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.
 sf« is is is is is is is is is is is is is is is is is is is is is is is is is is is is is is is

ISSN 0705 - 0348

CRUX MATHEMATICORUM

Vol. 8 S No. 9
November 1982
Sponsored by
Carleton-Ottawa Mathematics Association Mathematique d*Ottawa-Carleton
Publie par le College Algonquin, Ottawa
Printed at Carleton University
The assistance 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.
ftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftft
* * * * * * ft it ft * is ft ft ft * ft ft ft * * ft ft ft * * * * * ft ft ft ft ft ft ft it it 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:
VolSo 1&2 (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: I&o SauvS, Algonquin College, 281 Echo Drive, Ottawa, Ontario, Canada KlS 1N3.
Managing Editor: F.G.B. Maskell, Algonquin College, 200 Lees Ave., Ottawa, Ontario,
Canada KlS 0C5.
Typist-compositor: Nghi Chung.
Second Class Mail Registration No. 5432. Return Postage Guaranteed.
ft * ft

CONTENTS

Ducci's Four-Number Problem: A Short Bibliography . . . . . . Leroy F. Meyers 262

Some Notable Palindromic Triangular Numerals . . . . . . . . . Charles W. Trigg 267

Homage to Herman Nyon ........... a 268

The Olympiad Corner: 39 . . . . . . . . . . . . . . . . M.S. Klamkin 269
The Puzzle Corner 275
Mathematical Swifties ............................ 275
Problems - Problemes . ......... . . . . 276
Solutions .............. . . . . . . . . . 279

261
 - 262 -
DUCCI'S FOUR-NUMBER PROBLFN: A SHORT BIBLIOGRAPHY
LEROY F. MEYERS

In a recent note, R.V. Andree T20] suggests that a computer may be useful for
investigation the properties of a certain operation on ^-tuples for various values
of k. For each k-tuple A = (ai,...,a,), let TA, the transform of A, be the k-tuple

(|a2-ai(, |a 3 -a 2 | ^ak~ak-l^ ^al~ak^)m

As usual, self-composites of T are indicated by superscripts: if A is any fc-tuple,
then
T°A = A, TlA = TA, and r"+14 = Ti^A) f or n > 0.

A k-tupleAis desirable just when there is a nonnegative integer n such that

^ A is the zero fc-tuple 0 = (o,...,o). The order of a desirable fc-tuple is the
smallest such «.
Andree raised the question: "For which positive integers k (in particular, for
4 < k < s) is it true that every k-tuple is desirable (in particular, if the entries
in the k-tuple are digits: o, l, ..., 9)?"
The question is not new, as can be seen from the bibliography appended to this
note. In fact, most of Andree's questions (and many more) about the operation T
are answered in what is apparently the first paper on the subject Cl], wherein it
is stated that the problem is due to E. Ducci.
The definition of the operation T9 as given above, does not specify the nature
of the entries in the k-tuple A. In most of the papers mentioned in the bibliography,
these entries are required to be nonnegative integers, or they may be any real num-
bers. A fc-tuple (in a given discussion) will be called admissible just when its
entries are of the proper kind (for that discussion).
Many properties of the operation T are stated and proved in the literature.
Most of those that I could find are listed below. In all of them, A - (ai,...,a^)
is an admissible k-tuple. Properties A, B, and C are obvious.
A. The cyclic permutations (a. ., ..., a,, a l 5 ..., a.) for o < 3 < k9 and the
reversal (a*, ..., a \ ) , are admissible ^-tuples, and the w-fold repetition
(a\, .. . , a* , a\ , ... , a* , ..., a\ , ... , a.)

is an admissible mfc-tuple. Each of these tuples is desirable if and only if A is

desirable, and in that case has the same order as A.
B. If r and s are given, and r * 0, then (rai+s, ..., ra.+s) need not be ad-
missible. However, if it is admissible, then it is desirable if and only if A is
 - 263 -
desirable, and in that case has the same order as A9 unless A = (s/r, ..., - s / r ) * 0 .
C. If k < 2, then A is desirable, and the order of A does not exceed k.
D. If ^ is a power of 2 and all of the entries in A are nonnegative integers,,
then:
Di. A is desirable, and
D 2 . the order of A does not exceed kh-h+l9 where h is an integer, 2h > a/d9
a = max {a\,... 9a^} 9 and d = qcd {ax ,...,a^}; (let h be 0 if A = 0 ) ;
D 3 . more precisely, iffc= 4, then the order of A does not exceed 3[g/2] + i 5
where (7 is the smallest integer such that a i d < t , and the Tribonacoi number t
is defined so that
tnu = o, J = tz2 = l,
ti and t - n-1
n t .+tn-2n + tn-3
„ for n > 3.

E. If k i s not a power of 2, then 4 need not be desirable. In fact, (o,o9...9l)

and its repetitions are undesirable.
F. If k = 4 and none of the cyclic permutations of A is a monotonic k-tuple,
then the order of A does not exceed 6.
G. If k = 4 and o < a \ < a 2 £ a 3 and a^ = ai+a2+a3 > o, then
B = (a^-ai, a$+ai s a^+2a2+a\ 9 3a^+2a2+cci )
is a 4-tuple whose order (if A is desirable and not 0) is l more than the order of
A9 and whose entries satisfy inequalities and equality like those of A. (Note that
TB = 2A.) If all entries in A are nonnegative integers, then so are the entries in
B. By beginning with (o,o,l,l), we see that there are desirable 4-tuples of integers
of arbitrarily hiqh order. By use of repetitions of 4-tuples, it is easily shown
that there are 2s-tuples of arbitrarily high order, where s > 2 .
H. If A = ( a k " 1 9 a k ~ 2 , ..., a, l ) , where k > 3 and a is the positive real num-
ber satisfying
afc-l= ak~2 + ak-3 + . . . + a + l s
then TA = aA * A9 and A is undesirable. However, l i m ^ = 0. (Note that a is ir-
rational.)
I. If all the entries in A are integers, then the sequence (T n A)^ = 0 is even-
tually periodic, and one of thefc-tuplesin its period consists of at most two dis-
tinct integers, one of which is o. (Ifkis a power of 2, then the period is l, and
the only fc-tuple in its period is 0.) The number of preliminary terms before perio-
dicity sets in may be arbitrarily large, but in no case larger than approximately
q / ( k - l ) 9 where q is the smallest integer for which a/d < u , and

u 0 = ... =u f e _ 3 = 0, «fc_2 = 1. «n = V l % _ 2 + . . . + V f c + 1 f o r n * f c - l .
(Compare D 3 .)
 - 264 -
There is a short proof of Di and E which uses some nice ideas from the theory
of polynomials. Most of this proof is in ["181, and equivalent proofs using matri-
ces are found elsewhere.
We first treat the case where each entry in A is o or l. Since \a-b\ = a+b
(mod 2) whenever { a » M c {0,1}, it is easily seen that TA = A + (a 2 ,... »a^,,a1) (mod 2 ) .
With each fc-tuple A = (ai,...,a,) we associate the polynomial
k-l k-2
PAx) = a\X + a<i& + ... + CL-,.X + a,
in the indeterminate x. Then we have, modulo 2,
k-l k-2
P
mA(x) = (ai+a2)x + (a2+a3)x + ... + (a^.+a^x + (a-^+ai)
k
= (ar+l)P-(ar) + a\{x +1).

If we now take remainders on division by x +1, we can write this as

PT/x) E (x+l)PA(x) (mod 2 , arfe+l).

Repeating the operation then gives

P (x) E (a:+l)nP (a?) (mod 2, ar*+l) for n > 0. (1)

A
^A
Now O + 1 ) 2 = #2+2a:+i E a?2+i (mod 2 ) , so t h a t an easy induction gives

(x + l ) 2 E x2 +1 (mod 2 ) . (2)
s
Using (2) in (1) with n := 2 yields
9s k
P (x) E (x + l)PA(x) (mod 2, a +1).
T2 A
Iffe= 2 s , then
P , (x) = (xk + DP Ax) = 0 (mod 2, a^+l),
A
'TA
and so T^4 = 0
On the other hand, if k is not a power of 2, let A be (0, ... 9 o, l), so that
p (a?) = 1- If f'A = 0 , then by (1) we have (x+l)n = 0 (mod 2, aM-l). Multiplic
by (a?+l)2 ~n, where s is an inteqer such that 2s >rc,then yields, from (2),

x2 + 1 = (x + l ) 2 = 0 (mod 2, tfk+l). (3)

Since fc is not a power of 2, the division algorithm yields

2s = kq + r9 where 0 < r < k
 - 265 -
f o r integers q and r . Then

x2 + l E (xk)qxr + l = a?r + i £ o (mod 2, ^ + 1 ) ,

which contradicts (3). Hence (o,...,o,l) is not desirable.
We now return to the case in which the entries of A are arbitrary nonnegative
integers and k = 2 s . Let A1 be the k-tuple whose entries are the remainders on
dividing the entries of A by 2. The conclusion of the first part of the proof,
namely that T A1 = 0, can be interpreted as stating that all entries of T A are even,
so that \TA is an admissible fc-tuple. Repetition of the argument shows that
(1/2^)2^ A is an admissible k-tuple for every nonnegative integer m. Let a be the
maximum entry in A. Then the maximum entry in TA is not larger than a, and so the
maximum entry in 'f1 A is not larger than a. But every entry in "f1 A is divisible
by 2m. If m is chosen so that 2m > a5 then every entry in "f1 A must be o, so that
A is desirable, and its order does not exceed mk.
Acknowledgments. Many thanks to Martin Gardner, who responded immediately to
a request for references, which enabled the search tree to be extended. Thanks are
also due to Laszlo Csirmaz, who supplied references [6] and [7].

REFERENCES
Letters in brackets at the end of a reference indicate that the similarly
lettered items from the list of properties of T given above are proved or mentioned
in the article or book.

1. C. Ciamberlini and A. Marengoni9 "Su una interessante curiosity numerica",

Periodico di Matematiche, (4) 7 (1937) 25-30. CDj 9 E s G, H]
2. G. TJekuthiel Ginsburg], "Curiosa 30. An interesting observation", Scripta
Mathematiaa, 5 (1938) 135. rDi 3 Report on ["13.
3. Benedict Freedman, "The four number game", Sovipta Mathematical V\ (1948)
35-47. TD l9 E, G3
4. Moshe Lotan, "A problem in difference sets", American Mathematical Monthly,
56 (1948) 535-541. rD x , E, F, G9 H (k = 4)3
5. Boris A. Kordemskii, Matematicheskaja Smekalka, 5th ed., Moscow, 1958,
problem 345x, pp. 302-304; solution pp. 555-557. TD l9 F3
6. D.0. Shkljarskii, N.N. Chencov, and I.M. Jaglom, Izbvannye Zadachi i Teovemy
Elementarnoi Matematiki, 3rd ed., part I, Moscow, 1959, problem 89: statement, p. 28;
solution, pp. 177-178. rDj, E3
7. D.0. Shklarsky, N.N. Chentzov, and I.M. Yaglom, The USSR Olympiad Problem
Book: Selected problems and theorems of elementary mathematics, translation of [63
by John Maykovich, edited and revised by Irving Sussman, Freeman, San Francisco, 19629
problem 89: statement, p. 22; solution9 pp. 166-168. fDj, E3
 -.266 -
8. Roland Sprague, Recreation in Mathematics, Blackie, 1964; Dover, 1963;
item 4, pp. 5, 26-28. ["I've not seen this.]
9. A. Lindgren, "Mathematical note 319", Australian Mathematics Teacher, 22
(1966) 61-63; answer to question proposed to the Problem Bureau of The Mathematical
Association (New South Wales Branch), which appeared in Australian Mathematics
Teacher 19 (1963) 25. [dx , D 2 , E, (F), H (k = 4)]
10. John M. Hammersley, "On the enfeeblement of mathematical skills by 'modern
mathematics1 and by similar soft intellectual trash in schools and universities",
Bulletin of the Institute of Mathematics and Its Applications,4 (1968) 68-85, esp.
problem 6, p. 84. TProposes Dx (all k), E (k = 3), F (k = 4)]

11. L. Carlitz and R. Scoville, "Solution to problem 69-1°. Sequences of absolute

differences", SIAM Review, 12 (1970) 297-300; proposed by John M. Hammersley, ibid.,
11 (1969) 73-74. rD2 , D 3 , E, F, (G), H, I]
12. Ross Honsberger, Ingenuity in Mathematics, New Mathematical Library, No. 23,
1970, pp. 73-74, 80-83, and 86. [dx (k e {4,8})]
13. Boris A. Kordemsky, The Moscow Puzzles, translation of some edition of [5]
by Albert Parry, edited by Martin Gardner, Scribner8s, New York, 1972, problem 342(A):
"Repeat the operation", p. 159; solution, pp, 292-293. TDi, F]
14. Elwyn R. Berlekamp, "Design of slowly shrinking labelled squares", Mathe-
matics of Computation, 29 (1975) 25-27. \k = 4 : D!, F, G]
15. Hans Helmut Lemmerich, "Ouadrupelfolgen", Praxis der Mathematik (1975), pp.
219-223. ri've not seen this.]
16. M. Burmester, R. Forcade, and E. Jacobs, "Circles of numbers", Glasgow Mathe-
matical Journal, 19 (1978) 115-119. CD!, E, F, G, H]
17. Robert Miller, "A game with n numbers", American Mathematical Monthly, 85
(1978) 183-185. TDX, E, G]
18. P. Zvengrowski, "Iterated absolute differences", Mathematics Magazine, 52
(1979) 36-37. [Dx, D 2 , E, G]
19. Michel Dumont and Jean Meeus, "The four-numbers game", Journal of Recreational
Mathematics, 13 (1980-1981) 89-96.. TDX, E, F, G, H (k = 4)] An interesting sketch
shows the orders of 4-tuples (o, a, 3, 1), where a and 3 are rational and o< a < 3 < l .
20. Richard V. Andree, "An interesting recursive function", this journal, 8 (1982)
69; and "Postscript to ...", ibid., p. 96. ["Conjectures D ls E]
21. William A, Webb, "The length of the four-number game", Fibonacci Quarterly,
20 (1982) 33-35. [k = i+: Di , D3 , G? H]
22. Fook-Bun Wong, "Ducci processes", Fibonacci Quarterly, 20 (1982) 97-105,
CDlf E] Generalization in which [a^a^ is replaced by g(a€9aUl\ where g is some
given function. The most interesting example is the one in which g(a,b)= <f>(a) +^(2?).
Math. Dept., The Ohio State University, 231 West 18 Ave., Columbus, Ohio 43210.
 - 267 -

SOME NOTABLE PALINDROMIC TRIANGULAR NUMERALS

CHARLES W. TRIGG

The numerals representinq a triangular number, Tin) = nin+l)/2, have different

appearances in different bases, except that Til) = 1 in every base. For example,
T(5) = 10-. _ = 15 t = 17 . . = 23 . .
fifteen ten eight six

Numbers palindromic in several bases*

Within the ranges studied, ten triangular numbers are palindromic in two systems
of notation, and three are palindromic in three bases. They are (the n in Tin) being
always in base ten unless otherwise indicated):

Tib) = 1111,. = 33^ Tilk) = 1221^ = 151

• u,
two four four eight
Tie) = 10101 = 111 £ TilS) = 333 = 171
two four seven ten
TiS) = 121^. Ti36) = 22122^ = 666^
eight
five four ten
Ti9) = 101101 ^(90) = 333333^ = 7777 . ,
= 5 5
• 1.4- eight
two eight four
Til2) = 303£. = 141 T(132) = 21112 . ^
eight
= 8778^
five seven ten

mo) =313^four = 131 .

six
= 55^
ten
T(2*) = 606 = 363 .
eight
seven nine
T(25) = 101000101 = 11011- = 505 , .
eight
two four
Observe that T(2*\) is palindromic in three consecutive bases.
In Ti5) and T(90), both of the given representations are repdigits, and each
digit is b-l, one less than the base b. In general, if two equal repdigits in bases
v and s have digits r^T and s-l occurring p and q times, respectively, then

ir-DliiP-D/ir-Dl = (s-l)C(^-l)/(s-l)]9
so 2^ = s^. It follows that rand s are powers of some common factor t9 with
iPu = tqv and p/q = v/u. If the repdigits are triangular, then tpu-l = n(n+l)/2,
so 20?u-±) is the product of two consecutive integers.
Workinq in the decimal system, the product, P, of two consecutive integers ends
in o, 2, or 6, and P = o, 2, 3, or 6 (mod 9 ) . If a? has the form 3k, thenP =2(ar/-i) = 7
(mod 9 ) , and if x ends in 5, P ends in 8. Hence for x < 10 it is necessary to examine
only x - 2 and x - 7, since x - h and x = 8 are special cases of x = 2. For p < io 8 ,
there are only two solutions:

P = 2(2^-1) = 5-6 and P = 2(2 12 -1) = 90-91.

 268
These lead to

1111. 33
two four

and
111111111111 = 333333^ 7777 . u = 15 15 15 . , 63 63
sixty-four9
two four eight sixteen
one of the situations noted above and an expansion of the other.
Numerals palindromic in several bases.
A triangular number has the same appearance in all systems of numeration with
bases greater than the number, which will be a "digit" in that system.
There are also a few "look-alike" palindromic numerals of different magnitudes,
since they are expressed in different bases. Some of these are

T{IO^ ) = ii , T{3-. ii_„..., T{H = 11

five ) five9
two two
and all fix.) = 11, when b (x-l)(x+2)/2;

Till. ) 33 T{10 . ) 33
four four

and all T(x.) = 33, when b = (x+3)(x-2)/§, an integer-,

T(H = 44eight9
six six ^"eight5
T ljL
( • *,J
eight = 55
eight9
r 10
( teJ = 55^. ,
ten
T(12. ) = Ill- T(l4 . )
nine
= 111nine
. ,
four
four
T(l01 ) = 1111,. T{W . ) = 1111 .
v
two' two» nine
nine
T(llO^ ) = 10101,. , = 10101
two two ^"three* three'
7
5 (12143£8 ) = 102121201 T(l3401 . _ J = 102121201 . _ .
five five9 eight eight

2404 Loring Street, San Diego, California 92109.

HOMAGE TO HERMAN NY0N

Over the years, some of our long-term subscribers have become more than
mathematical comrades, they have become personal friends. Such was Herman
Nyon. We have just been informed by his wife, Beatrix Nyon, that Herman died
unexpectedly on October 3, 1982, in Paramaribo, Surinam.
Since he retired in 1972, he derived much enjoyment from problem-solving
and recreational mathematics, and he was generous in his moral and financial
support for this journal. A few short weeks before his death, he renewed his
subscription for 1983, added a generous supplement, and wrote of his intention
to include Crux in his will. His death leaves us all greatly saddened. We
extend our deepest sympathy to his wife and family.
 - 269 -
THE OLYMPIAD CORNER: 39
M.S. KLAMKIN

We give below the problems of the 1962 9 1963, and 1964 Peking Mathematics
Contest (Grade 12), Paper II. The translations were provided by Andy Liu. As usual,
I invite readers to send me elegant solutions.

1962
1, Evaluate
. (777+1)! (777+2)1 (777+n)!
m + + + +
' 1! 2! '•• n\ '

2, Six circles in a plane are such that the center of each circle is outside
the other circles. Show that these six circles have empty intersection.
3, A car can carry fuel which will last a distance of a. A distance d > a is
to be covered with no refueling station in between. However, the car may go
back and forth transporting and depositing fuel en route. What is the most economical
scheme to get to the destination if d = 4a/3? What if d - 23a/l5?
i\t A group of children forms a circle and each child starts with an even number
of pieces of candy. Each child then gives half of what he has to his right-
hand neighbor. After the transaction, if a child has an odd number of pieces, he or
she will receive an extra piece from an external source. Show that, after a finite
number of such steps, each child will have the same number of pieces of candy.
1963

1, A polynomial P(x) with integral coefficients takes on the value 2 for four
distinct integral values of x. Show that P(x) is never equal to 1, 3, 5, 7,
or 9 for any integral value of x,
2, Nine points are randomly selected inside a square of side 1. Show that three
of the points are the vertices of a triangle of area at most 1/8,
3, Given are 2n+3 points in the plane, no three col linear and no four concyclic.
Is it possible to construct a circle passing through three of the points so
that exactly half of the remaining 2n points lie inside the circle? Justify your
answer.
£|, A set of 2n objects is partitioned into a number of subsets. A move consists
of transferring from one of the subsets to an equal or smaller subset a number
of objects equal to the cardinality of the second subset. Prove that, irrespective
 - 270 -
of the initial partition, all the subsets can be combined into a single set by a
finite number of moves.
1964

1, ABC is a triangle in which angle A is nonacute. Let B 1 DEC 1 be any inscribed

square of triangle ABC with Bx on AB, C! on AC, and DE on BC. Let B 2 D 1 E 1 C 2
be any inscribed square of triangle AB1C1 with B 2 on ABi, C 2 on ACi, and D ^ on B ^ .
The process is continued in the same way for a finite number of steps. Prove that
the sum of the areas of all the inscribed squares is strictly less than half the area
of triangle ABC.
2, Let (a x , a 2 , ..., a , ...) be a sequence of positive real numbers such that
a2 < a -a A for all n . Show that a < 1/n for all n .
n n n+1 n
3, A round track has n fueling stations (some possibly empty) containing a
combined total of fuel sufficient for a car to travel once around the track.
Prove that, irrespective of the initial distribution of fuel among the stations, it
is always possible for a car with an empty tank to start from one of the stations
and complete a round trip without running out of fuel on the way.
£}, Show that a circle of diameter D can cover a parallelogram with perimeter
2D. Show that the same circle can cover any planar region with perimeter
2D.

We now present solutions to some problems that have appeared earlier in this
column.
4, [1981: 42 ;1982: ioo] Prove that 2:{l/(^1i2.. .i, )} = n, where the summation
is taken over all nonempty subsets H i 9i2>...ȣ,} of {l,2,... 9 n}.
II. Solution by David Singmaster3 Polytechnic of the South Barik3 London,
England.
We have
m/(Mi2...^)} = (i+j)d+|)...d+£) - i
= 2 . 3# % n+1
1*2 n

= n.

3, [1981: 43 : 1982: 102] Let p be a prime number and n a positive integer. Prove
that the following statements (a) and (b) are equivalent:
(a) None of the binomial coefficients (?) for k = o,l,...,n is divisible by p.
 - 271 -
(b) n can be represented in the form n = psq - l, where s and q are integers,
s > 0, 0 < q < p .
II. Comment by David Singmaster, Polytechnic of the South Bank, London, England,
This result follows from Corollary 14.1 given in my paper "Divisibility of
binomial and multinomial coefficients by primes and prime powers" 9 18th Anniversary
Volume of the Fibonacci Association.
1, [1982: 99] Proposed by Jack Brennen, student, Poolesville, Maryland.
Sum the series
00 ,«

A (36^-1)2-
^-l
I. Solution by the proposer.
Ifs is the required sum, then we have

2S = +
^XT^TV (6t+i)4= ^ + ^ + II2 * &
Now, from the well-known value of C(2),
TT 2
T
2* 3 Z x ••" = "e"9
we obtain successively
l . i +i 2
2* + IIF
F + 6e s" +
••• =
24 9
2
1. . 1 . . _ 1., *2 *2
1 +
s2" + s2" + ••• =
Te " 2? =
T!
1 1 1 TT 2
+
F g2" +
15 2 " +
"•• =
72s

and
1 1 1 1 TT 2 7T 2 TT 2
= 9
25 + 1 = 1 + 57 + ^ + jp- + jjz + ... = T " 72 T

from which
w2

s= 18

II. Solution by Henry E. Fettiss Mountain View, California.

Consider t h e double e q u a l i t y

O 00 00 ,n n

TT2 = y 1 = 1 2 y iff.,
i=l
The first equality is well known (see [1] or [2]) 9 and the second merely transforms
 - 272 -
the series into a form convenient for our purpose. With z = 1/6, we obtain from the
above

from which we find the required sum to be (TT 2 -9)/18.

REFERENCES

1. Einar Hille, Analytic Function Theory, Chelsea, New York, 1973, Vol. I,
p. 261.
2. V. Manqulus, Handbook of Series for Scientists and Engineers, Academic
Press, New York, 1965, p. 78.

2, [1982: 99] Proposed by Gregg Patruno3 student3 Princeton University.

How many eauations must be considered in order to maximize a qiven differ-
entiate function F{x\ 9X2,...9x ) in the region o < x. < l, i = l,2,...,n, by dif-
ferential calculus?

Solution by the proposer.

In addition to the 2n endpoint equations, there are 2 (?) (n-fc)-dimensional
cubes to be considered for k = o,l,...,n-l. Thus, in general the number of equa-
tions to be considered is

Now let

Fix) = ^l (l)(2x)k = (1+2*)n .

k=0
Then we have

Sn = 2n + {nF(x) - xFlW}x=1 = 2 n + n.3n_1.

3, [1982: 99] Proposed by Gregg Patrunos student, Princeton University.

A quick proof that the rationality of p, q, and Jp + /q implies the ration-
ality of /p is furnished by the identity

/p+Jq

Prove in similar fashion that if p, q9 r9 and Jp + /q + /r are rational, then so is

 - 273 -
Solution by the proposer.
The desired result follows from

2ps2 + j-(s2-p-q-r)2 - (qr+rp+pq)

P ~ s(s2+p-q-r) 9

where s = /p + /q + /r.

Comment by M.S.K.
It would be interesting to have a "simple" proof of the following generali-
zation: if p., i - 1,2 n and i Jp7 are all rational, then so is each /p7.

t\t [1982: 99] Proposed by Brian Hunt, student, Montgomery Blair H,S,, Silver
Spring3 Maryland,
If a.9b. > o for i = l,29...,rc and
n n
I a. = I b. = 1,
' A I* • A I*

^=l ^=l

prove that, for all m > l,

Solution by Jeff Soesbe, student. Keystone School, San Antonio, Texas,

With all summations for i = l,2,...,n, we have, by the power mean inequality,

la. 1 m lb. A -m
and n K
n' '
n n

hence
, m^m+1 ^ 1-m2 and
f^-mK171'1
[tb^ )
^
>n
m2-l
(laj) >n

(la™) • (Zb™) * 1.

as required.
5, [1982: 100] Proposed by Noam D. Elkies, student, Stuyvesant H.S,, New York,
N.Y,
One solution of the Diophantine equation
7x2 + 8a: - 3 = y2

is (x9y) = (-3,6). Are there solutions for which x is positive?

 -m-
Solution by the proposer.
We first note that the niven equation is equivalent to
(3a:)2 + y2 = (Mar-l)(iur+3). (1)
Our proof will be based on the following well-known result |~l]: a positive
inteqer N is representable as a sum of two squares if and only if, in the prime
power canonical factorization of N9 all primes of the form 4n+3 occur (if at all)
to an even exponent. In particular, if an odd inteqertfis representable as a sum
of two squares, then, in any factorization N = ab into two relatively prime factors,
each of the factors a and b must be of the form 4w+l. For the product of two numbers
of the form 4n+l, or of two numbers of the form *m+3, is of the form *m+l; and,
because a and b are relatively prime, each prime factor of N of the form 4n+3 must
occur (if at all) an even number of times in a or an even number of times in b.
It is now immediately apparent that (i) has no solution (x9y) with x > o,
since the two factors on the right are relatively prime and each is of the form
4n+3. However, there may be solutions (x9y) with x < o, for then each factor in (1)
would be of the form 4n+i 9 and in fact the proposal gives one such solution. •
It follows from the above discussion that any equation equivalent to
(rx)2 + y2 = (**sx+l)(iitx+3),
where r3 s> t are positive integers, has no solution in integers (x9y)9 either for
x > o or x < 0.

REFERENCE

1. William J. LeVeque, Topics in Number Theory, Addison-Wesley, Readinq, 1956,

Vol. I, p. 126.
R, T1982: 1001 Proposed by Noam D. Elkies, Stuyvesant H.S.3 New York3 N.Y.
Let Si,S29S39Sh be four spheres in 3-space such that Si9S29S3 intersect
at points Ai* and B^; Si9s29S^ intersect at A 3 and B 3 ; Sx 9S3 ,£4 intersect at A 2 and
B 2 ; and, finally, S29S3,Sn intersect at Ax and Bx. If P is any point distinct from
A.,B. ( i - 1,2,3,4), prove that the circumcenters of triangles PAxBj, PA 2 B 2 , PA3B3,
and PA4B4 are coplanar.
Solution by the proposer.
[This solution tacitly assumes that P t A.B. for i = 1,2,3,4, a restriction
which should have been mentioned in the proposal. (M.S.K.)l
The four spheres have a radical center which will be denoted by R. Then we
have
RAx-RBx = RA 2 «RB 2 = RA 3 «RB 3 = RAi^RBi* = k2;
 - 275 -
hence, for i - 1,2,3,4, R,A.,B. are collinear and the circumcircles of PA.B. are
invariant under inversion with respect to a sphere with center R and radius k.
Under the inversion, P is transformed into some other point P \ The four circumcen-
ters must all be equidistant from P and P'; hence they must lie on the perpendicular
bisectinq plane of PP'. By continuity, this argument can also be used if P lies on
the sphere of inversion.
Editor's note. All communications about this column should be sent to Professor
M.S. Klamkin, Department of Mathematics, University of Alberta, Edmonton, Alberta,
Canada T6G 2G1.

THE PUZZLE CORNER

Puzzle No. 23: Homonym (5; 3)

"We sailed (the skipper wrote his log)

A PRIMAL line."
"We formed (the sailor said) for grog
A FINAL line."

Puzzle No. 24: Rebus (10)

+H
If this puzzle
Strikes you dumb,
This is what
You may become.

Puzzle No. 25: Rebus (6)

f(D
This graph, in mathematics view.
In art they'd say, "It's ALL, won't do."
ALAN WAYNE, Holiday, Florida

Answer to Puzzle No. 22 ["1982: 260]: 057 = OAK,

NUTCRACKER = 281345376'+ = 53042 2 = ACORN 2 .

MATHEMATICAL SWIFTIES
"Two lines always intersect", Tom proclaimed projectively.
M.S. KLAMKIN
"Its value is approximately 3.14159", Tom murmured piously.
EDITH ORR
 - 276 -

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
April 1, 1983, although solutions received after that date will also be considered
until the time when a solution is published.

781, Proposed by Alan Wayne3 Holiday3 Florida.

The Philosophic Father
Your brother's first to bathe, my son,
So do not look so W R Y .
You'll have to use the T U B in turn,
There's room for only O N E .
There's very little room, I guess.
Aye, there's the R U B! But try.
Use soap on cloth to R U B . Don't spurn
My counsel; answer "Y E S".
In time, you'll find it's true—
Much later, by the B Y E —
As ashes resting in an U R N ,
There will be room for T W O .
Regard the preceding three patterns of three capitalized words as interdepen-
dent arithmetic additions in the decimal system. Restore the digits.

782! Proposed by H.S.M. Coxeter, University of Toronto.

(a) Sketch the plane cubic curve given by the parametric equations

x - a(3-y) 2 9 y = 3(Y~a) 2 9 « = y(a-&)2, a + 3 + y = 09

where (x9y,z) are barycentric (or area!, or trilinear) coordinates, referred to an

equilateral triangle. In what respect do its asymptotes behave differently from
those of a hyperbola?
(b) Eliminate the parameters a,3,y so as to obtain a single equation

#3+z/3+33 + a(x2y+x2z+y2z+y2x+z2x+z2y) + bxyz = 0

for certain numbers a and b.

(c)* What equation does the curve have in terms of polar coordinates?
 - 277 -
783, Proposed by R.C. Lyness, Southwold, Suffolk, England,
Let n be a fixed natural number. We are interested in finding an in-
finite sequence (v0,vi 9v29...) of strictly increasing positive integers, and a
finite sequence (u0,ui,...9u ) of nonzero integers such that, for all integers
m > n,

UQV + U\V , + ... + U2V „ = UnV2 + U\V2 . + ... + U V2 . (1)

u l u L
m m-1 n m-n m m-1 n m-n
(a) Prove that (l) holds if

u = coefficient of xr in (l-x)n
and
coefficient of xr in (l-x) n
.
(b) Find other sequences (u ) and (v ) for which (l) holds.
r r
784 t Proposed by Stanley Rabinowitz, Digital Equipment Corp., Merrimack,
New Hampshire,
Let F = a./b.9 i = 1,2,... 9m9 be the Farey sequence of order n9 that is, the
ascending sequence of irreducible fractions between o and 1 whose denominators do
not exceed n. (For example,

5 9
4 5' 4' 3 9 5' 2 s 5' 3' 4 9 5' l'9
with m - 11.) Prove that, if P0 = (0,0) and P. = {a.9b.)9 i - 1,2,. ..,m, are lat-
tice points in a Cartesian coordinate plane, then P0P!...P is a simple polygon of
area (m-l)/2.

785 a Proposed by J. Chris Fisher, University of Regina.

Suppose a closed differentiate curve has exactly one tangent line par-
allel to e\/ery direction. More precisely, suppose that the curve has parametriza-
tion f(e): [O,2TT] -*• R2 for which
(i) dv/de = ( H e ) cose, H e ) sine) for some continuous real-valued function
H e ) , and
(11) F(e) = i/(e+7T).
Prove that the curve is described in the clockwise sense as e runs from o to 7r.
786 • Proposed by 0. Bottema, delft, The Netherlands.
Let ri,r2,r3 be arbitrarily chosen positive numbers. Prove that there
exists a (real) triangle whose exradii are ri9r2*r39 and calculate the sides of this
triangle.
 - 278 -
787, Proposed by J. Walter Lynch3 Georgia Southern College.
(a) Given two sides, a and b9 of a triangle, what should be the length
of the third side, x9 in order that the area enclosed be a maximum?
(b) Given three sides, a9 b9 and c9 of a quadrilateral, what should be the
length of the fourth side, x9 in order that the area enclosed be a maximum?

788, Proposed by Meir Feder3 Eaifa3 Israel.

A pandigital integer is a (decimal) integer containing each of the ten
digits exactly once.
(a) If wand n are distinct pandigital perfect squares, what is the smallest
possible value of \Jm-Jn\?
(b) Find two pandigital perfect squares m and n for which this minimum value
of \Jm - Sn\ is attained.

789, Proposed by H. Kestelman, University College, London3 England.

If A and B are square matrices (any orders), then they have a common
eigenvalue if and only if AX - XB for some X * 0.
790, Proposed by Roland H. Eddy 3 Memorial University of Newfoundland.
Let ABC be a triangle with sides a9b9a in the usual order, and let I ,
Z,,Z and Z',Z',Z' be two sets of concurrent cevians, with I ,Z,,Z intersecting
D Q a D Q a b a
a9b9o in L,W,N, respectively. If
Z n Z' = P, Z, n Z ' = Q 9 Z nZ'=R9
a b b c ^ c a

prove that, independently of the choice of concurrent cevians z',z',Z!, we have

AP.BQ.CR _ aba
PL QM RN " BL • CM • AN ~ a s
with equality occurrinq just when Z ,Z, ,Z are the medians of the trianqle.
a D c
(This problem extends Crux 588 ri981: 3061.)
7fi?, T1982: 2091 Correction. Part (b) should read as follows:
(b) If a = 5 and 3abc/*+(a3+b3+c3) = 5/24, determine b and c, given that
they are integers.
J'C s'c &

(continued from page 294)

identify it. In that reference, the problem, strictly equivalent to our own, was
to find the common area accessible to four goats tethered at the vertices of a
square garden. The editor looked into his crystal ball and found one possible
source for this solution: Mathematical Wrinkles , by S.I. Jones, published by the
S.I. Jones Co., Nashville, Tennessee, in 1930.
 - 279 -
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.

%, [1975: 85 ; 1976: 34] Proposed by Leo Souve5 Algonquin College.

(a) Determine, as a function of the positive integer n9 the number of
odd binomial coefficients in the expansion of (a + b ) n .
(b) Do the same for the number of odd multinomial coefficients in the ex-
pansion of (a\ + a2 + ... + a ) n .
II. Solution of part (a) by Leroy F. Meyers> The Ohio State University.
We will use the congruence
(x + l)2 = x2 + 1 (mod 2), (l)
which is easily shown by induction to be valid for s = 0,1,2,... .
Let n be a nonneqative inteqer whose binary expansions with only the nonzero
terms listed, is
n = 2 S 1 + 2 S 2 + ... + 26m,

where o < si < s2 < ... < s . Then by (l) we have

n 2Sl 2S? 2Sm

(x + 1) = (x+l) (x+l) ...(x+ir
9si 9s2 9sm
= (xz +l)(x +l)...(x +1) (mod 2).
Now all the terms obtained by multiplying out the last line are distinct and there
are exactly 2™ such terms, each with coefficient l. This can be interpreted as
saying that:
The number of odd coefficients in the expansion of (x+l) by the binomial
theorem is 2 3 where m is the number of 1 fs in the binary expansion of n.
it & &

P39. T1981: 146 ; 1982: 145] Proposed by Hayo Ahlburg, Benidorm, Alicante,
Spain.
If x + y + z = 0 , prove that
x5+y5tz5 _ x3+y3fz3 a x2ty2tz2
5 3 ° 2

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

In a comment following the solution of this problem [1982: 148], the editor
asked for a solution of the following related problem:
r r r ~p
Let S - x +y +z +u . Find all integer pairs {m,n} such that
 - 280 -

s s s
m+n _ JH 9 _n ,. x
m+n m n
is meaningful and true for alt quadruples (x,y,z9u) for which S\ = 0^ having given
that one solution is {m,n} = {3,2}.
I show that im,n} - {3,2} is the only solution. Another solution {m9n}9 if
one exists,must remain valid whether or not u = o, and it follows from my solution
of the original problem that the only candidate is {m9n} = {5,2}. However, in
this case (l) does not hold for the quadruple (x9y9z9u) = (3,-1,-1,-1).
Editor's comment.
It would be interesting to know if there are other solutions {m9n} valid
for all quadruples (x9y9z9u) for which Si = o and xyzu * o. Readers will recall
that Klamkin had found that {m9n} = {3,-1} was also a solution of the original
problem when the restriction xyz * o was added.

665. C1981: 205 ; 1982: 221] Proposed by Jack Garfunkel, Queens College,
Flushing3 N.Y.
If A,B,C,D are the interior angles of a convex quadrilateral ABCD, prove that

/ 2 E C O S — - < ZCOt^

(where the four-term sum on each side is cyclic over A,B,C,D), with equality if
and only if ABCD is a rectanqle.
II. Generalization by M.S. Klamkin3 University of Alberta.
The published solution of this problem depended in part on the convexity of
cote for acute angles e. We use the same property, together with the convexity
of -cose for acute anqles e, to prove a generalization.
Let AiA2... hn be a convex n-qon. We apply successively to each vertex A^
weights xv > o and yr > o such that ix = ly = 1 (summations throughout are for
r - l,2,...,n) to form the weiqhted sums

B =X AT +X . Ao + . . . + X .k ,
l z
r r r+1 r+n-1 n r = 1 2 ... n
Cr = y r Ai
l + yr+1,A*2 + ... + 2/r+n-1„An ,

where all subscripts are reduced modulo n. By Jensen's inequality for convex func-
tion, we have

Icot^>ncot^f^ and J cos Cf <- n cos &£&.

 - 281 -
Hence

with equality if and only if the polygon is equiangular.

The original problem is the special case for which n - 4, (x ) = (1,0,0,0),
and (yp) = (J,J,o,o).
* * *

677. C1981: 240] Proposed by E.J. Barbeaus University of Toronto.

Prove that there do not exist four distinct square integers in arithmetic
progression.
(This problem is apparently due to Fermat (1640), but the proofs mentioned in
Dickson's History of the Theory of Numbers (Vol. II, p. 440) either have obscure
references, are unclear, or appear unsatisfactory.)

The problem was solved by the proposer. Comments were received from W.J. BLUNDON,
Memorial University of Newfoundland; ANDY LIU, University of Alberta? BOB PRIELIPP,
University of Wisconsin-Oshkosh; and KENNETH M. WILKE, Topeka, Kansas.

Editor's comment.
All the references given below were supplied by readers. The most satisfactory
proof of this theorem of Fermat appears to be that of Pocklington [l], which can be
found in Sierpinski [2]. It is based on the fact, also established by Pocklington,
that the equation

ar1* - x2y2 + y^ = s2

has no solution in natural numbers (x9y,z) apart from the trivial one (x9x,x2).
Essentially the same proof is given in Mordell T3].
The proposer noted that the problem had appeared in 1898 in the Monthly [4].
No complete proof was forthcoming at that time (for natural numbers) but G.B.M.
Zerr found the imaginary solution

(6 + / ~ L i ) 2 , 529 (6-/^TT)2, {/(25+^g6l)/2 - / ( 2 5 - / 6 9 6 1 )/2 } 2

,

with another solution obtainable by taking complex conjugates.

REFERENCES

1. H.C. Pocklington, "Some diophantine impossibilities", Proo. Cambridge

Philos. Soc3 17 (1914) 108-121.
2, Wacjaw Sierpinski, Elementary Theory of Numbers, Hafner, New York, 1964 9
pp. 74-75.
 - 282 -
3. L.J. Mordell, Diophantine Equations, Academic Press, New York, 1969, pp.
20-22.
4. Problem 62 (proposed by John M. Arnold), American Mathematical Monthly,
5 (1898) 180.
* * ft

678, [1981: 240] Proposed jointly by Joe Bellinger and Ferrell Wheeler,
students, Texas ASM University, College Station, Texas,
For a qiven fixed integer n > 2, find the greatest common divisor of the in-
tegers in the set {an-a | aeZ}, where Z is the set of all integers.

Solutions were received from FRIEND H. KIERSTEAD, JR., Cuyahoga Falls, Ohio;
and the proposers. Comments were submitted by W.J. BLUNDON, Memorial University
of Newfoundland; and STANLEY RABINOWITZ, Digital Equipment Corp., Merrimack, New
Hampshire. In addition, two incorrect solutions, based on a misreading of the prob-
lem, were received.

Editor1 s comment.
The two comments received alerted the editor to the fact that essentially the
same problem appeared, with two proposers different from ours, in the August-Septem-
ber 1981 issue of the American Mathematical Monthly (Mo.E 290l). We received the
problem on September l, 1981, and published it immediately in the October 1981 issue
of Crux, which was then under way. So, although the problem was published almost
simultaneously in the two publications, it is clear that it must have been in the
Monthly "pipeline" long before we received it. Our proposers must have been unaware
of the impending publication of the problem in the Monthly when they sent it to us.
It seems only fair to let the Monthly carry the ball on this problem. When
a solution appears in the Monthly, a few months hence, we will reopen the problem
here and qive our readers the reference.
* * a
679, C1981: 240] Proposed by Bob Trielipp, University of Wisconsin-Oshkosh.
The equation x2+y3 = 2zh has the solutions (1,1,1) and (239,1,13). Does
it have infinitely many solutions (x9y9z) consisting of relatively prime positive
integers?
I. Solution by Friend H. Kierstead, Jr., Cuyahoga Falls, Ohio.
With considerable expenditure of time on a fast computer, I found only the
following six solutions with z < 2770 and y < 416000:
(1, 1, 1 ) , (4479, 161, 59),
(13, 7, 4 ) , (4445, 263, 66),
(239, 1, 13), (490013, 11047, 944).
 - 283 -
In view of these results, it is likely that solutions are few and far between,
although I would hesitate to say that the number of solutions is finite. It also
appears that the solution (239, 1, 13) is atypical, and it is likely that if an
infinite family of other solutions is found, each will satisfy the relation x>y>z.
There are also many solutions like

(64, 16, 8 ) , (875, 25, 25), (4174281, 20079, 1890)

in which x9y9z are not relatively prime. Since the given equation is not homoge-
neous9 these do not represent multiples of more primitive solutions.
II. First part of the solution by Kenneth M. Wilkes Topekas Kansas,
We first find all the solutions for which y = l. The given equation then
reduces to
x2 = 2zh - 1.

According to Mordell [l], it has been known for two centuries that (x, z) = (l, l)
and (239, 13) are solutions of this equation, and it was proved in 1942 by Ljung-
gren [2] that these are the only positive integer solutions. Mordell adds that
"the proof is exceedingly complicated". Thus the two solutions given in the pro-
posal are the only ones for which y = 1.

Editor's comment.
If the proof for y = l is "exceedingly complicated", that for y * l can hardly
be less so. The rest of Wilke's solution was long and complicated, though not
"exceedingly" so, which leads us to suspect that his solution is not complete. He
found tv/o infinite families of solutions in relatively prime integers x9y9z with
sc>09y*l9z>0 (thus allowing y to be negative). They are

ix = \rk - 2^sh\(rB + 8 1 6 2 ^ + 576s8)

Q
\y = -r + 3362*^ - 576s8 (1)
Iz = 6rs(rh + 24s4)

and
ix = |32^ - Ssh\(9r8 + eiBrk8k + 64s8)
B h h 8
\y = -9r + 336r s - 64s (2)
\z - §rs{3rh + 8^)

where (r, s) = 1, r is odd, (r9 3) = 1 in (l), and (s9 3) = 1 in (2).

Of the four solutions with y * l found by Kierstead, only (4445, 263, 66)
belongs to one of these families. So there are other families of solutions lurking
in the underbrush, and finding them all will probably be an "exceedingly complicated"
business.
 - 284 -
Finally, Wilke noted that families (l) and (2) each contain infinitely many
solutions with y > o. In family (l), these correspond to pairs (r9 s) for which

1.145721 « ^168-96/3 < - < ^168+96/J « 4.275891,

and in family (2) to pairs for which

o.661483 « ^ i g § s < | < ^ n i p s w 2-468687-

REFERENCES

1. L.J. Mordell, Diophantine Equations, Academic Press, New York, 1969, p. 271.
2. W. Ljunggren, "Zur Theorie der Gleichung x2 + 1 = Dyk", ^/z. Norske Vid.
Akad. Oslo, No. 51 (1942).
^ ^ ft

6801 C1981: 240] Ppoposecf by W.J. Blundon3 Memorial University of Newfoundland.

Find interesting sets of three distinct real numbers such that their
product is equal to their sum.
Editor's comment.
The following remarks were culled from the solutions of several solvers. If
{a9b9o} is a satisfactory set, then so is {-a9-b9-c}\ and {1,2,3} is the only
satisfactory set of natural numbers (see [1] and T2]). If a * o, then {o,a,-a} is
a satisfactory set.
To save space, in all solutions given below it will be tacitly assumed that
suitable restrictions are imposed upon the numbers concerned to make them real and
distinct, as the problem requires. Host solvers submitted several solutions, but
only the ones that the editor judged to be particularly "interesting" are given
below.
I. Solution by Clayton W. Dodge3 University of Maine at Orono.
It is easy to show that the set {a9b9c} satisfies abc - a+b+c if and only if
ab * l and c = (a+b)/(ab-l). So the most general satisfactory set is

{*•»• &) •
II. Solution by Stanley Rabinowitzs Digital Equipment Corp.3 Merrimaek3 New
Hampshire.
Krechmar [3] gives the following satisfactory set:

b-o a-a a-b \

1+bo9 1+ea' 1+abf '
 - 285 -
II3L Solution by J.A.H. Hunter, Toronto, Ontario.
The following set is satisfactory:
/ mfl
fe^r ^Pf J™^)-
IV„ Solution by Kesiraju Satyanarayana, Gagan Mahal Colony, Hyderabad, India.
If a+b+c = l, the following set is satisfactory:

}lbo9 %a9 \db\ '

V. Solution by Fred A. Miller, Elkins, West Virginia.

The following set is satisfactory:
iq_ _£__ ao \
\b> a-b> b*-db+c*f '

VI. Solution by Leon Bankoff, Los Angeles, California.

Elementary trigonometry yields the following satisfactory sets, where A s B s C
are the angles of a triangle:
CtanA, tan B 9 tan C}
and
|cot^ 9 c o t | , cot I } .

VII. Solution by the proposer.

I offer two satisfactory sets consisting of an arithmetic progression,
{a + /a z -3 5 a, a - /a z -3} 9

and a geometric progression,

Sa(a2-±) t a/(az+l)(az-3) a(a2-l) - a/(aztl)(az-3)
( 2 2
Also solved by FRIEND H. KIERSTEAD, JR., Cuyahoga Falls, Ohio; and KENNETH M.
Wilke, Topeka, Kansas.

REFERENCES

1. E.P. Starke,, Solution of Problem E 2262, (proposed by G.J. Simmons and

D.B. Rawlinson), American Mathematical Monthly, 78 (1971) 1021-1022.
2. M. Misiurewicz, "Ungeloste Probleme", Elem. Math., 21 (1966) 90.
3. V.A. Krechmar, A Problem Book in Algebra* Mir Publishers, Moscow, 197M-,
p. 25.
 286
681, C1981: 274] Proposed by J.A.H. Hunter, Toronto, Ontario.

Of all the girls that are so smart,

There's none like pretty Sally.
She is the darling of my heart,
And she lives in our alley.
And when our alley the sunlight dapples,
My darling SALLY
SELLS
RIPE .
APPLES
(With apologies to Henry Carey (c. 1687 - 1743).)

Solution by Charles W. Trigg, San "Diego, California.

Proceeding from the right, the columns establish the following eguations:
Y + E = lo, (1)
2L + P + 1 = E + l o f c , k € {0,1,2}, (2)
L + I + k = 10, (3)
A + E + R + 1 = P + 10m9 77? e { 0 , 1 , 2 } , (4)
2S + 77? = P + 10A.

If A = 2, then S = 9, P = o, and m = 2. However, the maximum for A + E + R + l

is 2+8+7+1 < 20, so A = l and 777 * 2. Hence, from (1), E * 0, l, 5, or 9. In (4),
if A + E + 1 =10, then R = P, so E * 8. In (2), if L = 0, then k = 0, and then (3)
gives the impossible I = 10. Thus L * 0.
We now try in succession the possible values of E: 2, 3, 4, 6, and 7. The
details for E = 2 are given below, with a star to indicate where the attempted solu-
tion fails.

W I (777)

(l)
(l) (1) 6 (solution!)

(2)

(2)

Proceeding in like manner with E = 3, 4, 6, and 7,no additional solutions

are found. Therefore, the unique solution is
61448
62446
9532 .
133426
 - 287 -
Also solved by E.,C. BUISSANT DES AMORIE, Amstelveen, The Netherlands; ALLAN
WM. JOHNSON JR., Washington, D.C.; SID KRAVITZ, Dover, New Jersey; J.A. McCALLUM,
Medicine Hat, Alberta; LISA PETROCCO, Erskine College, Due West, South Carolina;
STANLEY RABINOWITZ, Digital Equipment Corp., Merrimack, New Hampshire; RICHARD
RHOAD, New Trier H.S., Winnetka, Illinois; ROBERT TRANQUILLE, College de Maisonneuve,
Montreal, Quebec; KENNETH M. WILKE, Topeka, Kansas; and the proposer.
s't j'; ft

682. C1981: 274] Proposed by Robert C. Lyness3 Southwold3 Suffolk, England.

Triangle ABC is acute-angled and Aj is its orthic triangle (its vertices
are the feet of the altitudes of triangle ABC). A 2 is the triangular hull of the
three excircles of triangle ABC (that is, its sides are the external common tan-
gents of the three pairs of excircles that are not sides of triangle ABC).
Prove that the area of triangle A2 is at least 100 times the area of triangle Ai.
I. Solution by George Tsintsifas, Thessaloniki, Greece.
Let A = ABC, Ai = DEF (with D on BC, etc.), A 2 = A'B'C (with A' opposite A,
etc.), and let K9 #i, K2 be the areas of A, Ai, A 2 , respectively, with square brack-
ets denoting the areas of other triangles as required. (Make a figure!) We must
show that
K2 > 1 0 0 ^ . (1)
Let BC intersect A'B' and A*C* in M and N, respectively, let AC intersect A'B !
in P, and let AB intersect A ' C in Q. Since
/FDB = /QNB = /A and /CDE = /CMP = /A,
and AB = BN, AC = CM, we have
[BNQ] = CCPM] = K.
Also, MN = 2s, where s is the semi perimeter of A, so that [A'MN] = s2 tan A. With
these and similar results, we get
K2 = s 2 (tanA + tanB + tanC) - 2K

and (l) is equivalent to

s 2 (tanA + tan B + tanC) > 1 0 0 ^ + 2K. (2)
We establish the triple in inequality
s 2 (tanA + tanB + tan C) > 3/3s2 > 21K > 1 0 0 ^ + 2K9 (3)
from which (2) follows. The first inequality follows from
tan A + tan B + tan C > 3/3,
which holds for all acute-angled triangles ri9 p„ 261; the second follows from
 - 288 -
2
s > 3/3X, which holds for all triangles [l, p. 42]; and the third is a consequence
of the well-known inequality
K > HKi. (4)

Equality holds throughout in (3) just when A is equilateral.

II. Outline of the proposer's solution.
It is easy by angle-chasing, using symmetry about each line joining a pair of
excentres, to show that corresponding sides of Ai and A 2 are parallel. It is not
difficult to show that the linear enlargement is
l + cos A + cos B + cos C .,
2cos A cos B cos C " 1U lb;

for acute-angled triangles, and (l) follows, with equality just when A is equilat-
eral . •
The nomothetic centre of h\ and A 2 has trilinear coordinates
(tan A, tan B, tanC). (6)
It is well known that (cos A, cosB, cosC) is the circumcentre of A , and that
(sin A, sinB, sinC) is its Lemoine point. But (6) I have never seen referred to.
Perhaps in view of the publication in your journal it should be christened as the
Crucial Point of the triangle.
On the other hand, the result (like most of my amazing discoveries) may be
well known to pundits of triangle geometry.
Also solved by JORDI DOU, Barcelona, Spain; J.T. GROENMAN, Arnhem, The Nether-
lands; and KESIRAJU SATYANARAYANA, Gagan Mahal Colony, Hyderabad, India.

Editor's comment.
Our first solver did not give a reference for inequality (4), but it follows
from the relation
Kx = 2XCOS A COS B COS C
recently established in this journal [1982: 230] and from
cos A cos B cos C < -,
o

which can be found in [l, p. 25]. Note that (M-) holds even for nonacute-angled
triangles if signed areas are used.
The proposer gave a calculus proof of inequality (5). The proof was awkward,
as are most calculus proofs of triangle inequalities. The inequality is interesting
in its own right. Can some reader prove it without using calculus?
 - 289 -
Altshiller Court [2] proves that Ax is also homothetic to the tangential
triangle of A, formed by the tangents to the circumcircle of A at its vertices, but
he does not give a name to their homothetic centre. If that homothetic centre is
the same as that of Ai and A 2 (can some reader prove or disprove this?), then it is
doubly crucial that this point be given a name. In that case, we modestly suggest
that it be called instead the Lyness Point of the triangle.
REFERENCES

1. 0. Bottema et al., Geometric I n e q u a l i t i e s , Wolters-Noordhoffs Groningen,

1969.
2. Nathan Altshiller Court, College Geometry, Barnes & Noble, New York, 19529
p. 98.
}'; A ${

6831 C1981: 274] Proposed by Kaidy Tan3 Fukien Teachers University 3 Foochow3
China.
Triangle ABC has AB > AC, and the internal bisector of angle A meets BC at T.
Let P be any point other than T on line AT, and suppose lines BP,CP intersect lines
AC,AB in D,E, respectively. Prove that BD > CE or BD < CE according as P lies on
the same side or on the opposite side of BC as A.
Solution by Jordi Dou3 Barcelona3 Spain. fr
Our proof is based on the following elementary result.
If a point P is equidistant from the sides of an angle A,
and if a transversal through P cuts the sides of angle A
in X and Y, then the minimum length of XY occurs when it
coincides with the transversal segment MN which is per-
pendicular to the bisector of angle A. (This is illus-
trated in Figure l.) Furthermore, if XY makes an acute
angle w with MN, then XY = §(u) increases strictly from
MN to « as 03 increases from o to a = (TT-A)/2. For

XY = • ( « ) - A P c Q S t t j s i n c t o o ) + S 1 n ( « * » ) }

and
i \
t*w =AP cos(a-o)? _ cos(«+coW
T
|sinz(a-a)) Sinz(a+a>) J
Coming now to our problem, we assume that P is an interior point of the segment
AT. Let B'.C'.D'.E' be symmetric to B,C,D,E, respectively, with respect to line
AT, and let MN and a be as in Figure l. One of the segments BD or B'D' makes with
 - 290 -
MN and acute angle (less than a ) greater than the acute angle between MN and CE or
CE'. Hence
BD = B'D' > C E ' = CE.

Also solved by E.C. BUISSANT DES AMORIE, Amstelveen, The Netherlands; KESIRAJU
SATYANARAYANA, Gagan Mahal Colony, Hyderabad, India; GEORGE TSINTSIFAS, Thessaloniki,
Greece; and the proposer (three solutions).

Editor's comment.
It is clear from the proposer's solutions that he intended the point P to be
restricted to the segment AT, as in our featured solution. The proposer noted that
when P is at the incentre of triangle ABC, the resulting BD > CE provides a well-
known indirect proof of the Steiner-Lehmus Theorem. (See [1976: 19-24] in this
journal for information about and many references to this theorem.)
It was the editor who, thinking he could espy a more general theorem, modified
the proposal to allow the point P to range over the entire line AT. Satyanarayana
and Tsintsifas showed that BD > CE continues to hold when P lies beyond A on the
half-line TA. But nearly all solvers gave
counterexamples to show that BD < CE does
not hold for all positions of P beyond T on
the half-line AT.
Let CCX || AB and BB^ |J AC, as shown in
Figure 2. Tsintsifas outlined a proof showing
that f[P) = BD - CE goes from o to -« as P qoes
from T to Cj; that it qoes from -°° to +« as P
goes from Cx to B!; and that it remains posi-
tive when P lies beyond Bx on the half-line
AT. Hence, by continuity, there is a point
S between Cx and Bx such that /(P) vanishes
and BD = CE when P is at S. This point S is
a remarkable point of a nonequilateral tri-
angle whose existence has apparently been
hitherto unsuspected. It would be interesting
to have this point more precisely identified,
and to know some of its properties and how
to Construct it. Figure 2
For certain special nonisosceles triangles
ABC with AB > AC (called pseudoisosoeles triangles), the location of the point S
is known: it is the excentre l\ opposite angle A (see the Steiner-Lehmus reference
 - 291 -
given above). These triangles are characterized by the relation
. A .B . C
s m Z22 = sin-sin-.

One well-known example has A = 36° s B = 12°, C = 132°.

& ft ft

68^, C1981: 275] Proposed by George Tsintsifas, Thessaloniki, Greece.

Let 0 be the origin of the lattice plane, and let M(p,<?) be a lattice
point with relatively prime positive coordinates (with q > l). For i = 1,2 ^-l»
let P. and 0. be the lattice points, both with ordinate i, that are respectively
the left and right endpoints of the horizontal unit segment intersecting 0M„ Final-
ly, let P.Q. n 0M = M..
(a) Calculate
q-l
* - I P^.
r=l
(b) Find the minimum value of P.M. for 1 < i < q-l.
(c) Show that P.M. + P Jh . = 1, 1 < i < q-l
1 i q-r q~ 1
(d) Calculate
q-l P.M.
n
i=l i%

(e) Show that the area of a simple triangle is i. (A simple triangle is one
whose vertices are lattice points and which has no other lattice point in its in-
terior or on its perimeter.)
Solution by the proposer.
(a) Since (p,q) = l, the segment 0M has no lattice point in its interior.
Its equation is y = (q/p)x\ so when y = i we have x = ip/q and
(1)
PTH. =iZ-k.=2.
r v q 1 a
where k. is the integer [ip/ql and r. is the remainder when ip is divided by q.
For i = 1,2,...,q-l, the remainders r.are all different and less than q. The
sequence (r.) is therefore a permutation of the sequence (-O, and we have

q - V PTT -1 + 2 f '" + (^) - £±

(b) Since the smallest r. is 1, the minimum value of O L is 1/q.

(c) With the help of (l), we get
 - 292 -

p M = (^-i}P - r({?-^)p] = v _ IE _ L _ IE]

q-i q-i q L <7 J ^ <7 L ^ J
= P - &. - •— - p - & . —

^ ^ <7 ^ r

*i-'4.
from which follows
P.M. + P .M . = l.

(d) Since
P .M. r. / r. r.
^ t _ __l/(i _ _!) - 1_

we have
<7-l P.M. <7-l P .
q v
i=i M.q. A -i

Also solved by FRIEND H. KIERSTEAD, JR., Cuyahoga Falls, Ohio.

Editor's comment.
We have omitted the proposer's proof of part (e) because this result is an
immediate consequence of Pick's Theorem (see, e.g., Cl]), and because, as Kierstead
noted, it is equivalent to a result stated and proved in Hardy and Wright [2],

REFERENCES

1. A. Liu, "A Direct Proof of Pick's Theorem", this journal, 4 (1978) 242-244.
2. G.H. Hardy and E M. Wright, An Introduction to the Theory of Numbers, Fifth
Edition, Oxford University Press, 1979, p. 29, Theorem 34.

585. C1981: 275] Proposed by J.T. Groenmans Arrihem, The Netherlands,

Given is a triangle ABC with internal angle bisectors ^a^,9t meeting
a9b9c in U,V,W, respectively; and medians m 9m.9m meeting a9b9e in L,M,N, respec-
tively. Let
ma n t,b = P,
' m*
b n t c = Q,
^' mc n ta = R .

Crux 588 [1981: 306] asks for a proof of the equality

 - 293 -
AP BQ CR
PL * QM # RN "
Establish here the inequality

AR BP 0
RETPT
with equality if and only if the triangle is equilateral.

Solution by Kesiraju Satyanavayana3 Gagan Mahal Colony, Hyderabad, India.

If we apply the theorem of Menelaus to triangle
ABU and transversal CRN, we get (see figure)
A
AR UC BN
RU * CB ' NA " _1*
Since BN = NA, we have

AE - §C _ BU _ c b+c
m~uc~1 +
uc = 1+
bz:~b~'
With this and two similar results,we now get
AR s BP w CQ b+c c+a a+b ^ 0
RU PV QW " ~1 o a~ ~ 8 '

since (b+c)(c+a)(a+b) > 8abc9 with equality just when a = b = c (easily shown, or
see [l]).

Also solved by E.C. BUISSANT DES AMORIE, Amstelveen, The Netherlands; JACK
GARFUNKEL, Flushing, N.Y.; V.N. MURTY, Pennsylvania State University, Capitol Campus;
GEORGE TSINTSIFAS, Thessaloniki, Greece; and the proposer. One incorrect solution
was received.

Editor's comment.
The present proposal points out the superficial similarity between this problem
and Crux 588. See Crux 790 in this issue for still another related problem. Note
that 790 is a true extension of 588 but not of the present problem; for here equal-
ity occurs if and only if the trianqle is equilateral, whereas in 790 equality can
occur even for nonequilateral triangles and strict inequality can occur even for
equilateral triangles.

REFERENCE

1. 0. Bottema et al., Geometric Inequalities, Wolters-Noordhoff, Groningen,

1968, p. 12.
 - 294 -
686, C1981: 275] Proposed by Charles W. Trigg, San Diego, California.
Without using calculus, analytic qeometry, or trigonometry, find the
area of the region which is common to the four quadrants that have the vertices
of a square as centers and a side of the square as a common radius.
[A solution using analytic geometry appears in School Science and Mathematics,
78 (April 1978) 355.]
Solution by Jordi Dou, Barcelona, Spain.
It will simplify matters if we assume that the square has sides of unit length.
If certain regions have areas A9 B9 C, as shown in Figure l, then we have

A + 45 + 4C = 1,
A + 35 + 2C = H
4'
1 _ /|
A + 2B + C =
3 4

Figure 1 Figure 2

(Fiqure 2 explains the formation of the third equation.) Solvinn this system, we
find the required value to be

For good measure, we give also

/3 7T
* = -*•?•& C = 1-

Also solved by E.C. BUISSANT DES AMORIE, Amstelveen, The Netherlands; CLAYTON
W. DODGE, University of Maine at Orono; J.T. GROENMAN, Arnhem, The Netherlands; ANDY
LIU, University of Alberta; LEROY F. MEYERS, The Ohio State University; STANLEY
RABINOWITZ, Digital Equipment Corp., Merrimack, New Hampshire; KESIRAJU SATYANARAYANA,
Gagan Mahal Colony, Hyderabad, India; DAN SOKOLOWSKY, California State University at
Los Angeles; and the proposer. M.S. KLAMKIN, University of Alberta, sent in a comment,
and one incorrect solution was received.

Editorfs comment.
A photocopy of a relatively complicated solution by S. Jones, apparently taken
from a book, was sent byKlamkin, who found it in his files but could not further
(continued on page 278)