Crux

Mathematicorum
Published by the Canadian Mathematical Society.

http://crux.math.ca/

The Back Files

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

Journal title history:

➢ The first 32 issues, from Vol. 1, No. 1 (March 1975) to
Vol. 4, No.2 (February 1978) were published under the
name EUREKA.
➢ Issues from Vol. 4, No. 3 (March 1978) to Vol. 22, No.
8 (December 1996) were published under the name
Crux Mathematicorum.
➢ Issues from Vol 23., No. 1 (February 1997) to Vol. 37,
No. 8 (December 2011) were published under the
name Crux Mathematicorum with Mathematical
Mayhem.
➢ Issues since Vol. 38, No. 1 (January 2012) are
published under the name Crux Mathematicorum.
 ft ft -: ft ft ft ft j' ft ft ft ft ft ft ft ft ft ft ft ft & ft & ft s\ & & ft ft ft ft ft ft ft
ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft s'f & ft JV 5»f ft

ISSN 0705 - 0348

CRUX MATHEf^ATICORU

Vol. 8 9 No. 3
March 1982

Sponsored by
Carleton-Ottawa Mathematics Association Mathematique d'Ottawa-Carleton
Publie par le College Algoiqum, Ottawa
Printed at Carleton University
The assistance of the publisher and th*3 support of the Canadian Ilathematical Olympiad
Committe-1, the C a r l e t o n n \ersity Department of Mathematics and Statistics, the U n i v e m t y
of Ottawa Department or Mathematics, and the eidorsem^nt of the Ottawa Valley Education
Liaison Council are gratefully acknowledged.
J
ft ft t t « { c <• « ft t J t *• ft
» t i ' *

CRUX MATHhMA^lCORUM is a problem-solving journal at the senior secondary and university

undergiadua^e levels for those \*ho practise or teach mathematics Its ourpo^e i s p r x i r a n l y
educational, *?ut it saives also those who read it for professional, cu tural, or recreational
reasons
It is published monthly (except Julu ard August). The yearly subscription rale for ten
issues is $20 in Canada, US$19 elsewhere Back issues $2 each. Bound volumes with irdex
Vols. 1&2 (combined) and each of Vols. 3-7, $16 in Canada and US*15 elsev/here. Cheques and
money crders, payable to CRUX MATHEMATICORVrf, should be sent to the managing editor
All communications about the content (articles, problems, solutions, etc.) should be
sent to the editor. Ail changes ot address and inquiries about subscriptions and back issues
should be sent to the managing editor.
Editor Lao Sauve, Algonquin College, 281 Echo Drive, Ottawa, Ontario, Canada K1S 1N3.
Managing Editor- r.G.B. Maskell, Algonquin College, 200 Lees Ave., Ottawa, Ontario,
Cana.ua i ±S uC^.
Typist-compositor Nghi Chung.
Second Class Mail Registration No. 5432. Return Postage Guaranteed.
* * ft

CONTENTS
A New Inequality for R9 r, and s . . . . Vedula N. Murty 62
An Interesting Recursive Function . . . . . Richard V. Andree 69
The Olympiad Corner: 33 . . . . . . . . . . . . . M.S. Klamkin 70
The Puzzle Corner . . . . . . . 76
Problems - Probl£mes . . . . . . . . . . . . . . 77
Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

- 61
 - 62 -

A NEW INEQUALITY FOR R, R , AND s

VEDULA N. MURTY

1. Introduction.
Given an ordered t r i p l e [R3r3s) of p o s i t i v e real numbers, does there always
e x i s t a t r i a n g l e w i t h circumradius R9 inradius r , and semiperimeter s (which we
w i l l b r i e f l y c a l l an R-r-s triangle)! I f n o t , what i s a necessary and s u f f i c i e n t
c o n d i t i o n f o r the existence of an R-r-s triangle? An i n e q u a l i t y representing such
a c o n d i t i o n was given by Blundon T i l . L a t e r , Bottema [ 2 ] derived anew Blundon's
fundamental inequality and wrote t h a t , because i t i s a necessary and s u f f i c i e n t
c o n d i t i o n , " i t i s the best c o n d i t i o n a v a i l a b l e ; i t can not be improved and, i n
p r i n c i p l e , a l l other i n e q u a l i t i e s f o r Rs r and s are consequences of t h i s one".
(Bottema then went on to discuss several such i n e q u a l i t i e s . )
In t h i s paper we w i l l f i r s t , f o r completeness, give a b r i e f o u t l i n e of
Bottema's d e r i v a t i o n of Blundon's fundamental i n e q u a l i t y and give a geometric
interpretation. We w i l l then derive an i n e q u a l i t y f o r R3 r arid s which we believe
to be new and, f i n a l l y , use i t to prove Garfunkel's inequality proposed e a r l i e r
i n t h i s journal fu],

2. Blundon's fundamental inequality.

I f a3b,o are p o s i t i v e real numbers, i t i s well known t h a t there e x i s t s a
t r i a n g l e w i t h sides a3b3o i f and only i f u- > o (i = 1 , 2 , 3 ) , where

u\ - b+c-a, U2 = c+a-b, u$ - a+b-o.

And if the u. > 0 only are known, then the sides a>b9e are easily recaptured:

a = J(W2+W3)> b = J(w3+wi), c - J(wi+W2).

I t can be shown t h a t the elementary symmetric functions o f the u^ i n terms

o f R3r,s are

u\ + Ui + U3 = 2 s ,

U2U3 + W3W1 + U\U2 = 4r(4i?+p) 9

W1W2W3 = 8 r 2 8 .

Hence the cubic equation w i t h the roots u. is

u3 - 2su2 + ^r(^R+r)u - Sr2s = 0 (1)

and (assuming R>r*s > 0) an R-r-s t r i a n g l e e x i s t s i f and only i f (1) has three
 - 63 -
positive roots. It is clear from the signs of the coefficients that (l) has
neither negative roots nor a zero root. Hence a necessary and sufficient con-
dition for the existence of an R-r-s triangle is that (l) have real roots.
The substitution u = 2v + 2s/3 transforms (l) into
V3 + pv + q = 0, (2)

where

p = |(12f?r+3r 2 -s 2 ) 5 q = ~s(18Rr-9r2-s2) $

and the roots of (2) are real i f and only i f

4p 3 + 27q2 < 0 .

Hence the required condition is

(12Rr+3r2-s2)3 + s2(18Rr-9r2-s2)2 < 09

and this is equivalent to the following, which is Blundon's fundamental inequality:

(r2+s2)2 + 12Rr3 - 20Rrs2 + 48i? 2 r 2 - 4i? 2 s 2 + 64i? 3 r < 0 . (3)

3. Geometric interpretation.
We observe that (3) is a homogeneous polynomials so only the ratios of Rs
r, and s are of interest. Therefore we introduce variables x > o and y > o
defined by
Rx = r and Ry = 8. (4)

This transforms (3) into

(x2+y2)2 + 12# 3 - 20xy2 + 48# 2 - 4z/2 + e^x < 0 . (5)

To each R-r-s triangle there corresponds a point (x,y) of the graph of (5) in
(the first quadrant of) the Cartesian plane; and, conversely, to each point
(x,y) of the graph of (5) there correspond infinitely many R-r-s triangles (one
for each R > o, for which r and s are then given by (4)). The graph of (5) is
the shaded region in the figure on page 65.
Since the left member of (5) can be written

{y2 - (2+lOx-x2)}2 - 4(l-2tf)3s

arc 0AX in the figure is the graph of
y = /(2+10tf-#2) - 2(l-2aj) 3/2 $ 0 < x < J9

and arc AjD is the graph of

 - 6'! -
(Bottema T2] has shown that arcs OAx and AiD are part of a hypocycloid of three
cusps, or deltoid.) The points of arcs OAi and A X D correspond to isosceles R-r-s
triangles; the point Ax corresponds to all equilateral R-r-s triangles; and
the points on segment OD (which are not part of the graph of (5)) correspond to
degenerate triangles. Arc OAi is concave downward, arc A X D is concave upward, the
line y = /3(i+#) is tangent to both arcs at A x , and the line x = o is tangent
to arc OAi at 0.
For every triangle ABC, we will henceforth assume without loss of generality
that A < B < C. Every triangle falls into one of two types:

type I: B > TT/3; type II: B < TT/3.

Bager [3] has shown that, with sums cyclic over A,B,C,

for type I : /3Z cos A < EsinA;

for type I I : /3£ cos A > zsinA.
(Both of these inequalities are mistakenly reversed on page 15 of Bager's a r t i c l e . )
Since
a?=| = zcosA-i and y = | = EsinA,

we obtain
f o r type I : y > / 3 ( l + a O ;

f o r type I I : y < / 3 ( l + a ; ) .

Hence the points in the region with vertical shading lines in the figure corre-
spond to triangles of type I , and those in the region with horizontal shading lines
correspond to triangles of type I I .
M-, A new inequality for R3 r3 and s .
We will prove the following
THEOREM. If an R-r-s triangle ABC is of type I3 or is an acute-angled
triangle of type IIs then

/3(r2+s2+2Rr-2R2) - Qrs > 0 . (6)

Before proving the theorem, we establish a few results we w i l l need. F i r s t ,

as we did with Blundon's inequality (3), we use (4) to obtain from (6) the equiv-
alent inequality
x2 - ^xy + y2 + 2x - 2 > 0 . (7)
/3

Next, observing that the left member of (7) is

 - 65 -

M(-|,0)
 - 66 -
(y~)2 - §(13*2-6*+6),

we can write (7) in the form

i(y~)
~ —Sl3x2-6x+6}{(y~) + —/l 3*2-6^+6} > 0. (8)
/3 /3 /3 /3
Now every point (x9y) in the shaded region of the figure lies above the line 0AX,
whose equation is y - 3/3#. Hence, for every R-r-s triangle, the corresponding
point (x9y) satisfies
y > 3/3x > — ,
/3
and the second factor of (8) is always positive. Thus (8) holds if and only if
its first factor is positive, and so (6) and (7) are both equivalent to

*\x +, ~1/ l 3 t f 2 - 6 # + 6 . (9)

Next, we observe that, for o < x < 1,

/3(l+a;) > — + — /13#2-6a;+6 9 (10)

/3 /3

since straightforward algebraic manipulations show that (10) is equivalent to

x2 < i. Finally, we w i l l need the following
LEMMA* If an R-r-s triangle ABC is acute-angled and of type J J , then the
corresponding (x9y) satisfies x > 4.
Proof. The graph of cose is concave downward for o < e < IT/2, SO we have
the simple inequalities

|i - ||. for o < e < \9

COS 0 > I.

|f(|-e). for|,eS|.

Since A < B < | < C < | i n our triangle, we have

cos A > 1 - |^ , cos B > l - |5 , cos C > ^(|-C).
2TT ^TT TT I
Hence
i« = E cosA*2-£<,-C)+jJ(f-C)=2-§jU£.
and so # > 5.
Proof of the theorem.
If triangle ABC is of type I, then the corresponding point (x,y) lies above
 - 67 -
the line MA2 in the figure; hence

y > /3(l+ar),

and then (9) follows from t h i s and ( 1 0 ) .

I f t r i a n q l e ABC i s acute-angled and o f type I I , i t follows from the lemma
t h a t the corresponding p o i n t (x,y) l i e s i n the shaded p o r t i o n o f the f i g u r e to
the r i g h t o f l i n e FG. Hence we have

y > /(2+10a?-a; 2 ) - 2 ( l - 2 a r ) 3 / 2 9 i < x < J. (11)

I t i s a simple e x e r c i s e to v e r i f y that

v^+lOar-a; 2 ) - 2 ( l - 2 a 0 3 / 2 ^ ^ § + ~ / l 3 t f 2 - 6 # + 6 , J < x <\ , (12)

w i t h e q u a l i t y holding f o r {x « 0.232 < J and) a? = J . Now (9) follows from (11)

and ( 1 2 ) , and the proof i s complete. •

Note t h a t we have only established t h a t c o n d i t i o n (6) i s necessary. It

i s not s u f f i c i e n t , as can be seen from the counterexample

A = 30° 9 B = 59° 9 C = 91°.

Here t r i a n g l e ABC i s of type I I but not acute-angled, y e t

X = ICOSA - 1 « 0.363611072
and
y = ZSinA « 2,357014996

satisfy (9).

5. Proof of Getrfunkel's inequality,

Garfunkel ! s i n e q u a l i t y , as given i n [ 4 ] , is t h a t

z c o s ^ > ~=isinA (13)

2 V3
when A+B+C = TT6 We w i l l show t h a t (13) holds whenever A9B,C are the angles of a
triangle, ( I t also holds f o r some, but not a l l , A,B,C which sum to IT but are not
the angles o f a t r i a n g l e . For example, i t does not hold when A = -u and B = C = TT;
but i t does hold when A = -n/is and B = C = 19TT/36.)
Bager has shown [ 3 , p. 10] t h a t , i f a,e,y ^re the angles of an B-r-s triangle
(with associated p o i n t (x9y))9 then

ECOS6COSY=^±f^ =£ + l(i-l. (W)

 - 68 -

EsinBsinY=^i^ = £ + ^ + , , ( 1 5 )

ESin2a = 4-IlSina = ^ = 2xy, (16)

and adding (1M-) and (15) gives

E C O S ( B - Y ) = l(x2+y2+2x-2). (17)

Now let ABC be any triangle. If it is of type I, so that

A < | < B < C,

then triangle A ' B ' C , where

B C (18)
A - 2 " J- " 2 " 2' - 2 " 2'

i s o f type I I and acute-angled. Hence, from ( 7 ) , the p o i n t (x,y) associated

with triangle A ' B ' C satisfies

l(x2+y2+2x-2) > ~xy. (19)

/3
Now, from (17) and ( 1 6 ) ,

zcos ( B ' - C ) > ~ z s i n 2 A ' , (20)

/3
from which (13) f o l l o w s .
I f ABC i s o f type I I , so t h a t

A < B < | < C,

then triangle A'B'C (again defined by (18)) is of type I and the point (x,y)
associated with triangle A'B'C again satisfies (19). So (20) holds and again
(13) follows.
We conclude that (13) holds for all triangles ABC.
REFERENCES
1. W.J. Blundon, "Inequalities associated with the triangle", Canadian
Mathematical Bulletin, 8 (1965) 615-626.
2. Oene Bottema, "Inequalities for R3 v and s", Publikaeije Elektrotehniokog
Fakulteta Univerziteta u Beogvadu, No. 338 - No. 352 (1971), pp. 27-36.
3. Anders Bager, "A Family of Goniometric Inequalities", ibid,, pp. 5-25.
4. Jack Garfunkel (proposer), Problem 613, this journal 7 (1981) 79 9
8 (1982) 55.
Department of Mathematics, Pennsylvania State University, Capitol Campus,
Middletown, Pennsylvania 17057.
 - 69 -
AN INTERESTING RECURSIVE FUNCTION
RICHARD V. ANDREE

The recursive f u n c t i o n o f the t i t l e has been around f o r some time. I have

used i t w i t h undergraduates f o r a t l e a s t twenty y e a r s , but I have no r e c o l l e c t i o n
of seeing i t discussed e a r l i e r i n the l i t e r a t u r e (although i t surely must have
been). I t i s discussed, as f a r as I know f o r the f i r s t time i n p r i n t , i n Lesson
17 Of EXPLORE COMPUTING with THS-BO and Computer Sense (Prentice-Hall, 1982).
The recursive f u n c t i o n i n question generates, from an i n i t i a l p o s i t i v e integer
n 0 , a sequence { n 0 , nl9 n 2 9 . . . } defined as f o l l o w s : i f n . = ABC...KL ( i n decimal
X % l
n o t a t i o n ) , then n^ = A'B C\ . ,K L , where

A' - \A-B\3 B' = \B-C\S C' = \C-D\j ..., K1 = \K~L\S L! = \L-A\.

Here are a few examples (where n. + n. ):

n 0 = 2468 ->• 2226 + 0044 + 0404 -*- 4444 -> 0000 + 0000 ->•...

n 0 = 3410 -> 1313 -*• 2222 -> 0000 -* 0000 -*-...

n 0 = 4721 -> 3513 ^ 2420 + 2222 -> 0000 + 0000 -> . . .

n 0 = 7269 -* 5432 -> 1113 -> 0022 -^ 0202 -v 2222 -»• 0000 -^ ...

A preliminary problem is to show that, if the initial number n 0 has four

digits, then the sequence converges to oooo and attains this limit in at most
8 steps. Formally,
1000 < no < 9999 => n. = 0000 for i > 8.

Since there are not that many four-digit numbers, this is an ideal question that
can be settled by a programmable calculator or small microcomputer.
The surprise comes when we take for the initial no a number having more than
four digits. When n 0 has five digits, for example, one rapidly concludes that
the sequence does not always converge (or, even, seldom converges) to ooooo.
A similar phenomenon occurs if no has six or seven digits. But it appears that
all sequences with an eight-digit no converge to oooooooo.
Readers are invited to show thats if n0 has any number of digits, the se-
quence will eventually enter a cycle containing ooo...ozz , where XX represents
a repeated digit not necessarily 00. It is conjectured (but this is only a wild
conjecture with little evidence to support it) that one can guarantee that xx - 00
for all &-digit n 0 only if k is a power of 2.

Department of Mathematics, University of Oklahoma, Norman, Oklahoma 73019.

 - 70 -
THE OLYMPIAD CORNER: 33
M.S. KLAMKIN

A couple of months ago in this Corner [1982: 12], I was able to give, through
the courtesy of Bernhard Leeb, the problems of the First and Second Rounds of the
West German Mathematical Olympiad for 1981. This month, again thanks to Bernhard
Leeb, I am able to give, soon after they were issued, the problems of the First
Round of the West German Mathematical Olympiad for 1982. I hope to be able to
publish the problems of the Second Round as soon as they are available. I under-
stand that the problems are distributed by mail and that the students have ll
weeks to complete each round. As usual, I solicit elegant solutions from all
readers for possible publication later in this column.

WEST GERMAN MATHEMATICAL OLYMPIAD 1982

First Round

1, Let S be the sum of the greatest odd divisors of each of the numbers
1,2,3,...,2n. Prove that 3S = 4 n + 2.
2, In a quadrilateral ABCD, the sides AB and CD are each divided into m
equal parts, the points dividing AB being labeled (in order from A)
S l f S 2 , ..., S , and those dividing CD being labeled (in order from D) T x ,
T 2 , ..., T . Similarly, the sides BC and AD are each divided into n equal
parts by the points U l 5 U 2 , ..., U (in order from B) and V l f V 2 , ...» V
(in order from A ) , respectively. Prove that each of the segments S.T.d <i <m -l)
is divided into n equal parts by the segments U.V.d < j < n-1).
v d

3, A 1982-gon is given in the plane. Let S be the set of a l l triangles

whose vertices are also vertices of the 1982-gon. A point P lies on
none of the sides of these triangles. Prove that the number of triangles in S
that contain P is even.
l\t A set of real numbers is called simple i f i t contains no elements x,y3z
such that x + y = z. Find the maximum size of a simple subset of
{l,2,...,2n+l}.

Now follows, through the courtesy of Mark E. Saul, an English translation of

the VI All-Russian Mathematical Olympiad 1979-80.
 - 71 -
VI ALL-RUSSIAN MATHEMATICAL OLYMPIAD 1979-80

V I I I Grade

1, A group o f t o u r i s t s decided to s i t i n a set o f buses i n such a way t h a t

each bus would contain the same number o f t o u r i s t s . At f i r s t they t r i e d
to s i t 22 on each bus, but i t turned out t h a t one t o u r i s t was l e f t over. But
then one bus l e f t empty, and the t o u r i s t s were able to d i v i d e themselves equally
amonq the remaining buses. I f each bus holds fewer than 33 people, how many buses
and how many t o u r i s t s were there (originally)?

2, Along a segment AB, 2n points are chosen which are symmetric i n pairs
w i t h respect to the midpoint of the segment. Any n of these points are
colored b l u e , and the r e s t are colored r e d . Prove t h a t the sum of the distances
from the red points to A i s equal to the sum o f the distances from the blue points
to B.

38 In a reqular hexagon ABCDEF, points M and K are the midpoints of CD and

DE, r e s p e c t i v e l y , and L i s the i n t e r s e c t i o n of segments AM and BK. Prove
t h a t the area of t r i a n g l e ABL i s equal to t h a t of q u a d r i l a t e r a l MDKL. Also, find
the measure o f the angle between l i n e s AM and BK.

L\t I f {x} denotes the f r a c t i o n a l p a r t o f x ( e . g . , { 7 / 5 } = 2 / 5 ) , how many

d i s t i n c t numbers are there i n the sequence

{12/1980}5 {22/l980}9 {32/l980}, ...9 {19802/l980>?

5, From a point M on the circumcircle of a triangle ABC, perpendiculars MN

and MK are drawn to lines AB and AC, respectively. For which point M
will NK be longest?

IX Grade

1, Can the natural numbers from 1 to 30 be arranged in a 5x6 rectangular

array in such a way that (a) all columns have the same sum and (b) all
rows have the same sum?

2, For which natural numbers n is 2 8 + 2 1 1 + 2n a perfect square? [This

problem also appeared in the 1981 Hungarian Mathematical Olympiad. See
[1981: 267; 1982: 46].]

3, Each vertex of a convex (2n+i)-gon i s colored w i t h one o f three d i f f e r e n t

colors. No two adjacent v e r t i c e s are col ore1 d the same. Prove t h a t the
polygon can be p a r t i t i o n e d by nonintersecting diagonals i n t o a set of triangles
 - 72 -
each of which has its three vertices of different colours. [Diagonals which
meet at a vertex are considered nonintersecting.]

4, In expressing the fraction m/n as a decimal, where m and n are natural

numbers and n < 100, a student found, at a certain place after the
decimal point, the sequence of digits 167. Show that the student must have made
an error.

5, Equilateral triangles ABC and A'B'C are drawn in a plane (both sets of
vertices being labeled clockwise). The midpoints of segments BC and B ' C
coincide. Find
(a) the angle between the lines AA' and BB';
(b) the ratio AA'/BB'.

X Grade

1, For each vertex of a tetrahedron, the point symmetric to that vertex with
respect to the centroid of the opposite face is chosen. Find the ratio
of the volume of the tetrahedron whose vertices are these new points to that of
the original tetrahedron.
2, The map of a city has the shape of a convex polygon. Each diagonal of
the polygon is a street, and the intersections of the diagonals are
intersections of the streets (but the vertices of the polygon are not considered
to be intersections of streets). Streetcar lines go through the city. Each line
goes from one end of a street to the other end, and has stops at each intersection
as well as at the endpoints. At each intersection only two streets cross, and a
streetcar runs along at least one of them. Show that one can transfer from any
intersection to any other, making no more than two transfers. (A transfer may be
made whenever two streetcar lines have a common stop.)

3, Consider the 2k numbers

2 1 - 1, 2 2 - 1 9 ...9 2 - 1,
where k > l. Show that at least one of them is a multiple of 2k + l.
I|, R being the set of all real numbers, find all functions F: R -+R which
satisfy
pF(a) + (l-p)F(b) > F{pa+(l~p)b)

for all asb3p e R.

5, The squares ABCD, A 1 B 1 C 1 D 1 , and A 2 B 2 C 2 D2 are coplanar (and their vertices
 - 73 -
are labeled counterclockwise). Vertices A and Ax coincide, and so do vertices
C and C 2 . Show that D ^ 1 BM, where M is the midpoint of B 1 B 2 , and that D1d2 = 2BM.

I now present solutions to the problems, presented here last month, of the
1982 Alberta High School Prize Examination in Mathematics. The problems and
solutions were prepared by a joint committee of the Departments of Mathematics of
the University of Alberta and the University of Calgary, consisting of G. Butler,
M.S. Klamkin, Andy Liu, J. Pounder, W. Sands, and R. Woodrow.
1, [1982: 40] A 9xi2 rectangular piece of paper is folded so that a pair of
diagonally opposite corners coincide. What is the length of the crease?
Solution.
Let ABCD be the r e c t a n g l e , w i t h AD = 9 and AB = 12,
and suppose t h a t v e r t i c e s A and C coincide a f t e r the
f o l d , which produces the crease EF, as shown i n the
figure. The Pythagorean Theorem applied to t r i a n g l e
ADC gives AC = 15. I t i s clear t h a t t r i a n g l e FCG i s
r i g h t - a n g l e d and s i m i l a r to t r i a n g l e ACD, so
AD-CG h 45
FG 4 and EF = 2FG
~DT 4 '

2, [1982: 40] Let a = s i n A , b = sin B, and c = sin (A+B). Determine cos (A+B)
as a q u o t i e n t o f two polynomials i n a3b3c w i t h i n t e g r a l coefficients.

Solution.
We have
cos (A+B) = cos A cos B - ab - abx,
x = cot A cotB - l. Since
„«+ A , ™-f. D cos A cos B a cos B + b cos A a
cot A + cotB = +—T— = T ' ab
a b ab
i t follows t h a t

( c o t A + c o t B ) 2 = ( c o t 2 A + l ) + ( c o t 2 B + l ) + 2x
a2b2
1 1
+ + 2X
" o2 b~2 >
2 2 2
X + b< + 2a b x and
-b2
cos (A+B) abx -
2ab
 - 74 -
3, [1982: 40] A cylindrical tank with diameter 4 feet and open top is partially
filled with water. A cone 2 feet in diameter and 3 feet in height is
suspended (vertex up) above the water so that the bottom of the cone just touches
the surface of the water. The cone is then lowered at a constant rate of 10 feet
per minute into the water. How long does it take until the cone is completely
submerged, given that the water does not overflow?
Solution.
The volume of the cone is (i/3)«ir«i2«3 = TT cubic feet. A cylinder of diame-
ter 4 feet and equal volume would have a height of 7T/(TT-22) = i foot. Hence the
cone is lowered through a distance of 3 - i = 11/4 feet. This process will take
(n/4)/io = n/40 minute.
l\t [1982: 40] John added the squares of two positive integers and found that
his answer was the square of an integer. He subtracted the squares of the
same two positive integers and again found that his answer was the square of a
positive integer. Show that John must have made an error in his calculation.
Solution.
Suppose there exist positive integer solutions of the system
a2 + b2 = o2 (1)

a2 - b2 = d2. (2)

We consider in particular the solution (a3b3c3d) in which o is minimal. It

follows that a3b3o3d are pairwise relatively prime. Moreover, a3o3d are odd and
b is even. Adding (l) and (2), we obtain
fl2 = (S +d } 2 + {q^\ (3)

Both (o+d)/2 and (e-d)/2 are integers, one odd and one even. Let them be denoted
by p and q9 with p odd and q even. Note that a3p3q are pairwise relatively
prime. From (3), we have
q2 = (a+p)(a-p). (4)

Now both a+p and a-p are even, so we may write a+p = 2h> a-p = 2k, and (4) becomes
q2 = i*Hk. (5)

Since a = h+k and p = h-k9 it follows that h and k are relatively prime; hence,
from (5), each is a square, say h = m2 and k = n2. Thus
2p = 2{m2-n2) (6)
 - 75 -
and, from ( 5 ) ,

q = 2mn. (7)

Now from ( l ) 9 ( 2 ) , ( 6 ) , and ( 7 ) ,

b2 = l(c2-d2) = 2pq = ^mn(m+n)(m-n).

Since ms ns mm, and w-rc are pairwise r e l a t i v e l y prime, each i s a square,

say

m - w2 9 n = x2 9 m+n = y2 s /7?-rc = s 2 .

Thus (w3xsy,z) i s also a s o l u t i o n o f the o r i g i n a l system ( i ) - ( 2 ) . Furthermore,

y2 = w2 +x2 = m+n < 2mn - q < ~- < a < a2 s

so t h a t y < o9 a c o n t r a d i c t i o n .

5, C1982: 40] Twenty-five Knights gather a t the Round Table f o r a j o l l y

evening. They belong to various Orders, w i t h every two Orders having
a t l e a s t one common member. Members of the same Order occupy consecutive seats
a t the Round Table.,
(a) I f each Order has a t most nine members, prove t h a t there i s a Knight
who belongs to no Order and there i s a Knight who belongs to every Order.
(b) Without the r e s t r i c t i o n on the sizes o f the Orders, prove t h a t there
are two Knights such t h a t between them they hold membership to every Order.

Solution.
Both statements are vacuously true if there are no Orders. Let the Knights
be labeled consecutively from 1 to 25 around the Round Table. For each Order,
the member with the lowest label will be designated its leader.
(a) We may assume that the largest Order K consists of 9, 10, ...9 ks
where k < 17. It is easy to see that no Order can contain both l and 25, as
otherwise it cannot intersect K. Suppose l belongs to some Order and 25 to another.
These two Orders cannot intersect. Hence either l or 25 belongs to no Order.
Let n be the leader with the highest label and let N be an Order of which he is
the leader. Suppose there is an order X which does not contain n. Now its leader
must have a label less than n9 and it follows that every member of I has a label
less than n. This is a contradiction, for then X and N do not intersect as re-
quired by the problem. Hence n belongs to every Order.
(b) Again let n be the leader with the highest label and N an Order of
which he is the leader. If n = 1, then he belongs to every order and the condition
 - 76 -
of the problem is trivially satisfied. We now assume that n > 1. Suppose there
is an Order X which contains neither l nor n. Again, its leader must have a
label less than n, and so has every member of X. This is a contradiction, for
then X and N cannot intersect as required. Hence l and n between themselves
hold membership to every Order.

Editor's Note. All communications about this column should be sent to Pro-
fessor M.S. Klamkin, Department of Mathematics, University of Alberta, Edmonton,
Alberta, Canada T6G 2G1.
s'f & it

THE PUZZLE CORNER

Puzzle No. lit Rebus (7)

L: y = x2

In this WHOLE way

Descartes, Reng,
A curve would say.

Puzzle No. 12: Rebus (5 M- 3 12)

\x\ < 6, 6 •> 0

So said the Indians on the strand:

(An ethnic slur, you understand—
Columbus was about to land.).
ALAN WAYNE, Holiday, Florida

Puzzle No. 13: Alphametic

SQUARE is a cube and CUBE is a square

And THREE is not prime... Stop pulling your hair.
I'll wager some money the answer you'll get
But never again such a perfect bet!

Puzzle No. 14: Alphametic

The road to solutions is not always clear;

From left to right you may often veer.
My own advice is: Stay close to the center;
Approaching a "One-Way", DO
NOT.
ENTER"
HANS HAVERMANN, Weston, Ontario

Answer to Puzzle No. 10 T1982: 35]: (a) 10; (b) no.

 - 77 -

PROBLEMS--PROBLEMES

Problem proposals and solutions should be sent to the editor, whose address
appears on the front page of this issue. 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 submit-
ted 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 his permission.
To facilitate their consideration, your solutions, typewritten or neatly hand-
written on signed, separate sheets, should preferably be mailed to the editor before
October 1, 1982, although solutions received after that date will also be con-
sidered until the time when a solution is published.

721« Proposed by Alan Wayne , Pasco-Hernando Community College, New Port

Riohey, Florida.
A propos of the editor's comment following Crux 611 T1982: 30], verify that,
with decimal integers,
(a) uniquely, TRIGG is three times WRONG;
(b) independently, but also uniquely, WAYNE is seven times RIGHT.

722. Proposed by Paul R. Beesack, Carleton University, Ottawa, Ontario.

A very large prison has loooo cells numbered from l to loooo (each
occupied by one prisoner), an eccentric warden, and an ingenious electronic device
for opening and closing cell doors. Early one day the warden announces that all
those prisoners whose cell doors are left open at the end of the day will be free
to leave the prison. During the day he presses buttons which open or close cell
doors as follows. First he opens all cell doors beginning with cell l. Next,
beginning with cell 2, he operates on every 2nd cell door, closing those that are
open and opening those that are closed. This operation is repeated throughout
the day so that at the nth step (n = 1,2,... ,10000), every nth cell beginning with
cell n is closed if it was open, or opened if it was closed. How many prisoners
are freed at the end of the day?
723, Proposed by George Tsintsifas, Thessaloniki, Greece.
Let G be the centroid of a triangle ABC, and suppose that AG,BG,CG
meet the circumcircle of the triangle again in A ' J B ' J C , respectively. Prove that
(a) GA' + GB' + G C > AG + BG + CG;
(b) AG/GA' + BG/GB* + CG/GC = 3;
(c) GA'-GB'-GC > AG-BG-CG.
 - 78 -
7241 Proposed by Hayo Ahlburg, Benidorm, Alicante, Spain.
Let ABC be a triangle (with sides a,b,c) in which the angles satisfy
C + A = 2B (that is, the angles are in arithmetic progression). Such a triangle
has many interesting properties. Establish the following (and possibly others):
(a) sin (A-B) = sin A - sin C.
(b) a2 - b2 = o(a - c).
(c) Vertices A and C, circumcentre 0 , incentre I , orthocentre H, and excentre
L a l l l i e on a c i r c l e of radius R9 where R i s the circumradius o f the t r i a n g l e .
Furthermore, i f t h i s c i r c l e meets the l i n e s AB and BC again i n A' and C , respec-
1
t i v e l y , then AA = CC* = \o-a\.

7251 Proposed by H. Kestelman, University College, London, England.

An ttxn matrix i s c a l l e d simple i f i t s eigenvectors span (f1 and i s
c a l l e d deficient i f they do n o t . I f A_ and B are simple, can A+B and AB both be
deficient? I f ^ i s simple, show t h a t adj A i s simple; i s the converse true?
(The (r,s) element of adj A i s the (s,r) cofactor o f A.)

7261 Proposed by Kesiraju Satyanarayana, Gagan Mahal Colony, Hyderabad, India.

Show t h a t , in a r e g u l a r n - s i m p l e x ,
(a) the dihedral angle i s Arccos ( l / n ) ;
(b) the product of the a l t i t u d e and the circumdiameter i s h a l f the square
of an edge (independently o f n).

7271 Proposed by J.T. Groenman, Arnhem, The Netherlands.

Let t , and t be the symmedians issued from v e r t i c e s B and C of t r i a n g l e
ABC and terminating i n the opposite sides b and c, r e s p e c t i v e l y . Prove t h a t
t, = t i f and only i f b = c.
b e
7281 Proposed by Stanley Rabinowitz, Digital Equipment Corp., Merrimack,
New Hampshire.
Let tf(P,Q,R) denote the ellipse with foci P and Q which passes through R.
If A,B,C are distinct points in the plane, prove that no two of #(B,C,A), #(C,A,B),
and #(A,B,C) can be tangent.
729 i Proposed jointly by Dick Katz and Dan Sokolowsky, California State
University at Los Angeles.
Given a unit square, let K be the area of a triangle which covers the square.
Prove that K > 2.

7301 Proposed by G.C. Giri, Midnapore College, West Bengal, India.

Prove that D = o for all real e if D = \a..\ is a determinant with

a.. - cos(i+j)e, i,j = 0 9 l s 2 5 ... 9rc; n > 2.

 - 79 -
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.

492, T1979: 291; 1980: 291; 1981: 50, 117, 277] Proposed by Ban Pedoe,
University of Minnesota.
(a) A segment AB and a rusty compass of span r > JAB are given. Show how
to find the vertex C of an equilateral triangle ABC using, as few times as possible,
the rusty compass only.
(b) Is the construction possible when r < JAB?
IV. Further comment by the proposer.
Part (b) has been solved when the entire segment AB is given, but no complete
solution has been forthcoming when only the points A and B are given. I am now
happy to communicate a beautiful solution by the professors L. Yang and J. Zhang
of the China University of Science and Technology, Hefei, Anhui, People's Republic
of China. A translation was sent to me by Geng-zhe Chang, Visiting Scholar at
Brown University, Providence, R.I.
I have been corresponding on geometrical matters with Drs. Yang and Zhang
for some time. They have expressed their pleasure at their solution being pub-
lished in this journal. I submitted to them proofs of the theorems used in their
construction and a simplification of one part of their proof, and they have agreed
to both proofs and modification.
Without loss of generality, we assume that the fixed radius of the compass
opening is r = l, so that AB > 2. We shall simply say "draw a circle" instead
of "draw a circle of radius 1". We use the notation of Stanley Rabinowitz
[1981: 277], according to which X v Y = Z means that XYZ is an equilateral tri-
angle with vertices labelled X,Y,Z in counterclockwise order.
LEMMA 1. If A,B,P are distinct points3 and the points P',B',C are defined
t
^AvP=P',B B= P"M\ and P' V B' = C, respectively, then A V B = C.
The proof given below, by transformation geometry, is not easily replaced
by a direct Euclidean proof (see fi]). We will denote by (X)T the effect of a
transformation T on a point X, and by #(Y,e) a rotation with centre Y through
an angle e.
Proof. A given translation can be composed of two opposite rotations. We
have (see Figure l)

(P')/?(P'fir/3) = P',

(P')i?(A,-TT/3) = P,
 - 80 -
and we also have
(B')i?(P',7./3) = C.
Since B'B = P'P, we must therefore have
(C)tf(A9-ir/3) = B. D

We now consider a rhombus PQ X BQ 2

with sides PQi = QXB = BQ 2 = Q 2 P of
unit length, with PB < 1, and show:
LEMMA 2. (a) We can rotate this
unit rhombus about the vertex P through
an angle TT/3, resulting in a rhombus
PQJB'Q^.
(b) We can translate the original
Figure 1
unit rhombus to a unit rhombus
P,Q{,B,,QJ' , where PP' = 1.
Proof, (a) For i = 1,2, draw circles with centres at P and Q., denoting
their intersections by 01 (chosen so that P v Q. Then draw circles with
V-
centres at 0'., taking their intersection other than P as B' The required unit
rhombus PQJB'Q 2 results. (Note that this construction is essentially the five-
circle construction given earlier T1980: 291] for an equilateral triangle PBB'
given P, B, and a rusty compass of radius greater than JPB.)
(b) We are given that PP' = l. For i = 1,2, draw circles with centres at
P' and Q., taking the intersections, other than P, of these two circles as Q'.'.
Then draw circles with centres at Q'.'s taking their intersection, other than P',
as B". The required rhombus P'Q{'B n Q^ results. D
These unit rhombuses are a tool for rotating PB through TT/3 or translating
it.
Our final construction is of a lattice of points covering the plane, using
our rusty compass to draw equilateral triangles, starting from an arbitrary equi-
lateral triangle ARS such that A v R = S (see Figure 2 ) .
Assume that a = AR and t = AS. Then in the affine system {A; a3t} we can
construct any lattice point (m9n) by counting.
LEMMA 3. There is a lattice point P such that PB < 1.
Proof. The system of closed circular disks with centres at the lattice points
covers the entire plane, so at least one of these disks covers the point B. If P
is the centre of this disk, then PB < 1.
 - 81 -

Figure 2

LEMMA 4. There is a lattice point P! such that A V P = P'. If P = (k,l)9

then P' is either (-1, k+l) or (k+l, -k).
Outline of proof'. If we rotate the whole lattice around the point A through
TT/3, all the lattice points fall on lattice points, so that P falls on a lattice
point P 1 .
The second part of the lemma shows that we can determine P' by counting. To
prove that P' is either {-I, k+l) or (k+l, -k)9 we can use the Bankoff result
stated as a Lemma by Rabinowitz [1981: 277], using only equilateral triangles,
with vertices P,Q,R,S at lattice points. This gives an elegant proof. An alter-
native is to work out the transformation of coordinates formulae for rotations of
TT/3 in oblique coordinates. •
We can now give the construction for the point C such that A v B = C. Draw
a unit rhombus P0iBQ2 and translate it step by step, with P going from lattice
point to lattice point until it reaches P \ where A v P = P \ Then we also have
B'B = P'P, defining B', and rotating the unit rhombus through TT/3 about P', we
have the situation shown in Figure l, and B' falls on C, the desired point making
ABC equilateral.

REFERENCE
l. Daniel Pedoe, "A Key Theorem in Transformation Geometry", The Mathematics
Teacher, 67 (December 1974) 716-718.
 - 82 -
617, ri981: 80] Proposed by Charles W. Trigg, San Diego* California.
The sum of two positive integers is 5432 and their least common mul-
tiple is 223020. Find the numbers.

Solution by W.J. Blundon, Memorial University of Newfoundland.

Suppose the required integers, if they exist, are ad and bd9 where a3b3d
are positive integers and (a9b) = l. It easily follows that also (ab9a+b) = l.
Since
a+b = ad+bd = 5M-32 = 28-194 _ 194 ( .
U ;
ah abd 223020 28*7965 ~ 7965'

and the first and last fractions in (l) are both in lowest terms, we must have
a + b = 194 and ab - 7965. Thus a and b are the roots of the quadratic
X2 - 194x + 7965 = 0,
that is, ia,b] = {59, 135}. It follows from (l) that d - 28, from which

{ad, bd] = {1652, 3780}.

These numbers, which are satisfactory in all respects, constitute the only
solution.

Also solved by HAYO AHLBURG, Benidorm, Alicante, Spain; JAYANTA BHATTACHARYA,

Midnapur, West Bengal, India (2 solutions); JAMES BOWE, Erskine College, Due
West, South Carolina; CLAYTON W. DODGE, University of Maine at Orono; N. ESWARAN,
student, Indian Institute of Technology, Kharagpur, India; BIKASH K. GHOSH, Bombay,
India; G.C. GIRI, Midnapore College, West Bengal, India; J.T. GROENMAN, Arnhem,
The Netherlands; J.A.H. HUNTER, Toronto, Ontario; FRIEND H. KIERSTEAD, JR.,
Cuyahoga Falls, Ohio; F.G.B. MASKELL, Algonquin College, Ottawa, Ontario; J.A.
McCALLUM, Medicine Hat, Alberta; LEROY F. MEYERS, The Ohio State University;
FRED A. MILLER, Elkins, West Virginia; KESIRAJU SATYANARAYANA, Gagan Mahal Colony,
Hyderabad, India; DAVID R. STONE, Georgia Southern College, Statesboro, Georgia;
RAM REKHA TIWARI, Radhaur, Bihar, India (2 solutions); ROBERT TRANQUILLE, College
de Maisonneuve, Montreal, Quebec; ALAN WAYNE, Holiday, Florida; KENNETH M. WILKE,
Topeka, Kansas; DAVID ZAGORSKI, student, Stuyvesant H.S., New York, N.Y.; and the
proposer.

Editorrs comment.
Meyers reported that this problem appears, with answer but without solution,
in Uspensky and Heaslet ri].
REFERENCE
1. J.V. Uspensky and M.A. Heaslet, Elementary Number Theory, McGraw-Hill,
New York and London, 1939, p.4i, problem 9.
* * *

618, ri981: 80] Proposed by J.A.H. Hunter, Toronto, Ontario.

For i = 1,2,3, l e t I . be the centres and r. the radii of the three
 - 83 -
M a l f a t t i c i r c l e s of a t r i a n g l e ABC, as shown i n the f i g u r e below. Calculate the
sides a = BC9 h - CA, and c = AB of the t r i a n g l e i n terms of the r..
TThe n o t a t i o n i n the problem has been revised to conform w i t h the s o l u t i o n
which f o l l o w s . ]

Solution by George Tsintsifas, Thessaloniki, Greece.

We adopt the usual notations I , r9 and s f o r the i n c e n t r e , i n r a d i u s , and

semiperimeter of the t r i a n g l e . From the s i m i l a r t r i a n g l e pairs BEI 29 BDI and
CFI 39 CDI (see f i g u r e ) , we have

ED
(r-r2)(s-b) and DF =
(r-r^)(s~o). (1)
r r
and from right triangle PI2I3 arises
ED + DF = EF = 2Jr~ir~i . (2)

Now, from (1) and (2) and s i m i l a r r e l a t i o n s , we obtain the system of equations

(r-r2){s~b) + (r-r3)(s-c) = 2Wr2r3

(r-r3)(s-c) + (r-ri)is-a) = 2r^r^r\
(v-v\)(s~a) + (r-v2)(s-b) = 2v^¥\r2
from which we get

\s-a = ——(Sr2r$ + ^r3rl + ^r\r2)

{s-b =•„v\ ( /r2r$ - Jr$r\ + / r 1 r2) (3)
r-r2 '
r
~° = r^^ y/r2T3 + "/r3ri ~ / p l p 2 ) -
 - 34 -
The well-known formula for r,
2 _ (s-a)(s-b)(s-c)
r
s
is equivalent to
r2{(s-a)-i(s-b)+(s-e)} = (s-a)is-b){s-c),

and substituting (3) into this yields an equation equivalent to

(-—+--=+-—:)r2 - 2(/r^Wr~2~Wr^)r + 2/rlr2r3 =0. (4)

With t h i s remarkable equation, we can solve our problem. The appropriate r o o t

o f (4) i s

^rYFs+y/r^rJ+^rJrJ

Since a = ( s - f c ) + ( s - c ) , we obtain from (3)

with r given by (5). The values of b and o can be obtained from (6) by cyclic
permutations of ri,r2,r3.

Also solved by KESIRAJU SATYANARAYANA, Gagan Mahal Colony, Hyderabad, India;

and the proposer. In addition, two incorrect solutions and one unhelpful comment
were received.

Editor's comment.
Our featured solution was the only really satisfactory one received. The
other two solutions credited above were technically correct, but they did not,
as the problem seems to require, give explicit formulas for the sides in terms
of the Malfatti radii. Instead, they gave the sides in terms of other functions
in which the Malfatti radii were implicit and very well concealed. One of the
incorrect solutions was a two-line production giving the sides in terms of the
Malfatti radii and the angles of the triangle, the last of which are not known.
This deserves a booby prize of some sort.
The Malfatti problem, which dates from 1803 and has been extensively treated
in the literature, runs as follows: given a triangle (or its sides), construct
the (Malfatti) circles (or calculate their radii). Our problem is just the oppo-
site: given the Malfatti radii of a triangle, calculate its sides. It seems
improbable that this inverse Malfatti problem has never appeared in the literature
since 1803. If it has, the editor would appreciate receiving a reference.
Our second incorrect solution (incorrect only because it solved the wrong
 - 85 -
problem) gave one of the known solutions of the Malfatti problem (and references
to several others), and our unhelpful commentator decided to throw the editor a
bone by giving him one reference to the Malfatti problem. Thank you very much,
but that is not what our problem is about.
*
619, C1981: 80] Proposed by Robert A, Stump, Hopewell* Virginia,
If k is a positive integer, find the value of

y I
.LAi(i+k)°

I. Solution by Kesiragu Satyanarayana, Gagan Mahal Colonys Hyderabad, India,

For n > k9 we have
n n n
1 1 1 1 1 1 1
S E = =
n X.UUk)
^=l ' k\M
i=l ~ T^ kS\ji=l ~ t=l
\AU
m k
1 n 1 n+k 1 i ^ i i

t-1 ^=K+l ^=l t=n+l

k.LAi n
^=l
where

A 1 J i <1 J i 1£ = I.

Since i? -*- o as n •+ °°, we have

n

lim^ = \ LI i = i ( i + i + i+ ... +i).

n -V<» n k. Ai k 2 3 fe
^=l
II. Solution by Friend H, Kierstead, Jr., Cuyahoga Falls, Ohio,
TJust as in solution I], we can show that, more generally, for suitable
a9 b9 and positive integer k9
°° k
V i _ i r _A_ CD
.L (a+bi){a+bU+k)} " bk^±a+bi9
from which our problem follows when a = o and 2? = l . The more general form ( 1 ) ,
with appropriate values for a, b9 and fc, will give many of the r e s u l t s tabulated
by Jolley [ l ] , in particular his relations numbered (204), (205), (208), (221),
(224), (231), and (233).
 - 86 -
Also solved by PAUL R. BEESACK, Carleton University, Ottawa, Ontario; JAMES
BOWE, Erskine College, Due West, South Carolina; S.C. CHAN, Singapore; CLAYTON
W, DODGE, University of Maine at Orono; MICHAEL W. ECKER, Pennsylvania State
University, Worthington Scranton Campus; BIKASH K. GHOSH, Bombay, India; RICHARD
A. GIBBS, Fort Lewis College, Durango, Colorado; PETER A. LINDSTROM, Genesee
Community College, Batavia, N.Y.; F.G.B. MASKELL, Algonquin College, Ottawa,
Ontario; BOB PRIELIPP, University of Wisconsin-Oshkosh; LAWRENCE SOMER, Washington,
D.C.; DAVID R. STONE, Georgia Southern College, Statesboro, Georgia; ALAN WAYNE,
Holiday, Florida; KENNETH M. WILKE, Topeka, Kansas; DAVID ZAGORSKI, student,
Stuyvesant H.S., New York, N.Y.; and the proposer. Two incorrect solutions were
received.

Editor's comment.
Two solutions were labeled incorrect for the good and sufficient reason that
they arrived at the wrong answer. A couple of other solvers figuratively skated
over a patch of nonexistent ice when they wrote

OO CO 00

^-l ^=l ^=l

but they managed to land on their feet and arrive at the correct answer. For this
singular feat of equilibrium they were credited with a correct solution.

REFERENCE
l. L.B.W. Jolley, Summation of Series , Second Revised Edition, Dover,
New York, 1961.
ft ft ft

620, H1981: 80] Proposed by Fred A. Miller3 Elkins3 West Virginia.

Using the digits o and l, express each of the following in the nega-
binary system of notation (base - 2 ) :

i 1 1 1 1 i i i i 1
2'3'4S5'6' 2' 3' 4' 5 * ~6'
Solution de Robert Tranquille, College de Maisonneuve3 Montreal, Quebec.
Dans ce qui suit, seules les representations qui contiennent un "point
decimal" sont des representations nggabinimales et, comme d'habitude, un groupe
de chiffres surlignes indique une representation negabinimale infinie peYiodique.
Par exemple,

11.101 = (-2)1+(-2)°+{(-2)""1 + (-2)"3}+{(-2)"\(-2)"6}+... = - ~ .

On trouve immediatement
1 n -1 1 - 1
i = (_2)u + (-2) L = 1.1, -± = (-2) "• = 0.1,

I = (-2)~2 = 0.01, -| = (-2)"1 + (-2)"2 = 0.11.

 - 87 -
Tous les autres nombres proposes ont des representations negabinimales infinies
peYiodiques. Ces representations correspondent done 5 des progressions g£om§-
triques infinies ayant comme raison v un des nombres (-2)"1, (-2)~2, (-2)~3, ... .
Si n est un de ces nombres et si g est le premier terme de la progression ggome-
trique correspondante, on doit done avoir

Trouvons d'abord les nombres n pour lesquels v = (-2) . On a alors n = 2g/3.

On obtient n = -1/3 pour g = (-2)"1 et n = 1/6 pour g = (-2)~2; done

4 = 0.1 et \ = 0.01.
o b
-2 -2
Pour r = (-2) , il vient n = 4#/3; alors n = 1/3 correspond a # = (-2) et
n =-1/6 correspond a # = (-2)" , d'oO
I = 0.01 et -| = 0.001.
o b
Ces deux dernieYes representations ne sont pas uniques. En effet, puisque

I = (-2)°+(-2)(l/3) et -| = (-2)_1+(l/3)s

on obtient des representations precedentes

i = 1.10 et -i = 0.110.
3 o

Enfin, pour r = (-2)"49 il vient n = 16^/15. On obtient alors

n = 1/5 pour # = 3/16 = (-2)~2+(-2)"3+(-2)"4 = 0.0111

et
n = -1/5 pour g = -3/16 = (-2)_1+(-2)"2+(-2)~4 = 0.1101;
d'oD
i = 0.0111 et -| = 0.1101.
5 b
Also solved by CLAYTON W. DODGE, University of Maine at Orono; BIKASH K.
GHOSH, Bombay, India? RICHARD A. GIBBS, Fort Lewis College, Durango, Colorado;
FRIEND H. KIERSTEAD, JR., Cuyahoga Falls, Ohio; PETER A. LINDSTROM, Genesee Commu-
nity College, Batavia, N.Y.; LEROY F. MEYERS, The Ohio State University; KENNETH
M. WILKE, Topeka, Kansas; and the proposer.

Editor's comment.
According to Gilbert and Green fi], for an integer s > 1, the numbers with
two different representations in base -s are those of the form

(s)k(a + — ) ,
 - 88 -
where a and fr are i n t e g e r s . Readers who wish to pursue the subject f u r t h e r are
i n v i t e d to look also i n [ 2 ] and [ 3 ] .

REFERENCES
1. William J. Gilbert and R. James Green, "Negative Based Number Systems",
Mathematics Magazine, 52 (September 1979) 240-244.
2. Wacjaw Sierpifiski, Elementary Theory of Numbers, Warszawa, 1964, Chapter
VII: Representation of Numbers by Decimals in a Given Scale.
3. David Hale and Peter Wells, "Base Negative Two", Mathematics Teaching,
No. 60 (September 1972), pp. 32-33.
* * *

6211 [1981: 80] Proposed by Herman Nyon3 Paramaribo3 Surinam.

For the adjoining alphametic, there is unfortunately THREE
THREE
no solution in which SQUARE is a square, but there is one in
THREE
which the digital sum of SQUARE is, very appropriately, the EIGHT
EIGHT
square
SQUARE
3 + 3 + 3 + 8 + 8 = 25.

Find t h i s s o l u t i o n .

Solutions were received from J.A.H. HUNTER, Toronto, Ontario; ALLAN WM.
JOHNSON JR., Washington, D.C.; CHARLES W. TRIGG, San Diego, California? DAVID
ZAGORSKI, student, Stuyvesant H.S., New York, N.Y.; and the proposer.

Editor's comment.
All solvers are agreed that

45066
45066
45066
68354
68354
271906

i s the only s o l u t i o n w i t h 25 as the d i g i t a l sum o f SQUARE, and Hunter noted t h a t

i n t h i s s o l u t i o n , i n t e r e s t i n g l y enough, the d i g i t a l r o o t o f THREE i s 3 and t h a t
o f EIGHT i s 8.
Apparently no solver was able to get a good mathematical "handle" on the
problem and thereby avoid extensive use o f brute f o r c e . Two sent i n only the
answer, and the other three c a r e f u l l y swept most o f t h e i r extensive brute force
c a l c u l a t i o n s out of s i g h t . I f i n f a c t no such "handle" e x i s t s , some may think
t h a t t h i s i s the mark o f a good alphametic, b u t , as f a r as we are concerned, i t
makes t h i s a problem f o r computers or a low-grade amusement f o r c h i l d r e n and
other people who have nothing b e t t e r to do w i t h t h e i r t i m e , about on a par w i t h
r e c o n s t r u c t i n g a 5oo-piece jigsaw puzzle. I t i s not mathematics.
 - 89 -
622. C1981: 115] Proposed by tr.Tn Groenmans Arnhem5 The Netherlands,
The Diophantine equation
8x3 - 21x2y + 35#z/2 - 83y3 = z3

has the obvious s o l u t i o n s (x,y9z) = (k9o92k)9 where k i s any i n t e g e r . Find a t

l e a s t two solutions f o r which y * o.

I. Solution by Friend H. Kiersteads Jr.3 Cuyahoga Falls, Ohio.

3
I f we s u b t r a c t (2x-3y) from both sides o f the given equation, we get
2
15x y - 19xy2 - 56z/3 = z3 - (2x-3y)3 %

or
y(3x-8y)(5x+7y) = z3 - (2x-3y)3.

For z = 2x~3y one o f the f a c t o r s on the l e f t must vanish. The t r i v i a l solution

set given i n the proposal r e s u l t s i f we set y = o. I f 3x-Sy = o, then

2x
*£ = &. = ~3y
8 ~ 3 7

and we have the solution set

(x9 y, z) = (8k, 3k9 7k); (1)

and 5x+7y = o gives

•E - JL 2a?-3i/
7 " -5 29
and the solution set
(x9 y, z) = (7k, -bk9 29k). (2)
2
A computer search reveals no other solution (x9y9z) with M »l#l»l l < loo.
II. Solution by Kenneth M. Wilke3 Topeka3 Kansas.
[Solution sets (1) and (2) having been found], we will obtain other solutions
(x9y9z) with y # o from rational solutions (u9v) of the equation

8u3 - 21u2 + 35w - 83 = V3 9 (3)

where u - x/y and v - z/y. We describe a technique of Fermat [l, p. 566] which
we will use to find rational solutions of (3). Consider the equation

Au3 + Bu2 + Cu + D = v3. (4)

If A = a3, set v = au + B/3a2; if D - d3, set v = d + Cu/3d2; if both 4 = a3 and

3
D = d , set t; = aw+d. Each of these substitutions, when appropriate, will reduce
(4) to a linear equation in u.
Only the Fermat substitution v = 2u-(7/k) is appropriate for (3), and it
 - 90 -
yields u = 4969/1064, from which we get v - 8076/1064 and the solution set
(x, y9 z) = (4969k, 1064k, 8076k). (5)
To find additional rational solutions of (3), we can use the "tangent method"
of Lagrange, which is described in [i, p. 595] and in [2], [3], This involves
starting from a known solution (UQ, VQ) of (3), substituting u = t+uQ in (3), and
applying the Fermat technique to the resulting equation.
With w 0 = 8/3 from (l). for example, we substitute u = £+(8/3) in (3) and
obtain
8t3 + H3t2 + mt + m = v3.

Now three Fermat substitutions are possible for this equation:

7 o^^S 7 281t
V = 2* + -, v = 2* + n , V = - + — .

I t can be verified that the f i r s t and second substitutions lead to solution sets
(2) and ( 5 ) , respectively. But the third substitution yields
7341572 _ 3848172 = -8525447
U V
7082283 s ~ 2360761 * 2360761 *

and we have the new solution set

(x9 y, z) = (3848172/:, 2360761k, -8525447k). (6)
_7
Similarly, starting with u 0 = /5 from ( 2 ) , the tangent method yields solution
sets ( l ) , (5), and the new solution set

(a?, y9 z) = (8555030953k, 3393217903k, -1111978813k). (7)

Readers who love to crunch numbers will enjoy using the tangent method of Lagrange
to find still more solution sets from (5), (6), and (7).
Also solved by W.J„ BLUNDON, Memorial University of Newfoundland; BENGT
MANSSON, Lund, Sweden; BOB PRIELIPP, University of Wisconsin-Oshkosh; and the
proposer.

REFERENCES
1. Leonard Eugene Dickson, History of the Theory of Numbers3 Chelsea, New
York, 1952, Vol. II.
2. L.J. Mordell, diophantine Equations, Academic Press, London and New York,
1969, pp. 66-70.
3. H.D. Grossman, "Investigation of Lagrange's Tangent Method on Diophantine
Binary Cubics", American Mathematical Monthly, 47 (1940) 305-306.
s'c ii "k
 - 91 -
623 • C1981: 116] Proposed by Jack Garfunkel3 Flushing3 N.Y.
If PQR is the equilateral triangle of smallest area inscribed in a
given triangle ABC 9 with P on BC S Q on CA, and R on AB, prove or disprove that
AP, BQ„ and CR are concurrent.,

Solution by J.T. Groenman3 Amhem3 The Netherlands.

We disprove the theorem and show that AP, B0 S and CR are concurrent if and
only if triangle ABC is isosceles.
[t is well known that the three Apollonius circles of a triangle have two
common points, J and J', called the isodynomio points of the triangle; that the
distances from either isodynamic point to the vertices are inversely proportional
to the sides of the triangle; and that the pedal triangle of either isodynamic
point is equilateral. (See Johnson [l].)
The pedal triangle of one of the isodynamic points, say J, is inscribed in
triangle ABC. Let this (equilateral) triangle be PQR (with P on BC, etc.). We
show that PQR is the inscribed equilateral triangle of smallest area. Let P'Q'R'
be any inscribed equilateral triangle (with P' on BC, etc.), and let e be the angle
of smallest absolute value through which PQR must be rotated around J to make its
sides parallel to those of P'Q'R'. This rotation makes PQR nomothetic to P'Q'R',
with J the nomothetic center, and we have the nomothetic ratio

JP
QR
QTT
jpr = cosi
hence, with brackets denoting area, we have

[PQR] = [P'Q'R']cos 2 e < [P'Q'R'],

with equality i f and only i f = o, and so PQR is the inscribed equilateral triangle
of smallest area.
As noted above, we have
JA : JB : JC = bo i oa : ab,

where a - BC, etc., and so

JA = kbc9 JB = kca9 JC = kab
for some ^ o. Referring to the figure, we have
a = BP + PC
= JB cos 3 + JCcos y
= koa COS 3 + kab COS ys

and so
 - 92 -

1 = kc COS 3 + kb COS y. (1)

From (kb cosy)2 = d - fo?cos3)29 we obtain

k2b2 - k2b2sin2y = 1 - 2ko cos 3 + fc2e2 - fc2<?2sin23. (2)

Now, observing that

JP = JBsin 3 = JCsin y
= kca sin 3 = kab sin y,

so that c s i n 3 = frsiny, we see that (2) reduces to

k2b2 = 1 - 2fo?cos 3 + k2c2;

hence
kc cos 3 = lil-k2(b2-c2)}9
and
kb cosy = lil+k2(b2-c2)}
then follows from ( l ) . Finally, we have

BP = JBCOS3 = kca cos & = iail-k2(b2-c2)},

PC = JCcosy = kab cosy = ia{l+k2(b2-c2)},
and similar expressions can be found for CQ, QA and AR,RB.
Now by Ceva's Theorem and i t s converse, AP, BQ, and CR are concurrent i f
and only i f
BP-CQ-AR = PC-QA-RB.

With the expressions found above, the necessary and sufficient condition is
equivalent to
{l-k2(b2-c2)}{l-k2(c2-a2)}{l-k2(a2-b2)}
= {l+k2(b2-c2)}{l+k2(c2-a2)}{l+k2(a2-b2)} s

which is itself equivalent to

(b2-c2)(c2-a2)(a2-b2) = 0.
We conclude that AP, BQ, and CR are concurrent if and only if b = c or e -a
or a = b9 that is, if and only if triangle ABC is isosceles.
Also disproved by W.J. BLUNDON, Memorial University of Newfoundland? O.
BOTTEMA, Delft, The Netherlands; BIKASH K. GHOSH, Bombay, India; DMITRY P. MAVLO,
Moscow, U.S.S.R; VADIM V. MUZYCHENKO, Moscow, U.S.S.R.; KESIRAJU SATYANARAYANA,
Gagan Mahal Colony, Hyderabad, India; and GEORGE TSINTSIFAS, Thessaloniki, Greece.

REFERENCE
1. Roqer A. Johnson, Advanced Euclidean Geometry (Modern Geometry), Dover,
New York, I960, pp. 294-296.