Crux

Mathematicorum
Published by the Canadian Mathematical Society.

http://crux.math.ca/

The Back Files

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

Journal title history:

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

ISSN 0705 - 0348

CRUX HATHEPATICORU

Vol. 7, No. 4
April 1981

Sponsored by
Carleton-Ottawa Mathematics Association Mathematique d'Ottawa-Carleton
Publie par le C o l l e g e Algonquin

The assistance of the publisher and the support o f the C a n a d i a n M a t h e m a t i c a l Olympiad

Committer, the C a r l e t o n U n i v e r s i t y M a t h e m a t i c s Department,, t h e O t t c v a V a l l e y E d u c a t i o n Liaison
C o u n c i l , and the University of Ottawa M a t h e m a t i c s Department are gratefully acknowledged.
* ft * ft *ftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftftft
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 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 $12.00. Back issues: $1.20 each. Bound volumes with index; Vols. 1&2 (combined),
$12.00; Vols. 3 - 6, $12.00 each. Cheques and money orders, payable to CRUX MATHEMATICORUM
(in US funds from outside Canada), should be sent to the managing editor.
All communications about the content of the magazine (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.
Editci: Lcc Szuve, ."rchitcz^arc Ztcpzzzmcnt, Algonquin College, 281 E„ko Drive, Ottawa,
Ontario, KlS 1N3.
Managing Editor: F.G.B. Maskell, Mathematics Department, Algonquin College, 200 Lees Ave.,
Ottawa, Ontario, KlS 0C5.
Typist-compositor: Lillian Marlox*.
ft * ft

CONTENTS
A Lattice Point Assignment Theorem Harry D. Ruderman 98
An Angle Trisection Method which (Usually)
Does Not Work Ed Barbeau and John Im loo
Notes on Notation: II Leroy F. Meyers J O I
Areas of Triangles Inscribed in a Triangle M.S. Klamkin 102
The Olympiad Corner: 24 M.S. Klamkin 105
115
Problems - Probl^mes
Solutions H" 7
A Pandiagonal Sixth-Order Prime Magic Square . . . Allan Wm. Johnson Jr. 130

- 97 -
 - 98 -

A LATTICE POINT ASSIGNMENT THEOREM

HARRY D. RUDERMAN

Let I be a rectangular array of lattice points having at least two rows and two
columns. Two points of L will be called adjacent if they agree in one of their rec-
tangular coordinates and differ by l in the other. Thus the points (a, b) and (a, btl)
are adjacent if they are In L, and the same can be said of the points (a, b) and
(a±ls b). A small square is one that has each of its vertices adjacent to two other
vertices. It will always be oriented in the counterclockwise sense, as is the square
ABCD in Figure 1, where B is said to follow A,
C to follow B, etc, Thus, if two small squares
have a common side, such as AB in Figure l»
then B follows A in one square, but A follows
B in the other.
Let E = {1, 2, 3, ^ } . A mapping f:L+E
assigns to each point of L one of the numbers in
E, and we will say that /(B) follows /(A) for Figure 1

a given square if B follows A in that square.

The mapping f is arbitrary except for the
3 1 2 1
following two restrictions: C
4 1 1
(i) If A and B are on the boundary of
EC 'Vjump
the array, B following A, and if f(A) = i, 3 3 2
• • • « V
then /(B) * 4. We say that ^ may not follow
EC C
1 4 3 M- 3
l on the boundary. jump
(ii) If A and B are adjacent, then Q •
C
• •
EC
«
2
If (A) - /(B) | * 2. This means that M- may follow 2 3 2 1
• • • «
l (except on the boundary, as noted in (i)) or Figure 2
l may follow *+, but otherwise adjacent points are assigned numbers differing by at
most l.
Such a mapping / is illustrated in Figure 2. A jump is said to occur if 1 follows
M- on the boundary. A small square is said to be complete (C) if its vertices' assign-
ments are l, 2, 3, M- in counterclockwise order. If the orientation is reversed, we
call the square reversed complete (EC), Figure 2 shows 2 jumps, 3 reversed complete
squares, and 5 complete squares. It is no accident that here 2 + 3 = 5 , for we
have the
 - 99 -
THEOREM. If the map-ping f: L -*- E adheres to the restrictions described above,
then the number of jumps plus the number of reversed complete squares equals the
number of complete squares.
Proof, We assign a number from the set 5 = {-2, -l, o, 1, 2} of absolutely
least residues modulo 5 to each directed edge of each (counterclockwise-oriented)
small square as follows: if A and B are adjacent, then to the directed segment M
is assigned the number r € s if /(B) - /(A) = r (mod 5), It follows that the assign-
ments for AB" and for M have a sum of zero. We now find, in two ways, the sum of all
the assignments given in this manner to all the directed segments.
Clearly, all segments having at most one endpoint on the boundary will yield a
sum of zero because each such segment will have associated with it two oppositely
directed segments with opposite assignments. The segments on the boundary will have
but one associated directed segment, oriented counterclockwise. The sum of the
boundary segment assignments, which equals the sum of all directed segment assign-
ments, will be 5 times the number of jumps (since the sum must increase by
1+1+1+2 = 5 from one jump to the next).
Next, we sum the directed segment assignments by adding the sums found separately
for each square. If a square has fewer than four distinct numbers assigned to its
vertices, the assignments for its directed segments will total zero. See Figure,3
for some examples. That the statement is true in all cases can easily be checked.
2 - 1 3 2 0 2 if. -2 1 * 1 3

2 1
Figure 3

Otherwise the square must be either complete

or reversed complete. The sum for each com-
plete square is 5, and it is -5 for each re-
versed complete square, as can be seen from
Figure 1.
As the sum of the directed segment as-
signments must be the same for the two ways
Figure M-
of summing, we conclude that 5 times the number
of jumps must equal 5 times the number of complete squares minus 5 times the number
of reversed complete squares. Hence the number of jumps plus the number of reversed
 - 100 -
complete squares equals the number of complete squares. Q
It is an obvious corollary that if the boundary has no jumps, then the number
of complete squares equals the number of reversed complete squares.
I feel very sure that this theorem can be generalized to graphs more general
than those involving squares. Probably other assignments may be used in the general-
ization.
In closing, I want to give credit to O.P. Lossers, of Eindhoven University of
Technology in The Netherlands. His solution to a problem of mine [1] contains the
main ideas of this theorem and its proof.
REFERENCE
1. O.P. Lossers, Solution I to Problem 6192 (proposed by Harry D. Ruderman),
American Mathematical Monthlys 86 (1979) 598.

Mathematics Department, Hunter College, 94th Street & Park Avenue, New York, N.Y. 10028.
* * &

AN ANGLE TRISECTION METHOD WHICH (USUALLY) DOES NOT WORK

ED BARBEAU and JOHN IM

Problem 298 in Book 3 of 1001 Problems in High School Mathematics (collected

and edited by E. Barbeau, M. Klamkin and M. Moser, and published in 1978 by the
Canadian Mathematical Society), asks the reader to determine, for the following
proposed construction for a straightedge-and-compasses trisection of the arbitrary
acute angle POQ, those angles for which it works, and to show that it fails for
other angles:

From any point B on 0Q., drop a

perpendicular to meet OP at A. Con-
struct an equilateral triangle ABC
with C and 0 on opposite sides of
line AB. Then /POC = § /POQ. (See
Figure l.)

Two relatively complicated tri-

gonometric solutions are given in Book
«+ of the collection (published in l98o) 9
one of them requiring the analysis of a
cubic equation. At a meeting of the Metro Toronto Mathematics Club, E. Barbeau
suggested that it would be more satisfactory to have an argument using only
 - 101 -
Euclidean methods and offered a two-dollar prize
for the first student to find one. There were
two takers: John Im, then at the Toronto French
School, and David Atwood, then a Level 6 student
at Northern Secondary School. Although dupli-
cate prizes were awarded, we give here only
Im's solution, since Atwood's proved rather
less than the problem required, merely that
the construction fails when /POQ = 60°.
Solution by John Im.
We show that the construction works if and
only if £P0Q = 4-5°. It is quickly apparent
from Figure l that /PGC = 15° if /POQ = M-5°,
so the condition is sufficient.
To prove necessity, suppose that, in Figure 2, /POC = x (degrees) and [QQQ = 2x.
If D is the point symmetric to C with respect to line OP, and OM = OC, it is easy to
see that triangles ACD and ABC are equilateral and congruent, and that triangles OCD
and OMC are isosceles and congruent. It now follows that CB = CM. Now

/CBM = 2x + (30°-Hc) = 3x + 30°

and
/CMB = /OCM = l(180°-2a?) = 90°-x.
Since /CBM = /CMB, we have 3^+30° = 90°-x, soa: = 15° and /POQ = 45°, as required.
University College, University of Toronto, Toronto, Canada M5S lAl.
* & *

NOTES ON NOTATION: II
LEROY F. MEYERS
The rules for omitting parentheses are incompletely specified. How do you in-
terpret
sin 2x3 sin TT/3, sin a(x-y), sin x/x ?
The usual interpretation (as in this journal) seems to be to associate to the right
— u p to the next "outside" +, -, or / siqn or the end of the expression—except pos-
sibly for the second example. Compare, for example: sin | and sin (a/2) [1979: 261],
 - 102 -
cos \k/F [1981: 51]. However, I have seen the third expression used to mean
($ina)(x-y). How is the customary use of association to the right reconciled with
the usual (even in CM) interpretation of sin#sin# as (sinx)(sin#) and not as
sin (x siny)? What If an explicit multiplication symbol is used, as in sina'(x-y)?
Recently, on a student's homework paper, I noticed the following confusion, which
I have only lightly edited:

F(x9y) = cos m/ i - ira? sin iry j = P(x,y)i + Qix^y)^,

Since &1&M1 s CQS v s _t & W^u± s _T s in iy ,

ay dx
1 1
F{x9y) is not a gradient unless sin try = -, or y « ~ '
What is needed is an explicit convention about the omission of parentheses in
the neighborhood of function symbols written with several Roman or Gothic symbols
(such as sin, log, sgn, max), where the argument of the function is not usually en-
closed in parentheses. Can you make up a rule which agrees with common practice,
insofar as there is agreement?
Department of Mathematics, The Ohio State University, 231 West 18th Avenue,
Columbus, Ohio ^3210.
ft ft ft

AREAS OF TRIANGLES INSCRIBED IN A TRIANGLE

M.S. KLAMKIN
This note is a postscript to two notes published earlier in this journal by Dan
Pedoe [1979: 191] and Howard Eves [1979: 280]. The well-known problem considered by
Pedoe and Eves was the following:
TUJO triangles whose vertices lie on the sides of a given triangle at equal dis-
tances from their midpoints are equal in area.
Thus, in each of the Figures l to M-, lV\?z?$l - [QiQ2°33» where the brackets
denote area. Observe that "mixed orientation" of the P. and Q. on each side is
^ t
allowed, in general giving (except for labeling) four distinct pairs of equal tri-
angles. Eves appropriately calls such pairs isotomia triangles because pairs of cor-
responding vertices are isotomic points on the sides of the given triangle (i.e.,
each pair is equidistant from the midpoint of the side).
The problem is given and solved in Johnson [1], a reference given by Pedoe. The
proofs of Pedoe and Eves are simple and interesting, but Johnson's proof is also
quite simple and it has the advantage of leading most naturally to other interesting
results. Here, essentially, is Johnson's solution:
 - 103

Figure 2

We have given a triangle AiA 2 A 3 , with sides au aZj a3j and two points ?i and
qo (not necessarily distinct) on each side a^ and we wish to determine conditions
under which
CP1P2P33 = CQ1Q2Q3] • 0)
If we use signed areas for triangles, the points ?i and Q^ need not be restricted
to the segment a. but can be anywhere on its line of support. Regardless of the
positions of the ?i and Q^, (l) is equivalent to

CAlp3p2] + CA 2 P!P 3 ] + CAsPaPxD = [AxQaQa] + CA 2 Q 1 Q 3 3 + CA3Q2Qil - <2>

We assign numerical values to directed segments as follows:

0TA3 = Vis ^ 1 = #2, §3^2 = 2/3.

Since the areas of two triangles having a common angle are proportional to the pro-
ducts of the including sides, we have
 - 10'4 -

X3(a Zl rA
CA^Pz] = a\'a2 ' iA2«3]. etc.,
and (2) is equivalent to

l l
cyclic £3' CyC

When the points P. and Q. are isotomic on a. for i = l s 2, 3, then we have

£L -ZL* &2 - U.» a?3 - £3.,

a\ ai <*2 a3
57 <*3

and (3) follows immediately. D

But (3) also follows if

x
l ^3 %2 &1 #3 _ #2
a\ a3 * a2 a\ ' a3 a2

In this case, Q 2 P 3 II A 2 A 3 , Q3pi ||A3Ai, and Q1P2II A X A 2 . This result appeared sub-
sequently as a problem in the Hungarian Kiirscha'k prize competition for high school
students and graduates in the following form [2]: In the convex hexagon ABCDEF,
suppose that every pair of opposite sides are parallel. Prove that the triangles
ACE and BDF have equal areas, A solution by orthogonal projection was given by this
author in a joint paper on problem solving [3]. But the solution given here is
simpler and shows the general pattern of which the problem is a special case.
The remaining "cyclic" equality case,

fl = MjL , Ez = £1 * ^ = i±
a\ a2 a2 a3 a3 a\
also gives three pairs of parallel lines.
Among the remaining cases for which (3) holds, there are at least the three
"noncyclic" equality cases, one of which is

x
fLL = Ux * fi. = \LL » .$ = yz -
a^ a^ a2 a^ a3 a2

Finally, we mention a more special result, also given in Johnson, which makes
the triangles P1P2P3 and Q1Q2O3 congruent:

If the three lines QA|, QA 2 , ^A3 rotate as a rigid system about fl, and in-

tersect the sides AjA^» A 2 A 3 , A3Aj respectively at P\9 P 2 , P3; while Q'Aj, Q'A2,

Q'A3 rotate through the same angle 0 about Q' in the opposite direction^ meeting
AjA3, A 3 A 2 , A2Aj respectively at Qi, Q2» Q 3 ; then triangles PiP2P3 and Q1Q2Q3 are
congruents and similar to the given triangle ... .
 - 105 -
Here Q and Q 1 are the two Brocard points of triangle A 1 A 2 A 3 , and Johnson shows that
the ratio of similitude is sinw/ sin (w+G), where w is the Brocard angle.
REFERENCES
1. Roger A. Johnson, Modern Geometry, Houghton Mifflin Co., Boston, 1929,
p.p. 80, 276. (Reprinted as Advanced Euclidean Geometry, Dover, New York, I960.)
2. J. Acze"!, "A Look at Mathematical Competitions in Hungary," American Math-
ematical Monthly, 67 (1960) 435-t*37,
3. M.S. Klamkin and D.J. Newman, "The Philosophy and Applications of Trans-
form Theory," SI AM Review, 3 (1961) 10-36.
Department of Mathematics, University of Alberta, Edmonton, Alberta T6G 2G1.
* rt *

THE OLYMPIAD CORNER: 24

M.S. KLAMKIN
Problem 3 of the Twelfth Canadian Mathematics Olympiad (1980) reads as follows:
Among all triangles ABC having (i) a fixed angle A and (ii) an inscribed circle
of fixed radius r, determine which triangle has the least perimeter.
A solution by elementary triqonometry was published earlier in this column
[1980: 2^2]. It showed that both the perimeter and side BC are minimized when
/B = /C. Subsequently, in the Canadian Mathematical Society Notes 13 (1981) 16-17,
J. Aczgl, W. Gilbert, and C.T. Nq gave two rather nice elementary geometric solu-
tions. Solution l below essentially reproduces one of their arguments. I shall
later give an alternate geometric solution.
Solution 1. Figure 1 shows an isosceles triangle ABC and its incircle tangent
to BC at its midpoint P, and a general triangle AB'C with the same incircle. Note
that AB'C will have a minimum perimeter if and only if B'C is a minimum. (This
follows immediately from the equality of tanqents drawn to a circle from an ex-
terior point.) If B"C" is drawn through P parallel to B ' C , then, from similar
triangles,
B'C7B"C" = AB'/AB" z 1,

so B'C > B"C" and it suffices to show that B"C" > BC. Equivalently, if PD is the
reflection of PC" in the bisector AP of /A, we have to show that B"P + PD > 2PB.
Mow PB is the bisector of the angle at P in triangle DPB" so, if PM is the cor-
responding median, the well-known inequality involving adjacent sides and the in-
cluded median of a triangle yields B"P+ PD > 2PM > 2PB, as required. D
 The second geometric solution I shall give is a practically immediate conse-
quence of the following theorem which has long been well-known and which, in ad-
dition, has already been proved twice in this journal (Problem 120 [1976: 103-106]
where references to the theorem in the literature are given, and Problem 2 in an
article by Howard Eves [1980: 232-237])t
Given a point P inside an angle XAY, let y be the unique circle through P that
is tangent to the sides of the angle and such that A and P lie on the same side of
the chord of contact (= ST in Figure 2). Then the segment BC through P terminating
in the sides of the angle such that triangle ABC has minimum perimeter is the one
tangent to y at P.
For an elegant method of constructing circle y for an arbitrary point P in-
side angle XAY, see Sokolowsky's comment in this journal [1976: 139].
Solution 2. Figure 2 shows an isosceles triangle ABC and the midpoint P of
its base (determined by the given incircle of the triangle). Since P lies on the
bisector of /A, the circle y described in the above theorem is an excircle of the
trianqle. If B"C" is any segment through P, then it suffices to show (as in solu-
tion 1 above) that

perimeter of AB,fC'! > perimeter of ABC.

But this follows immediately from the theorem cited above.
***
We now give solutions received from readers for some problems published earlier
 - 107 -
in this column. Solutions to the remaining unsolved problems, or improved solu-
tions to those already solved, will be gratefully received,

0-3, C1980: m 6 ] Which is larger, sin(cosx) or cos (sin*)?

Solution by Gall Salvatore, Perkins, Quebec,
The result is that sin(cos#) < cos (sin x) for all real x. For

cos (sinx) - sin (cos a:) = cos (sin x) + cos ((ir/2)+cos*)

- orns (7T/'2)+cosa;+ sinx (7r/2)+C0S3:-sin3:

> o,
since

|cos# ±sina:! = /I ±sin 2x <» /2 < TT/2. 0

This proof is based on that given in Cl3. The problem was later proposed in [2],
but the solution given in a later issue C3] 1s not valid for all real x. But the
problem is much older, having already appeared (without solution) in Hobson CM-]. The
problem is also given and solved in [5].
It is natural to wonder whether, for example,
sin (cos (sina:)) < cos (sin (cosa?))
holds for all real x. The answer is no, for this inequality is sometimes true (e.g.,
for x = i) and sometimes false (e.g., for x = o ) .
REFERENCES
1. D.O. Shklarsky, N.N. Chentzov, and I.M. Yaglom, The USSR Olympiad Problem
Book, W.H. Freeman and Co.; San Francisco, 1962, pp. 4-3, 297-299,
2. Sidney H.L. Kung (proposer), Problem 835, Mathematics Magazine, 4-5 (1972) 167,
3. Leon Bankoff, Solution of Problem 835, Mathematics Magazine, M-6 (1973) 109.
M-. E.W. Hobson, A Treatise on Plane and Advanced Trigonometry, Dover, New York,
1957, p. 135, Problem 7 (originally published in 1891).
5. E. Barbeau, M. Klamfcfn, W. Moser, 1001 Problems in Sigh School Mathematics,
Book 3, Canadian Mathematical "Society, 1978, pp. 12, 7M-.

J - 2 . C1980: 14-5, 312] Solve in rational numbers x,„ y, z, and t:

(x+yfl)1 + (stt/?) 2 = 5 + U-/2".

Solution by B.C. Rennie, James Cook University of North Queensland, Australia.

 - 108 -
A rational solution [xsyszst) of the given equation would also satisfy

(x-yfl)1 + (z -t/2)2 = 5 - «+/?".

This gives a c o n t r a d i c t i o n since 5 - ty/2 < o, and hence the given equation has no
rational solution.

J-8, C1980: 116] Prove t h a t x c o s . r < 0,71 f o r a l l xelo9u/2l.

Solution by K.S. Mtwray, Brooklyn^ N,Z.

Let x = TT/2 - y; then o < y £ TT/2 and ( w i t h ( s i n * / ) / * / = 1 f o r y = 0 , as usual)
we have

a: cos a? = y(Tt/2 - y)($iny)/y

< {max?/(7r/2 -z/)Hmax ($iny)/y]

= 7T2/16 w 0.617 < 0 . 7 1 .

J-9. C1980: 146] Is it possible to cut two arbitrary squares

Into polygons which would form a new square?
Comment.
That the answer is yes follows from any dis-
section proof of the Pythagorean Theorem. In par-
ticular, the dissection proof indicated in the ad-
joining figure was given by H. Peri gal in 1873 [!]•
Here the sum of the two squares on the legs a and h
of a right triangle is cut into three pieces which,
by rearrangement, can be formed into the square on
the hypotenuse c.
More generally, we have the Bolyai-Gerwin Theorem [2]: Two polygons which have
equal areas are equideaomposable. Note that two figures are said to be equidecompost-
able if it is possible to decompose one of them into a finite number of parts which
can be rearranged to form the second figure. It is an unsolved problem in general
to determine the minimum number of cuts required. See also [3],

REFERENCES
1. Howard Eves, A Survey of Geometry> Revised Edition, Allyn and Bacon, Boston,
1972, p. 197.
2. V.G. Boltyanski, Equivalent and Equidecomposable Figures, D.C. Heath, Boston,
1963, p. 7.
 - J09 -
3. H. Lindgren, Recreational Problems in Geometrical dissections & How to Solve
Them, Dover, New York, 1972.

J-10, [1980: 146] Solve the system of equations:

y(x+y)2 = 9, (1)

y(x3 -y3) = 7. (2)

Solution by Gali Salvatore, Perkinss Qugbec.

It is easy to verify that (xsy) = (2,1) is a solution. We show that it is the
only real solution using the fact, obvious from (l) and (2), that x > y > o for any
solution (x3y).
Suppose o < y < It then,from (1),
(x+l)2 ;> (x+y)2 > 9,

from which we get x > 2. Also,

2 2
i - (fty)
5 3
< to-D
d
*
7 ~ x -y a? -l

an inequality equivalent to (x~2)(9x2+llx+B) < o which yields x < 2. Thus x = 2, and

then y - l from (l)
If we assume y > 1, all the inequalities in the steps above are reversed; so
we obtain again x = 2, and then y = 1.

Comment by M.S.K.
That there are other complex solutions follows by l e t t i n g x = tys giving

y3{Ut)2 = 9, yh(t3-l) = 7,

and then e l i m i n a t i n g */ gives

7 3(1+f)8 = 9^(^3-1)3.

J-lli £1980: 3161 Which i s l a r g e r , fa13 or 6 + ^3 ?

Solution by Donald Cross, University of Exeter, England,

We use t h e e q u i v a l e n c e

a 3 - £ 3 - e 3 > 3abc <=> a > b+c (1)

mentioned and proved in Klamkin's comment on Problem J~6 f~1980: 315]. If we set
a = #fl3, b = 6, and c = f5, the left side of (1) becomes 194 > 18^1239, which
is equivalent to 912673 > 903231. Hence, from the right side of (1), fai3 > 6 + fi.
 - no -
Comment by M.S.K.
One can also use the power-mean inequality:

,(I)3+(i)3+..^(6)3f3,i/3 £+6 + 6 + ^
v ;
yn' n V__( > n n n ,
J w+1 | rc+1
which is equivalent to
{(n+l) ? (3+6 3 /n 2 )} 1/3 > 6 + ?$.
Setting rc=4 yields ^412.5 > 6+^3, and the desired result follows.

J—12. C1980: 316] Solve the equation $ = yx in positive integers.

Comment.
If x < y9 the only solution is (x,y) = (2 9 4). Note that this problem is
very old and has already appeared in this journal. See Problem 188 [1977: 73-74]
for two elegant solutions and many references. The most recent discussion of the
equation = y is in Knoebel [1].

REFERENCE
J. R. Arthur Knoebel, "Exponentials Reiterated", American Mathematical Monthly,
88 (April 1981) 235-252.

J~13i £1980: 316] Prove that every convex polygon of area 1 contains a tri-
ancjle of area J.
Solution by K.S* Murrays Brooklyn, N.Y.
More generally, we show that every closed convex figure of area l contains a
triangle of area J. Let F be such a figure. The areas of all the triangles in-
scribed in F form a compact set of positive real numbers. Consequently, there
exists an inscribed triangle t of maximum area |t|. At each vertex of t, there must
'be a line of support of F which is parallel to the opposite side of t (otherwise we
can obtain a larger triangle by moving the vertex). The three lines of support form
a triangle T whose area is \T\ - 4|t|. Since clearly \T\ > 1, it follows that
\t\ > i and, by continuity, F contains a triangle of area J. D
For related results, see the comment and the references at the end of the next
problem (J-14).

J-Mt [1980: 316] Show that the following statement is false: A convex poly-
hedron of volume 1 contains a tetrahedron of volume 1/8.
 - Ill -
A natural estimate is not (1/2) 3 = j/8 but d / 3 ) 3 = 1/27, Obtain this esti-
mate and try to prove it.
Solution by K.S. Murray, Brooklyn, N.Y.
We prove, more generally, that every closed 3-dimensional convex region of vol-
ume 1 contains a tetrahedron of volume 1/27, and that the statement is false if 1/27
is replaced by 1/8.
That 1/8 is too large follows by considering the ratio of the volume V of a
regular tetrahedron to the volume v of its circumscribed sphere. Here we have
Vt/V = 2/3/97T < 1/8.

Our proof of the first part parallels that of J-13 above. Let R be a closed
3-dimensional convex region of volume 1. R contains an inscribed tetrahedron T of
maximum volume |x|. At each vertex of x, there is a plane of support of R which is
parallel to the opposite face of x. The four planes of support form a tetrahedron
T whose volume is \T\ - 27|x| (we will prove this later). Since clearly \T\ > 1,
it follows that |x| > 1/27 and, by continuity, R contains a tetrahedron of volume
1/27.
To prove that |^| = 27|x| let T = A-BCD and let 3a, 3?, 3?, 3? be the position
vectors of A, B, C, D, respectively, from some common origin. It is easy to see that
the vertices of x are the centroids of the faces of T, with position vectors
b +c +dj c +a H a9 a +a + b, a + £> + c.
With the usual notation [w, v, w] for the scalar triple product u -vxw, we have

6|r| = [[3(?-£), 3(?-J), 3(34)]!

and
6[x( = \la-t, a-c, a-d[~\\,

so that |r|/|x| = 27. D

With respect to this and the preceding problem (J-13), a stronger result is that
the circle minimizes the maximum-area inscribed triangle over all convex sets of the
same area, with similar results in n dimensions. See references [l]-[4], which con-
tain all the information available at this time.
REFERENCES
1. W. Blaschke, Leipzig Gesellschafte Berichte, 69 (1917).
2. Gross, ibid., 70 (1918).
3. A.M. Macbeath, "An Extremal Property of the Hypersphere", Proceedings of the
Cambridge Philosophical Society, 47 (1951) 245-247.
4. H. Groemer, "The Average Size of Polytopes in a Convex Set," (preprint).
 - 112 -
J~15i F1980: 3161 Does there exist an infinite family of pairwise noncon-
gruent right triangles such that the lengths of the sides are natural
numbers and the lengths of the two short sides differ by 1?
Comment.
Bob Prielipp, University of Wisconsin-Oshkosh, refers to the solution in Sier-
pifiski [11. See also Crux 460 [1980: 1591.

REFERENCE
1. W. Sierpifiski, Elementary Tlieory of Numbers^ Hafner, New York, 1964, pp.
44-46.

vJ-17. T1980: 3161 Prove the inequality (sinx)" 2 £ x"1 + 1 - 4/TT2 for o<x< TT/2.
Comment.
The inequality is incorrect as stated. The inequality sign should be reversed.
A proof follows by considering the function

Fix) = (sinrc)"2 - x~2.

We have

Mm Fix) = 1/3, F(TT/2) = 1 - 4/TT2,

x-±0
and it follows from

/Sin#\ 3 ^ „ n t .
(——) * cos a:
x
(proved i n Problem J - l t l 9 8 0 : 3121) t h a t

_,, . -2cos# 2 . .
F ix) - - — n — + r3 ^ °»
sin% x°
so that F i s nondecreasing. Thus Fix) < Fiu/2) for o < x < TT/2.
ft

J-18. [1980: 3161 Let ABCDE be a convex pentagon with the property that each of
the five triangles ABC, BCD, CDE, DEA, and EAB has area 1. Find the
area of pentagon ABCDE.

Comment.
The answer is (5 + /s)/2. This problem was posed in the First U.S.A. Mathematical
Olympiad (1972). For a solution see Greitzer [ll. A more general problem was given
by Mobius and solved by Gauss; for solutions and references, see our Problem 232
[1977: 2381.
 -113-
REFERENCE
1. S. Greitzer, "The First U.S.A. Mathematical Olympiad," American Mathematical
Monthly, 80 (1973) 276-281.
*
J~2Q* T1980: 3161 Let {x} denote the fractional part of x. Find
lim {(2 + /i) n }.

Solution by Gali Salvatore, Perkins, Quebec.

For any positive integer n, we have

(2 + / 3 ) n = C(2 4 / 3 ) n ] + fn, 0 < fn < 1,

where / is the function whose limit we are seeking. Since o < 2 - /5 < l, we also
have
0 < ( 2 - / i f = gn < 1.
n n
Now (2 + / 3 ) + ( 2 - / 3 ) is an integer; hence / + g is an integer strictly between
o and 2, and thus equals l. Since clearly ljjg# = o s it follows that Ifyf = 1.

J-21, T1980: 316] Prove that the first thousand digits after the decimal point
in the decimal expansion of (6 + / 3 5 ) 1 9 7 9 are nines.
Solution by I.S. Pressman, Carleton University, Ottawa, Ontario.
It is clear that
(1)
I
= (6 + /35f + (6 - /35f
n ,
is an integer for n = 1,2,3,... . Also, o < 6 - /35 < o.i, so that
0 < (6 - /35) n < 10" W , n = 1,2,3,... . (2)
n n
By (1) and (2), (6 + /35) is less than an integer (j ) by at most lo~ . Taking
n = 1979 gives the required result.

I cannot help wondering how the professors who set the above "Jewish" problems
would have fared if they themselves had been asked to solve these or similar pro-
blems under the same time and psychological constraints that the students who took
these tests had to face.
I have two new problem sets to present this month. As usual, I solicit solu-
tions from all readers, and will publish some of the more elegant ones from time to
time. The first set is taken from Rbze'piskolai Matematikai Lapok 60 (1979) 140
(for Problems 1-4) and 61 (3 980) 99 (for Problem 5). I am grateful to Frank Papp
who translated them from Hungarian into English.
 - im -
a
H-li ^ * b, e are the sides of a triangle with a < b < c^ determine the
best possible upper and lower bounds for the expression {afb\c)2/bc.

H-2. Let n be a positive integer. As a first step, we have given the se-
quence {a\y a?s ...3 ay\s where k - 2n and each a. is 1 or -1. As a
second step, we form the new sequence {a\ais a2a^ ...., a.a.}, and continue to re-
peat this process to generate new sequences. Show that, by at most the 2nth iterated
step, we arrive at a constant sequence with every term equal to 1.

H-3. Given that a. = ±1 for I - 1,2,...,«, prove that

? sin^Cai + a\ai/7 + ... + a\a2>. .a /2n ) = a\Jl + a2Jl + ... + a /2.

H-4, Show that, for every natural number n > 4, there is always at least one
inteqer between n\ and (w+l)! which is divisible by n 3 .

H-5, Let "o" denote a binary operation on the integers such that
(i) o°a = a for all integers o\
(ii) (aob)oc ~ c°(ab) + {a°c) + (2>°<?) - 2c for all integers a3 b, o.
Determine 8°9.

The second set consists of problems which have appeared in some recent math-
ematical competitions in Bulgaria. I am grateful to J. Tabov and P. Kenderov for
sending them to me in English.
Two
B-li given circles K\ and K2, with centers Oj and O2 and different radii,
are tangent externally at point A. Also given inside one of the circles
is a point M which does not lie on the line of centers OjC^. Show how to determine
a line £ through M such that, for some triangle ABjE^ with Bx e InKi and B2e£n# 2 »
the circumcircle of the triangle is tangent to line OjOa-
g-2, Let n be a positive integer, and let x and a be real numbers such that
0 < a < 1 and a n f l < x < 1. Prove that

k=l k-±
B~3i Given is a truncated triangular pyramid with lateral area s and parallel
bases of areas B\ and B 2 . Prove that, if there is a cross section parallel
to the bases such that it divides the given truncated pyramid into two truncated
 - 115 -
pyramids in which spheres can be inscribed, then
- -_ - _ 2
S = (/#i f /B1){VBl + VB2) .

B-4. Let the vertices A and C of a quadrilateral ABCD be fixed points on a

circle K9 while B moves on one and D moves on the other of the two arcs
of K with endpoints A and C, in such a way that always BC •= CD. If M is the point
of intersection of AC and BD and F is the circumcenter of triangle ABM, prove that
the locus of F is an arc of a circle.

B-5i N points are given in general position in space (i.e., no four in a plane).
There are ( ) possible tetrahedra with vertices only at the given points.
Prove that, if a plane does not contain any of the N given points, then it can in-
tersect the ( ) tetrahedra in at most N2(N-2)2/m plane quadrangular cross sections.

B ~ 6 . Let K be one of the arcs into which a given circle is divided by a chord
AB and let C be the midpoint of K. Let P be an arbitrary point on K and
let M be a point on segment PC such that PM = |PA - PB|/2. Find the set of all pos-
sible points M for all points P of K.

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.
5*i »V i"f

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 submitted
without a solution.
Original problems are particularly sought. But other interesting problems may
also be acceptable provided they are not too well known and references are given as
to their provenance. Ordinarily, if the originator of a problem can be located, it
should not be submitted by somebody else without his permission.
To facilitate their consideration, your solutions, typewritten or neatly hand-
written on signed, separate sheets, should preferably be mailed to the editor be-
fore October 1, 1981, although solutions received after that date will also be con-
sidered until the time when a solution is published.

6221 Proposed by J-T. Gvoenman, Arnhem, The Netherlands.

The Diophantine equation
8a*3 - 7\s£2\j + 35#i/ 2 - 83z/3 = z3

has the obvious solutions {x3y3z) = (k9o92k)s where k is any inteqer. Find at least
two solutions for which y * o.
 - 116 -
623* Proposed by Jack Garfunkels Flushing, N.Y,
If PQR is the equilateral triangle of smallest area inscribed in a
given triangle ABC9 with P on BC, Q on CA9 and R on AB, prove or disprove that AP,
BQ, and CR are concurrent.
]
62^4 • Proposed by Dmitry P. Mavlo, Moscow, U.S.S.R.
ABC is a given triangle of area K9 and PQR is the equilateral triangle
of smallest area K0 inscribed in triangle ABC, with P on BC, Q on CA, and R on AB.
(a) Find the ratio
X = K/K0 = .f(A,B,C)
as a function of the angles of the given triangle.
(b) Prove that X attains its minimum value when the given triangle ABC is equi-
lateral.
(c) Give a Euclidean construction of triangle PQR for an arbitrary given tri-
angle ABC.
6251 Proposed by Gali Salvatore, Perkins, Quebec,
(a) Let R denote the real field and let P be a polynomial in Rlxl.
Prove that if there are positive definite polynomials Qi,Qz c R(xl such that
P - Q\-Q2* tnen there are infinitely many such pairs (^1,^2)-
(b) Exhibit one such pair (Qi,Q2) for the polynomial P defined by
P(x) = a$ + a\x + ... + a x , a.zR.

6261 Proposed by A. Liu, University of Alberta,

A (i? J Jb J r J ^X)-configuration on a set with v elements is a collection
of b fc-subsets such t h a t '

( i ) each element appears in exactly r of the k-subsets;

( i i ) each pair of elements appears in exactly X of the k-subsets.
Prove that kr > v and determine the value of b when equality holds.

6271 Proposed by F. David Hammer, Santa Cruz, California.

Consider the double inequality

6 < 3^ < 7.

Using only the elementary properties of exponents and inequalities (no calculator*
computer, table of loqarithms, or estimate of 1/3 nay be used), prove that the first
inequality implies the second.
 - 117 -
6281 Proposed by Roland H. Eddy, Memorial University of Newfoundland.
Given a triangle ABC with sides a,b,c, let T ,Th,T denote the angle bi-
sectors extended to the circumcircle of the triangle. If R and r are the circum-
and in-radii of the triangle, prove that

s
a b c
with equality just when the triangle is equilateral.

629? Proposed by S.C. Chan, Singapore.

For which constants A and m does the infinite series

1 3 A -m 2 3 A 2 -2m 3 3 A 3 -3m 4 3 A 4 -Urn

converge, and to what sum?

6301 Proposed by Charles W. Trigg, San Diego, California.

Partition the palindrome 2662 into two integers, one of which is divisible
by 29 and the other by 43.

6311 Proposed by Sidney Kravitz, Dover, New Jersey.

Richard Bedford BENNETT (1870-1947) and Sir Wilfrid LAURIER (1841-1919)
are former Canadian prime ministers; and Pierre-Elliott TRUDEAU is, of course, the
current one. They have buried their political differences and find themselves united
in the following alphametic, which you are asked to solve:

BENNETT
TRUDEAU .
LAURIER

6321 Proposed by Leroy F. Meyers, Tlie Ohio State University.

Let ABCD be a plane quadrilateral in Euclidean 3-dimensional space.
Find a simple formula for the area of ABCD.
*V & »V

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.
^92. Clq79: 291; 1980: 291; 1981: 50] Late solution: BIKASH K. GHOSH, Bombay,
India.
s*; »V «

525, C1980: 78; 1981: 86] Late solution: F.G.B. kASKELL Algonquin College,
Ottawa.
 - 118 -
5331 F1980: 1131 Proposed by James Gary Propp* Harvard College3 Cambridge3
Massachusetts.
Consider the following products over the complex field:
n n
^±(x + 2k - 1) and fc0 (a? - 2fc -i l ) .

(a) For n- 1,2,3,4,5, find all complex x such that each product is real and
rational.
(b) Are there, for any n > 5, any real irrationalx such that each product is
rational?

Solution of part (a) by the proposer.

Let A and B denote the first and second products, respectively.
For n = l. we have A\ = x + 1 and B\ - x- 1 « These are both rational if and
only if i(A\ +£j) = x e Qs where Q is the set of all rationals.
For n - 2, we have A2 - (x+l)(x+3) and B2 = (x-l)(x-3). These are both rational
if and only if %(A2-B2) = x e Q.
For n - 3, we find from the expressions for A3 and B 3 that these are both rational
if and only if
l(A3+B3) = x(x2 ^-23)

and
3~0?3-£3+38iO = x2 + 23
and both rational, and this occurs just when
x e Q or x = ±t/23.
For w = H9 it turns out that ^ and B^ are both rational if and only if

l(AH+Bn) = #** + 96* 2 + 105 (1)

and
^(Ah-Bh) = xix^i-ll) (2)

are both rational. For any r e Qs (1) equals r just when x2 = - i m ^ , where
q = 1744+r is rational; and (2) will then be rational only if
ac2(tf2+ll)2= (~H3±i/q)(-32±Jq)2 (3)

is rational. Since the coefficient of Jq in (3) is ±(3776+<?), either q = -3776 or

Jq is rational. If q = -3776, then # 2 (rc 2 +ll) 2 = 600 2 from(3), and (2) is rational.
This yields the solutions x = ±/-43+-£/3776. If v^ is rational, then a?2 = -K3±Jq is
rational and so is x2 + 11; and then (?) is rational if and only if x e Q or
 - 119 -
x - ±-t/ll. All the solutions for the case n = 4 are therefore qiven by

x € Q or x ~ ±i/H or x = ±/-43±£/3776.

For n = 5, i t t u r n s o u t t h a t A$ and Bs a r e both r a t i o n a l i f and only i f

2
l(As+B5) = ^ O ^ + 230# +1689) (4)

and

J^M5-B5) = 5a;1* + 190a;2 + 189 (5)

2
are both rational. For any r e Qs (5) equals r just when x - -i9±Jq3 where
q = (i6l6+r)/5 is rational; and («+) will then be rational only if
ar 2 (;c lf +23aT 2 +1689) 2 = ( - 1 9 ± v ^ ) ( - 2 3 2 0 + 1 9 2 ^ + q ) 2 (6)

is r a t i o n a l . Since t h e c o e f f i c i e n t of Jq in (6) i s ±(^ 2 +24928g+22309l20), either

2
q + 2*+928q + 22309120 = 0 (7)

or Jq is rational. Since (7) has no rational solution, for (4) to be rational it

is necessary that Vq be rational and that x = ±Jsa where s = -19±Jq is rational. The
condition is also sufficient if x e Q. But if x i Q sufficiency requires the additional
condition

xh + 230a;2 + 1689 = s2 + 230s + 1689 = 0, (8)

a condition which is never realized since (8) has no rational solution s. So for
n - 5 only x e Q makes As and B5 both rational.

Part (a) was partially solved by J.T. GROENMAN, Arnhem, The Netherlands.

Editor's corvnent.
It is clear that A and B are both rational for all n if x e Q; and part (a)
n n
shows that, if 1 ^ n < 5 S A and B are never both rational for any (real) irrational
x. Part (b), which remains open, asks if the last statement remains true for all
n > 5.
* *
534• C1980: 1131 Proposed by Leroy F. Meyers, The Ohio State University.
Some time ago I noticed that the exponent of 2 in the prime factorization
of n! seems to be approximately twice the exponent of 3 in the same factorization,
at least for small values of n, say up to loo. Is this true in general? What about
the exponents of other primes? More precisely, if n is any positive integer and p
is prime, let e (n) be the exponent of p in the prime factorization of n\. Is it
true that 1 ime2(n)/e3(n) = 2 ? What about lime (n)/e (n) for primes p and ql
r r n
n\ n\ p q
 - 120 -
Solution de Marc Sackur, College Stanislas> Montreals Quebec.
D'apr£s un r £ s u l t a t bien connu de Legendre (voir la solution du Probleme 90
Cl976:3ti3 pour plusieurs references), si

n ~ ao + a\p + . . . + aJPP» 0<a.<p-l9 a *0 9

alors
n - (aQKZi + .. ,-HZ )
e(n)
pKn' ~ P-1
II r£sulte 1'encadrement

n - (r+l)(p-l) ^ < n
3
p-1 p p-1
d'oQ

p -1 ~n
i.i^^-L
n w p-1
(1)

Or il e s t c l a i r que p r <ns done r l o g p < logn et

0 < 11m£^ 1 . 1 1 m l 2 i 5 = 0.
nf n logp n+ n
On obtient done de ( t ) ]^pe (n)/n - l / ( p - l ) , d'oQ le r £ s u l t a t recherche:

)}Fe (n)/e (n) = ( ? - l ) / ( p - l ) .

n n 2
En p a r t i c u l i e r , Tirne 2 ( )/^3C ) = »
Also solved by JOHN T. BARSBY, St. Jchn's-Ravenscourt School, Winnipeg, Manitoba?
ERNEST W. FOX and MARIO D'ANGELO, Marianopolis College, Montreal, Quebec (jointly)?
J.T. GROENMAN, Arnhem, The Netherlands? FRIEND H. KIERSTEAD, JR., Cuyahoga Falls,
Ohio? DAVID SINGMASTER, Polytechnic of the South Bank, London, England? ROBERT A.
STUMP, Hopewell, Virginia? ROBERT TRANQUILLE, College de Maisonneuve, Montreal,
Quebec? KENNETH M. WILKE, Topeka, Kansas? KENNETH S. WILLIAMS, Carleton University,
Ottawa? and the proposer.
ft ft ft

[1980: 113] Proposed by Jack Garfunkels Flushing3

535. N.I.
Given a triangle ABC with sides asbsos let T ,T-.yT denote the angle
bisectors extended to the circumcircle of the triangle. Prove that
TT
a hT
b c > %/3abe
9 '9

with equality attained in the equilateral triangle.

I. Solution by Howard Eves, University of Maine.
 - 121 -
Denote the bisector of /A by t and let D and E denote the other ends of t and
*- a a
T, respectively. From similar triangles ABD and AEC it follows that * = ba/T ,
& a a*
with similar expressions for t^ and t . Hence
t t,t = a2b2a2/T TUT
aba aba
and it suffices to show that
t
aHta * P3abc* <*>
This is easily accomplished by means of inequalities 8.8 and 1,12 in the standard
reference [l]:

t <> Ss(s-a) and 6^s3(s-a)(s~b)(s-a) £ 21a2b2a2s (2)

where s is the semiperimeter of triangle ABC. For we have, using those inequalities
in turn,

tHH2 < s3(s-a)(s-b)(s~a) < ~ra2b2a2

aba 64
and (1) Follows. As in (2), equality holds if and only if the triangle is equi-
lateral. Inequality (l) deserves to be better known.
II. Solution by Roland H, Eddy, Memorial University of Newfoundland.
Let /A = 2a, /B = 23, /C = 2y. [With E as in solution I ] ,
BE = EC = a sin a/sin 2y.

Then, by Ptolemy's Theorem, we have

aT = £«BE + c E C
a
- a(b+a) s i n <Vsin 2y

and two similar results. These yield

T TiT = (b+c)(c+a)(a+b) s i n a s i n 3 s i n y / s i n 2 a s i n 23 s i n 2y
aba .
= ^;(b+a) (c va) (a+b ) / c o s a cos 3 COS y .
a

The required result is obtained immediately by applying inequalities 1.4 and 2.J28
of [1]:
(b+a)(a+a)(a+b) > Saba and cosacos3cosy < - / i , (3)

As in (3), equality holds if and only if the triangle is equilateral.

Also solved by LEON BANKOFF, Los Angeles, California; W.J. BLUNDON, Memorial
University of Newfoundland; ERNEST W. FOX, Marianopolis College, Montreal, Quebec;
G.C. GIRI, Midnapore College, West Bengal, India; J.T. GROENMAN, Arnhem, The Nether-
lands; V.N. MURTY, Pennsylvania State University, Capitol Campus; NGO TAN, student,
J.F. Kennedy H.S., Bronx, N.Y.; BOB PRIELIPP, University of Wisconsin-Oshkosh; KESIRAJU
SATYANARAYANA, Gagan Mahal Colony, Hyderabad, India; GEORGE TSINTSIFAS, Thessaloniki,
Greece; and the proposer.
 - 122 -
Editor's comment.
Nearly all solvers used inequalities from the "Bottema Bible" [l] in their
solutions. See Crux 628 in this issue for another inequality involving T , T,, T .

REFERENCE
l. 0. Bottema et al. Geometric Inequalitiess Wolters-Noordhoff Publishing,
Groningen, 1969.
* *
5361 C1980: 113] Proposed by B. Leeds, Forest Rills, New Xork.
Through each of the midpoints of the sides of a triangle ABC, lines are
drawn making an acute angle 9 with the sides. These lines intersect to form a tri-
angle A ' B ' C . Prove that A'B'C' is similar to ABC and find the ratio of similitude.
Solution by Dan Sokolowsky, Antioch College, Yellow Springs, Ohio.
More generally9 let P be any point in the plane of triangle ABC (it will become
apparent that the proposed problem corresponds to the special case when P is the circum-
center of the triangle) and let a, 3* Y be the vertices of the pedal triangle of the
point P with respect to triangle ABC, as shown in the figure. It is clear that
AyPp, BaPy, CBPa are all cyclic quadrilaterals, being inscribed in (say) circles
T J V
a b> Tc* res
Pectively-
 - 123 -
Let line I through a make a directed angle 9 with B?, with o <; J0| < 90°,
meeting r^ in B' and r in C ; let C'B meet r again in A', and join y to A' and to
B \ The angles marked 9 in the figure are clearly all equal, and it follows that
A',y,B' are collinear. It is now obvious that /A = /A', /B = /B f , /C = / C \ and
hence that AA'B'C* -AABC.
Now ABPC - AB'PC; so if Pa' i B ' C , which makes /oPa' = 0, then the required
ratio of similitude is
B
' C =_ Pa'= cos
„9n e 0
~W TST '
If e = o, then A'B'C coincides with ABC; and if |e| = 90°, then A'B'C collapses
into the single point P.

Also solved by JOHN T. BARSBY, St. John's-Ravenscourt School, Winnipeg, Manitoba;

W.J. BLUNDON, Memorial University of Newfoundland; HOWARD EVES, University of Maine;
G.C. GIRI, Midnapore College, West Bengal, India; J.T. GROENMAN, Arnhem, The Nether-
lands? LAX LANE LUEY, Willowdale, Ontario; KESIRAJU SATYANARAYANA, Gagan Mahal Colony,
Hyderabad,, India; GEORGE TSINTSIFAS, Thessaloniki, Greece; and the proposer.
3'C * ft

5371 C1980: 113] Proposed by Dan Sokolowsky, Antioch College, Yellow Springs, Ohio,
Find all pairs of integers (k,p) such that

(k-p)(2P+± -2) = (p-f-lM^ - 2 ) .

Solution by Kenneth M. Wilke, Topeka, Kansas (revised by the editor).

The given equation, which we write in the form

2(fc-p)(2P -1) = (p + D C ^ - 2 ) , (1)

has no solution for p = o, and the two sides vanish if and only if k - p = ±1, so we
have the two solutions (1,1) and (-1,-1). Suppose p > l. The two consecutive posi-
tive integers 2 ^ - 2 and 2^-1 are relatively prime, so 2^ - l divides p + l and
2^ -l <. p + l, an inequality which does not hold for any p > 2. And p = 2 yields the
solution (3,2). Suppose p < -l. Setting p = -r in (l) and multiplying both sides by
r
-2 yield the equivalent equation

(k+r)(2r+1 -2) = (1 - r ) ( 2 r + 1 - 1 ) ,

where r > l. The consecutive positive integers 2 r + -2 and 2 r + -1 are relatively

prime, so 2r+ - 2 divides l - r and 2 P + - 2 < |l-r|, an inequality which does not
hold for any r > 1.
The only integral solutions (k3p) are (-1,-1), (1,1), and (3,2).

Also solved by JOHN T. BARSBY, St. John's-Ravenscourt School, Winnipeg, Manitoba;

ERNEST W, FOX, Marianopolis College, Montreal, Quebec; J.T. GROENMAN, Arnhem, The
 - 124 -
Netherlands? FRIEND H. KIERSTEAD, JR., Cuyahoga Falls, Ohio? NGO TAN, student, J.F.
Kennedy H.S., Bronx, N.Y,? HERMAN NYON, Paramaribo, Surinam? BOB PRIELIPP, Univers-
ity of Wisconsin-Oshkosh? ROBERT TRANQUILLE, College de Maisonneuve, Montreal, Que-
bec? KENNETH S. WILLIAMS, Carleton University, Ottawa? and the proposer. Incomplete
solutions were received from G.C. GIRI, Midnapore College, West Bengal, India? and
MARC SACKUR, College Stanislas, Montreal, Quebec.
s*f ft sV

5381 C1980: 113] Proposed by Basil C. Rennie3 James Cook University of Worth
Queensland3 Australia.
Find
1 W ( 5 + 1 /(6 + 2/(7 + 3 /...•( (n+3) + (n-1 )An^men))...))).
n-K>°

Solution by the proposer.

For n > 2 and suitable x , let
R {x) - /m + H + mx, m = l,29...,w; (1)
then the sequence whose limit is sought is {/ }, where
fn = Ri°ni°..>°Rn(en)* n = 2,3,4,... .

It follows from (l) that

R (m+3) = m + 2 (2)
777
and i t is easily verified that
1 < x/y < k ==> 1 < J? (x)/R (y) < Jk. (3)
^ m m
Since n z 2 S we have
1 < en/(n+3) < ens (4)

and repeated use of (3) and (2) gives the successive implications
(4) => i < # (0n)/(n+2) < exp (n/2)
n
=> 1 <R °R (en)/(n+l) < exp (n/4)

=>
=> 1 < T ? ! ^ 0 . . . 0 ^ (en)/3 < exp(n/2 n ). (5)

We now have from (5)

3 < fn < 3exp(n/2 n ), (6)
and so n-*
limfc J
°n
= 3.
Also solved by LEROY F. MEYERS, The Ohio State University.
 - 125 -
Editor's comment,
Meyers showed more generally that the limit exists and is again 3 if

for any q > l. In fact, if q = ea, where a > o, we get in place of (6), for all suf-
ficiently large n,

3 < fn < 3exp (an/2n)J

which shows that the limit 3 is independent of the value of q.

In a subsequent letter, Meyers considered the still more general

£Jffl/(l+p + l / ( 2 + p + 2 / . . . / ( ( n - 1 ) + p + (n-1) / ( r c + p + n a ))...)),

where the sequence {a } is nonnegative and bounded by a geometric sequence, (Our
problem has p = 4 and a^ = en.) He conjectured that this limit, call it lip), de-
pends upon p alone and exists for all p > -1, The following table, computed by Meyers,
gives some values of lip).

JB1 E(P)» P= Up>* P= liv»

-1 1.7559485 4.5 3 .0949200 10 2 10.593024
-0 5 1.9207299 5 3 .1867400 10 3 32.151009
0 2.0714907 6 3 .3621498 10* 100.50881
0 5 2.2111897 7 3 .5280247 10 5 316.73054
6
1 2.3418786 8 3 .6857352 10 1000.5009
1. 5 2.4650490 9 3 8363555 10 7 3162.7779
2 2.5818250 10 3 9807453 10 8 10000.500
2, 5 2.6930786 20 5 1924418 10 9 31623.277
3 2.7995020 30 6 1542767 10 1 0 100000.50
3. 5 2.9016555 40 6 9761944 10 1 1 316228.27
12
4 3 50 7 7055891 10 1000000.5

It appears from the table that l(p) ~ /p for large p. It would be interesting
to have confirmation of Meyers' conjecture and a closed form expression for lip).
5*4 i'c &

5391 C1980: 114] Proposed by diaries W. Trigg, San Diego, California.

From among the three-digit primes less than 500, form four four-term arith-
metic progressions in which the first and last terms contain the same decimal digits.

Solution by the proposer.

Among the three-digit primes less than 500 can be found exactly 18 pairs iasb)
with a < b in which a and b are permutations of the same digit set. These are:
 - 126 -
(113, 131) (127, 271) (173, 317) (149, 491) (239, 293) (337, 373)
(113, 311) (137, 173) (139, 193) (419, 491) (241, 421) (349, 439)
(131, 311) (137, 317) (149, U19) (179, 197) (313, 331) (379, 397)

We wish to find all the pairs (a,b) for which

a + (b-a)/3 and a + 2(h-a)/3 (1)

are both primes. Now h-a is divisible by 3 in every case (divisible, in fact, by
9: see Crux 576 [1980: 251]), so the numbers (1) are always both integers; but only
for the pairs

(131, 311), (137, 317), (349, 439), (419, 491)

are they both primes. The four progressions are:

(131, 191, 251, 311),

(137, 197, 257, 317),
(349, 379, 409, 439),
(419, 443, 467, 491).

Also solved by JOHN T. BARSBY, St. John's-Ravenscourt School, Winnipeg, Manitoba?

ERNEST W. FOX, Marianopolis College, Montreal, Quebec; J.A.HR HUNTER, Toronto, On-
tario; ALLAN WM. JOHNSON JR., Washington, D.C.; FRIEND H. KIERSTEAD, JR., Cuya-
hoga Falls, Ohio; J.A. McCALLUM, Medicine Hat, Alberta; BOB PRIELIPP, University
of Wisconsin-Oshkosh; ROBERT TRANQUILLE, College de Maisonneuve, Montreal, Quebec;
and KENNETH M. WILKE, Topeka, Kansas.

Editor's comment.
It is a matter of editorial policy never to publish a proposer's solution unless
it is, in the editor's opinion, clearly superior to (or at least as good as and es-
sentially different from) all the other solutions received. This policy is even more
strictly adhered to when, as in this case, the problem is very easy and there is an
adequate supply of other solutions to choose from. But here the editor's hand was
forced. Three of the other solvers used the same approach as the proposer but care-
lessly omitted some of the 18 pairs (a,fr)-one had 17, one 16, and one 12-and it was
sheer luck that the omitted pairs did not lead to solutions. The remaining six sol-
vers submitted correct answers but with little or no indication of how the answers
were arrived at. This may be a good place to remind readers that when a problem has
an answer, numerical or otherwise, the answer itself does not constitute a solution
to the problem. A solution is a description (preferably in complete, grammatical,
properly punctuated sentences in the language of discourse, which is English or
French, not "mathematics") of the process by which the answer is arrived at. An-
swers alone belong at the back of textbooks.
it it it
 - 127 -
5401 [1980: 114] Proposed by Leon Bankoff, Los Angeles, California.
Professor Euclide Paracelso Bombasto Umbugio has once again retired to
his tour dHvoire where he is now delving into the supersophisticated intricacies
of the works of Grassmann, as elucidated by Forder's Calculus of Extension, His
goal is to prove Neuberg's Theorem:

If D, E, F are the centers of squares described externally on the sides of a

triangle ABC, then the midpoints of these sides are the centers of squares described
internally on the sides of triangle DEF.
Help the dedicated professor emerge from his self-imposed confinement and enjoy
the thrill of hyperventilation by showing how to solve his problem using only high-
school, synthetic, Euclidean, "plain" geometry.

Solution by J.T. Groenman, Arnhem3 The Netherlands.

It suffices to show that, say, the midpoint M of BC is the center of the square
on EF or, equivalently, that ME = MF and ME x MF (see Figure l) .

Figure 1.

A 90° rotation about A takes triangle PAB into triangle CAQ, so

PB = CQ and PB ± CQ.
Now ME - MF follows from
ME - iBP = JCQ = MF

and ME i MF follows from

ME |! BP i CQ (|MF.
 - 128 -
Also solved by JOHN T. BARSBY, St. John's-Ravenscourt School, Winnipeg, Manitoba?
W.J. BLUNDON, Memorial University of Newfoundland; CLAYTON W, DODGE, University of
Maine at Orono? HOWARD EVES, University of Maine• JACK GARFUNKEL, Flushing, N.Y.?
ANDY LIU, University of Alberta? LAI LANE LUEY, Willowdale, Ontario? LEROY F.
MEYERS, The Ohio State University? NGO TAN, student, J.F. Kennedy H.S., Bronx, N.Y.?
KESIRAJU SATYANARAYANA, Gagan Mahal Colony, Hyderabad, India? DAN SOKOLOWSKY, Antioch
College, Yellow Springs, Ohio; ROBERT TRANQUILLE, College de Maisonneuve, Montreal,
Quebec? JAN VAN DE CRAATS, Leiden University, The Netherlands? and the proposer.

Editor's comment.
This theorem is credited to Neuberg by Forder [l], who gives a two-line solution,
using the method and notation of Grassmann, which so far has thoroughly defeated (and
deflated) Professor Umbugio, the premier mathematician at the \
University of Guayazuela, despite frequent invocations to his \
patron saint and namesake Euclide, \
In an effort to help the good professor, most of our \ /
solvers submitted solutions (some quite lengthy) in "plain" \ /
geometry. Dodge and van de Craats gave proofs by transforms- .A /
tion geometry and expressed the hope that such proofs would ,'' \l
soon be considered "plain" geometry even by Professor <[ Q /\
Umbugio. In view of the parlous state of geometry in "^-* u •
Guayazuela (and North America-we don't know about the rest /
of the world), that time is not yet. Garfunkel gave a proof by /
complex numbers, which could only be considered "plain" geometry /
in an imaginary high school. /
There is only one statement in our featured solution which Figure 2.
is likely to cause Professor Umbugio some concern: that PB i CQ results from a 90°
rotation about A. Figure 2 is all that is needed to bring that statement into the .
realm of "plain" geometry.
REFERENCE
1, Henry George Forder, Tfte Calculus of Extension, Chelsea, New York, I960,
p. 40.

A * *
5411 C1980: 152] Proposed by Herman Nyony Paramaribo, Surinam.
Solve the following colourful alphametic, in which BLUE is a perfect
square:
RED
BLUE
GREEN .
BROWR
 - 129 -
Solution by Lai Lane Lueys Willowdale3 Ontario.
It is clear that we must have B = 8 or 9 S and examination of a table of squares
reduces the possible values of BLUE to 5 four-digit squares having no repeated digits:

8649, 9025, 9216, 9604, 9801.

The requirement D + E = 10 immediately eliminates 9025, 9604, and 9801. Since BLUE
= 9216 implies W = 4 = D, we must have BLUE = 8649, and D = l and W = 3 follow. Now
the hundreds' column requires R > 3, and of the available digits only R = 5 does not
result in a duplication. Finally, G = 7, 0 = 2, and N = o fall into place. The
unique solution is

591
8649 (= 93 2 )
75990
85230

Also solved by JOHN T. BARSBY, St. John's-Ravenscourt School, Winnipeg, Manitoba;

JAMES BOWE, Erskine College, Due West, South Carolina; CLAYTON W. DODGE, University
of Maine at Orono? J.A.H. HUNTER, Toronto, Ontario? ALLAN WM. JOHNSON JR., Washing-
ton, D.C.; J.A. McCALLUM, Medicine Hat, Alberta; NGO TAN, student, J.F. Kennedy H.S.,
Bronx, N.Y.; DONVAL R. SIMPSON, Fairbanks, Alaska; ROBERT TRANQUILLE, College de
Maisonneuve, Montreal, Quebec; CHARLES W. TRIGG, San Diego, California; KENNETH M.
WILKE, Topeka, Kansas; and the proposer.

ft ft ft

5^121 C1980: 153] Proposed by Michael W. Ecker, Pennsylvania State University3

Worthington Scranton Campus.
Let p be a permutation of the digits o through g. This induces, in a natural
way, a function p* defined on the interval (o,l) by writing (the sums are for
n - 1,2,3,...) x = lan/10 (using o's instead of 9's in the finite decimal case) and
putting p*(x) = I p(a )/10n. These functions were defined earlier in this journal
[1979! 184] where it was stated that some of them were bijective. Find a necessary
and sufficient condition for a function p * to be bijective.
Solution de Robert Tranquille, College de Maisonneuve_, Montreal, Quebec.
Nous all ons montrer que:
I. Une condition n^cessaire et suffisante pour qu'une de ces fonctions p* soit
une application de 1'intervalle (0,1) dans lui-meme est que la permutation p de
Tensemble {o,l,...,9} qui 1'engendre fasse correspondre {0,9} 3 {0,9}.
II. Cette condition Stant remplie, la fonction p * est ngcessairement une
permutation de 1'intervalle (o,l), c'est-a-dire une bijection de 1'intervalle (0,1)
sur lui-meme.
demonstration de I. La condition est n£cessaire. Car supposons, au contraire,
que p({0,9}) * {0,9}. Alors il existe un chiffre a d {0,9} tel que p(a) e {0,9},
 - 130 -
etpar suite* = Ia/lo € (o,l) mais p*(a?) = lp(a)/10n
n
= o ou .999... = l i (0,1).
La condition est suffisante. Car supposons que p({o,g}) = {o99} et soit x € (o,i).
Si x = .aaa..., alors a i {0,9} et p*(x) = Ip(a)/lon = .2>M>... € (0,1), car
b - p(a) t {0,9}. Si, par contre, x = .a 1 a 2 a 3 ... o0, disons, a. * a .s alors
p(a.) * p(a.) et p*(*) = Ip(a )/10n e (0,1).
^ y ft

Demonstration de II. Supposons que p({o99}) = {0,9}. En vertu de I, p* est une

application de I'intervalle (o.l) dans lui-m§me. Montrons qu'elle est infective. Soit
x*y « (0,1). leurs representations dgcimales x = ,aia2a3..9 et y - .Jnf>2j&3... ne
contiennent pas de suite infinie de 9, Supposons que x * y , done a. * b. pour un
certain i et, par suite, p(a^) * p ( K ) . S1 1'on a en m§me temps
p*(x) = lp(an)/lon = rp(bn)/10n = p*(z/),
alors p*(a?) et p*(y) contiennent, 1'un une suite infinie de o, 1'autre une suite
infinie de 9. Done x et y contiennent, T u n une suite infinie de j 9 1'autre une
suite infinie defc,et p({j3k}) = {0.9}. Puisque {j,k} = p-1({o,9}) = {o,g}, F u n
de x et y contient une suite infinie de 9, ce qui est faux. Done p*(x) * pHy) et
p* est injective.
Pour dgmontrer la surjectivite* de p* f soit y c (0,1). Si y = .bbb..., alors
£ ^ {0,9} et il existe un chiffre a £ {0,9} tel que p(a) = b; et pour
x = £a/10n € (0,1) on a

p*(x) = £p(a)/10n = Tb/10n = #.

Par contre, si y = . f c ^ ^ - . - avec, disons,fc.* £>., alors p~l(b.) *p'l(b.) et T o n

a p*(x) = y pour

a? = 2:p"1(^n)/10n c (0,1),

Also solved by JOHN T. BARSBY, St. John's-Ravenscourt School, Winnipeg, Manitoba?

and the proposer.

A PANDIAGONAL SIXTH-ORDER 23 ' 137 31 127 131 181

PRIME MAGIC SQUARE 173 83 163 103 89 19
263 7 M 67 193 59
This pandiagonal sixth-order magic square
is composed of distinct primes and has i*7 139 179 223 13 29
the magic sum 630. 107 ; 97 5 37 151 233

ALLAN WM. JOHNSON JR.

[ 17 1167 211 73 53 109 |