 is prime. Bob Prielipp suggested that the correct statement may have been that
25x + 3y and 31x + 7y are multiples of 41 for the same values of x and y. This is since
25x +3y 0 mod 41 i 16(25x +3y) 0 mod 41 i 31x +7y 0 mod 41, since 16 25 31
and 16 3 7 mod 41.
4. The celebrated Fibonacci sequence is dened by
a = 1; a = 2; ai = ai, + ai, (i > 2):
1 2 2 1
2 2 2 1 2
have
a n = ak (a n,k, + a n,k, ) + ak, a n,k,
2 2 1 2 2 1 2 1
= ak a n, k + ak a n,k, ;
+1 2 ( +1) 2 2
= an , 2anan, : 2
+1 1
233
* * *
We now turn to problems from the January 1990 number of Crux with solutions for
the problems of the Singapore Mathematical Society Interschool Mathematical Competition,
1988 (Part B) [1990: 4{5].
1. Let f (x) be a polynomial of degree n such that f (k) = k k for each k =
0; 1; 2; : : : ; n. Find f (n + 1).
+1
( +1)!
x+ (, n+1 x(x , 1) (x , n)
1)
f (x) = n
( +1)!
x+1 :
Thus (
f (n + 1) = n + 1 + ( , 1) n
= 1 n odd +1
n+2 n n even.
n +2
Editor's note. Murray Klamkin points out that this problem has appeared pre-
viously, for example in M.S. Klamkin, USA Mathematical Olympiads 1972{1986, MAA,
1988, pp. 20{21. Slightly less elementary solutions were sent in by Duane M. Broline,
Eastern Illinois University, Charleston; by Michael Selby, University of Windsor; and by
Edward T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario.
2. Suppose ABC and B``` Dh
DEF in the gure are congruent.
```
`` ` A
A
hhhh
h hhh
Prove that the perpendicular bisectors ( ( ( ( ( ((C
A
hhF
Pn 1 + 2n + 2n + 1 1 , 1 , 1 + 1 0;
and 5jPn. Finally suppose n = 4m. Then 2n (2 )m 1m 1 and Pn 1 + 1 + 1 + 1
4
4(6 0). Thus Pn is divisible by 5 just in case n is odd or twice an odd number.
Alternate Solution by Duane M. Broline, Eastern Illinois University, Charleston;
by Murray S. Klamkin, University of Alberta; and by Edward T.H. Wang, Wilfrid Laurier
University, Waterloo, Ontario.
Pn = 1 + 2n + (5 , 2)n + (5 , 1)n. If n is odd, then 5jPn by expansion of the last
two terms using the binomial theorem. If n is even, n = 2m, we get Pn = 2(1 + 2 m) + 5k,
2
ri 0 and mi odd for 1 i n +1. Since there are only n odd numbers up to 2n we must
1 2 +1
5. Find all positive integers x; y; z satisfying the equation 5(xy + yz + zx) = 4xyz.
Solutions by Seung-Jin Bang, Seoul, Republic of Korea; by Duane M. Broline, East-
ern Illinois University, Charleston, Illinois; by Stewart Metchette, Culver City, California;
by Bob Prielipp, University of Wisconsin{Oshkosh; by Michael Selby, University of Wind-
sor; and by Edward T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario. The
write-up given is Wang's.
The given equation can be rewritten as
1+1+1=4 ()
x y z 5
Without loss of generality, we may assume that 1 x y z . Since x, y and z are
positive x = 1 is clearly impossible. On the other hand, if x 4, then
1 + 1 + 1 3 < 4:
x y z 4 5
Thus () can have integer solutions only if x = 2 or x = 3.
When x = 3, () becomes y + z = . If y 5, then y + z < . Thus y = 3 or
1 1 7 1 1 2 7
4. In either case, we can easily nd that the corresponding value for z is not an integer.
15 5 15
y + z < . Thus y = 4; 5, or 6.
10 3 10
1 1 2 3
For y = 4, we solve and get z = 20, and a solution (2; 4; 20). For y = 5, we have
7 10
z = 10 and the solution (2; 5; 10). For y = 6 we nd z = 15=2, which is not an integer.
To summarize, the given equation has exactly 12 solutions obtained by permuting
the entries of each of the two ordered triples (2; 4; 20) and (2; 5; 10).
* * *
This completes the Corner for this month. Send me your nice solutions!
* * * * *
BOOK REVIEW
Edited by ANDY LIU, University of Alberta.
More Mathematical Morsels, by R. Honsberger, published by the Mathematical
Association of America, Washington, 1991, ISBN 0-88385-313-2, softcover, 322+ pages.
Reviewed by Andy Liu, University of Alberta.
This may be considered a sequel to the author's earlier problem anthology Math-
ematical Morsels. However, in the present volume, almost all of the 57 morsels appeared
previously in Crux Mathematicorum. The author has foraged in its fertile soil before. See
for instance his Mathematical Gems III, especially Chapter 7.
Let us examine rst the material from outside Crux. Morsel 51 presents a new
proof of a theorem of Moessner due to K. Post. Morsel 55 illustrates with an example the
236
Original problems are particularly sought. But other interesting problems may also
be acceptable provided they are not too well known and references are given as to their
provenance. Ordinarily, if the originator of a problem can be located, it should not be
submitted by somebody else without permission.
To facilitate their consideration, your solutions, typewritten or neatly handwritten
on signed, separate sheets, should preferably be mailed to the editor before May 1, 1992,
although solutions received after that date will also be considered until the time when a
solution is published.
1671. Proposed by Toshio Seimiya, Kawasaki, Japan.
A right triangle ABC with right angle at A is inscribed in a circle ,. Let M; N be
the midpoints of AB; AC , and let P; Q be the points of intersection of the line MN with
,. Let D; E be the points where AB; AC are tangent to the incircle. Prove that D; E; P; Q
are concyclic.
1672. Proposed by Walther Janous, Ursulinengymnasium, Innsbruck, Austria.
Show that for positive real numbers a; b; c; x; y; z ,
!
a (y + z ) + b (z + x) + c (x + y) 3 xy + yz + zx ;
b+c c+a a+b x+y+z
and determine when equality holds.
1673. Proposed by D.J. Smeenk, Zaltbommel, The Netherlands
0
.
Triangle ABC is nonequilateral and has angle = 60 . A is an arbitrary point of
line BA, not coinciding with B or A. C 0 is an arbitrary point of BC , not coinciding with
B or C .
(a) Show that the Euler lines of ABC and A0 BC 0 are parallel or coinciding.
(b) In the case of coincidence, show that the circumcircles of all such triangles
0 0
A BC meet the circumcircle of ABC at a xed point.
1674. Proposed by Murray S. Klamkin, University of Alberta.
Given positive real numbers r; s and an integer n > r=s, nd positive x ; x ; : : : ; xn
1 2
so as to minimize
!
1 + 1 + + 1 (1 + x )s (1 + x )s : : : (1 + x )s :
n
xr xr
1 2
xrn 1 2
constant.
(a) Show that Pnk PVk is also a constant.
4
(b) Does there exist a value of m 6= 1; 2 and a value of n 3 such that Pnk PVk m
=1
2
is independent of P ?
=1
238
* * * * *
SOLUTIONS
No problem is ever permanently closed. The editor will always be pleased to consider
for publication new solutions or new insights on past problems.
1448. [1989: 149; 1990: 222] Proposed by Jack Garfunkel, Flushing, N.Y.
If A, B , C are the angles of a triangle, prove that
2 X sin A X cos A; 2
3 2
with equality when A = B = C .
239
[Editor's note. The statement of this problem is false, but Murray Klamkin conjec-
tured on [1990: 222] that it should be true for acute triangles. Three readers have since
sent in proofs of this conjecture. Two of them contain dierent \best possible" results, so
here they are.]
II. Solution by G.P. Henderson, Campbellcroft, Ontario.
We choose the notation
p so that A B C . Then we prove:
(a) if B < 2 arcsin , 3
= 21:1, the inequality is false;
1
p
4
(b) if 2 arcsin , B < 30, it is false for small values of C and becomes true
3 1
as C increases toward B ;
4
2(1 , v) (8)
240
The rst factor can be omitted because x > 0 (from (5)) and v < 1. From the second part
of (5),
4x > 1 , v :
2 2
and
x = 1 + v2, 2v = (1 , v)(1 + 2v ) :
2
2
2
For these values of x, the expression under the radical sign in (8) is positive. Using (8),
(7) becomes
f (x) = 8(1 , 4v)x + 8v(1 , v)x , 8v + 2v + 9v , 3 0:
2 3 2
(10)
We are to determine the values of x and v that satisfy (1), (9) and (10). We nd
p
B
f (x ) = (1 , v)(4v 1 , v , 1) = 1 , sin 2 (2 sin B , 1);
1
2
p p
2
p
Case (a): B < 2 arcsin( 3 , 1)=4.
We have
p
0 < v < ( 3 , 1)=4 < 1=4; f (x ) < 0; f (x ) < 0; 1 2
is, as C increases from 0 to B . Therefore (10) is false for small values of C and becomes
true as C approaches B .
Case (c): B 30.
We have
p p p
1 < 6 , 2 v < 2 ; f (x ) 0; f (x ) 0;
4 4 2 1 2
= 1 , (1 +2 t) + (1 + t)(2x , 1):
2
2
After small manipulation, the inequality under investigation takes on the form
p
f (t; x) = 4(1 + 2t)x , 2(1 + t) 1 , t x , (2 + 7t + 2t ) 0:
2 2
For A < , equality requires t = 0 and x = 1, which occur when the triangle is
equilateral.
Klamkin's conjecture was also proved by JOHN LINDSEY, Northern Illinois Uni-
versity, Dekalb.
The editor did not succeed in combining the above proofs. Maybe some reader can
easily derive one result from the other. Note, by the way, that Kuczma's upper bound on
A and Henderson's lower bound on B are related:
p! p !
arccos 1 , 4 3 = 180 , 4 arcsin 3,1 ;
8 4
corresponding to the isosceles triangle with
p ! p !
A = arccos 1 ,84 3 137:8; B = C = 2 arcsin 3 , 1 21:1;
4
for which equality holds in the problem.
* * * * *
1553. [1990: 171] Proposed by Murray S. Klamkin, University of Alberta.
It has been shown by Oppenheim that if ABCD is a tetrahedron of circumradius
R, a; b; c are the edges of face ABC , and p; q; r are the edges AD; BD; CD, then
64R (a + b + c )(p + q + r ):
4 2 2 2 2 2 2
where E is the sum of the squares of all the edges emanating from one of the vertices and
0
0 1
4
(1)
The right-hand expression in (1) is less than the claimed one (n + 1) R =4, except for 4 4
jvij = R (i = 0; 1; : : : ; n);
w = n1 (v + + vn ) ; jwj R:
1
243
By denition,
n n n !
E = 12 jvi , vj j = (R , vi vj ) = n R ,
X X X 2
1
2 2 2 2
v i = n (R , jwj );
2 2 2
i;j =1 i;j =1 i =1
n
X n
X
E = 0 jv , vi j = (2R , 2v vi ) = 2n(R , v w);
0
2 2
0
2
0
i =1 i =1
dots denote inner (scalar) products of vectors. Since v0 w ,Rjwj, and using the
A.M.-G.M. inequality, we obtain
E E = 2n (R , v w)(R , jwj ) 2n (R + Rjwj)(R , jwj )
0 1
3 2
0
2 2 3 2 2 2
2 !=3
R + j j
3
2 3
= 8n R 4(R , jwj) =
3
2
w
5 1 3
" !#
1 2
8n R 3 (R , jwj) + 3
3
R + j wj
=
4n R : 3 3
4
2 3
Remark. Assume n 2. Then the estimate in (1) is sharp; it turns into equality if and
only if 3w = ,v . Note that, except for n = 2 and 3, there are many nonisometric optimal
0
congurations and that, except for n = 3, the regular simplex is not among them.
II. Solution by Walther Janous, Ursulinengymnasium, Innsbruck, Austria.
Let O be the circumcenter and G the centroid of the simplex A An. Then the 0
following identities are more or less familiar, at least for triangles and tetrahedra:
n
X n
X
GAi = (n + 1)(R , OG ) ;
2 2 2
E + E = (n + 1)
0 1 GAi 2
i=0 i=0
(see, e.g., various simplex-problems by Murray Klamkin and/or George Tsintsifas in Crux,
or Mitrinovic et al, Recent Advances in Geometric Inequalities, p. 493 and p. 502).
Therefore
E + E = (n + 1) (R , OG )
0 1
2 2 2
n
X n
X
E0 =0
w wj a j and E 0 =
0
2
0 1
wi wj aij 2
j =1 1 i<j
244
are weighted sums, W = Pni wi the sum of the weights, and where aij is the length of the
edge between vertices i and j .
=0
* * * * *
1554. [1990: 171] Proposed by Marcin E. Kuczma, Warszawa, Poland.
Describe all nite sets S in the plane with the following property: if two straight
lines, each of them passing through at least two points of S , intersect in P , then P belongs
to S .
Solution by Chris Wildhagen, Rotterdam, The Netherlands.
Suppose that a nite set F of points in the plane has exactly one of the following
properties:
I: the points of F are collinear;
II: jFj = 5, and the points of F are the vertices of a parallelogram together with
the point of intersection of the diagonals;
III: all points of F , except one, are on a straight line.
Clearly such a set has the \closure property" as required by the problem. Conversely, each
set F obeying the conditions of the problem satises I, II or III as we shall show.
So take a nite set F with the closure property. We may assume that jFj 4,
else F satises I or III and we are done. If A; B; C; D are 4 points of F , no 3 of which
are collinear, then we can group them in two pairs, say fA; B g and fC; Dg, such that the
two lines AB and CD intersect. This observation shows that F contains 3 collinear points
lying on a line `, say.
For each point L 2 F \ `, and each point P of F not on `, let (P; L) be the
non-obtuse angle between PL and ` (we assume that I doesn't hold, else we are done).
Choose P and L such that (P; L) is minimal and if
there are several choices for P choose the one with B B L
0 r r
0
AA
Also, since AB = CD, OM = ON , while B Q C
6 OMP = 6 ONR = 90. Hence OMP = ONR, so OP = OR. Then, since XP = XR,
OX is the perpendicular bisector of PR, as is QS . Hence the lines OX and QS coincide,
which implies that Q, S and O are collinear.
Also solved by JORDI DOU, Barcelona, Spain; L.J. HUT, Groningen, The Nether-
lands; WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria; KEE-WAI LAU,
Hong Kong; P. PENNING, Delft, The Netherlands; JOHN RAUSEN, New York; D.J.
SMEENK, Zaltbommel, The Netherlands; and the proposer.
* * * * *
1556. [1990: 172] Proposed by K.R.S. Sastry, Addis Ababa, Ethiopia.
Let and n be xed positive integers, not both 1. Prove that the equation
x +y =n
2 2
2
xy + 1
has innitely many natural number solutions (x; y).
246
or
(2x , n y) , (n , 4)y = 4n :
2 2 4 2 2
(1)
2
If n = 1 and = 2, the second term on the left vanishes and we have the innite set of
solutions x = y 1. With any other values for n and , the multiplier of y in (1) is not 2
shown by substitution into (1). And it is well known that a Pell equation with at least one
solution has an innite number of solutions.
Also solved by H.L. ABBOTT, University of Alberta; C. FESTRAETS-HAMOIR,
Brussels, Belgium; RICHARD I. HESS, Rancho Palos Verdes, California; WALTHER
JANOUS, Ursulinengymnasium, Innsbruck, Austria; MURRAY S. KLAMKIN, University
of Alberta; MARCIN E. KUCZMA, Warszawa, Poland; KEE-WAI LAU, Hong Kong; P.
PENNING, Delft, The Netherlands; KENNETH M. WILKE, Topeka, Kansas; and the
proposer.
The problem is related to (and inspired by) problem 6 of the 1988 IMO [1988 : 197]
(see Crux 1420 [1990:122] for another such problem).
The proposer also observes that, for n > 2, any solution of the given equation
2
* * * * *
1557. [1990: 172] Proposed by David Singmaster, South Bank Polytechnic, Lon-
don, England.
Let n be a positive integer and let Pn be the set of ordered pairs (a; b) of integers
such that 1 a b n. If f : f1; 2; : : : ; ng ! R is an increasing function, and g : Pn ! R
dened by
g(a; b) = f (a) + f (b)
is one-to-one, then g denes a (strict) total ordering on Pn by
(a; b) (c; d) if and only if g(a; b) < g(c; d) .
Moreover will have the property
(a; b) (c; d) whenever a c and b d (and (a; b) 6= (c; d)) . ()
Does every strict total ordering of Pn which satises () arise in this way?
247
(1; 1) (1; 2) (2; 2) (1; 3) (1; 4) (2; 3) (3; 3) (2; 4) (3; 4) (4; 4):
satises (). Suppose there exists f : f1; 2; 3; 4g ! R and g : P ! R as above such 4
that
(a; b) (c; d) () g(a; b) < g(c; d):
Then we have
f (2) + f (2) = g(2; 2) < g(1; 3) = f (1) + f (3);
f (1) + f (4) < f (2) + f (3);
f (3) + f (3) < f (2) + f (4):
By summing we get a contradiction.
A similar counterexample was found by MARCIN E. KUCZMA, Warszawa, Poland;
and by the proposer and the editor (jointly).
The problem was inspired by W.R. Ransom's problem 3471 of the Amer. Math.
Monthly, solution in Vol. 38 (1931) 474{475, which contains an example of an ordering
(for n = 4) which does arise in the above way.
* * * * *
1558. [1990: 172] Proposed by George Tsintsifas, Thessaloniki, Greece.
Let P be an interior point of a triangle ABC and let AP; BP; CP intersect the
circumcircle of ABC again in A0; B 0 ; C 0, respectively. Prove that the power p of P with
respect to the circumcircle satises
jpj 4rr0;
where r; r0 are the inradii of triangles ABC and A0B 0 C 0.
Solution by Murray S. Klamkin, University of Alberta.
We change the notation by letting (A ; A ; A ) = (A; B; C ) and (A0 ; A0 ; A0 ) =
(A ; B ; C 0). For an interior point P , p = R , (OP ) 0, where R is the circumradius
1 2 3
0 0
1 2 3
2 2
a0 = R0 = R R R0 = R p ; etc.,
1 2 1 2 2 1
a R RRR RRR
1 3 1 2 3 1 2 3
where ai , a0i are the sides of triangles A A A and A0 A0 A0 respectively, and hence
1 2 3 1 2 3
a0i = aiK
Rip
1 2 3
248
Inequalities, or other references given re Crux 1514 on [1991: 117]). From the same ref-
erences we see that, with F , F 0 the areas of triangles A A A and A0 A0 A0 and ri , ri0 the
1 2 3
0
F 0 = 21 a0i ri0 = 12 ai ri Ri and r0 = P2F ;
X X 2
ai Ri
so that, accounting for the factor p=K ,
P !
r0 = p P
a i ri Ri 2
:
K ai Ri
It is also known that X
ai ri Ri = 2pF 2
[equation (6) in M.S. Klamkin, An identity for simplexes and related inequalities, Simon
Stevin 48 (1974{75) 57{64]. Hence the given inequality can be rewritten as
P !
P 4Krp a i ri Ri 2
P
ai Ri
or X X
R R R ai Ri 4r ai ri Ri = 8rpF:
1 2 3
2
The latter now follows immediately from the product of the two known inequalities
R R R 2rp
1 2 3
occurring on [1990: 127]. Problems 1514 [1991: 116] and 1543 [1991: 190] are also closely
related.
Also solved by the proposer.
* * * * *
1559. [1990: 172] Proposed by R.S. Luthar, University of Wisconsin Center,
Janesville.
Find a necessary and sucient condition on reals c and d for the roots of x +3x + 3 2
cx + d = 0 to be in arithmetic progression.
249
p
Now the solutions to x + 2x + d = 0 are x = ,1 1 , d. In other words, if (2) holds,
2
Lau points out that a more general result (replace the 3 by an arbitrary coecient)
appears as problem 7, section 1:5 of E.J. Barbeau, Polynomials, Springer-Verlag, 1989,
with solution on p. 251.
* * * * *
1560. [1990: 172] Proposed by Ilia Blaskov, Technical University, Gabrovo, Bul-
garia.
The sequence a ; a ; a ; : : : of real numbers is such that, for each n, an > 1 and the
equation [an x] = x has just n dierent solutions. ([x] denotes the greatest integer x:)
2 3 4
Find limn!1 an .
Solution by Margherita Barile, student, Universita degli Studi di Genova, Italy.
The only solution of [a x] = x must be x = 0. Let now n > 1. If [anx] = x then
x 0, because x < 0 implies an x < x, since an > 1. As [an0] = 0, [an x] = x has exactly n
1
is true for a if a 2.
1 1
and q
z = x + y 4xy , 3:
To make z an integer, 4xy ,3 must be the square of an odd number 2n+1, i.e., 4n +4n+1 =
2
4xy , 3 or
n + n + 1 = xy;
2
where n can be any integer. We can choose x and y as factors of n + n + 1 (this can
2
equation (1).
A nice special group is the series
1; 1; 3; 7; 19; 49; 129; 337; 883; 2311; 6051; : : :;
where each term has the form F n , Fn Fn made up of Fibonacci numbers. Any three
2 +1 +1
Z
Z
! Z
Z
Z
Z
AP F Z
Z
P
e PP Z
e P PP O Z
e PP Z
e K P PP
PPE
Z
Z
H hhh
,
ehhh PP Z
De hhhh PP Z
e L hhhh
G h
PP
hhhh PP Z
Z
e hhh PPZ
hh
PZ
B e
M
h
P C
We have
6 FMH , 6 EMG = 6 FEH , 6 EHG = 6 ECH = 6 MCH = 6 MGH , 6 MHG;
therefore
6 MKG = 6 FMH + 6 MHG = 6 MGH + 6 EMG = 6 MLH:
Since 6 FME = 90, we conclude 6 MKL = 6 MLK = 45. Thus
6 MDC = 6 MKL , 6 DMK 45:
Equality holds only when A = F , i.e. A0BC is equilateral, i.e. 6 B = 60, 6 C = 30,
6 A = 90.
ZZ M
Note: the solution is based on the nice DPP Z
PP AZ
property illustrated at the right, namely, if PP AZ
CD is the bisector of 6 C , then KLM is an K PP A Z
P
A PZ
LP
isosceles right triangle. F A
PZPP
ZC
E
253
KM
HK = AH = sin C;
KC AC
donc
HC = HK + KC = sin C + 1
KC KC
et
KC = sinHC = b cos C :
C + 1 sin C + 1
Ainsi
KC MC () sinb cos C a = b sin A = b sin(B + C ) = b sin3C
C + 1 2 2 sin B 2 sin2C 2 sin 2C
() 2 cos C sin 2C (sin C + 1) sin 3C
() sin 3C + sin C sin C sin 3C + sin3C
() 1 sin 3C; A
Z
A
ce qui est vrai. FP AZ
PP Z
A PP Z E
III. Solution by the proposer. A PZ
Z
Note rst that
A
A
Z
Z
62 MDC = 2(6 MAC + 6 ACD) DAPP Z
P PP Z
= 26 MAC + 6 ACM
A
A
PPZ
PZPZ
= 6 MAC + 6 AMB: (1) B A
M
PC
n n nX
1 = a kN , 1 A kN , 1 < A k1N , 1 = 1;
1 1
X X +1
k
=1 n k n =1 +1 k n =1 +1
a contradiction. Thus fAng increases, and it suces to show An 3 for all n. So suppose
some An > 3. Then
n n n n
1 = A kN1 , 1 < 3kN1, 1 2k1N 12 k1
X X X X
k
=1 n k !
=1 k k !
=1 =1
2
n n
1 X 1 1 X 1 1 1 1
< 2 1 + (k , 1)k = 2 1 +
k
=2 k k , 1 , k = 2 1 + 1 , n < 1;
=2
a contradiction.
Also solved by MARGHERITA BARILE, student, Universita degli Studi di Genova,
Italy; MORDECHAI FALKOWITZ, Tel-Aviv, Israel; EMILIO FERNANDEZ MORAL,
I.B. Sagasta, Logro~no, Spain; ROBERT B. ISRAEL, University of British Columbia;
RICHARD KATZ, California State University, Los Angeles; MARCIN E. KUCZMA,
Warszawa, Poland; KEE-WAI LAU, Hong Kong; P. PENNING, Delft, The Netherlands;
and CHRIS WILDHAGEN, Rotterdam, The Netherlands.
Several solvers observed that the result is true for any N > 1 and/or found better
bounds for An. Falkowitz points out that problem E3348 of the American Mathematical
Monthly (1989, p. 735) can be shown to be similar to this problem.
255
* * * * *
1565. [1990: 205] Proposed by Marcin E. Kuczma, Warszawa, Poland.
From the set of vertices of the n-dimensional cube choose three at random. Let pn
be the probability that they span a right-angled triangle. Find the asymptotic behavior of
pn as n ! 1.
Solution by Shailesh A. Shirali, Rishi Valley School, India.
Coordinatise n-space and assume without loss that the vertices of the n-dimensional
cube are the 2n possible points all of whose coordinates are 0 or 1. Let us now count the
total number of 3-sets of vertices that can serve as the vertices of a right-angled triangle.
This equals 2nq(n) where q(n) is the number of such 3-sets in which the \elbow" of the right
angle is the origin O(0; ,!
; 0). ,!
Now if A, B are vertices of the unit cube and 6 AOB = 90,
then the scalar product OA OB = 0. Considering the placement of 1's in the two vectors
(the 1's must occur in disjoint positions if the scalar product is to be 0), it is clear that
q(n) can be equivalently dened as the total number of (unordered) pairs of nonempty
disjoint subsets of a given n-set. It follows that
,
nX
q(n) = 12 C (n; r)(2n,r , 1) ; where C (n; r) = r!(nn,! r)! :
1
r =1
This summation is easily evaluated, for Pnr , C (n; r) = 2n , 2 while by the binomial
=1
1
theorem ,
nX ,
nX
C (n; r)2n,r = C (n; r)2r = 3n , 2n , 1:
1 1
r =1 r =1
Therefore we nd that
q(n) = 12 (3n , 2 2n + 1):
Finally the required probability must equal
= 3(3n , 2 2n + 1) :
n ) n n
p(n) = C2(2qn(n; 3) (2 , 1)(2 , 2)
Thus for large n, p(n) is approximately equal to 3(3=4)n. In particular, p(n) tends to 0 as
n tends to 1.
Also solved by EMILIO FERNANDEZ MORAL, I.B. Sagasta, Logro~no, Spain; C.
FESTRAETS-HAMOIR, Brussels, Belgium; RICHARD I. HESS, Rancho Palos Verdes,
California; ROBERT B. ISRAEL, University of British Columbia; WALTHER JANOUS,
Ursulinengymnasium, Innsbruck, Austria; JOHN H. LINDSEY, Northern Illinois Univer-
sity, Dekalb; P. PENNING, Delft, The Netherlands; CHRIS WILDHAGEN, Rotterdam,
The Netherlands; and the proposer.
* * * * *
256
and so
8 cos A(2 cos A , 1 + cos B cos C ) = 0:
Therefore either cos A = 0 or 1 , 2 cos A = cos B cos C .
If cos A = 0 then A = =2, so A and H coincide. In this case the lengths
OA; AH; OH would be R; 0; R which are not in arithmetic progression.
Hence we must have 1 , 2 cos A = cos B cos C . Replacing cos A, cos B , cos C by
b +c ,a ; c +a ,b ; a +b ,c
2 2 2 2 2 2 2 2 2
4a bc , 4a (b + c , a ) = (a , (b , c ))(a + (b , c )) = a , (b , c )
2 2 2 2 2 2 2 2 2 2 2 4 2 2 2
or
3a , 4a (b + c , bc) + (b , c ) = 0:
4 2 2 2 2 2 2
CONTENTS
1
257
the edges 1; 2; 3; : : : ; k in such a way that at each vertex which belongs to two or more
edges the greatest common divisor of the integers labelling those edges is equal to 1.
[A graph G consists of a set of points, called vertices, together with a set of edges
joining certain pairs of distinct vertices. Each pair of vertices u, v belongs to at most
one edge. The graph G is connected if for each pair of distinct vertices x, y there is some
sequence of vertices x = v ; v ; v ; : : : ; vm = y such that each pair vi ; vi (0 i < m) is
0 1 2 +1
joined by an edge of G.]
5. Let ABC be a triangle and P an interior point in ABC . Show that at least one
of the angles 6 PAB , 6 PBC , 6 PCA is less than or equal to 30.
6. An innite sequence x ; x ; x ; : : : of real numbers is said to be bounded if there
0 1 2
is a constant C such that jxi j C for every i 0. Given any real number a > 1, construct
a bounded innite sequence x ; x ; x ; : : : such that
0 1 2
jxi , xj j ji , j ja 1
for every pair of distinct non-negative integers i; j .
*
As the I.M.O. is ocially an individual event, the compilation of team scores is
unocial, if inevitable. Team scores are obtained by adding up the individual scores of
the members. These totals, as well as a breakdown of the medals awarded by country, is
given in the following table.
Congratulations to the USSR for winning this year.
Rank Country Score Prizes Total Prizes
(Max 252) 1st 2nd 3rd
1 USSR 241 4 2 - 6
2 China 231 4 2 - 6
3 Romania 225 3 2 1 6
4 Germany 222 1 5 - 6
5 USA 212 1 4 1 6
6 Hungary 209 2 3 1 6
7 Bulgaria 192 - 3 3 6
8{9 Iran 191 2 1 2 5
8{9 Vietnam 191 - 4 2 6
10 India 187 - 3 3 6
11 Czechoslovakia 186 - 4 1 5
12 Japan 180 - 3 3 6
13 France 175 1 1 4 6
14 Canada 164 1 2 2 5
15 Poland 161 - 2 4 6
16 Yugoslavia 160 - 2 3 5
17 South Korea 151 - 1 4 5
18{19 Austria 142 - 2 3 5
18{19 United Kingdom 142 1 - 2 3
260
Rank Country Score Prizes Total Prizes
20 Australia 129 - - 3 3
21 Sweden 125 - 2 1 3
22 Belgium 121 - - 3 3
23 Israel 115 - 1 2 3
24 Turkey 111 - - 4 4
25 Thailand 103 - 1 1 2
26 Colombia 96 - - 2 2
27{28 Argentina 94 - - 3 3
27{28 Singapore 94 - 1 1 2
29{30 Hong Kong 91 - - 2 2
29{30 New Zealand 91 - - 2 2
31{32 Morocco 85 - - 1 1
31{32 Norway 85 - - 3 3
33 Greece 81 - - 2 2
34 Cuba 80 - - 2 2
35 Mexico 76 - - 1 1
36 Italy 74 - - 1 1
37{38 Brazil 73 - - 1 1
37{38 The Netherlands 73 - - 1 1
39 Tunisia (4) 69 - - 2 2
40{41 Finland 66 - - 1 1
40{41 Spain 66 - - 1 1
42 Philippines (4) 64 - - 2 2
43 Denmark (5) 49 - - - -
44 Ireland 47 - - - -
45 Trinidad & Tobago (4) 46 - - - -
46 Portugal 42 - - - -
47 Mongolia 33 - - - -
48{49 Indonesia 30 - - - -
48{49 Luxembourg (2) 30 - - 1 1
50{51 Switzerland (1) 29 - - 1 1
50{51 Iceland 29 - - 1 1
52 Cyprus (4) 25 - - - -
53 Algeria 20 - - - -
54 Macao 18 - - - -
This year the Canadian team slipped back from eleventh to fourteenth place, but
put in a good performance. The team members, scores, and the leader of the Canadian
team were
Ian Goldberg 39 Gold
J.P. Grossman 36 Silver
Adam Logan 32 Silver
Peter Milley 22 Bronze
Mark von Raamsdonk 21 Bronze
Ka-Ping Yee 14 Honourable Mention.
Team leader: Professor Georg Gunther, Sir Wilfred Grenfell College.
261
The USA team slipped from 3rd to 5th. The results for its members were
Joel Rosenberg Gold
Kiran Kedlaya Silver
Robert Kleinberg Silver
Lenhard Ng Silver
Michael Sunitsky Silver
Ruby Breydo Bronze
The team leaders were Professors C. Rousseau, Memphis State University, and Dan Ullman
of George Washington University.
* * *
Next we turn to \Archive" problems from the 1988 numbers of Crux.
3. [1988: 101] Tenth Atlantic Provinces Mathematics Competition.
Suppose S (X ) is given by
S (X ) = X (1 + X (1 + X (1 + X (1 + : : : )):
2 3 4
Is S (1=10) rational?
Solution by Murray S. Klamkin, University of Alberta.
We show that S (1=n) is irrational for any positive integer n > 1 where S (X ) is the
power series
X + X + X + = Xn n = :
X
3 6 ( +1) 2
Any rational number when expanded into a \decimal" in any base must eventually be
periodic. Here S (1=n) in base n is :1010010001 : : : : By considering the blocks of successive
zero digits, it is clear the number cannot be periodic.
8. [1988: 102] Tenth Atlantic Provinces Mathematics Competition.
Find the sum of the innite series
1+ 1 + 1 + 1 + 1 + 1 + 1 + 1 +
2 3 4 6 8 9 12
where the terms are reciprocals of integers divisible only by the primes 2 or 3.
Correction and generalization by Murray S. Klamkin, University of Alberta.
One must add \except for the rst term 1" at the end of the statement.
If we replace reciprocals of the numbers of the form 2k 3` with reciprocals of numbers
of the form P k1 P k2 : : : Pnk where P ; P ; : : : ; Pn are distinct primes, then the sum is given
1 2
n
1 2
by
1 1 !
1 1 1 1
!
P P : : : Pn :
!
X X X
P k P k P k = P ,1 P ,1
1
P ,1
2
[Editor's Note: The problem was also solved by John Morvay, Springeld, Missouri.]
* * *
262
Next is an interesting solution to one of the problems of the 29th I.M.O. in Australia.
Normally we don't publish solutions to these problems as the \ocial" solutions are readily
available. This warrants an exception.
6. [1988: 197] 29th I.M.O.
Let a and b be positive integers such that ab + 1 divides a + b . Show that 2 2
a +b 2 2
ab + 1
is the square of an integer.
Solution by Joseph Zaks, The University of Haifa, Israel.
If (ai ) is a doubly-innite sequence (i.e., ,1 < i < 1) of reals, satisfying
(1) ai = nai , ai for all i, and
+2 +1
a +a 2 2
(2) a ia +i 1 = n for one value of i;
+1
i i +1
where the last equality uses the assumption (2). Observe that ai = nai ,ai is equivalent +2 +1
to ai = nai , ai , thus it follows that (2) holds for all i.
+1 +2
To solve the problem, suppose a and b are integers, for which (a + b )=(ab + 1) = n 2 2
As a by-product, we can get all the triples (a; b; n) 6= (1; 1; 1) of integers satisfying
(a + b )=(ab + 1) = n: Such a and b will be any two consecutive terms of the sequence (ai )
2 2
p
x = m 2m , 4 ;
2 4
i p p i
m m m , m m , m ,4
! !
ai = p + 4 2
, p
4 2 4
m ,4 4 2 m ,4 2 4
* * *
Now we turn to problems from the January 1990 number of the Corner. First let me
apologize for leaving Murray Klamkin's name o the list [1991: 235] of people who solved
problem 5 of the Singapore Mathematical Society Interschool Mathematical Competition.
His solution was stuck to one for the Chinese contests discussed below.
1. [1990: 5] First Selection Test of the Chinese I.M.O. Team 1988.
What necessary and sucient conditions must real numbers A; B; C satisfy in order
that
A(x , y)(x , z ) + B (y , z )(y , x) + C (z , x)(z , y)
is nonnegative for all real numbers x, y and z ?
Solution by Murray S. Klamkin, University of Alberta.
Expanding out, we get
Ax + By + Cz + yz (A , B , C ) + zx(B , C , A) + xy(C , A , B ) 0:
2 2 2
6
4 B
7
5
C
where 2 = C , A , B , 2 = B , C , A, 2
= A , B , C , must be nonnegative denite. The
necessary and sucient conditions for this are that all the principal minors be nonnegative.
Thus A, B , C must be 0. All the principal 2nd order minors are the same and equal
1 (2(BC + CA + AB ) , A , B , C ): 2 2 2
4
Since
p pthis pexpression corresponds to 4 times the square of the area of a triangle of sides
A, B , C , these three numbers must satisfy the triangle inequalities. The third order
minor, which is the determinant of the matrix, is 0 (just add the 2nd and p 3rdp rowpto the
top one). Summarizing, the necessary and sucient conditions are that A, B , C are
possible sides of a triangle, possibly degenerate.
Comment: On replacing A; B; C by a ; b ; c , respectively, where a; b; c are sides of
2 2 2
where ; ;
are the angles of the triangle. Since x; y; z are arbitrary, so are ax; by; cz so
these can be replaced by arbitrary u; v; w to give
u + v + w , 2vw cos , 2wu cos , 2uv cos
:
2 2 2
conditions, but that did not extend to the observation about triangles. He sets X = x , y
and Y = x , z , and the expression becomes
AXY + B (X , Y )X + C (,Y )(X , Y ) = BX + (A , B , C )XY + CY :2 2
2(AB + BC + CA).]
2. [1990: 5] First Selection Test of the Chinese I.M.O. Team 1988.
Determine all functions f from the rational numbers to the complex numbers such
that
(i) f (x + x + + x ) = f (x )f (x ) : : : f (x )
1 2 1988 1 2 1988
for all rational numbers x ; x ; : : : ; x , and
1 2 1988
g(x + y) = f (fx(0)
+ y) = f (x)f (y)(f (0)) = f (x) f (y) (f (0)) = g(x)g(y):
1986
1987
and KG = (a , c) = CE . The
1
2
1
2 C
production of GO intersects IH (or its E
production) in L. Consider triangle E
E
OIL. Angle 6 ILO =
(= 6 C ), since E
OL ? CA, IL ? CB . It is easy to
E
E
verify that E
E
KG = CE
IL = sin
E
E
2 sin
E
GXXO ( (E F
and (
XX E
XXX
HF = CD :
OL = sin D E XX
EE XXX
2 sin
E
XXX
XXX
We see that triangles LIO and CED E
(
XXL
((((
E ( (
are similar, for LI : LO = CE : CD. K X X
XX((((E(
E H((((
((
Moreover LI ? CE and LO ? CD im- I
ply IO ? ED. Also
E
E
E
IO = 2DE A EB
sin
;
and for
= 30 this gives IO = DE , as desired.
*
The last solution to a problem in the January number received from readers is the
following.
1. [1990: 5] Second Selection Test of the Chinese I.M.O. Team 1988.
Dene xn = 3xn, + 2 for all positive integers n. Prove that an integer value can
1
be chosen for x such that 1988 divides x .
0 100
0 1
so that
xn = (x + 1)3n , 1:
0
where 0 < r < 1988. Since 1988 is not divisible by 3, r is relatively prime to 1988. It now
suces to show the existence of integers x and ` such that r(x + 1) , 1 = 1988`. Since
0 0
(r; 1988) = 1 there are an innite number of desired integer pairs (x ; `). Indeed since 0
1988 = 4 7 71, we must have x 0 mod 4, x 0 mod 7 and x 0 mod 71; it
100 100 100
follows that x 0 mod 4, x 1 mod 7, and x 45 mod 71. By the Chinese Remainder
0 0 0
Theorem we have x 400 mod 1988.
0
266
* * *
We now turn to problems from the February 1990 number of the Corner. Here are
solutions to the rst ve problems from the XIV \ALL UNION" Mathematical Olympiad
(U.S.S.R.) [1990: 33{34]. A frequently occurring team solver is JACL, an acronym for two
Edmonton students, Jason A. Colwell of Old Scona School and Calvin Li of Archbishop
MacDonald School, and Andy Liu of The University of Alberta.
1. All the two-digit numbers from 19 to 80 are written in a row. The result is read
as a single integer 19202122 : : : 787980. Is this integer divisible by 1980?
Solutions by Seung-Jin Bang, Seoul, Republic of Korea; by JACL; by Stewart Metch-
ette, Culver City, California; by John Morvay, Springeld, Missouri; by Bob Prielipp,
University of Wisconsin{Oshkosh; by Don St. Jean, George Brown College, Toronto; and
by Edward T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario.
Since 1980 = 2 3 5 11, we only need check divisibility by 4; 9; 5 and 11. Since
2 2
the last two digits are 8 and 0, the number is divisible by both 4 and 5. The sum of the
digits in the odd positions is 1 + (2 + 3 + 4 + 5 + 6 + 7)10 + 8 = 279. The sum of those in
the even positions is
9 + (0 + 1 + 2 + 3 + 4 + 5 + 6 + 7 + 8 + 9)6 + 0 = 279:
Since 279 + 279 = 558 is divisible by 9, so is the number. Since 279 , 279 = 0 is divisible
by 11, so is the number. Hence the number is divisible by 1980.
2. Side AB of a square ABCD is divided into 1A D
n segments in such a way that the sum of lengths of the ,
,
even numbered segments equals the sum of lengths of 2 ,
the odd numbered segments. Lines parallel to AD are ,
,
drawn through each point of division, and each of the 3 ,
n \strips" thus formed is divided by diagonal BD into .. ,
,
a left region and a right region. Show that the sum of .
,
the areas of the left regions with odd numbers is equal ,
to the sum of the areas of the right regions with even n ,,
numbers. B C
Solutions by JACL; and by Edward T.H. Wang, Wilfrid Laurier University,
Waterloo, Ontario.
We show that the conclusion holds when ABCD is any rectangle. Let AD = a
and AB = b. Denote by hi the height of the ith strip, so that i hi = i hi and
P
odd
P
even
P n h = b. Also let L and R denote the area of the ith left region and right region,
i =1 i i i
respectively, i = 1; 2; : : : ; n. Then
Li + Li = 12 ab
X X
i
odd i even
and
Li + Ri = a hi = 12 ab:
X X X
i
even i even i even
267
Solution by JACL.
Denote by [P ] the area of polygon P . Since BM = CM , [ABM ] = [ACM ].
Similarly [ADP ] = [ACP ] so that [ABCD] = 2[AMCP ]. Now MP is parallel to BD, and
C is at the same distance from MP as BD is from MP . Since ABCD is convex, A is on
the opposite side to C of BD. Now AMP and CMP have the same base. Since AMP
has the greater altitude, [AMP ] > [CMP ]. Finally, note that
[AMP ] = 1 AM AP sin MAP 1 AM AP 1 a 2
2 2 8
by the Arithmetic-Mean Geometric-Mean Inequality. Hence
[ABCD] = 2[AMCP ] = 2[AMP ] + 2[CMP ] < 4[AMP ] 1 a 2
2
as desired.
5. Does the equation x + y = z have solutions for prime numbers x; y and z?
2 3 4
z , x = y and z + x = y .
2 2 2
If z , x = 1 and z + x = y , then
2 2 3
2z = y + 1 = (y + 1)(y , y + 1):
2 3 2
It follows that y 6= 2, so that y is an odd prime number. Thus y > 2, making y +1 > 3 and
y , y +1 > y +1. Since z is a prime number we now obtain z = y +1 and y , y +1 = 2z .
2 2
2z = y + y = y(y + 1):
2 2
It follows that y 6= 2, so y is an odd prime. Thus y > 2. Since y < y + 1 and z is a prime
we get y = z and y + 1 = 2z . This gives z = 1, a contradiction.
Comment by Murray S. Klamkin, University of Alberta.
There are however an innite number of solutions if we remove the prime number
restrictions, and even for the more general equation
xr + ys = z t where (rs; t) = 1:
Just let x = 2ms, y = 2mr and z = 2p. Then we must satisfy mrs + 1 = pt. Since
rs is relatively prime to t, there exist an innite number of positive integer pairs (m; p)
satisfying the conditions.
* * *
That's all the space available for this issue! Send me your contests and solutions.
* * * * *
BOOK REVIEW
Edited by ANDY LIU, University of Alberta.
Mathematical Challenges, edited by the Scottish Mathematical Council, published
by Blackie and Son Limited, Glasgow, 1989, ISBN 0-216-92622-X, softcover, 180 pages.
Reviewed by Andy Liu, University of Alberta.
This book is a collection of the 213 problems set in the rst twelve years of the
\Mathematical Challenges", organized by the Scottish Mathematical Council. Unfortu-
nately, the preface gives no description of the organization and philosophy of the \Math-
ematical Challenges". The readers should bear in mind that this is not a conventional
contest, where participants write the papers in time-limited sittings. Instead, four sets of
problems are sent out during the school year, and the students have over a month to think
about them.
269
Many of the problems have a strong recreational
avour, a welcome relief from
the usual \solid and stolid" mode of British oerings. There is a great variety, includ-
ing cryptarithms, cross-numbers, logical inferences and number patterns. There is also a
healthy dosage of problems in geometry, as well as in diophantine equations, inequalities
and other Olympiad topics. Some of them are familiar classics, but most are new to the
reviewer.
Here is a sample problem. \Four players successively select two cards from four
numbered cards face down on a table. The total values of the cards drawn are 6, 9, 12
and 15. Then two of the four cards are turned over and their total value is found to be 11.
Determine the value of each of the other two cards."
The problems are primarily at the pre-Olympiad level, even though a few prove
rather dicult. The reviewer feels strongly that there is an urgent need for a book such as
this. Many aspiring young students who may be scared o by the Olympiad level problems
will nd the ones in this book attractive. Once they have gained some prociency and
condence, they will look for further challenges.
Although no names are associated with the book, it is no secret that the driving
force behind the \Mathematical Challenges" is the indefatigable Dr. David Monk of the
University of Edinburgh. To initiate such an enrichment program for the students is an
admirable feat, but to have sustained the eort over such a long period displays conviction
and dedication at the highest level.
For readers in North America, there is unfortunately no local distributor. All orders
should be placed directly with Blackie and Son Limited at 7 Leicester Place, London,
WC2H7BP, United Kingdom. The current price is $8.50 , plus $3.81 for air mail or $1.26
for surface mail.
* * * * *
PROBLEMS
Problem proposals and solutions should be sent to B. Sands, Department of
Mathematics and Statistics, University of Calgary, Calgary, Alberta, Canada T2N 1N4.
Proposals should, whenever possible, be accompanied by a solution, references, and other
insights which are likely to be of help to the editor. An asterisk (*) after a number indicates
a problem submitted without a solution.
Original problems are particularly sought. But other interesting problems may also
be acceptable provided they are not too well known and references are given as to their
provenance. Ordinarily, if the originator of a problem can be located, it should not be
submitted by somebody else without permission.
To facilitate their consideration, your solutions, typewritten or neatly handwritten
on signed, separate sheets, should preferably be mailed to the editor before June 1, 1992,
although solutions received after that date will also be considered until the time when a
solution is published.
270
that DA = DP .
1682. Proposed by Walther Janous, Ursulinengymnasium, Innsbruck, Austria.
For a nite set S of natural numbers let
Alt(S ) = x , x + x , ; 1 2 3
where x > x > x > are the elements of S in decreasing order. Determine
1 2 3
f (n) = Alt(S );
X
where the sum is extended over all non-empty subsets S of f1; 2; : : : ; ng.
1683. Proposed by P. Penning, Delft, The Netherlands.
Given is a xed triangle ABC and xed positive angles ; such that + < .
For a variable line l through C , let P and Q be the feet of the perpendiculars from A and
B , respectively, to l, and let Z be such that 6 ZPQ = and 6 ZQP = (and, say, the
sense of QPZ is clockwise). Determine the locus of Z .
1684. Proposed by Stanley Rabinowitz, Westford, Massachusetts.
Let
f (x; y; z ) = x + x z + ax z + bx y + cxyz + y :
4 3 2 2 2 2
Prove that for any real numbers b; c with jbj > 2, there is a real number a such that f can
be written as the product of two polynomials of degree 2 with real coecients; furthermore,
if b and c are rational, a will also be rational.
1685. Proposed by Clark Kimberling, University of Evansville, Evansville, Indi-
ana.
If equilateral triangles A A P ; A A P ; A A P are erected externally on the sides
2 3 1 3 1 2 1 2 3
of a triangle A A A , then A P ; A P ; A P concur at a point R called the isogonic center
1 2 3 1 1 2 2 3 3
(see p. 218 of R.A. Johnson, Advanced Euclidean Geometry). Prove that the line joining
R and its isogonal conjugate is parallel to the Euler line of the triangle.
1686. Proposed by Iliya Bluskov, Technical University, Gabrovo, Bulgaria.
The sequence a ; a ; a ; : : : is dened by a = 4=3 and
0 1 2 0
3(5 , 7an)
an = 2(10 an + 17) +1
n, n 2
p 1 1
Now p
(s , c) + s ,c c 2 s , 1;
p
as this inequality is equivalent to (c , s + s) 0. Therefore
2
ps
g(a; b; c) 2ps , 1 = 2 , 11=ps :
Since the term on the right hand side decreases, we get
g(a; b; c) 1p ;
2 , 1= k
p p p
as was to be shown. There holdsp equality for b = 0, c = k , k, a = k, i.e. for x = k,
1
x = = xn, = 0, xn = k , k.
2 1
For the original problem (k = 2) we get the better bound
p p
f (x ; : : : ; xn ) p 2 = 4 + 2:
1
2 2,1 7
Also solved by MARCIN E. KUCZMA, Warszawa, Poland; KEE-WAI LAU, Hong
Kong; JOHN H. LINDSEY, Northern Illinois University, DeKalb; EDWARD T.H. WANG,
Wilfrid Laurier University, Waterloo, Ontario; CHRIS WILDHAGEN, Rotterdam, The
Netherlands; and the proposer.
Kuczma, Lau and Lindsey all obtained the best possible bound given above.
* * * * *
1568. [1990: 205] Proposed by Jack Garfunkel, Flushing, N.Y.
Show that
sin A p2
X 2
cos A
X
3
where the sums are cyclic over the angles A; B; C of an acute triangle.
I. Solution by Vedula N. Murty, Penn State Harrisburg.
The inequality proposed can be restated as
y p2 (1 + x) 2
3
where
y = Rs = sin A; 1 + x = 1 + Rr = cos A;
X X
is a parabola opening upwardsp with vertex at the point M (,1;p0) shown in my picture.
p y-axis at (0; 2= 3), which is below the point E (0; 3), and passes through
It cuts the
A(1=2; 3 3=2). Thus the inequality clearly holds for all Type I triangles since the corre-
sponding region (vertical shading) is above the curve (1). It is also clear that there exist
obtuse Type II triangles for which the proposed inequality does not hold. However, from
the classic Steinig inequality
y 16x , 5x
2 2
(item 5.8 in Bottema et al, Geometric Inequalities), we see that the proposed inequality
holds for all x values for which
16x , 5x 34 (1 + x) ;
2 4
i.e.,
4x + 16x + 39x , 32x + 4 0:
4 3 2
The above polynomial has two real roots x 0:157 and x = 0:5, and two complex
0 1
conjugate roots. Hence for x x 1=2 the proposed inequality holds. [Editor's note.
0
From Murty's paper, all acute Type II triangles correspond to values x 1=4, thus all
acute Type II triangles satisfy the proposed
q
p inequality. Alternatively, since point F in
Murty's picture has y{coordinate 71 , 8 2=4, one need only check that
p
p2 (1 + 14 ) < 14 71 + 8 2;
q
2
3
since then the curve (1) must pass below F .]
II. Solution by the proposer.
From
cos A = 1 + 4 sin(A=2)
X Y
we have
X
cos A) = 1 + 8 sin(A=2) + 16 sin (A=2)
Y Y
( 2 2
= 1 + 10 sin(A=2):
Y
which simplies to
(R + 4r)(R , 2r) = R + 2Rr , 8r 0; 2 2
which is true.
Also solved by WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria;
MARCIN E. KUCZMA, Warszawa, Poland; and JOHN H. LINDSEY, Northern Illinois
University, DeKalb.
Kuczma also showed that the inequality holds for some obtuse triangles, namely
for any triangle whose angles do not exceed 132 400. Janous made the substitution
A ,! ( , A)=2, etc., in the given inequality to obtain
cos A p2 sin A
2
;
X X
2 3 2
valid for all triangles.
* * * * *
1569. [1990: 205] Proposed by Edward T.H. Wang, Wilfrid Laurier University,
Waterloo, Ontario.
Evaluate
a ln(1 + ax) Z
1+x
dx 2
0
where a is a constant.
I. Solution by Margaret Izienicki, Mount Royal College, Calgary.
Let F (a) be the given integral. By dierentiation with respect to a under the
integral sign, we get
d F (a) = a Z
x dx + ln(1 + a ) (F (0) = 0)
2
da (1 + ax)(1 + x )
0 1+a 2 2
= 1
a x + a , a dx + ln(1 + a )
Z 2
1+a 1 + x 1 + x 1 + ax
2
0
2
1+a 2 2
1 1
= 1 + a 2 ln(1 + a ) + a tan, a 2 1
= d 1 ln(1 + a ) tan, a :
2 1
da 2
276
II. Solution by Maria Mercedes Sanchez Benito, I.B. Luis Bunuel, Alcorcon, Madrid,
Spain.
If we made the change
x = 1a+,au
u;
then
dx = ,(1 +(1au+) ,au()a , u)a du = , (11++au
a du; 2
2
) 2
1 + ax = 1 + a(a , u) = 1 + a ;
2
1 + au 1 + au
and
1 + x = 1 + a , u = 1 + 2au + a u + a , 2au + u = (1 + a )(1 + u ) :
2 2 2 2 2 2 2
2
1 + au (1 + au) (1 + au) 2 2
Also,
x = 0 =) u = a;
x = a =) a(1 + au) = a , u =) u = 0:
Thus
Z
a log(1 + ax) dx = log[(1 + a )=(1 + au)] (,du)
Z 0 2
0 1+x 2
a 1+u 2
a log(1 + a )
Z 2 a log(1 + au)
Z
=
0 1+u du2
, 1 + u du
0
2
1+u 0
2
so a log(1 + ax)
dx = 12 log(1 + a ) arctan a:
Z
2
0 1+x2
Poland; KEE-WAI LAU, Hong Kong; JOHN H. LINDSEY, Northern Illinois University,
DeKalb; BEATRIZ MARGOLIS, Paris, France; JEAN-MARIE MONIER, Lyon, France;
P. PENNING, Delft, The Netherlands; CHRIS WILDHAGEN, Rotterdam, The Nether-
lands; and the proposer. There was one partial solution received.
The proposer remembered hearing the integral from his calculus professor in
Taiwan in 1961. Hess found the integral in Gradshteyn and Ryzhik, p. 556, no. 4.29.18.
Janous located it in the Russian book Prudnikov, Brychkov, Marichev, Integrals and Series
(Elementary Functions), p. 506, item 2.6.11.1.
* * * * *
1570. [1990: 205] Proposed by P. Penning, Delft, The Netherlands.
In n-dimensional space it is possible to arrange n + 1 n-dimensional solid spheres
of unit radius in such a way that they all touch one another. Determine the radius of the
small solid sphere that touches all n + 1 of these spheres.
Solution by John H. Lindsey, Northern Illinois University, DeKalb.
The (n + 1){tuples
(1; 0; : : : ; 0); (0; 1; 0; : : : ; 0); : : : ; (0; : : : ; 0; 1)
p
are mutually distance 2 apart and really lie in Rn since they lie in the hyperplane
x + x + + xn = 1. The center of mass
1 2 +1
1 ; ; 1
n+1 n+1
is distance s
n
s
1, 1 +n 1
2 2
=
n+1 n+1 n+1
p
from the original points. Blowing up the scale by a factor 2 to make the original n + 1
q
points mutually distance 2 apart, this becomes 2n=(n + 1) so the center ball should have
radius s
2n , 1:
n+1
[Lindsey ended by noting that the result is known and referring to a book of C.A.
Rogers, probably Packing and covering, Cambridge Univ. Press, 1964, p. 79. |Ed.]
Also solved by EMILIO FERNANDEZ MORAL, I.B. Sagasta, Logro~no, Spain;
MURRAY S. KLAMKIN, University of Alberta; MARCIN E. KUCZMA, Warszawa,
Poland; CARLES ROMERO CHESA, I.B. Manuel Blancafort, La Garriga, Catalonia,
Spain; D.J. SMEENK, Zaltbommel, The Netherlands; and the proposer.
* * * * *
278
2R 2R 2R
jR , OP jF = jpjF ;
area F 0 =
2 2
4R 4R 2 2
inequality, p p
3R 3R0 a + b + c a0 + b0 + c0 aa0 + bb0 + cc0;
2 2 2 2 2 2
n n =1
Assuming that plus and minus signs occur with equal frequency, i.e.
lim + + n = 0;
1
n!1 n
prove or disprove that the series necessarily converges.
279
1 n
X 1X 1 (,1)k
X
n=1 n = p2P p + k nk =1
diverges. But
(n) , 1 + + n (n) ;
1
n n n
where (n) is the number of primes n. Since (n) n= log n, we infer
lim + + n = 0:
1
n!1 n
[Editor's note : the above proof refers to the Prime Number Theorem ((n) n= log n),
but only needs the consequence (n)=n ,! 0.]
II. Generalization by Robert B. Israel, University of British Columbia.
It does not necessarily converge. More generally, the following is true.
THEOREM. Consider any sequence an 0 such that 1 n an = 1. Then there is
P
+ + n = 0; and X 1
nan diverges.
nlim
1
!1 n n=1
Proof. Take a sequence of integers (tk ) such that t0 = 0, and for all k,
tX
k+1
an 2k + 1;
n=tk +1
tk+1 , tk 9(tk , tk, ); 1
tX
k+1
Sk = an ( 2k + 1):
n=tk +1
We have Sk = k
j =0 Tjk where
P2
[Tjk is the sum of every (2k + 1)st term of Sk , starting at the j th.] Let Uk be the subset
of f0; 1; : : : ; 2kg corresponding to the k + 1 largest Tjk values (breaking ties arbitrarily).
Thus
Tjk 2kk++11 Sk :
X
j 2U k
n t
= k +1 j 2U j 62U k k
so that the series n an diverges. Next we will estimate Nn n . Note that in any 2k + 1
P P
=1
consecutive n's from tk + 1 to tk there will be k + 1 positive and k negative n , adding
+1
to 1. Thus t +1
n = tk2k +,1tk :
k
X
+1
n t = k +1
n n n + n
X
+
n
=1
j =0 n=tj +1 i=0 n=tk+i(2k+1)+1
n=tk +q(2k+1)+1
kX ,1 tj +1 , tj
+ q + r:
j =0 2j + 1
Note that for j k , 1,
tj , tj 9j ,k (t , t ) = 3j 3j , k (t , t )
+1
+1
2 +2
2j + 1 2j + 1 k k, 2j + 1 k k,1 1
k, j ,k
23k , 1 3j, k (tk , tk, ) = 23k , 1 (tk , tk, );
1 +1
2 +2
1 1
k k, 1
,! 0 as n ,! 1;
since N > tk tk , tk, 9k, for all k.
1
1
The proposer's solution was the same as Janous's. The other three solutions used
the fact that the series (n log n),1 diverges.
P
* * * * *
1573. [1990: 239] Proposed by Toshio Seimiya, Kawasaki, Japan.
Let M be the midpoint of BC of a triangle ABC . Suppose that 6 BAM = 6 C and
6 MAC = 15. Calculate angle C . O
I. Solution by Jordi Dou, Barcelona, Spain. C
Let
be the circle through AMC , C
C
with centre O and radius R. Since 6 MAC = A b
B@ C
15, 6 MOC = 30. Since 6 MCA = 6 MAB , B b @bb
C
C
AB will be tangent to
at A. The distance B @
@
b
15
b C
MM 0 of M to AC is R sin 30 = R=2, and B
B @
b
b
b
C
0
since CB = 2CM the distance BB of B to B @ b C
b C
AC will be R = AO. Thus AB is parallel to B @
b
b CB
0
B @ b
OC , so 6 AOC = 90 and 6 C = 6 BAM = B
@
b
b
C M
0
(see N. Altshiller Court, College Geometry, pp. 126 and 300). The lower bound e + f 2 2
As A and D get closer and 6 UIV approaches 180, we get as close as we want to this
upper bound. It would only be reached by a degenerate \quadrilateral", a straight line
BV AIDUC , with A I D.
Also solved by C. FESTRAETS-HAMOIR, Brussels, Belgium; JOHN G. HEUVER,
Grande Prairie Composite H.S., Grande Prairie, Alberta; MARCIN E. KUCZMA,
Warszawa, Poland; P. PENNING, Delft, The Netherlands; and the proposer. One
incorrect solution was received.
The proposer notes that the lower bound provides an immediate proof of a conjecture
of Tutescu, referred to on p. 411 of D.S. Mitrinovic et al, Recent Advances in Geometric
Inequalities.
* * * * *
283
t = 5 : 705232
= 220:375
t = 7 : 20516
128
= 160:28125
t = 9 : 6983312
512 = 13639:28125
t = 11 : 3781162256
2048
= 1846270:6328125
Two partial results:
LEMMA 1. There are no solutions for n = 2t where t is even.
Proof. Write the equation as m=n = x=10` where x is an integer containing precisely
the digits of m and n. Since each integer is congruent mod 9 to the sum of its digits, we
would have m + n x as well as m n x, so that
(n , 1) x = n x , x m , (m + n) = ,n mod 9:
But if n = 2t with t even, n , 1 is divisible by 3 while n is not, so this is impossible. 2
284
LEMMA 2. If there is at least one solution (m; n) for a given n = 2t with m n +8,
then there are innitely many.
Proof. Consider a solution
Ab = X ;
n 10`
where A is an integer and b a single nonzero digit, and X is an integer containing precisely
all digits of A, n and b (we could get along without the b except for the rather articial
restriction that m should not end in 0). Let A = n Y + r where 0 r < n. Then
X = 10` 10 nA + b = 10` Y + 10` 10 nr + b
+1
n
Then for a nonnegative integer C of d digits (with leading zeros allowed), we will obtain a
new solution
ACb = Y CZ ;
n 10`
i.e.
10d A + 10 C + b = 10d Y + 10 C + Z 10,`;
+1
+1
n
if
(10d , 1) r = (n , 1) C: (1)
Take d > 0 so that 10d 1 mod (n , 1) (d exists because (10; 2t , 1) = 1 when t is odd),
and let
C = r 10n ,,11 ;
d
since 0 r < n, C < 10d, so this C will have at most d digits, and (1) is satised.
The condition m n + 8 ensures that Y CZ has no leading zeros. For, if Ab = m <
10 n we get Y = 0 and r = A and the new solution would be
ACb = CZ :
n 10`
But then (since m = Ab = rb = 10 r + b)
r = m10, b n 10
,1
so that
C = r n(10, 1, 1) 10 10, 1 :
d d
Since C is an integer, this means that C 10d, and there are no leading zeros in C . 2
1
[Editor's note. The condition \neither m nor n ends in 0" was meant by the editor to
prevent such trivial solutions as 50=2 = 25:0; 500=2 = 25:00; and 5=20 = 0:25; 5=200 =
0:025; . There was probably a better way to do this.]
285
2 ,
Z A
center of , , ,
ZZ
AA
2
A,
C
Q
6 A = 6 BQC , 6 PBQ = BO C , PO Q ;
6 2
6 2
2
286
, 1
B A
B
1 2
,! ,! 1
,
2
q
A
allelogramme et AK =O O , quelle que soit A 1 2,
B A
C
la position de A sur , . Le lieu de K,!est1
Q
l'image du cercle , par la translation O O .
1 1 2
and
x cos
cos + x(sin
+ sin ) + 1 = 0:
2
Prove that
x cos cos + x(sin + sin ) + 1 = 0:
2
by multiplying the rst equation by cos A and the second by cos B and subtracting, they
give
x(cos A , cos B ) sin C = ,(cos A , cos B ) + x sin(A , B ): (2)
Note that both results become identities for A = B . Introduce
s = A +2 B ; v = A ,2 B ;
and use
,2 sin s sin v = cos A , cos B; 2 cos s sin v = sin A , sin B;
then (1) and (2) become, after division by sin v,
x sin s cos C = cos s;
x sin s sin C = , sin s , x cos v:
Elimination of C is now simple by squaring and adding:
x sin s = x sin s(cos C + sin C )
2 2 2 2 2 2
or
x (cos v , sin s) + 2x sin s cos v + 1 = 0;
2 2 2
or
x cos(v + s) cos(v , s) + x(sin(s + v) + sin(s , v)) + 1 = 0:
2
Q = (1 + b )(1 + b ) : : : (1 + bn);
1 2
v v 1 vn 2
hence
Q (1 + an=n)n; 1
n ; 1 i n , 1:
ai = ai=n (1)
We can conclude that
P (1 + an=n),n; 1