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.
 CRUX NATHEHATICORUn

Vol. 12, No. 2

February 1988
Published by the Canadian Mathematical Society/
Publie par la Societe Mathematique du Canada
The support of the Uoi¥ersity of Calgary Department of Mathematics and
Statistics is gratefully acknowledged*

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

and university undergraduate levels for those who practise or teach mathe-
matics. 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 subscrip-
tion rate for ten issues is $22.50 for members of the Canadian Mathematical
Society and $25 for nonmembers* Back issues: $2*75 each. Bound volumes with
index: Vols* 1 & 2 (combined) and each of Vols, 3-10: $20. All prices
quoted are in Canadian dollars. Cheques and money orders, payable to CRUX
MATHEMATICORUMf should be sent to the Managing Editor.
All communications about the content of the journal should be sent to the
Editor. All changes of address and inquiries about subscriptions and back
issues should be sent to the Managing Editor.
Founding Editors: Leo Sauve, Frederick G*B* Maskell*
Editor: G.W. Sandsf Department of Mathematics and Statistics, University
of Calgary, 2500 University Drive N*W*, Calgary, Alberta, Canada, T2N 114.
Managing Editor: Dr. Kenneth S. Williams, Canadian Mathematical Society,
577 King Edward Avenue, Ottawa, Ontario, Canada, KIN 6N5*
Assistant to the Editors: Mrs* Siu Kuen lam*
ISSN 0705 - 0348.
Second Class Mail Registration No* 5432* Return Postage Guaranteed*
* * *
CONTENTS
The Olympiad Corner: 72 * * * * * * * * * * * . M*S, llamkin 19
Strens Memorial Conference * 25
Problems - Problemes: 1111-1120 * * * * * * * * 26
Solutions: 982, 983, 986-989 . . . • • • • • 28
A message from the Editor * * 37

- 18 -
 - 19 -

THE OLYMPIAD CORNER: 72

M.S. KLAMKIN

All communications about this column should be sent to Professor

M.S. Klamkin, Department of Mathematics, University of Alberta, Edmonton,
Alberta, Canada, T6G 2G1.

I start out with the nine problems of the 1985 Austria-Poland

Mathematical Competition held in Hollabrunn, Austria. I am grateful to
Walther Janous for their translation and transmission. As usual, I solicit
from readers elegant solutions to these problems.
st
1 day, June 25, 1985 — time 4 1/2 hours.

2. If a, b, c are distinct real numbers whose sum is zero, prove that

l(h-c)/a + (c~a)/b + (a-b)/c][a/(h-c) + (b/(c-a) + c/(a-h)] = 9.
2. A given graph has n > 8 vertices. Is it possible for the vertices
to have the respective valences 4, 5, 6, — , n-4, n-3, n-2t n-2,
n-2, n-1, 12-1, /3-1?
3. Prove that for any convex quadrilateral of unit area, the sum of the
lengths of its sides and diagonals is > 4 + VS.

2 n d day, June 26, 1985 — time 4 1/2 hours.

4. Determine all real solutions x$ y of the system

x* + y2 - xy3 - 9x/8 = 0,
y* + x2 - yx3 - 9y/8 = 0.
5. We are given a certain number of identical sets of four integral
weights in grams. It is assumed that using all of these weights,
one can weigh all integral weights from 1 to 1985 grams inclusively. In how
many ways can such identical sets of weights be chosen if their total weight
is to be a minimum? Editorial note: There will be two answers depending on
whether or not one allows the (weighing) weights to be used on both sides of
the balance.

6. If P is an interior point of a tetrahedron ABCD^ show that the

volume of the tetrahedron whose vertices are the centroids of the
four tetrahedra PABCf PBCD9 PCDA, PDAB equals (vol. ABCD)/64.
 - 20 -

rd
3 dayf June 27, 1985 — time 4 hours.
Team Competition

7. Determine an upper bound for

(xy + 2yz + zw)/(x2 + y 2 + z 2 + nr2)
which is valid for all real quadruples (x,y,z,w) # (0,0,0,0). (The smaller
your upper bound is, the more points you are awarded.)
§• The consecutive vertices of a given convex n-gon are A Q , Aj,, ...,
A y The ji-gon is partitioned into n - 2 triangles by diagonals
which are non-intersecting (except possibly at the vertices). Show that there
exists an enumeration di9 d2$ ..., A __9 of these triangles such that A. is a

vertex of A . for 1 < i < n - 2. How many enumerations of this kind exist?

§• ^i> ^2> ..., P are consecutive vertices of a given convex i?-gon.

Show that there exists an interior point Q of the n-gon and three
vertices P., P., P, such that the angles QP.P.ll$ s QP.P.tl and Q P , P ? J 1 are
i9 j $ k ° i i+l j j+1 i Ar+1
acute (P , = Pi). Editorial note: I have edited this problem to remove an

ambiguity in the translation I received. It is quite possible that the

original version is a different problem. A related problem is to determine
the maximum number of angles QP.P. ,, i = l,2,...,n which can be acute for an

interior point Q.
*

Now through the courtesy of P.J* O'Halloran, I give the official

solutions of the 1983 Australian Olympiad [1983: 173].

March 15, 1983 — time 4 hours.

1. Each positive rational number occupies an infinite number of

positions in the following p a t t e r n :
1 2 1 3 2 1 4 3 2 1 5 4 3 2 1 .

For instance the number x2> occupies positions 9, 42, ... •

(a) Find the first five positions occupied by the number ^ .

(b) Find an expression for the nth occurrence of the number K- .

(c) Find an expression for the first occurrence of the number —, where p

and q are relatively prime and p < q.

 - 21 -

Solution •
1 I 1 1 3 2 1 4 3 2 1 5 4 3
2 21 1 6
T T ' 1 2" ' T ' S ' S ' T ' S ' S ' S ' T ' S ' S '
;

5 4 3 2 1 7 6 5 4 3 2 1 8 7 6 5
2 ' 3 ' 3 • S' B T'S'S'S'S'B^'T'S'S'^ ' 3 ' B '
2 1 9 8 7 6 5 4 3 2 1 10 9 8 7 6 5 4 3
; ,
7 ' 5 T '2 ' 3• 5 ' 5 ' B * 7 ' B ' 3 ~ I ' 2 3 ' 4 " ' 5 " ' B ' 7 ' B
2 1 11 10 9 8 7 63 5 2 1 12 11 10 B
3 ' 3 • 5 ' B *7 '
» 3 ' TO ' TT ' ~I ' 1
~2 ' ~5 '
9 8 7 6 5 4 3 2 1 13 14 13 12 11 10 9
4 ' B" ' B ' 7' 88 ' § ' TO" ' IT * 15 ' "T ' ... » -j , ~2 ' ~~3" ' ~4 ' ~5 ' B
8 7 6 1" 5 1 4 3 2 1
7' 8
' 3 ' LTS * TT ' 12 ' T3 ' T3

(a) The first 5 occurrences of 1/2 are at positions

3, 14, 34, 63, 101.
(b) Using finite differences
1 3
11
2 14 9
20
3 34 9
29
4 63 9
38
5 101
suggests the quadratic function
f(n) = 3 + l l ( n - 1) + | ( n - l ) ( n - 2)

= i ( 9 n 2 - 5n + 2 ) .
1 n
The nth occurrence of s i s w-. This occurs in the section
3n - 1 3n - 2 77
2~~
—r~
There are 3n - 2 sections preceding this one containing 1, 2, ..., 3n - 2
terms respectively, so the number of terms preceding this section is
7j(3n - 2)(3i? - 1). Therefore 75- first occurs in position

|(3n - 2)(3n - 1) + 2n = J(9n - 5n + 2).

(c) Similarly, the first occurrence of p/q (p < q) occurs in the section
P + q - I P
T q ' '"
There are p + q - 2 sections preceding this one so the total number of terms
preceding this section is
\{p + q - 2)(p + q - 1).
 22 -

Thus the position in which p/q (p < q) first occurs is

TF(P + q 2)(p + q - 1) + q.

Subsequent occurrences of p/q are of the form -£ (Jt > 1) which is in position
kq
kq in the string beginning
Hp + q) - 1

As k > 1, this string occurs after the stated first occurrence.

2, ABC is a triangle and P is a point inside it. Angle PAC=angle PBC.

The perpendiculars from P to BC, CA meet these sides at L, W
respectively, and D is the midpoint of AB. Prove that £?L = DM.

Solution.
Draw altitudes AX, BY. B is the
centre of circle AYXB9 with radius
r = AB/2, So
ID2 - r2 = LX-LB
2 2
MI? - r Jfr-/fA
A J M C
(see for example Corollary page 29 of Advanced Euclidean Geometry by
R.A. Johnson) so we have to prove LX*LB = MY'MA. Now IYBX = IYAX, so
subtracting the given equal angles we obtain IYBP = ZPAX = 8 (say) and thus
MY
sin@,
BP AP
Since dBPL and dAPM are similar,
BP BL
AP AM
Therefore
LX-LB = /fr-iVA.

3, A box contains p white balls and q black balls. Beside the box
there is a pile of black balls. Two balls are taken out from the
box. If they are of the same colour, a black ball from the pile is put into
the box. If they are of different colours, the white ball is put back into
the box. This procedure is repeated until the last pair of balls are removed
from the box and one last ball is put in. What is the probability that this
last ball is white?

Solution.
If the two balls taken out are both white, then the number of white balls
decreases by two, in the other cases it remains unchanged. Hence the parity
 - 23 -

of the number of white balls does not change during the procedure. Therefore
if p is even, the last ball cannot be white - the probability is 0. If p is
odd, the last ball has to be white - the probability is 1.

Editorial note: See also #30 in Olympiad Corner 63 [1985: 71].

4. Find all pairs of natural numbers (nfk) for which

1
(n + I ) - 1 = nl

Solution.
k
For n = 1, the equality 2 - 1 = 1 must be satisfied, whence 1 = 1 , so
one pair is (1,1).
k
For 13 = 2, the equality 3 - 1 = 2 must be satisfied, whence 1 = 1, and
(2,1) is a required pair.
Now if n > 2, then n\ is even; therefore the pair (n%k) for n > 2
satisfies the given equality only if n is even.
1
For n = 4, the equality 5 - 1 = 24 must be satisfied, so another pair is
(4,2)
For n > 4, an even number, take n = 2m and a pair (n9k) satisfying the
given equality. Then nl = (2n)! = 2m(2m - 1)!, and (2m - 1)! contains both
factors 2 and m where m # 2. Therefore nz divides nl and must also divide
1 k k k
(n + 1) - 1. But (n + 1) = n + ... + (0)ir + kn + 1, hence i? must divide
1 i?
1. That means 1 > n is true, which leads to (n + 1) > (n + 1) > n! + 1.
This contradicts the assumption that the pair (n9k) satisfies the given
equality. So there is no solution for n > 4, and the only solutions are
(n,k) = (1,1), (2,1), and (4,2).

5. (a) Find the rearrangement {ai,a2j...,a } of {1,2,...,J?} which

maximises
ai&2 + aa«3 + ••• + a ai = Q*

(b) Find the rearrangement that minimises Q.

Solution.
(a) Suppose we have a rearrangement {ai,a2,...,a }. Then if

1 < i < i + 2 < / < n and (a. ~ a . l 1 ) ( a . - a . l 1 ) > 0, we can increase Q with
- — •' i .7+1 .7 l+l
the rearrangement
{ai,... ,ai?a ;.,a/_1,... »« J+2 » a i+ i»« ;+i»® ;+2» • • • » a n J "
Indeed, by this alteration we increase Q by the quantity
 - 24 -

a .a . + a . tla .^ ~ a .a .,, - a ja ,%1 = (a . - a .,,) (a . - a . ,,) > 0.

Let A be a rearrangement that maximises Q (since the number of rearrangements

is finite, A exists). Using a cyclic permutation (which does not alter the
value of Q) we can suppose that A is of the form {1 = a1,a2?.--,a }. Let

2 = a.. Then, since (1 - a. _-.)(2 - a2) ^ 0, we must have i = 2 or i - n

because otherwise we could increase Q. We can suppose i = n for otherwise we

use a cyclic permutation and a reflexion (which does not alter Q either). If
3 = a ., then 7 = 2 since otherwise, from (1 - a .^-iKS - 32) > 0, we could

improve A again. Applying a reflexion, we get the rearrangement

{2f62f...,3,1}.
In the same way as above we can prove that £?2 = 4. Continuing this
argument we obtain that
A - {1,3,5,...,2^] + 1, 2[g],...,4,2}

(or a cyclic permutation and/or reflexion of this).

(b) A similar argument leads to the minimising rearrangement
{1,12,2,12-1,..., [|] 4 1}.

Editorial note-. See also #34 in Olympiad Corner 63 [1985: 71], and
Problem 1059 (Cyclic extrema) by D.B. Daykin, Math. Mag. 53 (1980) 115.

6. The right triangles ABC and AB\Ci are similar and have opposite
orientation. The right angles are at C and Ci, and ICAB = IC1AB1.
M is the point of intersection of the lines BCi and CB\. Prove that if the
lines AH and CC% exist, then they are perpendicular.

Solution0
Take A to be the origin for a
system of Cartesian co-ordinates. Let
the co-ordinates of C be (a, 5) and the „ , , L _ -- ^ ^ \ ^/ ,,
co-ordinates of Ci be (ufv).
If the ratio AC:CB = 1:1, then
the co-ordinates of £ are A(030)
k(-hfa) + (a,5) = (-15 + a$ka + £>),
and those of B\ are
i(v,-o) + (u,v) = (£v + u%~ku + v).
The equations of the lines BCi and B\C are
((*a + 5) - v)x - ((-15 + a) - u)y = (*a + 6)0 - (-15 «- a)v
 - 25

and
(b - (~-ku + v))x - (a - (kv + u))y = h(kv + u) - a(-Jru + v).
Rearranging these we have
£ax + kby = iau + kbv + bu - av - (b - v)x + (a - u)y
kux + Jfrvy = kau + *5v + bu - av - (b - v)x + (a - w)y.
The co-ordinates (s,t) of the point M satisfy both these equations, so
kas + AM = kus + **ft,
that i s ,
s fc - v
t a - u
which means the lines AM and CC\ are perpendicular.

THE EUGENE STRENS MEMORIAL CONFERENCE ON INTUITIVE

& RECREATIONAL MATHEMATICS & ITS HISTORY

July 27 to August 2, 1986

THE UNIVERSITY OF CALGARY

to mark the acquisition by the University Library of the Strens Collection•

There will be 15-20 principal talks. There will also be opportunities

for short contributed talks, and displays, both amateur and commercial, of
books, games, puzzles, and films.
For further information, write to:
Richard K. Guy or Bill Sands
Department of Mathematics and Statistics
University of Calgary
Calgary, Alberta T2N 1N4
Canada.

* t *
 - 26 -

PROBLEMS

IPwllem p^opo^tol^, trnd ^olutbom, ^A<m,l4, le, <k&nt to th& z4A,to^,9 wko$&
a44fieAA, top-paw^ on, the §*wmt, pa&& o§ tfH& 1&&U&* ^w/M^de&d- tJUwld, wfa&n&vesi
fw§AAslle>9 It, toccompanAstd, If to &ol(£&toa>9 ^&ie^em>e^9 <m4 othesi, oft<&ogAt& wttisch,
awe, ltkt,lf to le, o§ k€,lp> to t&e, edttw,. An u&teM>oJk (*) to§ttA, to fwmlest,
tndootote^, to fypoll&m <ui^mttt&d wttfoowt to ^olubtod,.
Gii,q,tfw& psw>llem&> wve, p4inttc4itw%tf ^owgM,. %u<b oth^v tatesi&&ttn$,
p^ollem^, maf, tol&o It, i^ooeptto^le pwy&td&d, th^f w%& tvot too well favowv toad
ts&§€s*e>fice>& w%t, q,ti>t>n, QA, to tHet's, p^uy^emuvoe. Gj%<ibiV(vuylf,9 t§ tfoe, osU,fta>totoji o§
to problem am le located 9 It <WUw£«t n*>t le &u<&mvtte>d If, &omelodf el<^e wvtfUtut
A^& OT, &€^> p>&%mi,bAAsOfi.
fo §toctlottote tb&bi coriAAs&e/*totL>on,i fowv iboluttoriA, 9 tfpew%i,tten, <yp oetotlp
fUmd4sntttea on, ^tp^ved9 i^epa^ate aAeefc^* ^JUuld, p^eie%allf It, mailed to tfoe
edoto^, le§o^e feptemltsv l f 1986, toltiUm§A &olutton<& t&cevwed a§tt/v tA&t dtote
moll tol&o le> oon^ode^ved un&vl the, tome wh^etvto&olubvon, i& pullt6&e>d>.

1111. Proposed by J.T. Groenm&n, Arnhem9 The Netherlands.

Let a, fi9 i be the angles of an acute triangle and let

f(a9p9y) = cos * cos C + cos £ cos je + cos « cos ^.

(a) Prove that f(a909t) > | 92.

# 1 —
(b) Prove or disprove that f(a909y) > ^ + 72.
1112. Proposed by Allan W. Johnson Jr., Washington, Z7.C.

Solve the synonyraical base 10 addition

LITHE
PLIANT
SUPPLE

1113. Proposed by Jack Garfunkel 9 Flushing, ^.F.

Consider two concentric circles with radii r and 2r, and a

triangle i4iBC inscribed in the inner circle. Points A 1 , B 1 , Cs on the outer
circle are determined by extending AB to B 1 , BC to C\ and CA to A 1 . Prove
that the perimeter of triangle A*B*C* is at least twice the perimeter of ABC.
Equality is attained when AW is equilateral.

1114. Proposed by George Tsintsifas, Thessaloniki, Greece.

Let ABC, A*B*C* be two triangles with sides a,h9c, a% ,h9 ,c* and
areas F9 Fl respectively. Show that
%
aa + bb% + cc1 > 473 7IT7.
 - 27 -

1115. Proposed by Helen Sturtevant and Edward T.H. Wang, Wilfrid Laurier
University9 Waterloo9 Ontario.

Determine all positive integers n such that an equilateral

triangle can be dissected into exactly n equilateral triangles.

1116. Proposed by David Grabiner, Claremont High School9 Claremont,

California.

(a) Let f(n) be the smallest positive integer which is not a

factor of 12. Continue the series f(n)9 f(f(n))9 f(f(f(n)))9 ... until you
reach 2. What is the maximum length of the series?
(b) Let g(n) be the second smallest positive integer which is not a
factor of n. Continue the series g(n)f g(g(n))9 g(g(g(n))), ... until you
reach 2. What is the maximum length of the series?

1117. Proposed by Jordi Dou9 Barcelona9 Spain.

Let ABCD be an isosceles trapezoid with bases AB > DC9 and let M
and N be points on AD and EC respectively so that UN is parallel to AB and IX!.
Let D* be the projection of D on AB9 let E = DD* f) BM9 F = BD fl AE9 and
P = NF n DC. Prove that PA is perpendicular to AB.

1118. Proposed by P. Erd'6s9 Hungarian Academy of Sciences.

Let ai, 52) 93» be a sequence of numbers such that

lim (a. , - a.) = °°. Construct an infinite sequence b% < bz < • • • so that

00

none of the sums 2 e .5 . (where e . = 0 or 1 for each i and all but finitely
i=l
many are 0) equals any of the a.'s. See also problem #85 [1976: 29].

1119. Proposed by Stanley Rabinowitz, Digital Eguipment Corp., Nashua,

New Hampshire.

The following problem, for which I have been unable to locate the
source, has been circulating around DEC. A rectangle is partitioned into
smaller rectangles. If each of the smaller rectangles has the property that
one of its sides has integral length, prove that the original rectangle also
has this property.
t
1120. Proposed by D.S. Mitrinovic, University of Belgrade, Belgrade,
Yugoslavia.
 - 28 -

(a) Determine a positive number A so that

(a + h + c)z(abc) > A(bc + ca + ah)(b + c - a)(c + a - 5)(a + b ~ c)
holds for all real numbers a, b9 c.
(b) As above, but a, 6, c are assumed to be positive.
(c) As above, but a9 b$ c are assumed to satisfy
b + c - a > 0 , c + a-5>0,a-f-ft~c>G.
* * *

S O L U T I O N S

to otm^ode^ §<w pu&btoabbwi, n&w ^obtetLom, <yv a&w tiv^Lfhti^ <m> fm&t fyvobbetm,.

982. [1984: 291] Proposed by George Tsintsifas, Thessaloniki, Greece.

Let P and Q be interior points of triangle AiA2A%. For i = 1,2,3,
let PA.=x.9QA.=y.9 and let the distances from P and Q to the side

opposite A. be p. and q,9 respectively. Prove that

Vxiyi + 7x2y2 + Vx 3 y 3 ^ 2(Vpi<7i + 7p2<?2 + Vp3g3)»

When P = Q, this reduces to the well-known Erdbs-Mordell inequality. (See the
article by Clayton H. Dodge in this journal [1984: 274-281].)

Solution by M.S. Klamkin, University of Alberta.

We prove a generalization of the Erdbs-Mordell inequality which includes
the inequality proposed here as a special case.
Let P. (j = l,2,...,iz) denote any set of n points lying in the interior

or on the boundary of a given triangle A1A2A3 with respective sides ai$ a2>
a*. Let R. . and r. . (i = 1,2,3) denote the distances from P. to the vertices

A. and to the sides a.9 respectively. It is known that

R r
lf " 2/a3/€il) + r
3y(a2/ai)»
R
2j ~ r3/**/**> + ^ (a3/a2),

P ;3 > r^ .(a2/a3) + r 2 (ai/a3),

with equality in all three inequalities if and only if P. is the circumcenter

of AiA2A3. (Incidentally, these inequalities can be used to provide an

elegant proof of the Erdbs-Mordell inequality. Just add the three
inequalities and note that (aa/aj) + (a3/a2) £ 2, etc. See 0. Bottema et al,
Geometric Inequalities9 12.13 and 12.16.)
 - 29 -

We now employ the following case of Holder's inequality:

where all H., y . > Q. With all products running from j = 1 to / = 12, we then

have
IttJ/" > n(r 2 .(a3/ax) + r 3 / ( a 2 / a i ) ) 1 / D > (aj/ai)IIr^n + (a 2 / a i ) n r ^ n ,

1/n
I»2^ " > n(r 3/ (ai/a 2 ) + ^ ; . ( a 3 / a 2 ) ) > ( a ^ J H r ^ " + (a 3 /a 2 )nrj/ n ,

1/n
riffJ/"
g > n(r x .(a2/a3) + r 2 . ( a i / a 3 ) ) > (a 2 /a 3 )IIr^ n + (ai/a,)IIr^D.

Since (32/^3) + (83/^2) - 2 with equality if and only if a2 = #3, etc., it

follows by addition that

z n i?;/n > 2 z n ri/l3, (i)

i=i ./=i '•' 1=1 7=1 1J
with equality if and only if the triangle is equilateral and all the points P .
./
coincide with its center.
The proposer's inequality corresponds to the special case n = 2. 0
By using the following more general version of Holder's inequality,
,, ,, 1/n n 1/n . n 1/n .
/Dl /n2 n
(«i + vi) (u2 + v2) .*.(u + ir ) > n u . •'+ n v. 7 ,
n n
;=1 J /=1 •'
where /?, > 1 for A = l,2,...,n and (1/ni) + (1/^2) + ... + (1/n ) = 1, we
obtain the following generalization of (1):
3 n 1/n . 3 n 1/n
Z n J?. . J > 2 Z n r. . •'. (2)
1J
J=l./=1 '•' 1=1 ./=1
By specializing the points P., (2) leads to a host of triangle

inequalities- For example, take n = 2 and let Pi and P2 be the circumcenter

and incenter, respectively, of an acute triangle A1A2A3. R and r being the
circumradius and inradius, respectively, we then have, for i = 1,2,3,
il = R ,' J?.
i?., il = R cos A.2 , r i2
i20 = r esc (A 1./2) , r... ,Q = r.

This leads to the interesting triangle inequality

3 3
Z csc 1/l?2 (A./2) > 2 Z cos 1/l?1 A., (3)
2
i=l * i=l
where i?i, J22 > 1 and (1/ni) + (l/n2) = 1. Actually, one can show that the
limiting cases n% —• » or 122 —• °° are valid. These lead respectively to the
 - 30 -

known inequalities
3 3
Z esc (A./2) > 6 and 3 > 2 Z cos A..
J J
i=l i=l
Using the last two inequalities, plus
8 sin (Ai/2) sin (A2/2) sin (A3/2) < 1,
and then applying the power mean inequality, we obtain the following chain of
inequalities (where all sums and products run from i = 1 to 3):
Jlz
Z esc (A ./2) Z esc1/fl2(A./2)
> 4 > n c s c 1 / 3 U ./2) > 2
3
»1
£ cos A £ cos 1 / n i A .
> 4 2: 4 A 1 > 4 n cos 1 / 3 A (4)

Here 12%, 112 - 1 but need not be otherwise related. The terms with the cube
roots correspond to the limiting cases of the power mean inequality for which
n 2 —# oo andn i —4 00.
We now compare
Z cos A ni/»2
1/l,2 2/l?a
Z cBc (Ai/2) > 3 • 2 (5)

=
from ( 4 ) ? where now ( l / i ? i ) + ( 1 / ^ a ) 1? w i t h i n e q u a l i t y (3), We show t h a t
(3) i s s t r o n g e r than (5) f o r n% > 2, t h a t i s ,

Z cos1/UlA. \ni/n2
1/ni >2/l2 2 (6)
2 Z cos A. > 3

When I/122 i s r e p l a c e d by 1 - 1/ni, (6) reduces t o

1 / n i
3>2 £cos1/niA. ,
1
which follows from (4) for m > 2, For 1 < m < 2$ (3) is weaker than (5).
Finally, it should be noted that inequality (2) holds more generally as
an n triangle inequality. Consider n triangles A-, A^ A j . of sides a, ., a^ ,

a« . (./ = 1, •. • ,12), and n points Pi9...9P where P . is an interior or boundary

<Sj ' n j
point of triangle A, A .A0 . for each L Then if 5. . and r. . denote the
J
1/ 2; 3/ ij ij
distances from P . to the vertex A and the side a. ., respectively, inequality
(2) is still valid. The same proof using Holder*s inequality goes through,
the only change being that instead of coefficients of the form a%/a2 + ^2/^1*
 - 31 -

we get
1/n . n
n '1/ J "2.7 1/n"./
n + n
L-2./J J - l 1-1./J
which are still > 2 since the product of the two products is 1. There is
equality in the generalized inequality if and only if each of the n triangles
is equilateral and each P . coincides with the center of its triangle.

Also solved by WALTHER JANQJS, Ursul inengywnasium, Innsbruck, Austria;

ami the proposer.
* # #

983f [1984: 291] Proposed by D.J. Smeenk, Zaltbommel, The Netherlands.

Let AQAI ... A be an n-simplex in IK .

(a) If m, is the median through A,, prove that

2 2
nS, - T, ,
k k k *
where S, is the sum of the squares of all the edges meeting in A, , and 7\ is

the sum of the squares of all the edges not passing through A, .

(b) Deduce from (a), or otherwise, that if the medians of the simplex
are all equal, 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?

Solution by G.P. Henderson, Campbellcroft, Ontario.

(a) Let A 0 be the origin and let the coordinates of A. be (x
v
k ~~ ~ir >Xnkh
* = 1,2,...,/2. Then JBQ is the distance from the origin to the point whose ith
coordinate is I x.. /n. Hence
ik

n2ml = Z(Z x ) 2 Z Z Z x. . x ...

. . * i / ik
i k
i ; k •'
We have

To ^ 1 1 %•'- x / =
Z?" (x
5/' 2x. .x.. + x2.,)
i.7 ii: xk
j k i i j k

= n Z Z x* - Z Z Z x. .x.. .

Hence (a) is true for k = 0 and similarly for £ = 1,2, >n.

(b) Let £ be the sum of the squares of all the edges of the simplex.
Then S + 2\ = E for all Jr, and

(n + 1)S. - ff.
J:
 - 32 -

Therefore m, is independent of k if and only if S. is.

Also solved by O. BOTTEMA, Delft, The Netherlands; ROGER CUCULIERE,

Paris, France; M.S. KLAMKIN, University of Alberta, Edmonton, Alberta;
BASIL RENNIE, James Cook University, Australia; and GEORGE TSINTSIFAS,
Thessaloniki, Greece. The proposer noted that part (a) in the case n = 3 is
known (see Theorem 187 in N. Altshi Her-Court, Modern Pure Solid Geometry,
Chelsea, 1964). Theorem 295 of the same book gives a generalization of part
(b) in the case n = 3.
% * %

986. [1984: 292] Proposed by Stanley Rabinowitz, Digital Equipment

Corp., Nashua, New Hampshire.
Let
3
| "Z M Z
x = 4p + 7r + 4g - Vr ,
where p, q, r are integers and r > 0 is not a perfect square. If x is
rational, prove that p = q and x is integral.

Comment and solution by Leroy F. Meyers, Ohio State University, Columbus,

Ohio.
If x = 0 then p and q need not be equal* Indeed, p = -g*

Assume x £ 0, and let a = Jp + 7r and 5 = Jg - 7? , so that x = a + 5.

If x is rational, then so is
x3 = a3 + 5 3 -f 3a&(a + &) = P + <? + Sabx.
Hence ab is rational, and so is
a3d3 = pq ~ r + (q - p)Jr.
But since 7r is irrational, it must be that q = p, in which case a353 is the
integer p2 - r, and x is a root of the polynomial equation
(x3 - 2p)3 = 27a353x3
with integer coefficients and leading coefficient 1. Every rational root of
such an equation must be an integer.

Also solved (for x # 0) by ANDREW CUSUMANO, Great Neck$ N.Y.;

KARL DILCHER, Dalhousie University, Halifax, Nova Scotia; WALTHER JANOUS,
Ursulinengymnasium, Innsbruck, Austria; MARK KANTRQWITZ$ Maimonides School,
Brookline, Massachusetts; FRIEND H. KIERSTEAD, JR., Cuyahoga Falls, Ohio;
M.S. KLAMKIN, University of Alberta, Edmonton, Alberta; KEE-WAI LAU, Hong
Kong; ROBERT C. LYNESS, Suffolk, England; J.A. McCALLUM, Medicine Hat,
Alberta-, KENNETH M. WILKE, Topeka, Kansas; ami the proposer.
jfc t *
 - 33 -

987* [1984: 292] Proposed by Jack Garfunkel, Flushing, N.Y.

If triangle ABC is acute-angled, prove or disprove that
r \ . A ^ . B . C ^ 4 / - . . A . £ . C 1
(a) s i n 7* + s i n ^ +• s m •& > 77 {1 + s i n ^ s i n ^ s i n ?j}
,u A ^ B ^ C ^ 4 fl ^ . A . B > Cx
(b) cos K + cos ?j + cos 7* > — {1 + sin ^ sin -& s m 77}.
Solution to part (a)fcyBasil Rennie, James Cook University, Australia.
~lfV3 - 11
Let /J = sin ^ — I • We prove a result slightly stronger than the one

suggested, namely: if the largest angle of triangle ABC is < n ~ 40

o
(approximately 94.12 ), then inequality (a) holds.
Lemma 1. I f 0 < v < u < 5 , then 4 cosv cos2u - 3 sinu > 0.

Proof. Let g(u,v) = 4 cosy cos2u ~ 3sinu. Then for 0 < u < pr and for fixed
v, g(u,v) decreases as u increases. Thus
g(u,v) > g(g,y) = 2 cosy - | > 2 cos g - | - y3 - | > 0. D

Lemma 2. If /? < u < g and s = sinu then 2s 2 + 2s - 1 > 0.

Proof. The positive root of the quadratic 2x2 + 2x - 1 = 0 is at

x = JS ~
- —1 - sin/} < sinu = s. 0

Now assume 0 < A < B < C < i r - 4 / $ . Then we can w r i t e A = 2u ~ 2v,

B = 2u + 2v, and C = ir - 4u, where 0 < y < a and u > 0. Furthermore, from
C > ^ we s e e t h a t u < ^ . Using elementary t r i g o n o m e t r y ,

sin 7> + sin ?r + sin 77 = 2 sinu cosy + cos2u

z z z
and
s i n 77 s i n ?j s i n 77 = 7j(cos2y - cos2u)cos2u.

Let
A H C1 A B C
F(u,v) = 3 ( s i n ^ + s i n 77 + s i n 77) - 4 - 4 s i n 77 s i n 77 s i n ?r
= 6 sinu cosy + 3 cos2u - 4 - 2(cos2y - cos2u)cos2u.
Then
&F
s~ = -6 sinu sinv + 4 sin2v cos2u
<?y
= 2 sinv(4 cosv cos2u - 3 sinu)
> 0 by Lemma 1.
Thus
F(u9v) > F(u,0) = 2 cos22u + cos2u + 6 sinu - 4.
Now put s = sinu; then cos2u = 1 - 2s 2 , so
 - 34 -

F(u90) - 2(1 - 2s 2 ) 2 + 1 - 2s 2 + 6s - 4
- 8s 4 - 10s2 + 6s - 1
= (2s - l)2(2s2 + 2s - 1)
> 0 by Lemma 2.
Hence F(u9v) > 0f and (a) follows.

Solution to part (b) by G.P. Henderson, Campbellcroft, Ontario; proof

adapted by the editor from Rennie's proof of part (a).
We prove that (b) holds if all angles of the triangle are
< n - 4cos (73fc)f where t is the unique real root of the polynomial
12x3 + 12x2 - 3x - 4.
Let 0 < A < B < C, and write A = 2a - 2v, B = 2a + 2v9 C = n - 4a, so
that 0 < v < a < 5 as before. Then
6
^ Q Q
cos ^ + cos H- + cos «• = 2 cosa COSF 4- sin2a
and again
A J3 C 1
sin x> sin «• sin * = iy(cos2^ - cos2o)cos2a.
Let
F(u9v) = 75(cos s + cos «- + cos ») - 4 - 4 sin ?? sin ?> sin ^

= 273 cosa cosv + 75 sin2a - 4 - 2(cos2vr - cos2o)cos2a.

Then
OF
-%- - -275 cosa sinv + 4 sin2v cos2ii
ov
= 2 sinv(4 cosv cos2o - 73 cosa)*
Letting g(u9v) = 4 cosv cos2a - 75 cosa? we see that g decreases as a function
of v, and thus for fixed u9
g(u9v) > g(u9u) = cosa(4 cos2a - 75)*
But
h(u) = 4 cos2a - 75
is positive for 0 < u < ^$ since h decreases and A(g) = 2 - 73 > 0* Thus

g(u9v) > 0 for 0 < v S u < 2 S so F(u9v) increases as a function of v. Hence

F(u,v) > F(u,0) = cosa(275 + 273 sino - 12 cosu + 8 cos3a).

Putting sina = Jl - cos2a and rationalizing! we get that

273 + 273 sina - 12 coso + 8 cos3a > 0
provided that
16 cos5a - 48 cos3a + 873 cos2a + 39 coso - 1275 < 0.
Set cosa = 75t; then this is equivalent to
 - 35 -

0 > 48ts - 48t3 + 8fc2 + I3t - 4 = (2fc - l)2(12t3 + 12t2 - 3t - 4)

which will hold provided t is less than or equal to the unique real root of
the polynomial
12x3 + 12x2 ~ 3x - 4,

approximately x = 0.551. Thus if 5 > u > cos_1(73x) « 17.34°, F(u9v) will be

o
0
> 0. This amounts to insisting that C < 110.64 .
Henderson found the strong solutions for both (a) and (b) using
-1 f72 + 41
noncalculus techniques. He also pointed out that if C > 2 sin ^—
o
* 128.94 , then (a) will fail for all A and B (assuming A < B < C), and that
if 94.12 < C < 128.94° then (a) will hold provided that B - A is small
enough. One partial solution was also received.
# # t

988. [1984: 292] Proposed by J.T. Groenmant Arnhem, The Netherlands.

Prove that
,~ 2trl , - 2nl A
+ sec =
Z s e c YT~~
U
** TT"
iJ
"•
1=1 1=1

Solution by Walther Janous, Ursulinengymnasium, Innsbruck, Austria.

We show more generally that
" 2ir* "I1 2TT* „
Z sec x—=r + 2 sec
. , 4n-l " 4n+l
1=1 1=1
12
2nl
for all n > 0. To do this we compute S = Z sec 0 ,., (a known series).
1=1
From the identity
n 2i2+l-21 . 21
(2n++l)*=
cos(Zi2 l)x = - D 1| ^
£K(-1)
1
"""rtl ]
L
1 cos
J
x sin x
1=0 ^
2nt
(provable with DeMoivre's theorem) it follows that cos 5~TT> * = 0,1,... ,2/2,

are all the roots of the polynomial

n
1 = I (-1)* [X1] *2n+1~2k (1 - - 2 ) j
k=0
2irt
Thus sec g <., t = 0,1,..., 2n, are all the roots of the polynomial

1
iZ <-»* PS1) B)
= *=0 V - h) •
which simplifies to
 36 -

2n+l "
x =
I (-D* (*£»] U* - l,'
k=0
or
2/3+1 n-1
X (-l)"(2n + l)(x2 - 1 ) " - Z (-1)* f2"!1] (x2 - 1)
L
*=0 ^*J
The negative of the coefficient of x in this polynomial will be the sum of
its roots, and hence
Z sec
jSr = (-D"(2n + 1).

«.. 2ir* 2w(2n+l-*) ,.*.!.,.

weg e t t h a t
Since s e c K-JT- = s e c — V ~ + T — '
2/7
2wA " 2irl
£ s e c ^ r = secO + 2 2 s e c ^ L = 1 + 2S ,
2
*=0 "+1 *=0 2 n + 1
so Sn = i|'(-l)"(2n + 1) - l ] = f n . i f
" i s e ^ n } . Hence S0 . + S 9 = 0 ,
J2 2[ J \ -ii-1 i f 12 i s odd | 2w-l 2J? '
as claimed.

Also proved by CURTIS COOPER, Central Missouri State University,

Warrensburg, Missouri; KARL DILCHER, Dalhousie University, Halifax, Nova
Scotia; M.S. KLAMKIN, University of Alberta, Edmonton, Alberta; ROBERT LYNESS,
Suffolk, England; LEROY F. MEYERS, The Ohio State University, Columbus, Ohio;
VEDULA N. MURTY, Pennsylvania State University, Middletown, Pennsylvania; and
the proposer. Most solvers did the general case* Meyers comments that there
m
9 k 2
is no such simple formula for Z esc 0 Ll , as may be seen from the sums —
2mhl
1=1 ' 73
242 + 75
and for m = 1 and m = 2, respectively.
V5

989» [1984: 292] Proposed by Kurt Schiffler, Schorndorf, Federal

Republic of Germany.
Let If be the orthocenter of triangle ABC. Prove that the Euler
lines of triangles ABC, BCH, CAH, and ABH are all concurrent. In what
remarkable point of triangle ABC do they concur?

Solution by Basil Rennie, James Cook University, Australia.

ff
For any triangle there is a pedal triangle" formed by the feet of the
perpendiculars from each vertex to the opposite side? and a "nine-point"
circle through these three feet. The four triangles mentioned all have the
 - 37 -

same pedal triangle and so they share the same nine-point circle. Since the
Euler line of any triangle contains the center of the nine-point circle, all
four Euler lines meet in this point.

Also solved by JORDI DOU, Barcelona, Spain; ROLAND EDDY, Memorial

University, St. John's, Newfoundland; H. FUKAGAWA, Yokosuka High School,
Aichi, Japan; J.T. GROENMAN, Arnhem, The Netherlands; WALTHER JANOUS,
Ursulinengymnasium, Innsbruck, Austria; M.S. KLAMKIN, University of Alberta,
Edmonton, Alberta; ROBERT C. LYNESS, Suffolk, England; D.J. SMEENK,
Zaltbommel, The Net her 1 amis; DAN SOKOWWSKY, Brooklyn, N.Y.; and
GEORGE TSINTSIFAS, Thessaloniki, Greece. Klamkin observed that the result is
a simple consequence of the theorem "The four triangles of an orthocentric
system have a common nine-point circle" (see R.A. Johnson, Advanced Euclidean
Geometry, Dover, 1960, p.197). Most proofs received were along these lines;
the most succinct is reproduced above.
% * *

A MESSAGE FROM THE EDITOR

With this issue the editorship of Crux Mathematicorum changes. I would

like to echo Ken Williams' words of last month in wishing Leo Sauve a quick
recovery and a long and happy retirement. And may we not have seen the last
of him in these pages!
In preparing this first issue, I have been struck by the immense amount
of knowledge and effort Leo has brought to Crux. Both are difficult to
replace, but knowledge the more so. Undoubtedly there will occur, for a while
at least, dubious problems, missed references, and other assorted booboos. I
hope you will patiently, but firmly, let me know of my errors.
As it has these past eleven years, Crux will continue to increase in
interest and value to you, its readers.

Bill Sands