Professional Documents
Culture Documents
Crux v10n09 Nov
Crux v10n09 Nov
Mathematicorum
Published by the Canadian Mathematical Society.
http://crux.math.ca/
C P UX FATHEfATICQRUM
Sponsored by
Carle ton-Ottawa Mathematics Association Mathematique d" Ottawa-Carle ton
Ptafalie par le College Algonquin,- Ottawa
The assistance of the publisher and the support of the Canadian Mathematical Society,
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.
***************************************
CRUX MATHEMATICOROM is a problem-solving journal at the senior secondary and university
undergraduate levels for those who practise or teach aathesatiGs„ 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} . S a yearly subscription rate for ten
issues is $22 in Canada, $24*50 (or 0SS2O) elsewhere^ Back issuess each $2,25 in Canada,
$2.50 (or US$2) elsewhere. Bound volumes with index: Vols. 1&2 (combined) and each of Vols.
3-9, $17 in Canada, $18.40 (or OS$15) elsewhere. Cheques and money orders, payable to CRUX
MATHEHRTXCOROM, should be sent to the managing editor.
All communications about the content (articles, problems, solutions, etc.5 should be
sent to the editor. All changes of address and inquiries about siabscriptions and back issues
should be sent to the managing editor.
Editor: Leo Sauve, Algonquin College, 140 Main Street, Ottawa, Ontario, Canada K1S 1C2.
Managing Editor: F.G.B. Maskell, Algonquin College, 200 Lees Ave., Ottawa, Ontario,
Canada K1S 0C5.
Typist-compositor: Kuen Huynh.
Second Class Mail Registration No. 5432. Return Postage Guaranteed.
CONTENTS
- 273 -
- 274 -
THE EXTENDED ERDOS-MORDELL INEQUALITY
CLAYTON W. DODGE
Ten years ago The American Mathematical Monthly published the following Prob-
lem E 2462 [81 (i97n) 281] s which is an extension of the earlier Problem 3740
proposed by Paul Erdos [>2 (1935) 396] and first solved by L.J. Mordell [44 (1937)
252-254]:
!f
E 2462. Proposed by Huseyin Demirs Middle East Technical University 5 Ankara^
Turkey.
Let P be a point interior to the triangle AiA 2 A 3 , Denote by R. the distance
from P to the vertex A., and denote by r. the distance from P to the side a. oppo-
^ ^ i
s i t e to A , , The Erdos-Mordell i n e q u a l i t y asserts t h a t
R\ + R2 + R3 ^ 2 ( r i + r 2 + 2*3).
Figure 2
Figure 1
rendering the argument inconclusive. The editors could find no simple remedy for
this flaw since the extended theorem requires that either one or two of the distances
v. be negative. We wrote to those who had submitted solutions, and Leon Bankoff
and I corresponded for perhaps a year in attempting to put together a satisfactory
proof. Over the next nine years I returned to the problem from time to time, fas-
cinated by its challenge.
Two cases were disposed of almost immediately.
Case 1. Point P lies inside the angle vertical to a vertex angle.
For example9 let P lie inside the angle vertical to Ai, as shown in Figure 3.
Then r 2 a n d ^3 are to be taken negative and we must prove that
i?l + i? 2 + i?3 ^ 2 ( r i - r2 - 1*3 ) ,
where the r. have all been taken nonnegative and we have inserted the appropriate
negative signs. Because i?2 and r\ are hypotenuse and leg of a right triangle, we
have
i?2 ^ r\> and similarly i?3 > r\.
Thus
R\ + f?2 + #3 ^ # 2 + i?3 Z r\ + 2»i £ 2 ( P I - V2 - r3 ) . D
_ 27^ -
A2 F3 A2
Figure 3 Figure 4
Now
R2 - PA2 > PA2{2 sin (A 2 +e) - 2 sin e) = 2r 3 - 22^ .
Hence
j?l + F 2 + i?3 + 2ri > r2 + (2r^-2r\) + r2 + 2r\ = 2v2 + 2 r 3 . •
a3 a2 a\ a3
2ri > 2r 2 + N.
279
Therefore s since we have 9 by the f i r s t two l i n e s o f (2),
tana? = COS 15
y = /(«)
2 - sin15C
.05
t h a t i s , when x « 29.019466 , at which p o i n t
fix) « 0.004419 > o. Since
3C 50 7Q
Figure 8
£ O 2 < y n 9 0 =
as sin 30
and so9 by Lemma 2S
a a
2 3
— + — < 2.5.
a
^3 2
Now ri z E3 sin (6-15 ), so s by Lemma l,
Then
+ 22, + c o s 5 +
R\ * i? 2 + ^ 3 1 = i?l + i?2 ^3 #3 0--COS 5 ) + 2v\
REFERENCE
I present three new problem sets t h i s month. The f i r s t two* which I obtained
through the courtesy o f D i m i t r i s Vathis 9 contain the problems set a t two stages o f
the annual High School Competition of the Greek Mathematical S o c i e t y , The t h i r d
contains a few i n t e r e s t i n g problems from the A p r i l 1984 issue of the Russian j o u r -
nal Kvant. I s o l i c i t from a l l readers elegant solutions to a l l o f these problems.
x3 - lax1 + (a2+l)x - 2a + 2 = 0
f o r a l l real a .
(c) Find the range of the f u n c t i o n
J K }
/x^+1 + x + 1
1, Prove or disprove that there exists in space a pentagon all of whose sides
are equal and all of whose angles are 90°.
- 283 -
2 B (a) Find the maximum and the minimum of 8x + 6y - 5 s , where x9y9z are
real and x2 + y2 + z2 - 5.
(b) Express the vector x in terms of the vectors a$ if
f o r a l l Z,n c tf.
Z|,The elements a = ( 1 , 1 , 0 , 0 ) , £ = ( i , o , - i , o ) , a = ( 0 , 0 , 0 , 1 ) , d = ( - 1 , 0 , 1 , 0 ) ,
<g = ( 0 , 1 , 0 , 1 ) , a n d / = ( o 9 o , i , o ) of Rh are given. Let 7 = <a9b,o and
F = <d9e9f> be the subspaces of Rk generated by the vectors a}> 9Q and d9e9f9
respectively,
(a) Prove that the s e t
V + W = {^1+5^2 I xl £
^^ ^2 €
^
is a subspace of tf*.
(b) Find a basis of the space F n w (and denote i t by A).
(c) Find a basis (denoted by B) of the v containing the A, and a basis C of
the fir containing the A as well.
(d) Prove that the set BvC is a basis of the space v+w and t h a t v+w = i?^
(e) Prove t h a t dim (V+W) + dim (VnW) = dim F + dim W,
J-25 s ri981: 143; 1984: 79] (Corrected) From a list of Russian "Jewish"
problems.
In a convex quadrilateral ABCD S the sides AB and CD are congruent and the mid-
point of diagonals AC and BD are distinct. Prove that the straight line through
these two midpoints makes equal angles with AB and CD.
more satisfaction.
Let Nz = n; then N2 - l = loois, where z is a three-digit integer. Let
ab = 1001 = 7«n«i3 and s = xy> Then (N-i)(N+l) = abxy, and we may take N-l = ax
and N+i = by, Then N = ax-tl = by-i9 from which
by = ax + 2.
This equation may be solved for positive integers x and y when a and b are given.
When a - i and b - 143, we obtain
(Grade 8)
I. Solution by Mark Kantrowitz^ student^ Maimonides School^ BrooklineJ
Massachusetts.
We weaken the hypothesis to a9bfc$d > o and a+b+c+d - l. By the A.M.-G.M.
inequality,
/ ^ I < ^ ^ + 1 = 2a + 1.
5 , [1983: 304] From the 1980 Leningrad High School Olympics ^cird Round,
A convex quadrilateral is divided by i t s diagonals intc t>ur triangles*
The sum of the squares of the areas of the t r i a n g l e s adjacent tr ^posite sides is
the same. Show that at l e a s t one of the diagonals is bisected a the point of
intersection* (Grade 8)
8 8 [1983: 304] From the 1980 Leningrad High School Olympian ~*Hrd Bound.
A triangle ABC has sides a,b,c in the usual order. If c':ie A is twice
angle B9 show that a2 = b(b+c). (Grade 9)
Solution by Gali Salvatore3 Perkinss Quebec,
Since A = 2BS we have sin A = 2 s i n B c o s B 9 and so
c2*z2-b2
a = 2b*
2ca
which is equivalent to (a 2 - b(b+c)}(b-e) = 09 If b * c% then c = b(b-bc)« If
b = c, then A = 90° and B = C = 45°, so a2 = 2b2 = b(b-tc) again
The converse s if a2 = b(b+c) then A = 2BS is also t r u e . Se olutions IV and
V to Problem 102 in t h i s journal [1976: 74].
10, [1983: 304] (Corrected) From the 1980 Leningrad High c^ool Olympiad^
Third Round.
Show t h a t 1f, for any value of x in the interval C o . i ] , thf ^ q u a l i t y
I ax2+bx+c\ < 1 is s a t i s f i e d , then | a | + |Z? |+|el < 17. (Grade 9)
[In the original version 17 had been replaced by U
- 288 -
Solution by Paul Wagner, Chicago, Illinois.
For x - o, i / 2 9 i , we obtain respectively
|c| < 1, |a+22?+4e| < 4S \a+b+c\ < 1.
If we l e t a + 22? + 4c - m and a + b + c = n, then
a = -m + 2n + 2e and £ = m - n - 3c.
Thus
|a | = |-m+2n+2<?| < |m | + 2 |n | + 2 |<? | < 89
H , [1983: 304] From £7ie 1980 Leningrad High School Olympiad, Tk>ird Round.
Find two different natural numbers whose arithmetic and geometric means
are two-digit numbers one of which is obtained from the other by interchanging
the d i g i t s . (Grade 9)
Solution by Daniel Ropp, student, Stillman Valley High School, Illinois.
Let a and b be the required natural numbers. Then there are nonzero decimal
d i g i t s c and d such that
and (a+b)/2 > fob implies that c > d. Thus a+b = 2(l0c-kf), ah = (10<i+c)2s and a
and 2? are the roots of
^ 2 - 2(10c+d)x + (lOd'+c)2 = 09
that i s ,
{a, b] = (10c + d ± 3/ll(e 2 -<Z 2 )}. (1)
Now (a-d)ic-d) = i i or 44 since c?2-d2 < 9 2 - i 2 ; and since c+d and c-d have the same
p a r i t y , we must have
(c + d - 11 (c + d = 22
or
Cc - c? = 1 (c - d = 2.
Only the f i r s t system has a solution in d i g i t s a and d: (c9d) = ( 6 , 5 ) , and then
( l ) gives the unique solution {r, b} = (32 9 98}.
- 289 -
1 2 , C1983: 304] From the 1980 Leningrad High School Olympiad^ Third Round.
We s h a l l c a l l a segment in a convex q u a d r i l a t e r a l a midline if it joins
the midpoints of opposite s i d e s . Show t h a t i f the sum of the midlines o f a
quadrilateral i s equal to i t s semi perimeter, then the q u a d r i l a t e r a l is a paral-
lelogram,, (Grades 9, 10)
\t f c - d\ + \t * a - t\
1 3 , [ 1 9 8 3 : 304] From the 1980 Leningrad High School Olympiad, Third Round,
Are there real numbers a and b such that the function fix) = ax + b
satisfies the inequality
{fix))2 - cosx°fix) < Jsin25?
for all x e [0 S 2TT]? (Grade 10)
and t h i s implies
If we sum the first and third of these relations, we get o < 2(air+b) < 2 9 and this
contradicts the second. So there are no real numbers a and b with the desired
property,
15, C1983: 305] From the 1980 Leningrad High School Olympiad^ Third Round,
How many different numbers appear in the sequence
' ±2 ' "22 " 19802j
1980^ • 1980 . .... _1980 j
where the square brackets denote the greatest integer function? (Grade 10)
S = {[&2/1980] I k = l 9 2 9 8 9 . s 1980}.
Every integer from o to C9902/1980] = 4-95 appears in S\. For if not 9 let n be
a missing integer. Then n < 495 and9 for some k9
k2 „„
< YI < nn+1 (k±±)
< ————
2
1980 *X ~ 1980 •
This implies that (2fe+i)/i980 > l , so k > 990 and n > 495, a contradiction.
No integer appears twice in 5 2 . For suppose m appears twice. Then there is a
positive integer k > 991 such that
^ k2 (k+l)2
m < < m+1
~ 1980 "l980" '
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-ELEVEN = FIFTY5
nhn\ = nSk ~ T^
where 5, is the sum of the squares of all the edges meeting in A^, and T^ is the sum
of the squares of all the edges not passing through A-.
(b) Deduce from (a), or otherwise9 that if the medians of the simplex are all
eaual, then the sum of the squares of all the edges meeting in a vertex is the same
for all vertices. Is the converse also true?
where p,q9r are inteGers and r > o is not a perfect square. If x is r a t i o n a l , prove
that p = q and x is i n t e g r a l .
987, Proposed by Jack Garfunkel3 Flushing3 N.Y,
If t r i a n g l e ABC is acute-angled, prove or disprove that
, * . A . B . C 4,„ . A . B . C,
(a) s i n - - s i n - + s i n - > - { i + sin - s i n - s i n -},
,,s A B C 4r„ . A . B . C,
(b) c o s - + c o s - + c o s - ^ - ^ i i + sin - s i n - s i n - } .
sec
I TT + I sec
lT = °-
989, Proposed by Kurt Schifflers Schorndorfs Federal Republic of Germany,
Let H be the orthocenter of triangle ABC. Prove that the Euler lines of
triangles ABCS BCH, CAH9 and ABH are all concurrent. In what remarkable point of
triangle ABC do they concur?
990, Proposed by Bob Prielipp3 University of Wisconsin-Oshkosh.
Find all pairs (u,v) of positive integers such that
U3 + (tt+1)3 = I?2.
ft ft ft
Editorfs comment.
This problem has also appeared in [2] with an outline of solution based on
Euler's formula v-e+f - 2a A Question now naturally a r i s e s : For which n is there
a convex polyhedron having exactly n edges? This problem has also appeared (without
solution) in [2] and in this journal (Problem 121 [1976: 113]) with solutions by
Meyers, Trigg, and Sauve*
REFERENCES
346, C1978: 134; 1979: 26] Proposed by Leroy F, Meyers^ The Ohio State Univer-
sity.
It has been conjectured by Erdos that every rational number of the form n/n9
where n is an integer greater than l, can be expressed as the sum of three or fewer
unit fractions (reciprocals of positive integers, also called Egyptian fractions),
not necessarily distinct As a partial verification of the conjecture, show that
at least 23/24 of such numbers have the required expansions.
- 294 -
I I I . Comment by J.L, Brenner3 Palo Altos California,
The proposal, as c l a r i f i e d in Wilke's solution I [1979: 26-29], in effect asks
for a proof t h a t <S = 1/24 w 0.0417 is an upper bound to the density of integers
n > 3 such that 4/n cannot be expressed as the sum of three d i s t i n c t unit fractions
(Egyptian f r a c t i o n s ) , that i s , such t h a t , for integers xfy5z with i < x < y < z,
the equation
H i l l M,
- = - i- - + - (1)
n x y z
has no solution* In f a c t , Wilke found the b e t t e r upper bound 5 = 1/140 « 0.007i 9
and Maskell (see the e d i t o r ' s comment [1979.- 30]) improved this s t i l l further to
1 9 10
1. Paul Erdos and Sherman Stein, "Sums of distinct unit fractions", Proc
Amer. Mavk, Soc* , 14 (1963) 126-131, MR 269 71,
2. B.M. Stewart and W,A. Webb, "Sums of fractions with bounded numerators",
Can. J". Math., 18 (1966) 999-1003, MR 33s 7297.
3. William Aa Webb, "On 4/n = 1/x + 1/y + l/s", Proo. Amer. Math, Soc, , 25
(1970) 578-584, MR 419 1639,
4. R.C, Vaughan, "On a problem of Erdos, Straus and Schinzel"* Mathematika,
17 (1970) 193-198. MR 44 9 6600,
58 Li Delang, "On the equation 4/n = l/x + l/y + 1/z", J, of Number Theory,
13 (1981) 485-494,
6, Xun Qian Yang, "A note on 4/n = l/x + l/y + 1/s", Proc. Amer. Math. Soc.t
83 (1982) 496-498,
851, [1983: 177] Proposed by Allan Wm, Johnson Jr.3 Washington^ B.C.
Solve the doubly true decimal alphametic
We note at the outset that A e (l 9 2 9 3}, for DECBF < 98765 implies ABC < 315; and that
(l) implies
(C,F) € { ( 1 , 0 ) , (3 S 6) S ( 5 9 0 ) 9 ( 7 , 2 ) } ; (3)
and t h a t
(C.F) = ( 7 9 2 ) = o j B = 3 = * n o V a l u e f o r A
*
(B = 8 = > A = l .
Editor?s comment,
Rabinowitz stated that, in addition to bases 7 and 10, the Hebrew alphametic
has a unique solution in bases 13 and m , but not in bases 6, 8 f 9, 12, 15, or 16;
and Feder stated Ci] that it has two solutions in base 11. Mills proved that the
- 296 -
corresponding English alphametic
TEN x TEN - TEN = NINETY (5)
has no solution in base 10 but does have one in base ii s and Feder stated [1] that
the next base in which (5) has a solution is 66.
REFERENCE
DP AD
PB = AB
PC W (1)
REFERENCE
x
Jn
7
o
where a > b > o$ and deduce t h e value of
°° ax
ax _ bx
/
x(e +l)(e +1)
(This problem was suggested by Problem A-3 on the 1982 Wil1iam Lowell Putnam
. Mathematical Competition J
I. Solution by Leroy F« Meyerss The Ohio State University.
By the integral property in (i), the integrals
sucn
Let / ( « ) denote t h e l i m i t in ( l ) . Given e > o, t h e r e e x i s t s a real number xQ
that
x > xQ => / ( « ) - e < fix) < f(°°) + £.
l\ = liml. = / ( » H n £ . (2)
Since z is arbitrary^ we obtain t-#» v
In particular,, if
1 g e
f(x) = - = . . I9 a- > o s
2(1 + e"*) e* + 1 ^
- 298 -
then /(«>) = 1/2 and, for every r > o9 the integral in (1) is a proper i n t e g r a l , for
Mm fix)/x = 1/4 by r H 6 p i t a 1 f s r u l e . For this f9 the integral Jx becomes i2. There-
in ^-o
fore
j 2 =/C»).lnf = i-1n|.
Tl = (f(oo) . / ( o ) ) - l n | .
j 2 = (f (c») -. / ( o ) H n f = (0 + | ) - l n | = § « l n | .
Editor's comment.
References [2] and [3] were provided by Meyers. In [2], Problems 132 and 134
give (without attribution) Frullani's Theorem and Elliott's extension thereof, respec-
tively. In [3] Elliott's extension is mentioned and misleadingly called Frullani's
Theorem, The most thorough (and accessible) discussion of the Frullani and Elliott
results is probably that in Edwards M ,
REFERENCES
and
5 = . JH . + _gg. + m ,
(y-Lx)(z+x) (z+y)(x+y) (x+z )(y+z)*
I t is easy to show t h a t A < n < Bs with equality if and only if x = z/ = 2 8
(a) Show that the inequality 4 < 3/4 is "weaker" than 3B > 9/4 in the sense
that
A + 35 > I + i = 3o
When does equality occur?
(b) Show that the inequality M < 3 is "stronger" than 8B > 6 in the sense
that
M^8B<3+6=9e
When does equality occur?
I. Solution by Vedula N9 Murtys Pennsylvania State Universitys Capitol Campus.
We will use the following easily established results:
(y+z )(z+x)(x+y) *
frorr which follows
and
Observe t h a t 4 < 3/4 £ 5 , mentioned in the proposal, follows from (1) and ( 2 ) .
Proof of (a). This inequality is equivalent to A > 3(i-B) 9 that is* to
yz zx xy 63%/s
(y+z)2 (z-tx)2 (x+2/)2 ~ (y+z)(z+x)(x+y)9
or to
ys(s +x)(x+y) _^ zx(x+y)(y+z ) xi/(y+zHz+x) > ^
by part ( b ) . Either g > o or g < o may occur if 2 < X < 3. For example, if
2 < X < 36i/i80 9 then g(2,i9i9\) > o and #(i 9 2 s 3 9 X) < o.
Putting i t another way, we have shown the following: Of the two inequalities
the first holds for all (x,y,z) if X > 3 S the second holds for all (x,y,z) if X < 2 9
and either one may hold if 2 < X < 3.
Also solved by the COPS of Ottawa? J a T. GROENMAN, Arnhem, The Netherlands? M.S.
KLAMKIN, University of Alberta? VEDULA N. MURTY, Pennsylvania State University,
Capitol Campus (second solution}? and the proposer.
* * *
Editorfs comment.
This problem has a long history, going back to Cauchy (1821), and there has
been much activity about it and related problems in the more recent literatures as
evidenced by the list of references given below, most of which were found through
information provided by readers.
In response to an anonymous query [l], Lew [2] gave a brief outline of the
history of the problems and he referred to his note r 3] where a much fuller account
is aivens with an abundance of references that augments the list given below. The
information in the next paragraph is taken from [3].
Relative to the polynomial bijections f : / 2 -> N with which we are concerned,
Chowla [4] has displayed one such polynomial for each positive integer nf and Lew
[5] has constructed o(n) essentially distinct such polynomials (i.e., distinct to
within a permutation of the variables}, where e(i) = i,
is Seauence lies in Sloane [ 6 ] , and, according to Knuth [-, pp. 2399 534, Ex. 12],
as n -> °°s
o{n) ~ k(3 + 2 / 2 ) " v f 3 / 2 f where k = J / ( 3 / 2 - a ) / ^ « 0.139.
n-1
The constructed polynomials have least and greatest decrees n and 2
Of all our solvers, only Meyers showed how to find a polynomial of least degree
n. He proved that g : iVo -*• N0s where N0 is the set of nonnegative integers and g
is defined by
respectively. Formula (3), which Lew [3] credits to Cantor (1878), is given in
the more recent references Barnard & Child [10], Mathews [ i l ] 9 and P61ya-Szego [12].
It was conjectured in 1923 by Fueter and Pdlya (reference 6 in [3]) that (3) is
essentially the only polynomial bisection N2 + N. It is known that (3) is essen-
tially the only quadratic polynomial with this property and that, if there is an-
other such polynomial, i t must be at least of fifth degree [2].
We now consider polynomial bisections f .- if1 + N of highest degree 2n~1. For
n = l and n = 2, such polynomials are qiven by (2) and (3), respectively. For
n > 2, Meyers (and all other solvers) argued that the required polynomials can be
defined recursively by
f (Xi 9X? > . . . ,a? ) = f 2(/ * (X\ 9Xo , . . , fX A ) . X ) ,
• oA • oB • oC 3
sin2j + smzj + s i n ^ ^ j .
SsinBsinC = s - L ^ f - ± j £ s l*-±_£ii. M
For each of the opposed inequalities, i t w i l l be seen that equality holds just
when R = 2PS that i s , just when the triangle is equilateral.
(a) From ( i and ( 2 ) 9
n re^2A 3
2ff-r 3 R-2r ^ R-2r t.
P - ^ - M = I c o s 2 | - z s i n B sin C - M
4J?
> +P (fl+r) 2 _ R-2v _ G?-2r)G?+r)
U
~ 2R " i?2 2/? 2i?2 ~ '
857, [1983: 179] Proposed by Leroy F, Meyers3 The Ohio State University,
(a) Given three positive integers, show how to determine algebraically
(rather than by a search) the row ( i f any) of Pascal's triangle in which these
integers occur as consecutive entries,
(b) Given two positive integers, can one similarly determine the row ( i f any)
in which they occur as consecutive entries?
(c)* The positive integer k occurs in the row of Pascal's triangle beginning
with i , k, . . . . For which integers is this the only row in which i t occurs?
Observe that, for any positive integer t9 (6) and (7) ensure that i < k < n. But
n and fc must also satisfy either one (and hence both) of equations (*o, Since
Editorfs comment,
Kierstead's terse answer t o p a r t ( c ) * was the two-word statement "Almost a l l " .
We await a more d e t a i l e d answer.
.•« A A
d , i . d ._ J_ : JL : i.
a b c n «YI n
a b a
*^CP) = £ . £ . £ ) • (i)
(a) It follows from (l) that P _ = t#(P }. Consequently, P„ being given, we
n~rZ n n
obtain P by c o n s t r u c t i n g the isotomic conjugate o f the isogonal conjugate o f P ,
- 307 -
(b) For n = o and n - i , we have
and these points are easily constructed. For n > l , we can use the method of part .
(a) and construct successively (and laboriously)
A much more practical method, however, is to calculate the actual distances d $d,9d
REFERENCE
> /3 - 1
y > 2 + x. (1)
The result
/3(l+ar) > y (2)
was established in this journal [1982: 64] for type II triangles. Final ly 9 from
(1) and (2) we obtain /3(i+#) > 2+x9 from which follows
/3 - 1 1
x>
Also solved by J«T. GROENMAN, Arnhem, The Netherlands; WALTHER JANOUS, Ursu-
linengymnasium, Innsbruck, Austria? M.S. KLAMKIN, University of Alberta; KESIRAJU
SATYANARAYANA, Gagan Mahal Colony, Hyderabad, India; and GEORGE TSINTSIFAS, Thes-
saloniki, Greece.
m-1 n-(m-l)
m=l m ,
so that
x
si = i1 = l, s2 = l^-2
= i2_ol = -1, s3 = l 3 -2 2 +3 ! o, Si, -2;J+3Z-41 = -2,
$8
= 28 s22 = 56929813933
s9 = -47 s23 = 615188040196
s = -525 s24 = 3794477505572
10
= -2056 525 = 12028579019537
*11
*12 = -3902 5 2 6 = -50780206473221
s
13 = 9633 s 2 7 = -1172949397924184
s
14 = 129033 s 2 8 = -10766410530764118
s
15 = 664364 S 2 9 = -61183127006113951
= 1837904 s30 = -102718668475675151
*16
sl7 = -2388687 331 = 2573781218763700380
5
18 = -67004697 S32 = 40137777624890418072