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.
* * * * * f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t * f t f t
ft ft ft & ft ft ft ft ft ft ft ft ft ft ft ft ft ft ft * ft ft A ft ft ft ft ft ft ft ft ft ft ft

ISSN 0705 - 0348

CRUX MATHEMATICORUM

Vol, 7, No. 10
December 1981

Sponsored by
Carleton-Ottawa Mathematics Association Mathematique d'Ottawa-Carleton
Publie par le College Algonquin
The asisstance of the publisher and the support of the Canadian Mathematical Olympiad
Committee, the Carleton University Department of Mathematics and Statistics, the University
of Ottawa Department of Mathematics, and the endorsement of the Ottawa Valley Education
Liaison Council are gratefully acknowledged.
ft*!***************************!***** ft ft W 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 ft ft ft

CRUX MATHEMATICORUM is a problem-solving journal at the senior secondary and university

undergraduate levels for those who practise or teach mathematics. Its purpose is primarily
educational, but it serves also those who read it for professional, cultural? or recreational
reasons.
It is published monthly (except July and August). The yearly subscription rate for ten
issues is $20 in Canada,, US$19 elsewhere. Back issues: $2 each. Bound volumes with index:
Vols. 1G2 (combined) and each of Vols. 3-7, $16 in Canada and US$15 elsewhere. Cheques and
money orders, payable to CRUX MATHEMATICORUM, should be sent to the managing editor.
All communications about the content (articles, problems, solutions, etc.) should be
sent to the editor. All changes of address and inquiries about subscriptions and back issues
should be sent to the managing editor.
Editor: Lio SauvS, Algonquin College, 281 Echo Drive, Ottawa, Ontario, Canada KlS 1H3.
Managing Editor: F.G.B. Maskell, Algonquin College, 200 Lees Ave., Ottawa, Ontario,
Canada KlS 0C5.
Typist-compositor: Nghi Chung.

CONTENTS
A Simple Proof of the Butterfly Problem . Kesiraju Satyanarayana 292
Biographical Note . . . . . . . . . . . . . . . . . . . . . . . . . . 294
The Puzzle C o m e r Hans Havermann 29*+
. ....,. Alan Wayne 300
More Nine-Digit Patterned Palindromic Primes . Charles W. Trigg 295
Mama-Thematics . Alan Wayne 297
The Olympiad Corner: 30 M.S. Klamkin 298
Problems - Probl§mes 301
Solutions . . B 303
The Dot Polka . , . . .Q . . . . . . . . . 320

- 291 -
 - 292 -
A SIMPLE PROOF OF THE BUTTERFLY PROBLEM
KESIRAJU SATYANARAYANA

The Butterfly Problem has been attracting attention at least since 1815, and
the number of known proofs is quite large (see the extensive list of references in
Cl], also [2] and [3]). Assuming that mathematical lepidopterists will always be
happy to meet a new specimen, we present here a simple elementary analytical proof
which we believe to be new.
In its simplest form, the problem may be stated as follows (see Figure l ) :
THE BUTTERFLY PROBLEM, Through the midpoint M of a chord AB of a circle, two
other chords, CD and EF, are drawn. ED and CF intersect AB in P and Qs respectively.
Prove that PM = MO.

Figure 1

Proof. Let rx be the given circle. We introduce a rectangular coordinate

system with origin M, a?-axis AB, and z/-axis MO, where 0(Q,<f) is the centre of the
circle. If the circle has radius r, its equation is

Ei = x2 + (y-d)2 - r2 = Q.

As the lines CD and EF pass through the origin, they form a degenerate conic r 2 whose
equation is of the form

ll = ax1 + 2hxy + by2 = 0.

 - 293 -
Now, f o r any fc3 l3

represents a conic r through the points common to Tx and r 2 , that is, through C,D,E,F;
and every conic through C,D,E,F 1s representable in this form.
Suppose the conic z = o intersects AB in V and W. The equation of AB is y = o,
and
l](x,0) = x2 + d2 - r23 Z2(^0) = ax2;

hence the abscissas of V and W are the roots of i{x3o) = o, that is, of
k(x2 + d2 ~ v2) + lax2 = 0.
Since this equation has no first-degree term, the sum of its roots is zero, so
W + W -- o, and
VM = MW. (1)

Now (1) holds for all conies through C,D,E,F, and the pair of lines ED,CF
is such a conic, so PH = MQ follows from (1). D
The pair of lines CE,DF is also a conic through C,D,E,F. If these lines inter-
sect AB in P ! and Q 1 , as shown in Figure 1, then P'M = MO' also follows from (i).
Suppose that rx is, instead of a circle, an arbitrary proper conic of equation

ll = Ax2 + 2Exy + By2 + 2Gx + 2Fy + C = 0.

With the rest of the notation as before, we have

Uxs0) - k(Ax2+2GxiC) + lax2.

If the coordinates of A and B are (~a,o) and (a,o),

respectively, then rx(-a,o) = ^(a.o) = o implies
that G - o; so the equation i(xso) = o has no
first-degree term, and the rest of the proof is
as before. We have thus proved
THE GENERALIZED BUTTERFLY PROBLEM.
Through the midpoint M of a chord AB of a
proper conic Tjj two other chords, CD and EF,
are drawn. A conic Y through C,D,E9F intersects AB in V and W. Prove that VM = HW.
This problem is illustrated in Figure 2. A proof by projective qeometry can be
found in Eves O ] .
 - 291 -
REFERENCES

1. Le"o Sauve\ "The Celebrated Butterfly Problem", this journal, 2 (1976) 2-5.
2. Leon Bankoff, Letter to the editor, this journal, 2 (1976) 90-91.
3. Dan Sokolowsky, "Another Proof of the Butterfly Theorem", this journal,
2 (1976) 189-191.
4. Howard Eves, A Survey of Geometry, Revised Edition, Allyn and Bacon, Boston,
1972, pp. 255-256.

c/o Sri K. Gourisankaram, 1-2-593/15 Gagan Mahal Colony, Domalguda, Hyderabad 500 029,
India.
* * *

BIOGRAPHICAL NOTE

Professor Satyanarayana, who is 84 years young, has been since 1978 one of the
most prolific contributors of solutions to Crux problems, mostly, but not exclusively,
in the field of geometry. He has contributed problems and solutions to other journals
as well; one of his problems, in fact, was recently published in the American Mathe-
matical Monthly (Problem E 2873 in the March 1981 issue). He has also had several
articles published in The Mathematics Student.

He was born on October 6, 1897 in Malakapalli, West Godavari District, India.

He holds First Class Honours B.A. and M.A. Degrees in Mathematics from Madras Univer-
sity. He lectured for 34 years to B.A., B.Sc, and B.Ed, classes of the Academic
and Teachers' Training Colleges at Rajahmundry, and retired in 1953 as Principal of
the Teachers' Training College. Since his retirement, he has published three re-
search-oriented books on geometry:
1. Angles and In- and Ex-Elements of Triangles and Tetrahedra (1962) . Rs 10/-.
2. Poristic Theory, Pedal Lines, Kantor Lines, Kantor Points and Allied Topics
(1976). Rs 30/-.
3. Dihedral Angles and In- and Ex-Elements of n-Space Simplexes (1979) . Rs 10/-.

These books are available from Visalaandhra Publishing House, Vijayawada 520 004,
Andhra Pradesh, India.

* * *

THE PUZZLE CORNER

Puzzle No. 5: Alphametic

FOUR is a square and FIVE is a prime,

SIX has been perfect for quite a long time.
Can you confirm when you've finished this poem
That SEVEN + NINE is a palindroem?

HANS HAVERMANN, Weston, Ontario

* * *
 - 295 -
TORE NINE-DIGIT PATTERNED PALINDROMIC PRIMES
CHARLES W. TRIGG

Delving once more (see f"l]) into the set of 5172 nine-digit palindromic primes,
a list of which was prepared by Jacques Sauve* on a PDP-11 A 5 at the University of
Waterloo, we find a number of subsets of these primes with the same characteristics
or patterns. Some selected subsets are given below. But first we do a bit of
juggling with the digits of the cardinal number of the set, 5172.
According to some points of view, all the digits of 5172 are primes, and
-(5+D+7+2 supplies the missing prime digit, 3. Furthermore,

5-1-7 + 2 = 0 = 5-1 '7 + 2

and
-5+1+7+2=5,
5 + 1 - 7 + 2 = 1,
5-1 +/7 + 2 = 7,
5 -A"- 7 + 2 = 2.

Now to our self-appointed task. Three of the primes have eight like digits,
namely:

111181111, 111191111, and 777767777.

If a nine-digit palindrome contains seven like digits, the other two digits must
be like., The twenty-two such palindromic primes having nine digits fall into four
different patterns. They are listed separately below according to the positions
of the two like digits.

188888881 121111121 110111011 111010111

199999991 131111131 1J2111211 111515111
322222223 181111181 11311J311 111616111
355555553 323333323 11511151:1 333434333
722222227 331333133 333535333
335333533
338333833
991999199

If a nine-digit palindrome contains six like digits and the other three digits
are like, then the palindrome is divisible by 3. So there are no palindromic primes
of this type.
 -296 -
The thirty-three palindromic primes with four like digits and five like digits
include the seven smoothly undulating primes [2]

323232323, 383838383, 727272727, 919191919, 929292929, 979797979, 989898989.

The other twenty-six primes fall into the five patterns separately exhibited below:

331111133 100111001 112212211 181888181 322323223

772222277 133111331 118818811 313111313 355353553
779999977 377333773 338838833 323222323 722727227
992222299 766777667 994494499 383888383 911919119
995555599 944999449 998898899 959555959
977999779
988999889

Thus all nine-digit palindromic primes composed of just two distinct digits
are accounted for.
There are thirty-seven palindromic primes which are permutations of five
consecutive digits. Each digit except the central one appears twice in the prime.
These primes are assembled below into columns according to their digit sets.

102343201 125343521 345262543 345676543 745686547 759686957

312040213 134525431 346525643 354767453 746858647 957686759
320141023 142353241 356474653 756848657 967585769
321404123 153424351 357646753 786545687 976858679
324101423 312545213 736545637 978656879
315424513 745363547 986757689
351242153 745636547
352141253 753646357
352414253 756343657
764535467
765343567

The first prime in the fourth column, 345676543, is the only peak [2] nine-digit
palindromic prime composed of consecutive digits [33.
The thirteen palindromic primes composed of five consecutive odd digits are:

135979531 319575913 719535917 913575319

153979351 371595173 759313957 971535179
157393751 395717593 791535197 973515379
157939751
 - 297 -
As a central digit, 3 and 7 each appear four times, l appears three times,
9 appears twice, and 5 not at all.
There are forty-two nine-digit palindromic primes composed of digits that
are powers of 2. The three composed of l*s and 2's, and the five composed of l"s
and 8's, have been mentioned previously. The twenty-nine primes composed of three
distinct powers, and the five containing the four digit-powers are:

121414121 121282121 111484111 144818441e 122484221

122444221 121818121 111848111 1448M8441/ 12844M821
141242141 128121821 114484411a J48414841e 142888241
144212441 128181821 114818411b 148444841/ 184212481
128282821 1J4848411<? 148818841a 184222481
118414811b 148888841
118848811a' 184414481s
141484141a 184818481a
lM1848141c? 1881M1881C?
j 1^484441 188414881a

Primes followed by the same lower-case letter are permutations of the same digit set.
There are twenty-eight nine-digit palindromic primes in which every digit is
a power of 3. The six primes composed of 1's and 3's, and the five primes composed
of I's and 9's have been previously mentioned. The other seventeen are:

113939311a 191939191 319131913a 319999913b 391999193b

139131931a 193191391c 319191913(? 331999133c? 399191993b
139999931b 199393991b 319393913d 391333193 913939319
191313191 313999313(i

Primes followed by the same lower-case letter are permutations of the same digit set.

REFERENCES

1. Charles W. Trigg, "Nine-Digit Patterned Palindromic Primes", this journal,

7 (June-July 1981) 168-170.
2. , "Special Palindromic Primes", Journal of Becreatiorial Mathematics,
4 (July 1971) 169-170.
3. l£o Sauve\ Editor's comment, this journal, 6 (November 1980) 289-290.

2404 Loring Street, San Diego, California 92109.

MAMA-THEMATICS
Frau Gauss, speaking of her son Carl Friedrich: "He likes to get down to
the roots of things."
ALAN WAYNE, Holiday, Florida
 - 298 -
THE OLYMPIAD CORNER; 30
M.S. KLAMKIN

The following problems, for which readers are invited to send me elegant
solutions, are from Kczepiskolai Matematlkai Lapok, 62 (May 1981) 208-209. They
are labeled "Olympiad Preparatory Problems'8 and were collected by Jozsef Szikszai^,
Miskolc. I am grateful to Frank Papp for supplying the English versions of these
problems.

1, Which of the following two numbers is larger:

/7+/7 - VT, VI - VI-VI 1

2, Justify the following assertion: If the positive numbers x^sx^.. ,sx

have product j, then

n n
) x, < ) x.
^=l ^=l

3, Determine the pairs (m9n) of natural numbers for which the equation

1 - sw2nx „.
—r-n— = smnx
1 - sir\zmx
has real solutions.
/}, Show that
w-1
I COt (k-n/n) • COS2(kir/n) - 0.
fe=l
5, If n is a qiven natural number, solve the equation
(2rc-l) + (1-x) =x .

6, If « is a given natural number, determine the largest and least values

of the expression
v n
IT (?-COS 2 cu) + II COS2cu,.

7, Show that, for nonnegative numbers a^b^c^d,

(aK3)(b+d)(2a+c+d)(2b+c+d) > Hed(2a+c)(2bi-d).

8, Let G denote the geometric mean of the n positive numbers a. and, for
natural numbers k, let p, denote the kth power mean, i.e.,
n v 1/k
Pk = (I fyn) •
 - 299 -
Show that

9, Show that, for an arbitrary pair (n3k) of natural numbers, there is a

unique natural number f(n,k) which satisfies the relation

(Sn+T + Jn)k = Vf(n,k)+1 + Sf(n,k).

1 0 , For which real numbers xsy is the following inequality satisfied?

A large number of problems published earlier in this column are still awaiting
a published solution. Space permitting, we would like to publish elegant solutions
to as many of them as possible, and readers are invited to collaborate in this project
by submitting their solutions to me. The following solution to one of the backlog
problems is not particularly elegant. Readers are urged to find a better one.

J-33. C1981:1M^] A straight line CD and two points A and B not on the line
are given. Locate the point H on this line such that /AMC = 2/BMD.
Solution,
We begin by showing that there is on CD a unique point M such that /AMC = fc/BMD
for any k > o. We may assume that A and B are on the same side of CD, for otherwise
we could replace one point by its mirror-image across CD.
When k = l, the construction of the-
point M is well known. It occupies the
position NT shown in Figure 1. As M moves
to the left on CD, the ratio p = /AMC//BMD
increases monotonically and becomes un-
bounded; and as M moves to the right on
CD, p decreases monotonically to zero.
Hence, by continuity, for ewery k > o there
is a unique point M on CD for which p = k.
This point M is to the left or to the right
of M1 according as k > 1 or k < l.
We show that when k - 2, as in our
problem, the point M can be constructed with
straightedge and compass. Let A' and B' be the feet of the perpendiculars from A
and B, respectively, upon CD, and set AA! = a and BB9 = bs as shown in Figure 2.
 - 300 -

We orient line CD so that A'B' = a is

positive or negative according as B is
to the right or to the left of A. In any
case we have a cot 29 + b cot 9 = c.
Using a familiar trigonometric
identity for cot 29, this equation
is easily shown to be equivalent to Figure 2

(a+2fc)COt29 - 2c? COt 9 - a (1)

from which we get

MB' = 2? cot 9 b{cWcz+a(a+2b)}9 (2)

which shows that the point M is constructive with straightedge and compass. Note
that we have used the positive root of (1) for cot 9 because /AMC = 29 < 180°
implies that /BMD = 9 < 90°. •
There may be an elegant construction that avoids the straightforward but tedious
Euclidean construction of (2). Readers are invited to find one and send it to me.
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.
* ft ft
THE PUZZLE CORNER

Puzzle No. 6: Rebus (*5 8)

EY, EY, EY, ...
The ALL we mathematicians may define
As certain fractions ordered in a line.
Puzzle No. 7: Rebus (*7)
H/0
Not everything, you see,
In ALL's philosophy.

ALAN WAYNE, Holiday, Florida

i'c

Readers are urged to verify on the front page of this issue that the addresses
of the editor (Leo Sauve) and managing editor (F.G.B. Maskell) are different, being
on different campuses of Algonquin College. The appropriate address should be used
in each case to ensure safe and prompt arrival of readers' communications.
 - 301 -

PROBLEMS — PROBLEI^ES

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
May 1, 1982, although solutions received after that date will also be considered
until the time when a solution is published.

6911 Proposed by J.A. MeCallum, Medicine Eats Alberta.

Here is an alphametic about the man who coined the word "alphametic",
the well-known author of the syndicated column Fun With Figures3 J.A.H, Hunter:

H E*S
THAT
FUN
MATHS.
HUNTER
The apostrophe has no mathematical significance and the answer, like the man himself,
is unique.
692 * Proposed by Ban Sokolowskys California State University at Los Angeles.
S is a set of n distinct objects. For a fixed k > 1, 2k subsets of
S are denoted by A.3B.3 i - l,...,fc. Find the larqest possible value of n for
which the following conditions (a)-(d) can hold simultaneously for i = l,...,k.
(a) A . v B . = s .
' ^ % n
(b) A i n B i = <f>.
(c) For each pair of distinct elements of S 9 there exists an i such that
the two elements are either both in A^ or both in B^.
(d) For each pair of distinct elements of S > there exists an i such that
one of the two elements is in A. and the other is in B..
t ^
593s? Proposed by Ferrell Wheelers student, Texas ASM University.
On a 4x4 tick-tack-toe board, a winning path consists of four squares
in a row, column, or diagonal. In how many ways can three X's be placed on the
board, not all on the same winning path, so that if a game is played on this partly-
filled board, X going first, then X can absolutely force a win?
 - 302 -
694* Proposed by Jack Garfunkel3 Flushing, N.Y.
Three congruent circles with radical center R lie inside a given
trianqle with incenter I and circumcenter 0. Each circle touches a pair of sides of
the trianqle. Prove that 0, R, and I are collinear.
(This generalizes Problem 5 of the 1981 International Mathematical Olympiad
[1981: 2231, where it was specified that the three circles had a common point.)
6951 Proposed by J.T. Groenrnan, Arnhem, The Netherlands.
For i = d,2,3, A. are the vertices of a trianqle with sides a. and
excircles with centers I, touching a. in B.. For j , k * ~i, M. are the midpoints
and m. the riqht bisectors of B .B,. Prove that the m. are concurrent.
^ ok ^
6 9 6 1 Proposed by George Tsintsifas, Thessaloniki, Greece,
Let ABC be a t r i a n g l e ; a,b,c i t s s i d e s ; and s,r,R its semiperimeter,
inradius and circumradius. Prove t h a t , w i t h sums c y c l i c over A,B,C S
(a) 3 + ^ c o s J ( B - C ) > E c o s A ;
(b) l a cos J(B-C) > s(l + 2r/B).

6971 Proposed by G.C. Giri3 Midnapore College, West Bengal, India.

Let
a - tan 9 + tan $ ,
b - sec 9 + sec <f> ,
c - esc 9 + esc (f> .
If the angles 0 and 4> are such that the requisite functions are defined and
bo * o, show that 2a/bc < 1.
698? Propose par Eippolyte Charles, Waterloo, Quebec.
Les sommes partielles de la se>ie harmonique (laquelle, on le sait
bien, est diverqente) sont definies par

s = i + | + ... + i .
•» n 2 n
La se>ie I - e s t - e l l e convergente ou diverqente?
n~\ n
6991 Proposed by Charles W, Trigg, San Diego, California.
A quadrilateral is inscribed in a circle. One side is a diameter of
the circle and the other sides have lengths of 3, u, and 5. What is the length of
the diameter of the circle?

7001 Proposed by Jordi Dou, Barcelona, Spain.

Construct the centre of the ellipse of minimum eccentricity
circumscribed to a given convex quadrilateral.
 - 303 -
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.

585, £1980: 284] Proposed by Jack Garfunkels Flushing, New York.

Consider the following three inequalities for the angles A,B,C of a
tr1angle:
B-C C-A A-B A B r
c o s ^ c o s ^ c o s Y * 8sin|sin|sin|, ft)

csc§cos5l£ + C s c | c o s ^ + c s c ^ c o s ^ > 6, (2)

A B C
CSC ^ + CSC - + CSC - > 6. (3)

Inequality (3) is well-known (American Mathematical Monthly 66 (1959) 916) and it

is trivially implied by ( 2 ) . Prove (1) and show that (1) implies ( 2 ) .
Solution by M,S. Klamkin, University of Alberta,
We will show that inequalities (1) and (2) are just disguised forms of the
two well-known elementary inequalities
(b+c)(c+a)(a+b) > Babe (l')
and
bc(b+c) + ca(c+a) + ab(a+b) > 6abc, (2r)
which are v a l i d f o r a r b i t r a r y nonnegative real values of a9b9c These are e a s i l y
established by the arithmetic-geometric mean i n e q u a l i t y (or see Bottema et a l . [ l ] ) .
I f asbjc arQ the sides of a t r i a n g l e , then d') and ( 2 ' ) are equivalent to
b+c c+a a+b > C1M)
a b e
and ' "
b+c c+a , a+b . _ , 0.«%
a b c

Now, by the law of sines,

b+c = s i n B + sinC _ 2 cos (A/2) cos (B~C)/2 . cos ( B-C)/2
a sinA " 2 sin (A/2) cos (A/2) " sin ( A/2)* (*)

If we substitute this and two similar expressions in (l") and (2"), we obtain
( l) and ( 2).
To show that d ' ) implies (2 ! ), and hence that (1) implies (2), we apply the
arithmetic-geometric inequality: with sum and product cyclic over asb,c3 we have
Ubc(b+c) > 3{Ubc(b+c)}1/3 > 6abc. 0

For the geometrical significance of (1), we refer to an exercise in Todhunter

[2] and an inequality of Gridasov [3], In the exercise, one is to show that if
 - 304 -
the bisectors of angles A 9 B,C of a triangle meet the opposite sides in D,E,F,
respectively, then
£ 5 | ^ = 2ll{sin (A/2)}/{cos (B-C)/ 2}, (5)
where [DEF] denotes the area of triangle DEF, etc. Gridasov's inequality is that
4CDEF] <; [ABC], which follows from (l) and (5). This inequality has been extended
by this author to arbitrary concurrent cevians for simplexes O ] . (See also this
journal [1978: 255-2563.)
To establish (5), we use an elementary result given in f>], viz.,
[DEF] _ 2xyz ( .
K }
[MCl " (y+z)(z+x)(x+y)>
where x 3 y 3 z are the barycentric coordinated of the point P of concurrency of the
three cevians. Here we have
P - xK + yt + zCt x3y,z ^ 0, x+y+z - 1,
where the vectors to P and the vertices are taken from an origin outside the plane
of the triangle. If the cevians are angle bisectors, then x - a s y - h s z - e
(assuming that the side lengths have been normalized so that a+b+c = 1 ) , and (5)
follows from (6) and (M-).
Also solved by W.J. BLUNDON, Memorial University of Newfoundland (two solu-
tions) j J.T. GROENMAN, Arnhera, The Netherlands? V.N. MURTY, Pennsylvania State
University, Capitol Campus (two solutions)? NGO TANf student, J.F. Kennedy H.S.,
Bronx, N.Y.? KESIRAJU SATYANARAYANA, Gagan Mahal Colony, Hyderabad, India;
GEORGE TSINTSIFAS, Thessaloniki, Greece? and the proposer.

Editor f8 comment.
The proposer gave without proof the following geometric equivalent of (l):
Let I be the incenter of triangle ABC, and let the bisectdrs of angles A,B,C meet
the circvmcircle again in R,S,T, respectively. Then
AR-BS-CT ;> 8-lR-IS-IT.
REFERENCES
1. 0. Bottema et al., Geometric Inequalities sWo}t£r$-Woor&\\off, Groningen,
1969, pp. 12, 13.
2. I. Todhunter, Plane TrigonometrysMacmi11 an, London, 1880, p. 309.
3. V. Gridasov, Matematika i fizihx, Sofia, 6 (1965) 52-53.
4. M.S. Klamkin, "A Volume Inequality for Simplexes", Publ. Fac. D'Elektrotehn.,
Univ. of Belgrade, No. 357 - No. 380 (1971) 3-5.
is is is
 - 305 -
586, [1980; 284] Proposed by Ngo Tan3 student, J.F. Kennedy H.S.* Bronx, N.Y.
(a) Given a natural number n, show that the equation
9n = 6abn + ab(a+b) (j )

has no solution in natural numbers a and b,

(b) Usinq (a), or otherwise, show that none of the following expressions is
a perfect square for any natural number n:
3
36n -} 36n? + 12rc + 1,
3 2
12« v 36w + 36?? + 9,
3 2
4n + 36n + 108« + 81.
Solution by the -proposer.
(a) Equation (1) i s equivalent to

(3rcf-a) 3 + (3tthfr) 3 = (3n+a+b)3

which, by Fermat's Last Theorem, has no solution in natural numbers (or even in
positive rationals). Hence, for any given n9 equation (1) has no solution in natural
numbers a and b.
(b) Solvinq (1) for a, we qet
„ - -b(b+Gn) t /ft ,9s
a - 2h , u;
where 2 2 3
A = & (fc+6w) + 36bn .
A necessary condition for the natural number b to be part of a solution (a3b) of
(1) is that fr|9?73. For such a natural number b9 A cannot be a perfect square;
otherwise ( a , b ) , with a given by (2), would be a positive rational solution of (1),
which is impossible. In particular, for b = l, 3, 9, the values of A, A/9, and
A/81, respectively, viz.,
3 2
36n + 36n + 12n + 1,
3 ?
12n + 36r? + 36?7 + 9,

*m + 36n f 108n + 81,

are not perfect squares for any natural number n.

also solved by KENNETH M. WILKE, Topeka, Kansas.

587. [1980: 317] Proposed by George Tsintsifas3 Thessaloniki, Greece.

Let o = AQAJ.-.A be an ^-simplex in if1. A straight line cuts the
(rc-l)-dimensional faces
o. = AnA,...A* .A. ....A s i - 0,1, ...n
 - 306 -
in the points B.. If M. is the midpoint of the straight line segment A.B., show
that all the points M. lie in the same (n-l)-dimensional plane.

I. Comment by M.S. Klamkin, University of Alberta.

(^incidentally, I had proposed the same problem in Elemente der Mathematiks
and a simple solution by I. Paasche was published in that journal [31 (1976) 14-153.
This problem extends the known results for n - 2, 3 for which the midpoints are
collinear and coplanar, respectively.

II. Comment by Hessel Pot, Woerden, The Netherlands.

The special case n = 2 brings a question to mind. Starting with a triangle
and a line, the three midpoints are again on a line, so the process can be repeated
with this new line. Is there any sort of convergence or regularity when the process
is repeated indefinitely?
So]utions were received from M.S. KLAMKIN, University of Alberta; LEROY F.
MEYERSr The Ohio State University; HESSEL POT, Woerden, The Netherlands; KESIRAJU
SATYANARAYANA, Gagan Mahal Colony, Hyderabad, India? and the proposer.
j'? & ft

5881 T1980: 3171 Proposed by Jack Garfunkel, Flushing, N.Y.

Given is a trianqle ABC with internal angle bisectors t . t,s t and
medians m , mh> m to sides a> bs a3 respectively. If

ma n t,b = P, m,b n tc = Q, mc n ta = R,
and L, M, N are the midpoints of the sides a, b, ca respectively, prove that
AE §2 £R R
PL'QM'RN " '
Solution by Poland H. Eddy, Memorial University of Newfoundland.
Since BP bisects anqle B in triangle ABL, we have AP/PL = o/(a/2) = 2o/a.
With this and two similar results, we have
i p 53 !±E - 2£ 2a 22? _ un
PL'OM'RN " a 'b *e ' *'
We show that if we replace the medians by the altitudes (when the triangle is
acute-anqled), the Gergonne cevians, or the Nagel cevians, we obtain
n _ AP.BO CR . 0

For the altitudes we have AP/PL = c/o cos B - sec B and two similar results,
from which
II = sec A secB secC > 8, (1)
The Gergonne cevians join the vertices to the points of contact of the
incircle with the opposite sides. They are concurrent in the Gergonne point of
 - 307 -
the triangle. The Magel cevians join the vertices to the points of contact with
the opposite sides of the excircles relative to those sides. They are concurrent
in the Nagel point of the triangle. In both the Gergonne and the Nagel cases, we
find AP/PL = c/(s~c) and two similar results, from which
a h o
TT = > Q
" (s-a)(8-b)(e-c) - b* (2)
Inequalities (1) and (2) can be found in 0. Bottema et al., Geometric Ine-
qualities, Wolters-Noordhoff, Groningen, 1968, pp. 25, 12.
M s o solved by W.J. BLUNDON, Memorial University of Newfoundland; CLAYTON W.
DODGE, University of Maine at Orono? JORDI DOU, Barcelona, Spain? G.C. GIRI,
Midnapore College, West Bengal, India? J.T GROENMAN, Arnhem, The Netherlands;
FRED A. MILLER, Elkins, West Virginia; NGO TAN, student, J.F Kennedy H.S., Bronx,
N.Y.; HESSEL POT, Woerden, The Netherlands; KESIRAJU SATYANARAYANA, Gagan Mahal
Colony, Hyderabad, India; MALCOLM A„ SMITH, Georgia Southern College, Statesboro,
Georgia; DAN SOKOLOWSKY, California State University at Los Angeles; ROBERT L
STUMP, Hopewell, Virginia; and the proposer.

Editor's comment.
See Crux 685 [1981*. 275] for a related problem.
i*e s'c sV

589 B T1980: 317] Proposed by Ngo Tan, student, J.F. Kennedy H.S., Bronx, N.Y.
In a triangle ABC with semiperimeter s , sides of lengths a, h, c, and
medians of lengths m , m , , m , prove that:
(a) There exists a triangle with sides of lengths a{s~a), b(s-b), c(s-c).
(lb) (m / a ) 2 + (m./b)2 + (mc/c)2 > 9/4, with equality if and only if the
triangle is equilateral.
Solution by M.S. Klamkin, University of Alberta.
(a) The desired result follows immediately if we set
x = s-a > 0, y - s-b > 0, z = s-c > 0,

for then the triangle inequality

b(s-b) + c(s-c) > a(s-a),

for example, becones y(z+x) + z(x+y) > x(y+z)9 which is equivalent to 2yz > o.
(b) With Hm2 = 2b2-\-2c2-a29 etc., the required inequality is easily found to
a
be equivalent to
(a2/b2+b2/c2+c2/a2) f (b2/a2+c2/b2+a2/c2) > 6,

which is true since each quantity in parentheses is at least 3 by the A.M.-G.M.

inequality, with equality if and only if a = b = c. D
 - 308 -
A dual inequality to (b) is

ia/m ) 2 4 (b/m*)2 i ie/m ) 2 > 4.

a b c
More qenerally, for any trianqle inequality
Ka.b.c.m
3 3 nu.m ) > 0
* as3 b* c
we have the dual inequality
Km ,m-t3rn s^a9%bs%c) > 0,
a D c
because the three medians of a trianqle also form a trianqle whose medians are I
the respective sides of the original triangle (see Nathan Alts"hiller Court, College
Geometry,Barnes and Noble, New York, 1Q52, p.66).
Also solved by W.J.BLUNDON, Memorial University of Newfoundland; S.C. CHAN,
Singapore; JORDI DOU, Barcelona, Spain; ROLAND H. EDDY, Memorial University of
Newfoundland; JACK GARFUNKEL, Flushing, N.Y.; G.C, GIRI, Midnapore College, West
Penqal, India; J.T. GPOENMAN, Arnhem, The Netherlands; LEROY F. MEYERS, The Ohio
State University; BOB PRIELTPr, University of Wisconsin-Oshkosh; KESIRAJU SATYA-
NARAYANA, Gagan Mahal Colony, Hyderabad, India; DAN SOKOLOWSKY, Califormia State
University at Los Angeles; ROBERT A. STUMP, Hopewell, Virginia; and the proposer.
it t'« s'c

590, F1980: 3171 Proposed by J.T. Groenman, Arnhem, The Netherlands.

Find all real solutions of the equation O 3 ! - 3fa:2] + 2[x~\ = o, where
the brackets denote the greatest integer function.
Solution by Hessel Pots Woerden3 The Netherlands.
The step function defined by fix) = fx*l - 3[x2l + 2[.r] for all real x obviously
satisfies f(x) < o for x < o. If x > 3, then H E 3 ! ^ C3a-2] > 3|>23 and f(x) > 2|>3 > 0.
So the solutions of the equation fix) - o all lie in the interval [0,3), to which
the subsequent discussion is restricted.
The step function is continuous (from the right) at x = o, continuous at x - 1
and x - 2 since, for example,
jTCl+O = fil) - /(i-e)+l-3+2 = fil-e)
when E is a small positive number, and continuous from the right at all points of
discontinuity, which are the square roots s and the cube roots o of inteaers in
the intervals (1,2) and (2,3). As x increases from l to 3, fix) decreases by 3
when x = s (i.e., when xs is a cube not a square), and it increases by 1 when
x = c (i.e., when x 6 is a square not a cube). It is now easy to evaluate mentally
fix) when xe ranges alonq the combined ascendinq sequence of squares and cubes
less than 3 G . Part of this sequence, with the corresponding values of fix), is
tabulated at the top of the next page.
 - 309 -
b 6 2 3
x = 0 l 2 2 32 42 52 33 62 72
fix) = 0 0 1 - 2 - 1 0 1 - 2 - 1 0

x6 - 2s 92 10 2 ll 2 S3 12 2 13 2 1«+2 63 152
fix) = 0 1 2 3 0 1 2 3 0 1
2 2 2
The rest of the sequence, 16 , 17 , ..., 26 , produces no more zeros for f because
it contains at least three /-increasing squares between two succeeding /-decreasing
cubes.
The above tabulation shows that the required solution set is
[0,?2) U C«\r5) U [77,?9) U [/5,?12) u C/6,^15").
Also solved by JORDI DOU, Barcelona, Spain? FRIEND H. KIERSTEADf JR., Cuya-
hoga Falls, Ohio? LEROY F. MEYERS, The Ohio State University; HARRY L. NELSON,
Livermore, California? DAN SOKOLOWSKY, California State University at Los Angeles?
ROBERT A» STUMP, Hopewell, Virginia? KENNETH M. WILKE, Topeka, Kansas? and the
proposer.
sf« * *

5911 C1980: 318] Proposed by Charles W, Trigg, San Diego, California.

December is a good month to solve the cryptarithm
SEER + SEER + SEER = MAGI,
which memorializes the start of a historical event. The three wise men, Pythagoreans
all, insisted that AG be twice the square of a prime. Find the unique solution.
Solution by Clayton W. Dodge, University of Maine at Orono.

Assume the addition is written in column form. Since 2p 2 = 08, 18, 50, and
98 for primes p with 2P 2 < 100, and since A and G can differ by only the carries
into their columns, we can have AG = 08 or 98 only. Thus G = 8, so (G,E,A) = (8,2,6),
(8,9,9), or (8,6,9). Only the third case is permissible since AG * 68 and A * E.
Now R * 2 since I * 6, and R * 3 since I * 9. Hence R = 1 and I = 3. Then S = 2
and M = 7. The unique solution is
2661 + 2661 + 2661 = 7983.
Also solved by T.A.H. HUNTER, Toronto, Ontario? ALLAN WM. JOHNSON JR., Washington,
D.C? HERMAN NYON, Paramaribo, Surinam? BOB PRIELIPP, University of Wisconsin-
Oshkosh (two solutions)? DONVAL R. SIMPSON, Fairbanks, Alaska? RAM REKHA TTWAJRI,
Radhaur, Bihar, India? ROBERT TRANQUILLE, College de Maisonneuve, Montreal, Quebec?
KENNETH M. WILKE, Topeka, Kansas? and the proposer.

Editor's comment.
One solver wondered why December was a good month to publish this cryptarithm.
The answer is, obviously, because the December 1980 issue of Crux was delivered
to readers in early January 1981, just around the Feast of Epiphany.
 - 310 -
592. C1980: 318] Proposed by Leroy F. Meyers, The Ohio State University.
(a) Given a segment AB of length Z, and a rusty compass of fixed opening
r, show how to find a point C such that the length of AC is the mean proportional
between r and Z, by use of the rusty compass only, if \l < r < I but r * \l.
(b) Show that the construction is impossible if r - }1.
(c)* Is the construction possible if r < \l or r > 11
(This problem was inspired by Dan Pedoe's Problem 492.)
Solution of parts (a) and (b) by the proposer.
Since only one radius is possible, it will be unambiguous and convenient to
denote by (P) a circle with center P and radius r.
(a) The circles (A) and (B) intersect the segment AB in unique points A'
and B 1 , respectively. Since \l < r < I and r * JZ, we have 0 < A'B' <; 2rs and
so the circles (A') and (B') intersect. Let C be one of the points of intersection,
and let D be the midpoint of AB. Then CD i AB and

AC ? = AD 2 + CD 2 = AD 2 + A'C 2 - A'D 2

= (H)2 + r2 - |JZ-r|2 = Ir.

Hence AC is the mean proportional between Z and r. The rusty compass was used
exactly four times. Note that if r = \l , then the triangle ACD is degenerate,
but the calculation goes through; if r = Z, then A' = B and B1 = A, so that C
is a point of intersection of the first two circles drawn.
(b) If r = ll3 then the circles (A) and (B) are tangent at the midpoint D
of AB. In the above notation, A* = B' = D, so that the circles (A') and (B') are
not distinct, and the construction, if possible, must be continued in a different
way. However, the only points which can be obtained successively as intersections
of circles with centers already determined are those of the triangular lattice of
side length r. The three smallest distances between any two of these points are
r3 r/3, and 2r,none of which is the mean proportional r/2. Hence the mean propor-
tional cannot be determined by rusty compass alone.
j'f j'j ft

593. [1980: 3181 Proposed by Andy Liu, University of Alberta.

Grandpa is loo years old and his memory is fading. He remembers that
last year — or was it the year before that? — there was a big birthday party in
his honor, each guest giving him a number of beads equal to his age. The total
number of beads was a five-digit number, x67y29 but to his chagrin he cannot recall
what x and y stand for. How many guests were at the party?
 - 311 -
Solution by J.T. Groenman, Arnhems The Netherlands.
Suppose the big party occurred last year, when Grandpa was 99. Then, with
o < x < 9 and o < y < 9, we have

#67?/2 = x + lOy + 69 = x + lOy - 30 E 0 (mod 9 9 ) ,

which implies that x+lOy = 30 and lo[x. So there is no solution and the big party
must have occurred the year before, when Grandpa was 98. Now we have
x67y2 = 4a? + lOy + 38 = hx + lOy - 60 = 0 (mod 98).

The acceptable values of x and y must therefore satisfy 2x+5y - 30, and the only
solution is # = 5, y - 4.
The number of guests was thus 56742/98 = 579.

Also solved by CLAYTON W. DODGE, University of Maine at Orono? RICHARD A. GIBBS,

Fort Lewis College, Durango, Colorado? G.C. GIRI, Midnapore College, West Bengal,
India? HANS HAVERMANN, Weston, Ontario? FRIEND H. KIERSTEAD, JR., Cuyahoga Falls,
Ohio? LEROY F. MEYERS, The Ohio Sate University? HERMAN NYON, Paramaribo, Surinam?
BOB PRIELIPP, University of Wisconsin-Oshkosh? KESIRAJU SATYANARAYANA, Gagan Mahal
Colony, Hyderabad, India? DAN SOKOLOWSKY, California State University at Los Angeles?
ROBERT A. STUMP, Hopewell, Virginia? RAM REKHA TIWARI, Radhaur,Bihar, India? ROBERT
TRANQUILLE, College de Maisonneuve, Montreal, Quebec (deux solutions)? CHARLES W.
TRIGG, San Diego, California? KENNETH M. WILKE, Topeka, Kansas? and the proposer.
s'f j'c &

5941 C1980: 318] Proposed by John Veness> Cremome3 N.S.W., Australia.

Let N be a natural number which is not a perfect cube. Investigate the
existence, nature, and number of solutions of either or both of the Diophantine
equations
x3 - Ny3 = ±1

in positive integers x and y.

Comment by Kenneth M. Wilke^ Topekas Kansas.
We will let x and y range over the set Z of all rational integers. The solutions
of the two equations in the proposal, if any, are then included among those of the
Diophantine equation

.r3 + < V = 1, (1)

which has been extensively studied, especially by B. Delaunay, whose methods were
later refined and extended by T. Nagell.
We quote from Mordell f"l3. "The integer solutions [of (1)] are trivial when
d is a perfect cube. Then if \d\ > l, the only solution is x = l, y ~ o, and when
\d\ = l, there is another solution x = o^ dy - l. We may suppose now that d > l
and is free from cubed factors since these can be absorbed in t/3. We consider the
 - 312 -
cubic field K = Q(M). The inteqers in K of the form x + yM + s&T2,where x,y,z
are rational integers, form a rinq Z[Ml, the units in which are those inteqers
r\ whose norms N(r\) - ±1. Let e be the fundamental unit in the rinq chosen so that
o < e < l. Then all the units in zC&Tl are qiven by n = ±e n 9 where n takes all
integer values." Mordell then goes on to discuss and prove Delaunay's result:
The equation x3 + dy3 = 1 (d > 1) has at most one integer solution with, xy * 0.
This is given by the fundamental unit in the ring when it is a binomial unit,
i.e., e takes the form e = x •* yW.
Mordell then qives Naqell's more comprehensive result as it applies to the
more qeneral equation ax3 + by3 = c. The Delaunay-Nagell Theorem as it applies
to equation (1) is qiven by LeVeque [2] as follows (adjusted only for notation):
The equation x3 + dy3 = 1 has at most one solution in integers x,y * 0.
Tf (x^y*) i-s a solution, the number x, + y $J is either the fundamental unit of
K = Q(Vd) or its square; the latter can happen for only finitely many values of d.
See Cohn [3] for a discussion of values of d for which (l) has no nontrivial
solution.
A comment was also received from HERMAN NYON, Paramaribo, Surinam.

REFERENCES
1. L.J. Mordell, Diophantine Equations, Academic Press, New York, 1969,
pp. 203-205, 219-225.
2. William Judson LeVeque, Topics in Number Theory, Addison-Wesley, Reading,
Massachusetts, 1956, Vol.11, pp. 112-120.
3. J.H.E. Cohn, "The Diophantine Equation x3 = dy3 + 1 " , Journal London Math.
Soc, 42 (1967) 750-752.
* * ft

595. T1980: 33 8] Proposed by G.C. Giri, Midnapore College, West Bengal, India.
Let f(x,y) = a2cosx cos y + a($inx + sin?/) + 1. Prove that
/ ( a , Y ) = 0 and / ( Y , a ) = 0 ==> f(a,3) = o.

Solution by the proposer.

It follows from the hypothesis that 9 = a and 0 = 3 are solutions of the
equation

a2cos y cos 6 + a(sin y + sin 8) 4 1 = 0

and hence of
a < f cos 2 ycos 2 e = ( a ( s i n y + s i n e ) + l } 2 (l)

as well as of
 - 313 -
a sin e = {a cos y cos 9 + a sin Y + l } 2 .
2 2 2
(2)

With cos2e = l - sin 2 e, (l) is equivalent to a quadratic in sine , for which the
sum of the roots is

With sin2e = l - cos 2 e, (2) is equivalent to a quadratic in cosefor which the

product of the roots is
(1
cos 9 cos 3 = IV^P2 if • 00
^ az(l + a 2 C0S^y)
Now, from (3) and (*+),
a2cos a cos 8 + a(sin a + sin 6) + 1 = o,
that is 5 /(qr,B) = 0.
Also solved by W.J. BLUNDON, Memorial University of Newfoundland? and KESIRAJU
SATYANARAYANA, Gagan Mahal Colony, Hyderabad, India.
ft & ft

5961 C1981: 18] Proposed by Leroy F. Meyers, The Ohio State University.
Automorphic numbers were discussed in my comment II to Crux 321 [1978:
2521. An automorphic number (in base ten) is a positive integer k whose square
ends in k, (Initial zeros are permitteda) Not counting the trivial solutions l,
01, 001, ..., there are exactly two n-digit automorphic numbers for each positive
integer n. Examples are
5, 6, 25 9 76, 376, 625, 0625, 9376.
For an arbitrary positive integer n, find explicit formulas for the two nontrivial
rc-digit automorphic numbers.
Solution by Robert A, Stump9 Hopewell5 Virginia (revised by the editor).
If k > 1 is an n-digit automorphic number, then, by definition,
k2-k = k(k-l) =0 (mod 10 n ); (1)

so, since {k3k-±) - l and lo/fc, e i t h e r

5n\k and 2 n |fc-l (2)

or
2 n |fc and 5 n |7<-l. ' (3)

There is at most one n-digit number k satisfying (2); for if kl is also such a
number, then k-V = o (mod 10 n ), and k = V since \k-V\ < lon. This number, if
it exists, will be denoted by a . Similarly, there is at most one n-digit number
b satisfying (3).
Having shown their uniqueness, we now show that a and b exist for every
positive integer n. In fact, we show that, explicitly,
 - 314 -
a - the number formed by the last n digits of 5 (M-)

and
b = the number formed by the last n digits of 2 . (5)

First, observe that if a positive integer k satisfies (2) or (3) for some
positive integer rc, then k has at least n digits, it satisfies ( 1 ) , and the number
77-1
formed by its last n digits is automorphic. Now 2 > n for e\/ery n and, from
Euler's generalization of Fermat's Theorem,
52 = 5*(2 }
=1 (mod 2 n ) ;
2 n-l
so (2) holds for k = 5 and (4) is established. The proof of (5) is similar.
It is based on
2^-5 = 2 *(5 )£ ± (|Mjd 5 n j

and on the fact that 4*5 > n for every n. D

It follows from (2) and (3) that
±On\a b and 1 0 n | ( a -1)(2> - 1 ) ;
hence
a Z> - (a -1)(Z> - 1 ) + 1 = a + £ = 1 (mod 1 0 n ) ;
wn n w n n
and since 1 < a < I 0 n and l < b < lo", so that 2 < a + Jt> < 2»lo n , we conclude
n n * n n
that
a + b = 10 n + 1. (6)
rc n
So when one of the two rc-digit automorphic numbers has been calculated, the other
can be found more easily from (6).
If the positive integer x has at least n digits, the number formed by its
last n digits is
x - io n O/io n 3. (7)
So if a more mathematically explicit formulation is required for (4) and (5),
e.q., for a computer who does not "speak English", one can always substitute
x =5 or 2 in (7) to obtain a or b .
n n
Also solved by the proposer. Comments were received from HAYO AHLBURG, Benidorm,
Alicante, Spain; BIKASH K, GHOSH, Bombay, India; J.T. GROENMAN , Arnhem, The Nether-
lands; ANDY LIU, University of Alberta; HEPMAN NYON, Paramaribo, Surinam; and
KENNETH M. WILKE, Topeka, Kansas.

Editor 's comment

was quite explicit in asking for explicit formulas for a and b .
The proposal was
Yet only our featured solver and the proposer addressed themselves specifically
 - 315 -
to that question. Of the other "comments" received, some gave references (the most
important of which had already appeared in this journal [1978: 254]) where limited
lists of automorphic numbers can be found, and others showed how to calculate a few
automorphic numbers (or even infinitely many, by recurrence). Strictly speaking,
all these discussions are beside the point here if they don't (and they didn't)
lead to explicit formulas for a and h .
n n

5971 ri981: 18] Proposed by Roland H. Eddy, Memorial University of Newfoundland,

Consider the equalities

The first occurs in W. Knight's item "...But Don't Tell Your Students" [1980: 240],
which inspired this problem. Find all positive integer triples (a3b3c)9 with b
and Q square-free and {b3a) = 1, that satisfy the second.
Solution by Leroy F, Meyers, The Ohio State University.
Suppose that a3bso are positive integers such that b and c are square-free
and relatively prime, and

fTf -afV (1)

Then ac - b{a2-l). Since (a3a2-l) = (b3o) = 1, we must have a\b and b\a9 so a = b
2
and a = a -l. Hence a3 like b, is square-free, and so are afl and a-1 since their
product o is square-free. Thus we have the necessary conditions:

a-1, a, a+1 are all square-free; b - a; c - a 2 -l. (2)

These conditions are also sufficient. For suppose a3b3c are positive integers
satisfying (2). Then (1) holds, b is square-free, (b3e) = (a3a2-l) = 1, and we
have only left to show that e is square-free. Observe that the square-free numbers
a+1 and a-1 must both be odd (otherwise one would be divisible by 4 ) , so their qcd
must be odd. Since this gcd divides their difference 2, it must be 1. The square-
free numbers a+l and a-1 are therefore relatively prime, and their product e is
square-free.
We conclude that the triple (a3b3c) is a solution to our problem if and only
if it satisfies (2). There are infinitely many solutions, for Sierpinski [l]
affirms: "One can prove that there exist infinitely many triples of consecutive
natural numbers such that each of the numbers is square-free." The first few values
of a leading to solutions are: 2, 6, 14, 22, 30, 34, 38, 42, 58, 66, 70, 78, 86,
94, 102, 106, 110, 114, 130, 138, 142, 158, 166, 178, 182, 186, 194.
 - 316 -
Also solved by HAYO AHLBURG, Benidorm, Alicante, Spain; LEON BANKOFF, Los
Angeles, California? JAMES BOWE, Erskine College, Due West, South Carolina; CLAYTON
W. DODGE, University of Maine at Orono; BIKASH K. GHOSH, Bombay, India; J.T. GROEN-
MAN, Arnhem, The Netherlands; J.A.H. HUNTER, Toronto, Ontario; FRIEND H. KIERSTEAD,
JR., Cuyahoga Falls, Ohio; ANDY LIU, University of Alberta; J.A. McCALLUM, Medicine
Hat, Alberta; HERMAN NYON, Paramaribo, Surinam; BOB PRIELIPP, University of Wisconsin-
Oshkosh; KESIRAJU SATYANARAYANA, Gagan Mahal Colony, Hyderabad, India; DAN SOKOLOW-
SKY, California State University at Los Angeles; ROBERT A. STUMP, Hopewell, Virginia?
ROBERT TRANQUILLE, College de Maisonneuve, Montreal, Quebec; CHARLES W. TRIGG, San
Diego, California; KENNETH M. WILKE, Topeka, Kansas; KENNETH S. WILLIAMS, Carleton
University, Ottawa, Ontario; and the proposer.

Editor's Comment,
The Sierpinski reference in the above solution was added for completeness by
the editor, who took it from the solution of Bob Prielipp.
Most of the other solvers arrived at the necessary conditions (2). But also,
the editor regrets to report, most did not seem to be aware that a proof of suffi-
ciency was also required for completeness. Even a bit of feeble arm waving in that
direction would have been welcome. They thus appeared to take an attitude towards
mathematical proofs that they would not tolerate in their students.
Ahlburg and Nyon hinted at the following more general result:
Let n > 2. If a3b,e are positive integers such that b and o are nth-power-
free and relatively prime, and
n
\~~T "lb
then
a-\3 a, a -f.. ,+a+l are all nth-power-free; b = a; c - a -1. (3)
The proof that conditions (3) are necessary follows the same pattern as in
the case n = 2, but the proof of sufficiency breaks down when we try to show that
e is nth-power-free. For example, when n = 3 and a - 10, then 9, 10, 111 are all
cube-free, but c = 999 = 33-37 is not. Thus conditions (3) would have to be strength-
ened to make them sufficient. In any case, this emphasizes the fact that a proof
of sufficiency was absolutely essential in the case n = 2.

REFERENCE
1. Wacjaw Sierpinski9Elementary Theory of NiAmbers, Warszawa, 1964, p.34.
* * *
5981 C1981: 18] Proposed by Jack Garfurikel, Flushing, N.I.
Given a triangle ABC and a segment PQ on side BC, find, by Euclidean
construction, segments RS on side CA and TU on side AB such that, if equilateral
triangles PQJ, RSK, and TUL are drawn outside the given triangle, then JKL is an
equilateral triangle.
 - 317 -
SoLutions or comments were received from JORDI DOU, Barcelona, Spain; BIKASH
K. GHOSH, Bombay, India; J.T. GROENMAN, Arnhem, The Netherlands? ANDY LIU, University
of Alberta; LEROY F. MEYERS, The Ohio State University; KESIRAJU SATYANARAYANA,
Gagan Mahal Colony, Hyderabad, India; and the proposer.

Editor 's comment.

As several readers pointed out, this problem is ill-posed and additional
conditions would have to be imposed to make it viable. As it stands now, with the
point J known, if any equilateral triangle JKL is drawn with K on the opposite side
of CA from B and L on the opposite side of AB from C s and then, by a trivial con-
struction, equilateral triangles RSK and TUL are drawn with RS on CA and TU on AB,
then the segments RS and TU constitute one of infinitely many solutions. Nothing
to write home about. This problem, it is clear, should have been diverted to the
circular file. But the proposer and the editor were both asleep at the switch.
A viable problem can, however, still be salvaged from the debacle. It is clear
from the proposer's solution that the problem he should have proposed is the follow-
ing:
Given a (not necessarily convex) hexagon PQRSTU in which two pairs of opposite
sides (hence also the third pair) are equal and parallel, equilateral triangles
PQJ, RSK, and TUL are drawn externally. Prove that triangle JKL is equilateral,

Solution adapted from the

proposer's.
We represent vectors by
complex numbers (denoted by Greek
letters). Let a = e2iri/3 (so
that o)3 - l) and
Pt) =a, flfc = 3, |fs = Y,
as shown in the figure. Then
Jk = Jt) + <$ + #k
= -aw 2 + (3 - yw
and
KL = l?S + St + TlL

= -yw 2 - a + iBw. j

Thus wJk = KL, which shows that triangle JKL is equilateral.

 - 318 -
599• C1981: 18] Proposed by Allan Wm. Johnson Jr., Washington, B.C.
Prove that 36 divides the sum of the 36 integers composing a sixth-
order magic square that is pandiagonal (magic also along the broken diagonals) or
symmetrical (pairs symmetrical with respect to the center have a constant sum).
Solution by the proposer.
It suffices to show that 6 divides the magic sum M of eyery sixth-order magic
square that is pandiagonal or symmetrical.

Proof for symmetrical squares.

Let T be the constant sum of pairs symmetrical with respect to the center. If
the top row contains the numbers A3 B, C, D, E3 F, then the bottom row contains the
numbers T-F3 T-E, ..., T-A. These twelve numbers form two complete rows whose sum is
6T = 2M. Hence T = M/39 which shows that 3\M.
To show that also 2\M9 we partition the sixth-order square
into four third-order squares, the nine numbers in each third-order [«1 Qz
square summing to Q\, Q2J Q-$% Q^9 repectively, as shown in Figure 1.
\Q3 Qh
Because the square is magic, we have
Figure 1
Q\ + Ql = Ql + Qh = 3M;

and because it is symmetrical, we have

Ql + Qh = 9T = 3M.

Hence Qi = Qk = 3M/2, and 2\M.

Proof for pandiagonal squares.

To show that 2\M9 we consider the square in Figure 2, which is assumed to be
magic and pandiagonal.

Mi »1 H B2 ^3 B3

Cl Dl c2 D2 c3 t>3

U, Bh H BS H BG

Ch Vk c5 »5 cG H
Al B7 ^8 Be ^9 B9

c7 »7 c* »8 c3 Pg |

Figure 2
Let
9 9 9
A = .1 Av B = I B D = I Vv
i-1 i-± i=l
 - 319 -
Because the square is magic, if we add rows 1,3,5, and then separately add columns
2,4,6, we get
A + B = B + D = 3M;
and because it is pandiagonal, adding the three northwest-southeast diagonals which
begin at A\^A2iA^ gives
A + D = 3M.

Hence A = D = 3M/29 and 2|A/.

To show that 3JAf, we consider the square of Figure 3, which is assumed magic and
pandiagonal.

p7r»ri~cTF nr | B2 | c2

°i E
I 1 ! D
2 *2
1 *2
k *1
J
l
G2
*2
C
J
2

\
"i f J
B C
3 3 B
i3 B
D
3 *3 F

J
3 J N J
Lfd g
3
Jj
3J
Figure 3
"1 \ |

Here we define 4 q.
>4 = I A v B= I B ^

with similar definitions for c3VsEsF5G3MsI« Adding separately rows 1,4, then rows
2,5, then rows 3,6, we get
A + B + C = 2M9 D + E + F = 2Af, G + B + I = 2M;

and adding serarately columns 1,4, then columns 2,5, then columns 3,6 gives
A + P + G = 2M, B + E + H = 2M, (7 -f F + I = 2M.
tnen
Finally,, adding separately the northwest-southeast diagonals beginning at d j . , ^
the northeast-southwest diagonals beginning at CisC2, we get
4 + E + I = 2Af, C + £ + C = 2M. r~ ' B
1
C 1
A
If follows from these results that the square in Figure 4 is ID E F
magic with magic sum 2M. Since every third-order magic square
\ G H I
has a magic sum equal to thrice the center number, we conclude
that E = 2M/3, and SO 3\M. D Figure 4
 - 320 -
By continuing to add broken diagonals in Figure 3 9 it is easy to show that the
square of Figure 4 is also pandiagonal, which is possible only if all its entries
are equal. This proves that the entries in a sixth-order pandiagonal magic square
can be partitioned into nine disjoint quartets each of which sums to 2M/3.
The 36 consecutive integers m3 /w+1, ...,, m+35 add up to 18(2^+35), so a magic
square composed of these numbers must have magic sum M = 3(2m+35). Because this
maqic sum is odd, parity prevents this magic square from being pandiagonal or
symmetrical. That 36 (more generally, (4p+2) 2 ) consecutive integers cannot be ar-
ranged into a magic square that is pandiagonal or symmetrical was first proved over
60 years ago by Planck Til.

A nearly complete solution was submitted by KENNETH M. WILKE, Topeka, Kansas?

and a somewhat inconclusive argument dealing with special cases was submitted by
BIKASH K. GHOSH, Bombay, India.

REFERENCE
l. C. Planck, "Pandiagonal Magics of Orders 6 and 10 With Minimal Numbers",
The Monist, 29 (1919) 307-316.

THE DOT POLKA

The editor has received, through the courtesy of Leon Bankoff, a generous
extract from a book entitled One Million, by Hendrik Hertzberg, published in 1970
by Simon and Schuster. The book contains one million dots.
We quote from the introduction. "There are 5000 dots to a page—10000 on
each double-page spread. ...Notes are scattered—like mileposts—here and there
in the inside margins. Each note corresponds to a number, and the dot signifying
that number is readily identifiable." The notes, of which there are several hun-
dreds, range from 2 (population of the Garden of Eden), through 32500 (number of
laps in Lapland) and 155024 (number of breasts in Brest), to 1000000 (number of dots
in the book). One of the most stirring passages in the book is reproduced below.

If the reader stares fixedly at the above for a few minutes, the dots will
soon begin to dance before his eyes. They are dancing, of course, a polka.
 - 321 -

INDEX TO VOLUME 7 , 1981

AUTHORS OF ARTICLES AND NOTES

AEPPL1, ALFRED. Area Characterizations of Curves: I and II, 34, 135.

BARBEAU, ED and IM, JOHN. An Angle Trisection Method which (Usually) Does Not
Work, 100.
CHARLES, HIPPOLYTE. Les mathematiques au service de Casanova, 9.
COLLINGS, STANLEY. Extensions of the Nine-Point Circle, 164.
CROSS, DONALD. Squares on Parade, 162.
DE STAEL, GERMAINE. De 1'enseignement des mathematiques, 196.
GIBBS, RICHARD A. Mathematical Swiftie, 134.
HAVERMANN, HANS. The Puzzle Corner, 237, 294.
HAUSMAN, MIRIAM. On (^-Perfect Numbers, 132.
IM, JOHN and BARBEAU, ED. An Angle Trisection Method which (Usually) Does Not
Work, 100.
JOHNSON, ALLAN WM., JR. A Pandiagonal Sixth-Order Prime Magic Square, 130.
. Pandiagonal Magic Square Equations, 258.
KLAMKIN, MURRAY S. On Equilateral and Equiangular Polygons, 2.
. The Olympiad Corner: 21 to 30, 11, 41, 72, 105, 139, 171, 220, 235, 267,
298.
and LIU, A. On Equiangular Polygons, 69.
. Areas of Triangles Inscribed in a Triangle, 102.
and LIU, A. Three More Proofs of Routh's Theorem, 199.
. Postcript to "Three More Proofs of Routh's Theorem", 273.
LIU, A. and KLAMKIN, M.S. On Equiangular Polygons, 69.
and . Three More Proofs of Routh's Theorem, 199.
MEYERS, LEROY F. Notes on Notation: I, II, III, IV, 40, 101, 170, 229.
. I Love a (Longer) Parade, 234.
MURTY, V.N. An Extension of an Identity of Feller, 226.
PEDOE, DAN. Geometry A Posterior(i), 232.
RUDERMAN, HARRY D. A Lattice Point Assignment Theorem, 98.
. Another Lattice Point Theorem, 144.
SATYANARAYANA, KESIRAJU. A Simple Proof of The Butterfly Problem, 292.
TRIGG, CHARLES W. A 1981 Gallimaufry, 6.
. Mama-Thematics, 19, 80.
Prime Arithmetic Progressions, 68.
". Nine-Digit Patterned Palindromic Primes, 168.
. More Nine-Digit Patterned Palindromic Primes, 295.
VAN DE CRAATS, JAN. Another "Proof" That 0 = 1, 39.
WAYNE, ALAN. Mama-Thematics, 49, 167, 297.
. The Puzzle Corner, 198, 219, 237, 300.

TITLES OF ARTICLES AND NOTES

ANGLE TRISECTION METHOD WHICH (USUALLY) DOES NOT WORK, AN. Ed Barbeau and John Im,
100.
ANOTHER LATTICE POINT THEOREM. Harry D. Ruderman, 144.
ANOTHER"PROOF" THAT 0 = 1 . Jan van de Craats, 39.
AREA CHARACTERIZATIONS OF CURVES: I, II. Alfred Aeppli, 34, 135.
 - 322 -

AREAS OF TRIANGLES INSCRIBED IN A TRIANGLE. M.S. Klamkin, 102.

BIOGRAPHICAL NOTE. 294.
CONTRIBUTION REOUEST FOR "OPEN QUESTIONS IN MATHEMATICS". 134.
DAN J. EUSTICE 1931-1981. 224.
DOT POLKA, THE. 320.
ENSEIGNEMFNT DES MATHFMATIQUES, D E L ' . Germaine de Stael, 196.
EpUIANGLULAR POLYGONS, ON. M.S. Klamkin and A. Liu, 69.
EQUILATERAL AND EQUIANGULAR POLYGONS, ON. Murray S. Klamkin, 2.
EXTENSION OF AN IDENTITY OF FELLER, AN. V.N. Murty, 226.
EXTENSIONS OF THE NINE-POINT CIRCLE. Stanley Collings, 164.
FIG. NEWTON. 17.
GEOMETRY A POSTERIOR(I). Dan Pedoe, 232.
I LOVE A (LONGER) PARADE. Leroy F. Meyers, 234.
INTERNATIONAL CONFERENCE ON TFACHING STATISTICS. 66.
LATTICE POINT ASSIGNMENT THEOREM, A. Harry D. Rudemtan, 98.
MAMA-THEMATICS. 19, 49, 80, 167, 276, 297.
MATHEMATICAL SWIFTIE. Richard A. Gibbs, 134.
MATHEMATICS IN THE (NEAR) FUTURE. 96.
MATHFMATIQUES AU SERVICE DE CASANOVA, LES. Hippolyte Charles, 9.
MORE NINE-DIGIT PATTERNED PALINDROMIC PRIMES. Charles W. Trigg, 295.
NINE-DIGIT PATTERNED PALINDROMIC PRIMES. Charles W. Trigg, 168.
1981 GALLIMAUFRY, A. Charles W. Trigg, 6.
NOTES ON NOTATION: I, II, III, IV. Leroy F. Meyers, 40, 101, 170, 229.
OLYMPIAD CORNER, THE: 21 to 30. Murray S. Klamkin, 11, 41, 7 2 , 105, 139, 171,
220, 235, 267, 298.
PANDIAGONAL MAGIC SQUARE EQUATIONS. Allan Win. Johnson Jr., 258.
PANDIAGONAL SIXTH-ORDER PRIME MAGIC SQUARE, A. Allan Wm. Johnson Jr., 130.
^-PERFECT NUMBERS, ON. Miriam Hausman, 132.
POSTCRIPT TO "THREE MORE PROOFS OF ROUTH'S THEOREM". M.S. Klamkin, 273.
PRIME ARITHMETIC PROGRESSIONS. Charles W. Trigg, 68.
PUZZLE CORNFR, THE. 198, 219, 237, 290, 294, 300.
SIMPLE PROOF OF THE BUTTFRFLY PROBLEM, A. Kesiraju Satyanarayana, 292.
SQUARES ON PARADE. Donald Cross, 162.
THREE MORE PROOFS OF ROUTH'S THEOREM. M.S. Klamkin and A. Liu, 199.

PROBLEMS AND SOLUTIONS

The numbers refer to the pages in which the corresponding name appears with
a problem proposal, a solution, or a comment.
AHLBURG, HAYO 19, 49, 146, 178, 189. DELLINGER, JOE: 240.
ALDINS, JANIS 145. DODGE, CLAYTON W.: 95, 154, 207, 211,
BANKOFF, LEON 25, 127, 204, 240, 243, 288, 309.
253. DOU, JORDI: 19, 25, 212, 239, 247, 248,
BARBEAU, E.J.: 240. 302.
BAPSBY, JOHN T. 161, 186. DUDLEY, UNDERWOOD: 29, 30.
BEESACK, PAUL R 179, 208. ECKER, MICHAEL W.: 54, 8 8 , 129, 178,
BHATTACHARYA JAYANTA: 146. 185, 205, 208.
BLUNDON, W.J 31, 179, 218, 240. EDDY, ROLAND H.: 18, 117, 121, 253,
BOTTEMA, O.: 25, 238. 306, 315.
BRENNER, J.L.: 20. ERDOS, P.: 186, 189, 241, 282, 284.
CHAN, S.C.: 13 7, 146, 251. EVERSON, TERRY R.: 183.
CHARLES, HIPPOLYTE: 215, 276, 302. EVES, HOWARD: 25, 56, 65, 120.
COVILL, RANDALL J.: 179. FISHER, BENJI: 241.
CSIRMAZ, L.: 280, 282. FISHER, J. CHRIS: 178, 241, 276.
 323 -

GARFUNKEL, JACK: 18, 25, 49, 6 1 , 79, PRIMER, JEREMY: 241.

116, 120, 147, 154, 178, 205, 211, PROPP, JAMES GARY: 84, 118.
276, 302, 303, 306, 316. RABINOWITZ, STANLEY: 277, 289.
GIBBS, RICHARD A . : 182, 249. RENNIE, BASIL C : 22, 124, 243.
GIRI, G.C.: 56, 57, 79, 86, 1 6 1 , 184, ROSEN, HYMAN: 190.
251, 302, 312. ROYTER, MATS: 249, 250.
GROENMAN, J . T . : 79, 9 1 , 115, 127, 147, ftUDERMAN, HARRY D.: 239.
180, 205, 275, 302, 308, 311. RURSA, IMRE Z.: 284.
HAMMER, F. DAVID: 116. SACKUR, MARC: 120.
HENDERSON, G.P.: 79. SALVATORE, GALI: 31, 49, 84, 116.
HUNTER, J.A.H.: 19, 32, 80, 274. SAMSOE PROBLEM GROUP: 247.
HURD, CARL: 20. SANYAL, ARUN: 95.
HUTCHINSON, JOAN P.: 19. SATYANARAYANA, KESIRAJU: 91, 155, 207,
ISAACS, RUFUS: 29.
290.
JOHNSON, ALLAN WM. JR.: 18, 48, 59,
SERVRANCKX, ROGER: 276.
94, 147, 205, 215, 246, 254, 288,318 SMITH, MALCOLM A.: 159.
KATCHALSKI, MEIR: 24. SOKOLOWSKY, DAN: 54, 122, 123, 146,
KIERSTEAD, FRIEND H.: 27, 59, 88, 190, 205, 207, 301.
245, 251. ST. OLAF COLLEGE PROBLEM SOLVING GROUP:
KILLGROVE, R.B.: 250. 209, 251.
KING, BRUCE: 27. STUMP, ROBERT A.: 80, 276, 313.
KLAMKIN, M.S.: 28, 51, 61, 65, 150, TAN, KAIDY: 180, 204, 274.
158, 161, 181, 217, 303, 306, 307. TRANQUILLE, ROBERT: 63, 129.
KLINE, JAMES S.: 145. TRIGG, CHARLES W. : 30, 52, 63, 80, 81,
KRAVITZ, SIDNEY: 81, 117, 180, 218. 95, 117, 125, 147, 156, 179, 192, 203,
LADOUCEUR, ANDRE: 189. 216, 238, 245, 250, 254, 256, 275, 302,
LEEDS, B.: 122. 309.
LINDSTROM, PETER A.: 238. TSINTSIFAS, GEORGE: 19, 23. 28, 48, J47,
LINIS, VIKTORS: 86, 213. 148, 150, 179, 205, 212, 214, 239, 275,
LIU, ANDY: 24, 56, 116, 190, 211, 310.
302, 305.
LUEY, LAI LANE: 129.
VENESS, JOHN: 311,
LYNESS, R.C.: 50, 274.
WAGON, STANLEY: 19, 145.
MANDAN, SAHIB RAM: 58.
WAYNE, ALAN: 20, 49, 80, 180, 192, 204,
M8NSSON, BENGT: 65, 160. 206, 254.
MASKELL, F.G.B.: 145, 290. WHEELER, FERRELL: 23, 53, 92, 146, 160,
MAVLO, DMITRY P.: 116, 177.
240, 301.
McCALLUM, J.A.: 52, 301.
WILKE, KENNETH M.: 30, 123, 193, 218,
McWORTER, WILLIAM A„; 81.
311.
MEIR, A.: 61.
WILLIAMS, KENNETH S.: 90, 213.
MEYERS, LEROY F.: 18, 30, 65, 81, 117,
WINTERINK, JOHN A.: 155.
119, 182, 184, 216, 252, 310, 313,
315.
MILLER, FRED A.: 80.
MOSER, WILLIAM: 240.
MURTY, V.N.: 64, 150, 181, 210, 217,
239r 248.
NELSON, HARRY L.: 157.
NGO, TAN: 49, 57, 203, 256, 305, 307.
NYBERG, CHRIS: 53.
NYON, HERMAN: 20, 51, 80, 128.
PEDOE, DAN: 50, 85, 277.
PENNER, SIDNEY: 27.
POT, HESSEL: 306, 308.
PRIELIPP, BOB: 56, 87, 156, 240, 279.