Hungarian Problem Book I (Number 11) (Bk. 1) (E. Rapaport)

HUNGARIAN PROBLEM BOOK I
PROBLEM BOOK I
based on the
Etvs Competitions
18981905
TRANSLATED BY
ELVIRA RAPAPORT
TRANSLATED BY
ELVIRA RAPAPORT
Jzsef Krschk (18641933) was born and educated

in Hungary. He was professor of mathematics at the
Polytechnic University in Budapest, member of the
Hungarian Academy and permanent member of the
Examination Board for prospective high school teachers of mathematics
His many contributions to mathematics include
work in the calculus of variations, in algebra and in
number theory. He used his great pedagogical skill in
developing and teaching an exceptionally good mathematics course for beginning engineering students.
He also gave courses for future high school teachers,
mainly in elementary geometry and in geometrical constructions. Several of his papers deal with the teaching
and popularization of mathematics. His devotion to intelligently guided problem solving
is illustrated by the famous problem book which forms the basis of the present volume.
HUNGARIAN
The Etvs Contests in elementary mathematics have been open to Hungarian students in their last year of high school ever since 1894. They are famous for the simplicity of the concepts employed, the mathematical depth reached, and the diversity of elementary mathematical elds touched. But perhaps their most remarkable
feature is the inuence that they, together with a mathematics journal for students,
seem to have had on the young people of that small country. Among the winners
of the rst eleven contests (i.e., those contained in this volume) many turned into
scientists of international fame; e.g., L. Fejr, T. von Krmn, D. Knig, M. Riesz.
Among the winners of the next twenty contests (i.e., those contained in volume
12) are G. Szeg, T. Rad, E. Teller; all three are well-known in the United States,
where they now reside. This translation of the Etvs Contest Problems from
18941928 is based on the revised Hungarian edition of J. Krschks original
compilation. Krschk combined his excellence in mathematics with his interest
in education when he supplied the elegant solutions and illuminating explanations.
MAA
TRANSLATED BY ELVIRA RAPAPORT
ISBN 978-0-88385-611-6
MATHEMATICAL ASSOCIATION OF AMERICA
9 780883 85611 6
11
This content downloaded from 130.64.11.153 on Sat, 27 Aug 2016 06:41:30 UTC
All use subject to http://about.jstor.org/terms
ANNELI LAX NEW

MATHEMATICAL LIBRARY

BASED ON THIZ EOTVOS COMPETITIONS, 1894-1905
NEW MATHEMATICAL LIBRARY

PUBLISHED BY
THEMATHEMATICAL
ASSOCIATION
OF AMERICA
Editorial Committee
Ivan Niven, Chairman (1978-77)
University of Oregon
Anneli Lax, Editor
New York University
W.G.Chinn (1977-79) City CoUege of Sun Francisco

Basil Gordon (1975-76) University of Califmkz, Los Angela
M. M. Schiffer (1976-78) Stunford U n i m i t y
The New Mathematical Libray (NML) was begun in 1961 by
the School Mathematics Study Croup to make avaihble to high
school students short expository books on various topb not
usually covered in the high school s y l b h . In a decade the
NML matured into a steadily growing se&?s of some twenty
titles of interat not only to the originally intended audience,
but to colkge students and teache7s at all levels. PTeoioUsly
published by Random House and L. W. Singer, the NlML
became a publicatiun series of the MathematicalAssodation of
America (MAA) in 1975. Under the awpices of the MAA the
NML will continue to grow and wiU remuin dedicated to its
ori@nal and erpanded purposes.
HUNGARIAN
PROBLEM BOOK I
BASED ON THE EOTV&3COMPETITIONS, 18944905
REVISED AND EDITED BY
G. HAJ65, G. NEUKOMM, J. S,IORIGINALLY COMPILED BY JbZSEF 'KORSCEIAE
translated by
Elvira Rapaport
Brooklyn Polytechnic InstitUte
11
MATHEMATICAL ASSOCIATION
OF AMERICA
Copyright 1963 by The Mathematical Association of America (Inc.)

All rights reserved under International
and Pan-American Copyright Conventions.
Published in Washington, D.C. by
The Mathematical Association of America
Library of Congress Catalog Card Number: 63-16149
Print ISBN 978-0-88385-611-6
Electronic ISBN 978-0-88385-927-8
Manufactured in the United States of America
Note to the Reader
his book is one of a series written by professional mathematicians in order to make some important mathematical ideas
interesting and understandable to a large audience of high school
students and laymen. Most of the volumes in the New Muthemuticul
Library cover topics not usually included in the high school curriculum; they vary in difficulty, and, even within a single book, some
parts require a greater degree of concentration than others. Thus,
while you need little technical knowledge to understand most of
these books, you will have to make an intellectual effort.
If you have so far encountered mathematics only in classroom
work, you should keep in mind that a book on mathematics cannot
be read quickly. Nor must you expect to understand all parts of the
book on first reading. You should feel free to skip complicated parts
and return to them later; often an argument will be clarified by a
subsequent remark. On the other hand, sections containing thoroughly familiar material may be read very quickly.
The best way to learn mathematics is to do mathematics, and each
book includes problems, some of which may require considerable
thought. You are urged to acquire the habit of reading with paper
and pencil in hand; in this way, mathematics will become increasingly
meaningful to you.
The authors and editorial committee are interested in reactions to
the books in this series and hope that you will write to: Anneli Lax,
Editor, New Mathematical Library, NEW YORKUNIVERSITY,
THE
COURANT
INSTITUTE
OF MATHEMATICAL
SCIENCES,
251 Mercer Street,
New York, N. Y. 10012.
The Editors
NEW MATHEMATICAL LIBRARY
1 Numbers: Rational and Irrational by Ivan Niven

2 What is Calculus About? by W. W. Sawyer
3 An Introduction to Inequalities by E. F. Beckenlnzch and R . Bellman
4 Geometric Inequalities by N . D. iluazarinoff
5 The Contest Problem Book I Annual High school Mathematics Examinations
1950-1960.Compiled and with solutions by Charla T. Sdkind
6 The Lore of Large Numbers, by P. J. Dwis
7 Uses of Infinity by Leo Zippin
8 Geometric Transformations I by I. M. Yaglom, translated by A. Shieldr
9 Continued Fractions by Carl D. Ol&
10 Graphs and Their Uses by Oystein Ore
Hungarian Problem Books I and 11, Based on the E5tvBS
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Competitions 1894-1905 and 1906-1928, translated by E. Rapaport

Episodes from the Early History of Mathematics by A. Aaboe
Groups and Their Graphs by I. Grossman and W. Magnw

The Mathematics of Choice by Ivan Niven
From Pythagoras to Einstein by K . 0. Friedrichs
The Contest Problem Book I1 Annual High School Mathematics Examinations
1961-1965.Compiled and with solutions by Charles T. Salkind
First Concepts of Topology by W.G.Chinn and N. E. Steenrod
Geometry Revisited by H. S. M.Coxeter and S. L. Greitzer
Invitation to Number Theory by Oystein Ore
Geometric Transformations I1 by I. M. Yaglom, translated by A . Shieldr
Elementary Cryptanalysis-A Mathematical Approach by A. Sinkou
Ingenuity in Mathematics by Ross Honsberger
Geometric Transformations I11 by I. M. Yaglom, translated by A. Shenirzer
The Contest Problem Book I11 Annual High school Mathematics Examinations
1966-1972.Compiled and with solutions by C. T. Salkind and J. M. Earl
Mathematical Methods in Science by George Ph'ya
International Mathematical Olympiads-1959-1977. Compiled and with solutions
by S. L. Greitzer
The Mathematics of Games and Gambling by Edward W. Packel
The Contest Problem Book IV Annual High School Mathematics Examinations
1973-1982. Compiled and with solutions by R. A. Artino, A . M . Gaglione arrd
N. Shell
The Role of Mathematics in Science by M . M . Schiffer and L. Bowden
International Mathematical Olympiads 1978-1985 and forty supplementary
problems. Compiled and with solutions by Murray S. Klamkin.
Riddles of the Sphinx by Martin Gardner
U.S.A. Mathematical Olympiads 1972-1986. Compiled and with solutions by
Murray S. Klamkin.
Graphs and Their Uses, by @stein Ore.Revised and updated by Robin J. Whon.
Other titles in preparation.
CONTENTS
Editors Note
From the Hungarkn Prefaces
Preface t Q the American Edition
Problems
solutiom
1894: 1,2,3
1895: 1,2,3
1896: 1,2,3
1897: 1,2,3
1898: 1,2,3
1899: 1,2,3
1900: 1, 2,3
1901: 1, 2,3
1902: 1,2,3
1903: 1,2,3
1904: 1,2,3
1905: 1,2,3
17, 21, 22
24, 29, 33
36, 38, 39
46, 4, 54
56, 58, 60
62, 68, 70
71, 72, 73
74, 76, 78
81, 86, 92
94, 95, 97
99, loo, loo
101,103,104
Classificationof Problems
107
List of Explanatory Notes
109
List of Biographical Notes
110
List of Contest Winners
111

BASED ON THE EOTVOS COMPETITIONS, 1894-1905
From the Hungarian Prefaces

The first of these contests was held in 1894 by the Mathematical and
Physical Society (of Hungary) in honor of its founder and president, the
distinguished physicist Baron LorAnd EGtvGs, who became minister of
education that year. To commemorate the event, the contests are given
every fall and are open to high school graduates of that year. The contestants work in classrooms under supervision; the Society selects the
two best papers, and the awardfirst and second EGt* P r i z e a r e
given to the winners by the president himseIf at the next Session of the
Society.
The present volume, appearing on the tenth anniversary of E6tvGs
death, contains the contests held to date. While it utilizes the winners
work, the solutions are in general not those found by the students.
The names of the winners are listed; their papers appeared in full in
Matmatikai I% Fizikai La$ok, the Journal of the Society; here, however,
the solutions were changed to suit the didactic aim of the book.
Some of my notes give definitions and proofs of theorems used in the
solutions. Others serve to point out the connection between problems
and famous results in iiterature. In some instances I was able to give EL
glimpse of the essence of an entire subject matter; in others the mere
statement of a general theorem had to suffice.. .
There are few prerequisites. A person who has learned to solve quadratic equations and knows plane geometry can solve many of the problems. If he also knows trigonometry, he can solve most of them. So little
of the material taught in the last two years of high school (in Hungary)
is needed here that a younger student can d y learn it from books.
...
HUNGARIAN PROBLEM BOOK
However, this book is meant not only for students and teachers. Anyone who retained an interest in mathematics in his adult life can find
things of note and value here and will be gratified to see how much can
be achieved with the elementary material to which high schools must
restrict themselves.
How should the reader use this book? All I can say is: without frenzy.
With a serious interest and perseverence, everyone will find the way best
suited to him in order to benefit from the varied material contained in
it ....
J6zsef Kiirschhk
Budapest, April 9th, 1929
Problems of the Mathematics Contests, edited by J6zsef Kiirschhk, was

published originally in 1929. The first edition was quickly sold out and
the Ministry of Education commissioned a new edition. We are undertaking this work with pleasure and in the hope of contributing to the
attainment of KiirschWs goal. This volume will soon be followed by
another containing a similar treatment of the contests held since.
The new edition required certain changes. By and large, Kiirschhks
approach was retained, especially his notes which were meant to widen
the readers horizon. We augmented these notes here and there, changing
them only to conform to present-day high school curricula. For example,
we added permutations, the binomial theorem, half-angle formulas, etc.
We added a few new notes. We included some new solutions that seemed
strikingly simple or ingenious, and so we deviated from the contestants
work more than the first edition did. ...
We made a few technical changes to facilitate handling. Thus, we
repeated the problems before giving the solution and we unified the
notation, thereby necessitating alterations in and additions to the
figures.. ..
We sincerely hope to have come up to expectation and that this work
will give pleasure and profit to many.
Gyorgy Hajds
Gyula Neukomm
J h o s Surhyi
Budapest, September, 1955
Preface to the American Edition

In recent times much effort has been devoted to the improvement of
mathematics teaching on all levels. Thus it is only natural that we search
for further stimulants and improvements in this direction, as well as for
new means of discovering and developing the dormant abilities which
may exist in our society. The Monograph Project of the School Mathematics Study Group has such a purpose. The present translation of the
Hungarian problem collection of the Eotvos Competitionserves this goal.
Since I am one of the few still existing links between the present mathematical generation and an older one that witnessed the first phase of the
interesting development of this competition, I was asked to write a few
introductory words to the English version of this collection.
The E6tv& Competition was organized in Hungary in 1894 and
played a remarkable role in the development of mathematics in that
small c0untry.f The competition was open to all freshmen entering the
university; the publication of the problems and the names of the winners
was from the beginning a public event of first class interest. Among the
winners during the period of the first decade of the competition were
such men as FejQ, von KtmnAn, Haar, Riesz, and numerous others who
later became internationally known. With some short interruptions due
to wars and related conditions the competition has been carried on to
the present day, though the name was changed, and the organization
and scope of the competition have become much broader in recent years.
The essence however has remained the same. The problems are almost all
t Before the h t world war Hungary had 19 million inhabitants;at present it has
about 10 million.
5
from high school material (no calculus is included), they are of an elementary character, but rather difficult, and their solution requires a
certain degree of insight and creative ability. Any amount of aid in the
form of books or notes is permitted.
Mathematics is a human activity almost as diverse as the human mind
itself. Therefore it seems impossible to design absolutely certain and
effective means and methods for the stimulation of mathematics on a
large scale. While the competitive idea seems to be a powerful stimulant,
it is interesting to observe that it was and is still almost completely
absent from academic life in Germany although mathematics has flourished in that country throughout the last two hundred years. The organization of the Eotvos Competition in Hungary was probably suggested
by British and French examples that had existed in those countries for
a long time. We mention in particular the Mathematical Tripos in
Cambridge, England and the Conco~rs~
examination problems for
admission to the Grandes ficoles in France. These early examples
suggest also that some sort of preparation is essential to arouse public
interest, to attract the best competitors and to give them proper recognition. In England the participation in the Tripos is preceded by systematic
coaching, and in France the public schools offer facilities to prepare for
the "Contours" examinations. In Hungary a similar objective was
achieved by a Journal published primarily for high school students as
another natural stimulant to the students preparation for participation
in the competition upon entering the university. t
The Journal was organized almost simultaneously with the competition, i.e. in 1894, by D h i e l Arany; for many years it was edited by the
able high school teacher L h z 6 Rhcz$ and later by various other teachers
of high quality. The articles were supplied partly by teachers and partly
by mathematicians affiliated with the university, mostly younger persons.
The Journal carried articles primarily from elementary mathematics,
much triangle geometry, some projective and descriptive geometry,
algebra and occasionally some number theory, later also some ventures
into calculus. But the most important and most fertile part was the
t A good account of the EdtvUs Competition and of the Journal is given in an article
by Tibor Rad6: On mathematical life in Hungary, American Mathematical MontMy,
vol. 39 (1932), pp. 85-90. (One slight correction has to be made on p. 87, line 6:
There was a girl winner, first prize, 1908.)
2 His name will go down in history for a second reason: RAcz was the teacher of J.
von Neumann in high school. Cf. the Obituary Note by S. Ulam, Bulletin ofthe American Mathematical Sockty, vol. 64 (1958), pp. 149; on p. 2 the name Rhcz appears in
distorted spelling.
PREFACE TO T H E AMERICAN E D I T I O N
problem section; it occupied a large part of the content and was essentially written for the students and by the students. The best solution
sent in was printed with the name and school of the author, and a list
of the others who sent in correct solutions was given.
I remember vividly the time when I participated in this phase of the
Journal (in the years between 1908 and 1912);I would wait eagerly for
the arrival of the monthly issue and my first concern was to look at the
problem section, almost breathlessly, and to start grappling with the
problems without delay. The names of the others who were in the same
business were quickly known to me and frequently I read with considerable envy how they had succeeded with some problems which I
could not handle with complete success, or how they had found a better
solution (that is, simpler, more elegant or wittier) than the one I had
sent in. The following story may not be accurate in all details but it is
certainly revealing:
The time is about 1940, the scene is one of the infamous labor camps
of fascist Hungary just at the beginning of its pathetic transformation
from semi-dictatorship to the cannibalism of the Nazi pattern. These
camps were populated mostly by Jewish youth forced to carry out some
perfectly useless tasks. One young man (at present one of the leading
mathematicians of Hungary) was in the camp; let us call him Mr. X.
He was panting under the load of a heavy beam when the sergeant
shouted at him in a not too complimentary manner, addressing him by
his last name. The supervising officer stood nearby, just a few steps
away, and said: Say, did I hear right, your name is X? Yes, was the
answer. Are you by chance the same X who worked years ago in the
High School Journal? Yes, was again the answer. You know, you
solved more, and more difiicult problems than any one of us and we were
very envious of you. The end of the story is that Mr. X received more
lenient treatment in the camp and later even had some mathematical
contact with the all-powerful officer.
The profound interest which these young men took in the Journal was
decisive in many of their lives. The intensive preoccupation with interesting problems of simple and elementary character and the effort of
finding clear and complete answers gave them a new experience, the
taste of creative intellectual adventure. Thus they were bound finally
and unalterably to the jealous mistress that mathematics is. There remained still the question of what special studies to undertake, whether
it should be mathematics or physics or engineering; but this was after
all a secondary matter; the main road was charted for life. We may think
H U N G A R I A N PROBLEM BOOK
of the adage of Kronecker who compares mathematicians with lotus

eaters: Wer einmal von dieser Kost etwas m sich genommen hat, kann
nie mehr davon lassen. (He who has once tasted of this fruit can never
more forswear it.)
And a final observation. We should not forget that the solution of any
worth-while problem very rarely comes to us easily and without hard
work; it is rather the result of intellectual effort of days or weeks or
months. Why should the young mind be willing to make this supreme
effort? The explanation is probably the instinctive preference for certain
values, that is, the attitude which rates intellectual effort and spiritual
achievement higher than material advantage. Such a valuation can only
be the result of a long cultural development of environment and public
spirit which is difficult to accelerate by governmental aid or even by
more intensive training in mathematics. The most effective means may
consist of transmitting to the young mind the beauty of intellectual work
and the feeling of satisfaction following a great and successful mental
effort. The hope is justified that the present book might aid exactly in
this respect and that it represents a good step in the right direction.
Stanford University, February, 1961
Ghbor Szeg8
Problems
1894 Competition
1894/1.
Prove that the expressions

2x
+ 3y
and 9x+ Sy
are divisible by 17 for the same set of integral values of x and y.

1894/2. Given a circle and two points, P and Q: construct an inscribed
right triangle such that one of its legs goes through the given point P
and the other through the given point Q. For what position of the points
P and Q is this construction impossible?
1894/3. The lengths of the sides of a triangle form an arithmetic
progression with difference d. The area of the triangle is t. Find the sides
and angles of thii triangle. Solve this problem for the case d = 1 and
t = 6.
1895 Competition
1895/1. Prove that there are 2(2n-1 - 1) ways of dealing n cards to
two persons. (The players may receive unequal numbers of cards.)
1895/2. Given a right triangle ABC, construct a point N inside the
triangle such that the angles NBC, NCA and N A B are equal.
9
10
1895/3. Given the following information about a triangle: the radius

R of its circumscribed circle, the length G of one of its sides, and the ratio
a/b of the lengths of the other two sides; determine all three sides and
angles of this triangle.
1896 Competition
1896/1.
Prove that
2 k log 2,
where n is a natural number and K the number of distinct primes that
log 72
divide n.
1896/2.
Prove that the equations
x2 - 3xy
and
+ 2y2+ x - y = 0
x2-2xy+y2-5x+7y=o
imply the equation
xy - 12x
+ 15y = 0.
1896/3. Construct a triangle, given the feet of its altitudes. Express

the lengths of the sides of the solution triangle Y in terms of the lengths
of the sides of the triangle X whose vertices are the feet of the altitudes
of triangle Y .
1897 Competition
1897/1. Prove, for the angles a,p and y of a right triangle, the following relation:
sin a sinp sin(a+)
1897/2.
then
+ sin p sin y sin(p-7) + sin y sin a sin(y-a)

+ sin (a- p) sin (p - y) sin (y - a) = 0.
Show that, if a,/3 and y are angles of an arbitrary triangle,
sin-sin-sin2
1
< -.
PROBLEMS
11
1897/3. Let ABCD be a rectangle and let M,N and P,Q be the points
of intersection of some line e with the sides AB, CD and AD, BC, respectively (or their extensions). Given the points M,N,P,Q and the length
p of the side AB, construct the rectangle. Under what conditions can
this problem be solved, and how many solutions does it have?
1898 Competition
1898/1.
by 3.
Determine all positive integers vz for which 2"
+ 1is divisible
1898/2. Prove the following theorem: If two triangles have a common

angle, then the sum of the sines of the angles will be larger in that triangle
where the difference of the remaining two angles is smaller.
On the basis of this theorem, determine the shape of that triangle for
which the sum of the sines of its angles is a maximum.
1898/3. Let A , B, C, D be four given p
i
n
t
s on a straight line e. Construct a square such that two of its parallel sides (or their extensions)
go through A and B respectively, and the other two sides (or their extensions) go through C and D, respectively.
1899 Competition
1899/1.
The points Ao, Al, A?, As, A4 divide a unit circle (circle of
radius 1) into five equal parts. Prove that the chords AoA1, A d 2 satisfy
(AoA1.AtJ2)2 = 5.
1899/2.
Let x1 and x2 be the roots of the equation
x2 -
(a
+ d ) x + ad - bc = 0.
Show that x i and x f are the roots of
r" - (a8 + (28 + 3abc + 3bcd)y + (ad - bc)* = 0.
1899/3.
Prove that, for any natural number n, the expression

A = 2903" - 803" - 464"
+ 261"
is divisible by 1897.
12
1900 Competition
1900/1. Let a, b, c, d be fixed integers with d not divisible by 5. Assume that m is an integer for which
ama
+ bm2 + cm + d
is divisible by 5. Prove that there exists an integer n for which

is also divisible by 5.
dn8
+ cn2+ blz + a
1900/2. Construct a triangle ABC, given the length c of its side AB,
the radius I of its inscribed circle, and the radius r, of its ex-circle tangent
to the side A B and to the extensions of BC and CA.
1900/3. A cliff is 300 meters high. Consider two free-falling raindrops
such that the second one leaves the top of the cliff when the t h t one has
already fallen 0.001 millimeters. What is the distance between the drops
at the moment the first hits the ground? (Compute the answer to within
0.1 mm. Neglect air resistance, etc.)
1901 Competition
1901/1.
Prove that, for any positive integer n,

1"
+ 2" + 3" + 4"
is divisible by 5 if and only if n is not divisible by 4.

1901/2.
Lf
u = cot 22'30',
v =
sin 22'30'
'
prove that u satisfies a quadratic and v a quartic (4th degree) equation

with integral coefficients and with leading coefficients 1.
1901/3. Let a and b be two natural numbers whose greatest common
divisor (g.c.d.) is d. Prove that exactly d of the numbers
a, 2a, 3a,
0 . 0 ,
( b - l ) a , ba
are divisible by b.
13
PROBLEMS
1902 Competition
1902/1.
Prove that any quadratic expression

Q ( x ) = AX2
+ Bx + C
a) can be put uniquely into the form
Q(x> = k
x(x-
1.2
1)
i-lx
+ m,
where k, 1, m depend on the coefficients A , B, C, and
b) Q ( x ) takes on integral values for every integer x if and only if K, 1, m

are integers.
1902/2. Let S be a given sphere with center 0 and radius Y. Let P be
any point outside the sphere S, and let S' be the sphere with center P
and radius PO. Denote by F the area of the surface of the part of 5"
that lies inside S. Prove that F is independent of the particular point P
chosen.
1902/3. The area T and an angle y of a triangle are given. Determine
the lengths of the sides a and b so that the side c, opposite the angle y,
is as short as possible.
1903 Competition
1903/1. Let 12 = 2F1(2p

l),and let 21, - 1 be a prime number.
Prove that the sum of all (positive) divisors of n (not including f i itself)
is exactly 12.
1903/2. For a given pair of values x and y satisfying x = sin a,
y = sin 0, there can be four different values of z = sin (a P).
a) Set up a relation between x, y and z not involving trigonometric
functions or radicals.
b) Find those pairs of values ( x , y) for which z = sin (a P) takes on
fewer than four distinct values.
1903/3. Let A , B, C, D be the vertices of a rhombus; let k1 be the

circle through B,C,and D; let kz be the circle through A , C,and D;let
H U N G A R I A N P R O B L E M BOOK
14
kg be the circle through A , B, and D;and let k4 be the circle through A ,

B, and C.Prove that the tangents to k1 and ks at B form the same angle
as the tangents to kz and k4 at A.
1904 Competition
1904/1. Prove that, if a pentagon (five-sided polygon) inscribed in a
circle has equal angles, then its sides are equal.
1904/2. If a is a natural number, show that the number of positive
integral solutions of the indeterminate equation
xi f 2%
(1)
+ 3x8 + +.nXn = U
*-*
is equal to the number of non-negative integral solutions of
[By a solution of equation (1),we mean a set of numbers {XI,x2,

which satisfies equation (1)
.]
0 ,
xn)
1904/3. Let &Az and BIB2be the diagonals of a rectangle, and let 0
be its center. Find and construct the set of all points P that satjsfy
simultaneously the four inequalities
AIP
> OP,
A2P
> OP,
BiP
> OP,
B2P
> OP.
1905 Competition
1905/1. For given positive integers n and p, find necessary and sufficient conditions for the system of equations
x+py=n,
x+y=p
to have a solution (x,y, z) of positive integers. Prove also that there is

at most one such solution.
1905/2. Divide the unit square into 9 equal squares by means of two
pairs of lines parallel to the sides (see Figure 38). Mow remove the central
square. Treat the remaining 8 squares the same way, and repeat the
process n times.
PROBLEMS
15
a) How many squares of side length 1/3" remain?

b) What is the sum of the areas of the removed squares as rc becomes
infinite?
1905/3. Let Ct be any point on side AB of a triangle ABC (see

Figure 39), and draw CIC.Let A1 be the intersectionof BC extendedand
the line through A parallel to CC1; similarly let Bl be the intersection of
AC extended and the line through B parallel to CC1. Prove that
AB'
1
1
-+-=AA1 BB1
CC1
'C',
c1
Solutions
1894 Competition
1894/1.
Prove that the expressions

2 x + 3y and 9 x + Sy
are divisible by 17 for the same set of integral values of x and y .
a) First let us determine those integral values of

x and y for which the expression 2x 3 y is equal to some given integer k.
From the equation
First Solution.
+
2x + 3y = k
(1)
we obtain
x=-- rk - 3 y - - y + - . k - Y
(2)
From ( 2 ) we see that x is an integer only if ( k

( k - y ) / 2 = S. H ~ C X
y = k - 2 ~
and ( 2 ) yields
X=
-y
- y ) / 2 is an integer,
+ s = - ( k - 2s) + s = 3s - k.
Therefore, if the integers x and y satisfy equation (l),they are necessarily of the form
(3)
x=-k+3~,
y=k--2S
where s is some integer. Conversely, for an arbitrary integer s, (3)
d&es integers x and y , which satisfy equation (1).
17
18
b) In a similar manner we can determine the integral solutions of
9x -k s y = 1.
(4)
where 1 is a given integer. From (4) we obtain

1 - 9x
1+x
y=5=
5
-2x+From (5) we see that y is an integer only if ( I
+ x ) / 5 is an integer, say
l+x
-- t.
5
Hence
x=5t-l
and, substituting in (5) ,
9 = - 2 ~ t = -2(5t
- 1) + t = -9t
+ 21.
Therefore, if the integers x and y satisfy equation (4), they are necessarily of the form
x = 5t - 1,
(6)
y = -9t -k 21,
where t is an integer. Conversely, for an arbitrary integer t, (6) defines

integers x and y which satisfy equation (4).
c) If 2x
to a),
+ 3y is equal to some multiple of 17, say 17*n,then according

x
-17% + 3s,
=
y = 17%- 2s.
(Here n and s are, of course, integers.) Then
9x
+ 5y = 9(-17n + 3s) + S(17rt - 2s)

=
17(--4.n
+s);
that is, if 2x 3y is divisible by 17, then so is 9x 5y. Ushg the result

of b) ,we can see in an analogous manner that for all integers x and y for
which 9x Sy is divisible by 17, 2x 3y is also divisible by 17.
SOLUTIONS
19
Note. Divisibility. Classification of the integers. Definition: We say

that the integer b divides the integer a if there exists an integer k such that
k times b is equal to a. In this case we also say that a is a multiple of b, or b
is a divisor of a.
The integers a, b, and k need not be positive. Any of them might be negative, and a and k might be zero.
If a is divisible by b and b is divisible by c then a is divisible by c.
If a and a are divisible by b then so is a a and a a. If b is a divisor
of the integer a and if a is different from zero, then the absolute value of b
is not greater than the absolute value of a. Thus a number different from zero
has only a finite number of different divisors. For example, the divisors of 12
are f l , f 2 , f 3 ,f4,f6,f12.
In enumerating the divisors of a number, we shall restrict outselves to the
positive ones;the negative divisors can be obtained by simply changing the
sign. (Zero cannot be a divisor of any integer.)
The only positive divisor of the positive and negative units (i.e. of 4-1
and -1) is 1.
Every integer whose absolute value is greater than 1has at least two positive divisors: its absolute value and 1.
If an integer p whose absolute value is greater than 1has no other positive
divisor except I p I and 1,then p is called a prime number.
If an integer a whose absolute value is greater than 1has a positive divisor,
say b, different from 1 a and 1, then a can be written in the form
a = kb.
Here not only b but also I k I is between 1 and I a I Hence a can be written
as a product of two numbers whose absolute values are greater than 1 and
less than I a I Such an integer a is called a composite number.
Zero is divisible by every integer.
We can now divide the integers into four different classes as follows:
the positive and negative units
the prime numbers
the composite numbers
zero.
The study of the divisibility properties of integers belongs to number
theory. A detailed discussion of its foundation can be found in Emlids
Elements, Books V I I I 4 X . t
t The Greek mathematician Euclid lived in Alexandria around 300 B.C. His most
important work, the Elements, is a systematic and unified presentation of the mathematics known in his time. It contains plane and solid geometry, number theory, the
theory of proportions, and the solution of second degree equations. This work is often
considered the pinnacle of rigorous logical thinking. Thus, Spinoza (Dutch philosopher, 1632-1677) wrote in his ethics more geometric0 demonstrata.
20
Second Solution. For the sake of brevity let

Ic =
Hence
+ 3y,
2x
2,
9x
+ 5y.
30 - 5% = 17~.
This can be written in the following two ways:
(1)
(2)
32,= 5%
+ 17x,
5u = 32,- 17~.
If x and y are integers such that u is divisible by 17, then from (1) 30 is
also divisible by 17. Since the first factor, 3, of the product 32, is not
divisible by 17, the above implies that 0 is divisible by 17.
Using equation (2), we can see in a similar manner that if the integers
x and y are such that 2, is divisible by 17, then so is u.
Note. An important property of prime numbers. The second solution of our problem is based on the following property of prime numbers:
a) If neither of h o numbers is divisible by a giuen prime number p, then their
product is not divisible by p .
I t is sufficient to consider natural numbers; for, on the one hand, any
number has the same divisors as its negative, and on the other hand, the
negative divisors of a number are -1 times the positive divisors of that
number.
We shall verify the following equivalent form of the statement to be
proved :
If a natural number a i s not divisible by a prime fi then ab is dievisible by fi
only i f b is.
Given a prime p and a number a' not divisible by p ; denote by B the set of
those natural numbers b for which ab is divisible by p . ( p and all its multiples
clearly belong to B.) Let us find the smallest number in B. Such a number
exists; to find i t we merely have to decide which of the numbers
(1)
a * l , a-2
***,
a($-
1), a * p
is divisible by p . Surely a*# is divisible by p . If, in the sequence (l),a - m is

the first number divisible by p , then m is the smallest number in B.
We shall show on the one hand that ewry number in B i s divisible by m and,
on the other hand, that m is equal to p .
Assume to the contrary that some element b in B is not divisible by m.
Let m * q be the largest multiple of m not exceeding b, that is
b
m.q+ r,
SOLUTIONS
21
where the remainder r is one of the numbers 1,2,

, m - 1. We shall show
that r belongs to B, which is impossible. Multiply the above equality by a.
Then
ab = amq+ ar
or
ar
ab - (am)q.
The first term on the right is divisible by p and so is the second, since the
factor in parenthesis is divisible by p. Thus ar is divisible by p , i.e. r is in B.
But this is impossible since 0 < r < m and m was the smallest number such
that its product by a is divisible by p. This contradiction shows that our
assumption was incorrect; thus b must be divisible by m.
We know that p belongs to B and therefore, according to the statement
just proved, also p is divisible by m. But the only divisors of p are 1 and p ;
rn cannot be 1 because we have assumed that a. 1 = a is not divisible by p.
Therefore, m must be equal to p. This proves our second assertion and thus
the proof of theorem a) is complete.
Theorem a) can be generalized as follows:
b) If none of the numbers

a , b, c, d,
0 . -
i s divisible by p , then the products

ab, abc = (ab)c, abcd
(abc)d,
*-.
are not divisible by p.

c) The following simple proposition is often useful:
A set A of positive integers contains a smallest number.
be have proved and applied this principle to the set B in the above proof
of theorem a). The same proof applies to the general case: Let a be any
element of A ; then it is sufficient to find, among the numbers 1, 2, * * * ,
a - 1, a, the smallest belonging to A. This will be the smallest element of A .
I t is essential that we are dealing here with positive integers. For instance,
among all even integers (not necessarily positive), or among the reciprocals
of all natural numbers (the set of reciprocals contains only one integer)
there is no smallest.
1894/2. Given a circle and two points, P and Q ; construct an inscribed
right triangle such that one of its legs goes through the given point P
and the other through the given point Q. For what position of the points
P and Q is this construction impossible?
H U N G A R I A N PROBLEM B O O K
22
u
Figure 1
Solution. The locus of points P for which angle PPQ is a right

angle is a circle k with diameter PQ; see Figure 1. The points of intersection of this circle with the given circle k are the (right-angle) vertices
of those inscribed right triangles whose legs, or their continuations, go
through the points P and Q. In order that P and Q lie on the legs themselves (as required by the problem), P and Q must be inside the circle
k. In this case, the problem has as many solutions as the circles k and k
have points in common. Thus, if r denotes the radius of k, and d the
distance from the center of k to the midpoint of the segment PQ, the
problem has
1. two solutions if 3PQ > I - d ,
2. one solution if 3PQ = r - d,
3. no solution if 3PQ < r - d,
because the shortest distance from k to k is exactly r - d.
189413. The lengths of the sides of a triangle form an arithmetic
progression with difference d. The area of the triangle is t. Find the sides
and angles of this triangle. Solve this problem for the case d = 1 and
t = 6.
23
SOLUTIONS
Solution. The sides a, b, c of the triangle may be represented by
a=b-d,
where 0
where
b,
c=b+d,
< d < b. We use Heron's formulat

P = S(S - U ) ( S - b) ( S - C)
In the present case,

b
s-a=-+d,
2 '
2
s = -3b
s-b=-
b
2'
s-c=-- b
d.
By inserting these values in the above formula, we obtain
or
3(b2)'
- 12dW - 1W = 0.
This is a quadratic equation in 62 and may be solved for b2 in term of d

and t:
In order to obtain a positive value for b2, it is necessary to take the positive square root; hence
b=1/2(#+dd4+m,
a=b-d,
c=b+d.
The angles a and 8, opposite the sides a and b, are necessarily acute:
they can be calculated from the formulas
bc sin a
t=2 '
,&a=-
2t
,
bc
UCSin/3
t=--'
2t
Sinp=--.
ac
? For a simple derivatiolqseee.g.N.D.Kaza.rinoff,GeometriG Imqu&ties,Chapt~2,

pages 35,36,in this series.
HUNGARIAN PROBLEM B O O K
24
If d = 1 and t = 6 , then from the above formula, b = 4; therefore

a = 3 and c = 5. Moreover,
so that
4=
36"52',
@ = 90" - LY = 53"8',
y = 90".
1895 Competition
1895/1. Prove that there are 2(2n-1 - 1) ways of dealing 1z cards to
two persons. (The players may receive unequal numbers of cards.)
Solution. Let us number the cards and, for the moment only, admit
the two cases that all cards go to one of the players. With two cards we
can then have the deals
(1)
A A , AB, B A , BB,
where, for example, B A means that B gets card 1, A gets card 2.

With three cards there are 2 4 = 8 possibilities:
A A A ABA B A A BBA
A A B ABB BAB BBB,

where, for example, BAB means that B receives card 1, A receives card
2, B gets card 3. The first row here is (1) with A attached and the second
is (1) with B attached to each entry.
Every additional card doubles the number of possibilities. Hence for
n cards there are 2" deals; they may be called permutations of two things,
A and B , n at a time, with repetitions allowed.
If no player may hold all the cards, there are 2" - 2 = 2(2n-1 - 1)
possibilities, as was to be proved. The excluded cases are the permutations containing either only A-symbols or B-symbols.
If we indicate the above permutations with the numbers 1 and 2 in
place of A and B y we obtain the following theorem:
THEOREM.
There exist 2%- 2 numbers that have n digits made up only
of the numbers 1and 2 and contain each at least once.
Note 1. Permutations with repetitions. Our theorem is in effect a
special case ( n = 2) of the following:
There are m" permutations of m things, n at a time, with repetitions allowed.
SOLUTIONS
25
It may be of interest, though beyond the scope of our problem, to discuss

such permutations here.
To find all permutations of m things, n at a time, we mark n places, ordered
as first place, second place, and so on, to be filled with the given elements
every possible way. If reNtition is allowed, we can use each element repeatedly.
We imagine these places in a horizontal row. This is not essential, for we could
choose, say, the vertices of a regular n-gon inscribed in a circle and numbered
consecutively clockwise or counter-clockwise. If we did this, we could rotate
the labels of the elements about the center of the circle, keeping the vertices
fixed; if the rotation is 360/n and in a direction counter to the numbering
of the places, we call the resulting permutation a cyclic fiermutation of the
original. For example, in Figures 2 4 the second arrangement is a cyclic
permutation of the first, the third of the second. One more cyclic permutation
would bring us back to the original 0ne.t Cyclic permutations are characterized by the fact that the place of every element is occupied by its righthand neighbor, the place of the last element by the first. If a combination V
is obtained from another, U, by cyclic permutations, then U can be obtained
from V likewise, so that they are obtained from each other by cyclic
permutations.
Figure 2
Figure 3
Figure 4
If V is obtained from U by k cyclic permutations, and Wfrom V by 1 such,

then W is obtained from U by k
1 cyclic permutations.
All this is clear if we view cyclic permutations as rotations. Let us classify
permutations ( n things at a time, repetitions allowed) : A set shall contain
those, and only those, fiermutations which can be obtained from each other by
cyclic +rmutations. Then the number of fiermutations in any such set divides
the number n.
To show this, let UObe a permutation in a given set. Each of its members
is one of the permutations
(1)
UQi
uli
."
ur,
.-- us, .-.*

1
where Uris obtained from UOby Y cydic permutations, and each permutation
in (1) belongs to the class under consideration. We have only to decide how
many of the permutations in (1) are distinct.
t In the usual notation, these permutations are: ACBACB, CBACBA, BACBAC.
26
If Ur and U, are equal, then Ur+l and U8+~

are also equal; if Ur and U, are
distinct, so are Ur+l and U8+pLet U, stand for thefirst permutation in (1)
that is equal to a previous one, say to Up.Then r = 0; otherwise U . 1 would
already equal Ur-l. The combinations
uo,
UI,
***,
U8-1
simply repeat in (1). Hence the class in question consists of just these s
permutations.
Only the permutations
UO, u*, UZS,
***
in (1) are all equal to UO.Clearly U, is also equal to UO.It follows that the
number n is equal to one of the numbers s, 2s, 3s, ,so that s indeed divides
n.
Note 2. Combinations. a) The problem can be attacked in yet another

way: We must deal 1, 2,3, * * * , n
2, or n
1 cards to A , and give the
rest to B. If A gets k cards (k = 1, * * * , n
1), it is immaterial in what
order he receives them (we may even hand out several cards at a time) ; two
deals are the same if and only if As holdings are identical.
Thus the question is: I n how many ways can one pick k elements (curds)
from among n distinct elements regardless of the order in which they were chosen.
The sets chosen are called combinations of n things, k at a time; the number of
possibilities is written
(3
(in words: n over k )
and is sometimes called binomial coeficient. (See Note 2 to 1902/1.)

In terms of these symbols, the number of possibilities is
This result is useful only if we can find a simple way to compute the binomial
coefficients.
b) To determine the number of distinct possible combinations, let us see
how to produce all combinations of n things, k at a time, if we already have
them for k - 1at a time where k > 1. We can add to a given combination
t These considerations date back t o Gauss, who treated the case where n is prime.
K . F. Gauss, pinceps mathematicorltm (prince of mathematicians), noted physicist
and astronomer, was born in 1777 in Braunschweig and died in 1855 in Mttingen,
where he taught from 1807 at the university and was director of the observatory. His
epochmaking number-theoretical work, the Disquisit(ones arithmeticue, appeared in
Leipzig in 1801. His statue, in his home town, stands on a pedestal having the shape
of a regular 17-gon to Signify that he invented the method for constructing the regular
17-gon by ruler and compass.
SOLUTIONS
of k - 1 objects any one of the remaining n

to give us
27
- k + 1 objects, and this seems
possibilities for a tentative answer. But in this number each combination of

n things, k at a time, has been counted k times. The reason for this is that
every member of a collection of k objects can be the added object.
This shows that the following products are equal:
i.e.,
(2)
c)
Obviously,
Thisand (2) give:
0 40
n- 3 n
(3) =
n.
n(n- 1)(n- 2 ) ( n- 3)
,
102-3-4
etc.
Continuing this way, with k = 1,2, * * * , n, we get
(3)
n(n - 1) *-. (n - k
1.2.3.
.k
...
+ 1)
c) The product 1-2-39 - - * -k is usually denoted by k! and called

k factorial. If we multiply numerator and denominator in (3) by
(n- k)!= (n- k ) ( n - k- 1).
**.
03.2.1,
then the binomial coefficient appears in a new form:
(4)
n(n- 1)
***
=
t
(n- k + 1)(n- k ) ( n - k- 1)
k!(n k) !
*..
3.201
n!
k!(n k) !*
H U N G A R I A N PROBLEM B O O K
28
This shows also that
This is the so-called symmetry property of the binomial coefficients.

The case k = n,
can be obtained from (3) as well as directly from the meaning of the binomial
coefficient.
Even though "n things, 0 at a time" is meaningless, it is customary to
write
(;)
1.
This convention is in accord with the symmetry property mentioned. Formula (4) agrees with this convention if we write O! = 1 by way of definition.
Let us finally compare the two solutions of our problem. The binomial
coefficients just introduced give the following interesting connection:
(;) + t)+ ...+ (n- 1) + t)
= 2".
(Cf. 1902/1, Note 2.)

d) A useful property of two consecutive binomial coefficients, namely
is easily derived by writing the left side with a common denominator:
n(n- 1)
***
(n- k + 1)
n(n- 1)
**.
k!
- n(n-
1)
- n(n- 1)
***
*..
(n- k + l)[(k+
(k+ I)!
1)
(n- k + 1 ) ( n - k)
(k
I)!
+ (n-
(n- k + 1 ) ( n + 1)
=
(k 1) !
k)]
c: 3.
This relation can be derived directly from the definition.
SOLUTIONS
29
1895/2. Given a right triangle ABC, construct a point N inside the

triangle such that the angles NBC, NCA and N A B are equal.
Figure5
First Solution. a) Analysis of the @roblem. Denote the angles of

the triangle ABC by a,8, y, with y = 90" (Figure S), and denote by N
a point for which
QNBC = Q N C A = QNAB.
Then
QBNC = 180'
- (QBCN + QNBC)
= 180" - (QBCN
= 180'
- y;
+ QNCA)
FIUNGARIAN PROBLEM BOOK
30
and similarly,
4 A N C = 180' - (Y,
4 A N B = 180'
- p.
b) Constructwn of the point N . With BC as diameter, construct a

circle k. At the point A , draw a ray perpendicular to A B and lying on the
same side of A B as the point C. It makes the angle 9 0 O - a with CA.
At C, draw another ray making an angle 90'-a with C A and intersecting the h s t ray in the point 0. With 0 as center and OC as radius,
draw the minor arc k1 of the chord CA. The intersection of k1 and k is
the point N .
c) Proof that N is the desired point. Since in our case,
4 B N C = 180'
- 7 = 90',
the point N must lie on the circle k. The arc k1 is the locus of all points
P such that 4 A P C = 180'--a and on the same side of line C A as B ;
see Note below. We must prove that the circle k and the arc k1 have a
common point inside the triangle. We note first that the point C cannot
be the only common point of the two circles; for, if it were, they would
have a common tangent at C. But the line AC is a tangent to k and a
chord of the other circle. Therefore, there exists another point N common
to the two circles. Since the circle R lies on the same side of AC as the
point B, the point N is also on that side of AC. N must therefore lie on
the arc k1 of the other circle. If we show that the entire arc k1 lies inside
triangle ABC, we shall have proved that N lies inside triangle ABC. To
this end, observe that A B is tangent to k1at A because A B is perpendicular to the radius A 0 at A . A B makes the angle -a with AC. Similarly,
the tangent t of the arc k1 at C makes an angle a with CA. Since a < 90',
this shows that the entire arc k1 is inside triangle ABC.
Angles NBC and NCA intercept the same arc CN of circle k. Angles
NCA and N A B intercept the same arc N A of the other circle. Therefore,
+NBC = 4 N C A = *NAB.
Note. The locus of points P at which a given line segment XY subtends a given angle. i) Let /3 be the given angle, 0' < < 180, and let
P be a point for which * X P Y = /3. Consider the circle about the triangle
X P Y. All but the end points of the arc X P Y (which we shall call k') of this
circle belong to the desired locus. (The end points do not form a triangle
with X and Y.) For, if Q is a point on the arc X P Y , the peripheral angles
4 X P Y and 4 X Q Y are equal. Clearly, the same holds for points on the
reflection k of k' in the segment X Y ;see Figure 6.
31
SOLUTIONS
t/
Figure 6
We shall now show that only k and k' belong to t..e locus sought. Clearly,
points of the segment XY do not belong. For reasons of symmetry, i t will
now suffice to investigate those points of the plane lying on one side of the
line through X and Y. Take the side containing k , and consider first a point
inside the circle containing k , such as the point R in the figure. Let Z be an
inner point of the segment XY; then the line through Z and R meets k in
some point S. SXRZ and +ZR Y are exterior angles respectively of A X R S
and AYRS; hence they are bigger than the interior angles a L S. Since
together they form the angle X R Y,
gXRY= 3XRZ
+ g Z R Y > +XSZ + q Z S Y = + X S Y = /3.
Thus R is not in the locus. Similarly, if a point is outside the circle, such as
R' in the figure, then the segment R'Z'contains an inner point S' which belongs to k: so that
QXR' Y = QXR'Z'+
QZ!R' Y < QXS'Z'+
gZ'S'Y = g X S ' Y = 8.
Thus R' is not in the locus.

We have thus shown that the locus in question consists of two arcs between
X and Y, symmetrical about the segment X Y. The points X, Y do not belong to
it. If B i s a right angle, the locus is a circle of diameter X Y.
ii) The locus is easy to construct. For example, we can use the symmetry
about X Y and about the perpendicular bisector of X Y, and the fact that the
tangent to k at X forms the angle @ with X.Y. Thus, we can draw the line
32
t at X for which Q Y X T = 0 ( T a point on t ) and the perpendicular p to t

a t X . Let the perpendicular bisector of X Y meet p in the point 0. Draw a
circle of radius OX about 0. Let k be that arc of the circle not intercepted
by 3 T X Y = 8. Reflecting k in X Y gives kt and so the entire locus.
iii) In the sequel we shall need the length of the radius OX of this circle.
Let M be the midpoint of X Y ;we can compute the length of OX by means
of the right triangle MOX as follows:
X M --X Y - sin Q M O X ;
--
ox
hence
20x
XY
ox = 2sinQMOX'
The central angle subtending k is 2p if 0 > 90' and 360' - 20 if

0 < 90'. Therefore q M O X = 4 Q X O Y is either 8 or 180' - 8. Since
sin (180' - 0) = sin p, in either case
sin 0:MOX = sin
QXOY
= sin 0.
2
I t follows that
XY
ox= 2sinp'
Second Solution. Again we observe that the point N must lie on
the circle k with BC as diameter. To find the exact position of N , we
shall determine the direction of the line A N by using the condition
<NBC = Q N A B ;
(1)
see Figure 7.
Figure 7
Any ray drawn from A into the interior of triangle ABC intersects
the circle k in two points, N and M. Angles NBC and AMC are equal
SOLUTIONS
33
because each intercepts the arc CN.Hence, by (1),

4 A M C = <NAB = Q M A B .
But this implies that AB is parallel to MC.

Thus, N may be constructed as follows:
Draw a line through C parallel to AB. This line must intersect the
circle K at some point other than C, say M . Draw M A ; the desired point
N is the intersection of the line MA with the circle K. By construction,
M lies outside triangle ABC and N inside, as required by the problem.
Note. Brocard points. Given an arbitrary triangle ABC; it is possible,
in a similar manner, to find two points Nl and NZinside triangle ABC such
that
* N i B C = QNiCA = SLNiAB,
and
+N2CB= %N2AC= QN&A.

N1 and N2 are called the Brocurd points of the triangle.?
1895/3. Given the following information about a triangle: the radius

R of its circumscribed circle, the length c of one of its sides, and the
ratio a/b of the lengths of the other two sides; determine all three sides
and angles of this triangle.
Figure 8
First Solution. The angle y opposite the side c is found from the
formula
siny = - *
2R
see Figure 8. If c < 2R, there are two possible values for y , and if c = 2R,
t After the French mathematician and meteorologist Brocard (1845-1922).
34
there is only one; c cannot be greater than 2R since the triangle is inscribed in the circle. For each value of y so found, the other parts of the
triangle can be determined as follows:
By the law of tangents (cf. the Note below)
tan3(a P) =-a b - (db) 1

tan $(a - p> a (a/b) - 1
where a and p are the angles opposite the sides a and b, respectively.
Since
Q
+ p - 180 -
--
we can express (a - ,8)/2 in tenns of a/b. Thus (a p) /2 and {a- p) /2

are known, hence a and p may be obtained from
(1)
a=-
+-f f 2- p
a+P
Finally, the law of sines yields

csina
a=---
Sin y
p=---a + P
2
a-B
c sin p
b=-.
sin y
Note. The law of tangents. We shall now prove the theorem used in
the above solution. The law of sines can be written in the form
U
-- -b
sins
sinp
Let us denote the common value of these fractions by X {A is equal to the

diameter of the circumscribed circle, i.e. X = 2R) SO that
then
a
b
-a-b
a = X sin a,
- X(sina
b = X sin p;
+ sinp) - s i n a + sin@
X(sina- sin@
sina- sinp
Let us write a and /3 as sums of two angles, see equation { 1) ,and let us express sin a and sin p by the formula for the sine of the sum of two angles.
Then we obtain
s i n a + s i n p - 2sin+(a+p) cos+(a-p) - sin+(a+p) c o s + ( a - ~ )
cos$(a+@) sin$(a-p)
sins- sin@ 2cos3(a+p) sin+(a-@)
-
tan +(a
+ p)
tan $ ( a - p>
SOLUTIONS
Thus
Second Solution. Let 0 be the center of the circumscribed circle of

triangle ABC; see Figure 9. Let Cl be the foot of the perpendicular from
0 to A B ; CI bisects AB. The theorems on inscribed and central
angles and intercepted arcs, when applied to the arc AB, show that
QBOCI = QACB = y. Hence
BCI
4 2
c
sin y = -= - = OB
R
2R'
If c < 2R,this equation has two solutions; one yields an acute angle,
the other an obtuse angle. If t = 2R, then y = 90".For each value of y
so determined, the rest of the triangle can be found as follows:
Drop a perpendicular from another vertex, say from B, to the opposite
side AC and denote its foot by Mz.Then
BM2 = asiny
and
CM2 = a cosy
(=a+F)
The positive root in the parentheses corresponds to an acute angle y,

the negative root to an obtuse y. Also
AM2= b - CM2= b - a c o s y .
Figure 9
36
We have
a sin y - (a/b) sin y

BM2
t a n a = -AM2 b - a cosy 1 - (a/6) cosy
This formula holds also if a is an obtuse angle. In this case AM2 is considered negative and b - a cos y is negative.
The angle /3 is obtained from
+ p + y = 180O.
Finally, from the law of sines,

c sin a
a=--
sin y
c sin p
b = -- 2R sin p.
sin y
- 2R sina,
1896 Competition
1896/1. Prove that
log n 2 K log 2,
where n is a natural number and K the number of distinct primes that

divide n.
Solution. Let n be a natural number greater than 1, let PI, p ,
be its primet divisors, and let
P?,
P?,
0 ,
fik
PT
.-*7
be the highest powers of these primes that divide n

Then (see Note below)
n = p ? 1 p $ z . . . PZk*
Since none of the p i is less than 2 and each ai 2 1,
n>
- 2a1+aZ+*k
2k.
This is true also for n = 1, for then
K=O
and
n=1=20=2k.
If the base a of the logarithm is greater than 1, for instance if a = 10,

then our inequality gives the desired result:
logan 2 K.10&2.
t By prime, we shall always mean positioe prime.
SOLUTIONS
31
Note. Factorization i n t o a product of powers of primes. (For definitions of prime and composite number, see Note to the First Solution of
1894/1.)
Every composite natural number i s the product of two or more primes.
The smallest composite number is 4 = 2.2; in order to prove the above
theorem it suffices to show that, if the theorem holds for all composite natural
numbers less than some composite natural number a, then it holds also for a.
(Complete induction.)
Indeed, according to the definition of composite number (see 1894/1,
page 19), a is a product of two factors, k and b, from among the numbers
2, 3, *.-, a
- 1.
If the theorem holds for all composite natural numbers less than a, then k
and b are either primes or products of primes. Therefore, a = kb can also be
written as a product of primes.
Every natural number is the product of non-negative powers of primes.
For a given composite natural number a , one finds the prime power factors
by first finding the prime factors, grouping equal ones together, and writing
the factors in each group as a power. For example
120= 2 . 2 . 2 . 3 - 5
= 25 31 51.
. .
Since the zeroth power of any number is 1, we may, of course, add primes to
the zeroth power in such a factorization, e.g.,
A prime p can be written as a product of the first power of p (that is, p) ,

and the zeroth power of any other primes.
Finally, 1 can be written as a product of any number of primes to the
zeroth power.
With these agreements, the factorization of natural numbers into powers
of primes is unique. In other words:
If
where the p , are distinct primes and the ai non-negative integers, then ai i s the
highest power of pi which divides a ; hence, the factorization (1) of a i s its only
factorization (exceptfor the order of thefactors and the appearance of an arbitrary
number of primes to the power zero).
The proof of this useful and by no means trivial theorem (known as the
Fundamental Theorem of Arithmetic) follows from the theorem: If none of
the factors of a product i s divisible by a prime number p, then the product itself
is not divisible by p [see 1894/1, Solution 2, Note, b)].
We shall prove that a number a cannot have two distinct prime power
factorizations, that is two factorizations in which a prime, say p , occurs with
two different multiplicities. (The multiplicity zero is also admitted here.)
38
Suppose, to the contrary, that

a=
(2)
d-w
T * Q =
with (Y > p > 0, and where the prime factorizations of v and w contain no
positive power of 9. Then, by the theorem quoted above, v and w are not
divisible by p. But it follows from (2) that
w=Q
- fl,
with
(Y
- 0 > 0,
i.e., that w is divisible by p. We are led to a contradiction and so our assumption is false; thus all factorizationsof a must contain the prime p to the same
power.
1896/2.
Prove that the equations
and
9 - 3xy
+ 2r + x - Y = 0
9- 2 x y + 4 -
imply the equation

XY
5 x + 7Y = 0
- 12%+ 15y = 0.
First Solution. The first of these equations can
form
(x
- y ) ( x - 2y + 1) = 0.
be written in the
Hence the given system of equations is satisfied by those pairs of values

( x , y ) which satisfy either
x-y=o
(1)
or
2)
d-2xy+y2-55X+7y=O
x-
x2 - 2xy
2y+ 1 = 0
+ r - 5x + 7y = 0.
From the first equation of system ( 1 ) ,we obtain

x = y;
substituting this result in the second equation, we obtain

2y = 0.
Hence the only solution of ( 1 ) is x = 0 and y = 0.

From the fxst equation of system ( 2 ) ,we h d
x = 2y-
1;
39
SOLUTIONS
substituting this result in the second equation, we obtain
- + 7~ = 0.
( 2~ 1)2 - 2(2y - l ) ~ f - 5 ( 2 ~ 1)
After some manipulations, this takes the form:
f - 5y
+ 6 = 0,
and is satisfied by y = 2 or by y = 3. The corresponding values of

- 1 are 3 and 5. Hence the solutions are
x = 2y
x=3, y=2,
and ~ = 5 ,y = 3 .
Therefore, the originally given system of equations has three solutions:
x=o, y=o;
x = 5 , y=3.
x=3, y=2;
A straightfoxward computation shows that these values satisfy the

equation
%y - 12x 15y = 0.
Second Solution. We consider the identity
+ + x - Y ) ( X - Y - 9)
+ (9-2xy+y2- 5 x +
(9- 3 ~ y 239
7y)(-x+2y+3)
=
2(xy - 12x
+ 15y).
This shows immediately that, if the iirst and second equations of our
problem are satisfied, the left hand side of the third equation is zero.
1896/3. Construct a triangle, given the feet of its altitudes. Express
the lengths of the sides of the solution triangle Y in terms of the lengths
of the sides of the triangle X whose vertices are the feet of the altitudes
of triangle Y.
Solution in the case that Y is an acute triangle. Definition. A

triangle X whose vertices are the feet of the altitudes of another triangle
P is called the orthk triangle of Y . j
f More generally, if from a point P inside a triangle Y three perpendiculars are
dropped to the sides of Y , then the triangle whose vertices are the feet of these pertriangle of Y . If the point P happens to be the orthocmh
pendiculars is called a @dal
(Le., the intersection of the altitudes of Y), the resulting pedal triangle 19 the &c
triangle.
40
Figure 10
a) THEOREM.
The altitudes of an acute triangle ABC bisect the vertex
angles of its orthic triangle.
If the altitudes AA1 and BB1 intersect in M,then AlMC and CMBI
are right triangles with hypotenuse CM (see Figure 10). Therefore A1
and B1 lie on a circle KC with CM as diameter. Moreover, if a,8, y denote
the vertex angles of AABC, then
QMCB1
90'
QCMBi
CY
- CY,
Q M B C i = 90'
<BMCi,
- CY.
The arc MBl of the circle kc is intercepted by the inscribed angles MA1B1
SOLUTIONS
41
and MCB1. Thus
QMAiBi = QMCBi = 90'

that is,
Q AA1B1 = 90"
- CY,
- CY.
Similarly, by considering the circle kB with diameter BM, we see that
QMA1C1
that is
QMBC1
Q AA1C1 = 90'
90'
- CY,
- LY,
so that A A1 indeed bisects the vertex angle at A1 of the orthic triangle;

the same reasoning yields the desired results for the other angles of the
orthic triangle. Incidentally, denoting the angles of the orthic triangle
AlBlC1 by LYI,
A , 71, we have the result
- = 900 a1
Pl
71
-=900-p,
The theorem just proved, together with the fact that the angle bisectors
of a triangle meet in a point, also shows that the three altitudes of AABC
meet in a point; this point is called the orthocmter of AABC, and is the
center of the circle inscribed in AAlBIC1.
b) Construction. According to a) there is just one solution to our
problem, namely that triangle ABC whose vertices are obtained by
bisecting the exterior angles of triangle A1B1C1 and determining the
intersections of these bisectors. In other words the points A , B, C are
just the centers of the circles which are tangent to one side of A1BIC1
from the outside, and to the other two from the inside.?
c) JustiJication of the Construction. Let us construct also the bisectors
of the interior angles of AAIBICl.The bisectors of two exterior angles
at two vertices and that of the interior angle a t the third vertex meet a t
one point. This can be seen, for example, by observing that the point A
(see Figure 10) is equidistant from the lines AIBl and BIC1 and also
from AK1 and BIG, so the locus of points equidistant from A1C1 and
AJ31 must also go through A . On the other hand, the bisector of our
exterior angle and that of an adjacent interior angle (the supplement)
t These circles (see Figure 10) are called exterior circles or excircles, and their centers are called excenters of triangle A&Q.
42
are perpendicular. Thus A A I , BB1, and CC1 are perpendicular, respectively, to BC, CA, and AB. Thus AI, BI, CI are indeed the feet of the
altitudes of AABC.
d) Compctatiolz. We saw in a) that
thus [cf. formula (12) in Note 2 below],
+ +
where sl is half the perimeter of AAlBlG, i.e., SI = *(a1 61 el).

Moreover, ACIABl is similar to ACAB because both have the angle
a! in common and
<CiBiA
90" - - = p = <CBA.
2
Bl
Hence
BiCi
CiA
-=-=
BC
CA
COS a!
that is
- = cosa!.
a1
With the aid of formula (12) of Note 2 below, this gives
a = - -a1
cos a!
Sl
- 61) (s1 -
Similarly
43
SOLUTIONS
Note 1. On the orthic triangle of an obtuse triangle. In an obtuse

triangle (just as in an acute triangle), the altitude from the obtuse vertex
bisects the interior angle of that vertex of the orthic triangle which is the foot
of that altitude. However, the altitudes drawn from the other two vertices
bisect the correspondingexterior angles of the orthic triangle.
If we admit also obtuse triangles as solutions, then we obtain, in addition
to AABC, the triangles BCM, CAM, ABM satisfying our conditions. The
altitudes belonging to these triangles meet, respectively, a t the points A , B,
C. In other words, the orthocenten of triangles BCM, CAM, ABM (and
ABC) are A , B, C (and M ) respectively.
All four solutions have this in common: The vertices and the orthocenter
of each solution triangle are centers of just those circles which are tangent
to every side of AAlBlCl.
For the acute solution, AABC, the orthocenter is the center of the circle
that touches every side of AAIBICI from the inside.
For the obtuse solutions, triangles B CM, CAM, ABM, the obtuse vertex is
the center of the circle that touches every side of AAIBICIfrom the inside.
Note 2. a) On circles tangent to the sides of a triangle. Let 0 be the
center and r the radius of the circle k tangent to the sides of AABC from the
inside, and let D, E , F be the points of tangency on BC, CA, AB, respectively; see Figure 11. Let ka be the circle tangent to the extensions of sides
AB and AC and to side BC and let the corresponding data for ka be Oa, ra
and D', E', F'. Moreover, denote the lengths of the sides AB, BC and CA
by c, a and b, and let the perimeter be 2s = a b c.
+ +
F i e 11
44
Since a pair of tangents from a point to a circle have equal lengths, we have
Hence
or
A E = AF,
BF = BD,
2s = A B + B C + CA = 2 ( A E
CD=S- (AE+BF)=s-
Similarly,
- a,
AE = s
For the circle h,
+BF+
CD),
CD
s= AE+BF+
so that
(1)
CD = CE.
(AF+BF)=s-c.
BF = s - b.
2AE'= AE'+ AF'= A C + CE'+ A B + BF'
+ AC + CD'+
= AB + A C + B C
= AB
so that
BD'
=2s
AE'= AF'= s
(2)
and
(3)
CE'= AE'- A C =
b,
S-
BF'=
S-
t.
Designate the area of a triangle X Y Z by T ( X Y 2 ) . We can compute
T(AB C ) by dividing up the triangle:

T(ABC) = T(A0B)
+ 3ra 4-&rb
= +rc
= &(a
that is
+b +
c),
T(AB C ) = YS.
(4)
On the other hand,

T(ABC) = T ( A O 3 )
(5)
+ T(AO,C) - T(B0,C)
+ arab = &ra(b+ c = $rat
Similarly,
(6)
+ T(B0C) + T(C0A)
= ra(s
- u).
U)
T ( A B C ) = %(s - b ) ,
SOLUTIONS
45
and
T ( A B C ) = G(S - c),
(7)
where a,r, stand for the radii of the circles kb, kc tangent to sides b, respectively c, and the extensions of the other two sides of A A B C.
b) Connection between the angles and the sides of a triangle.

We know that the center of the inscribed circle of a triangle is the point of
intersection of the angle bisectors, so that, for example 0:OA F = +a.Using
(4) and Heron's formula'f for the area of a triangle, we obtain
a OF = r
T(ABC)
tan-=
=
2 A F S - a S(S- U)
- z/s(s
s(s
- a)
s(s - a )
Similarly,
(9)
- a) (s - b) (s - c)
tan- =
2
(s - a ) (s - c)
S(S - b)
and
( 10)
But, using the fact that the area of a triangle is half the product of its base
and altitude and considering c as the base and expressing the altitude from
C as b sin a,we also have
1
T ( A B C ) = -bcsina
2
a
a
a
a
bcsin-as- = b c t a n - a s 2 2
2
2
2
which, together with (8) and Heron's formula gives
a
T(ABC)
bc a s s - =
= s(s- a).
2 tan+(a)
In this way we get, one after the other,
See footnote on page 23.
46
Since sin ( 4 2 ) = tan ( 4 2 ) cos (a/2),(11) and ( 8 ) , (9), (10) yield
1897 Competition
1897/1. Prove, for the angles a,p and y of a right triangle, the following relation:
sinasinpsin(a-@) +sinPsinysin(p-y) +sinysinasin (y -a)
+ sin (a- p) sin (a - y) sin (y - a) = 0.
Solution. Let a = 90' = p
+ y. Then
sina=l, sin(a-P) =sin?, s i n ( y - a ) = -&(a-y)

Therefore
= -s@
sinasinpsin (a- p) = sinpsiny,
sin p sin y sin ( p - y) = sin p sin y sin (@- y),
sinysinasin (y - a) = -sinysinp,
sin (a- p) sin (p - y) sin (y - a) = -sinysin (p - y) sinp,
and the s u m of these four products is zero.
Note. Relation between sums and products of trigonometric

functions. We claim that the formula of the above problem is valid for
arbitrary angles a,@ and y;that is, we claim that
sinasinpsin(a-@)
(1)
+ sinpsinysin@-y)
+sinysinasin(y-
+ sin (a- 0) sin (0- y) sin (y - a) = O
a)
holds, not only for the angles of an arbitrary triangle, but for any three angles
a,p, y. We shall prove this by transforming the products of the trigonometric
functions occurring in (1) into sums of trigonometric functions.
SOLUTIONS
47
a) The formulas for the sine and cosine of the sum of two angles lead to
the following relations between trigonometric sums and products:
sin(x+y)+sin(x-y)=2sinxcosy
cos(x-y)+m(x+y)
= 2mxccwY
cos(x-y)-ms(x+y)
=2sinxsiny.
Hence
sinxsiny=3[cos(x-~)-cos(x+y)],
cosxcosy = $[cos ( x
- y ) + cos ( x +
s i n x c o s y = +[sin(x+y)
in particular for x = y, these reduce to
sin2x = &[I
- cos 2x1,
cos2x = $[I
y)],
+sin(%-y)];
+ cos 2x1,
sin x cosx = & sin 2x.
By repeated application of these formulas, products of an arbitrary number

of sines and cosines may be transformed into sums.
Each term of (1) is the product of three sines. The product of three sines
can be transformed into a sum as follows:
- cos(x+y)]sinz
= ~ [ C O S ( x - y) sin z - cos ( x + y) sin 27
sinxsinysinz- &[cos(x= $[sin ( x
(2)
y)
- y + z) + sin ( - x
- sin ( x +
= $[sin ( - x +
b) Now,if x = a - p ,
x+ y
+ z = 0,
x-y+z=
- sin ( - x
- y+
s)]
+ sin ( x - y + z)
+ sin ( x + y - z) - sin ( x + y + z)].
y + z)
y = B-y,
z = y - a, then
-x+
-2(P-y),
y + z)
+ y + z)
y+
I=
x+y-z=
-2(a
- Is>,
-2(y-a).
Hence (2) yields

(3)
4sin (a-
a) sin (p - y) sin (y - a)
=
- sin 2(a - @)- sin 2(p - y) - sin 2(y - a).
48
Since this relation holds for arbitrary angles a,p and y, we may insert 0
successively for y, a,and B in (3) and obtain
(4)
4sinasinBsin(a-@) = s i n 2 ( a - B )
+sin2B-sin2a
(5)
4sinBsinysin(p-r)
= sin2(8- y) + s i n 2 y -
(6)
4sinysinasin(y-a)
=sin2(y-a)+sin2a-sin2y.
sin20
If we add equations (3)-(6), we obtain formula (1).

1897/2.
then
Show that, if a,B and y are angles of an arbitrary triangle,
+ +
First Solution. Since a

B y = 180, the angles a/2,@/2, r / 2
are acute angles, and we know that when an acute angle increases, its
sine increases. Hence the inequalities
a
- = go"-- P + Y
imply the inequality

a
sin-sin-
<
B
2
< 90"
9;
1
2
cos-sm- = -sin&
2
goo--
P . B
sin 90" - - sin- = cos-sm-;
and since
B . B
<
sinp
1,
cosp
5 1,
it follows that
(1)
sin-sin-
<
1
-sin8
2
1
-.
2
Suppose y is the smallest of the three angles; then y

5 30, so that
Y
1
sin- 5 sin30" = -
y/2
5 180/3 and
2'
SOLUTIONS
From (1) and ( 2 ) we obtain

1
< -.
a
a
r
sin-sin-sin2
2
2
(3)
Second So1ution.t a) It is known (cf. Note
1 below) that
sin-sin-sin- = 2
2
2 4R
where r is the radius of the incircle (inscribed circle) of the triangle with
angles a, y, and R is the radius of its circumcircle (circumscribed
circle). Since clearly r < R, the product on the left side is less than t.
a,
b) We shall sharpen the inequality (3) by showing that

(4)
a P r 1
sin-sin-sin- < 2
2
2-8
and we shall establish (4)by proving that r IR/2.

According to a theorem of Euler (cf. Note 2 below),
d2 = R2 - 2&,
where d is the distance between the center of the incircle and the center
of the circumcircle. Hence
2Rr I R2, and
IR/2.
If the triangle is equilateral, then d = 0 and the value of the product

sin (a/2) sin ( P / 2 ) sin ( y / 2 ) is exactly 1/8.
Note 1. The meaning of some trigonometric products. Let s denote
the semiperimeter $(a b
c) of a triangle, t its area, R the radius of its
circumcircle and r the radius of its incircle. Then the following formulas hold:
+ +
tan-tan-tan-= 2
2
2 . 9
or, since t = sr [see e.g. eq. (4) in Note 2 of 1896/3],

I1
tan-tan-tan2
-rs .
t while this solution of the problem requires much more knowledge, it gives greater
insight into the essence of the problem than the simple calculations of the first solution.
50
Also
s
4R
a
S
Y
cos-ms-ms-=
I11
and
sin-sin-sin- =
2
2
2 4R
IV
In order to prove these, we make use of Note 2 to the solution of Problem
1896/3. From formulas (8), (9) and (lo), see page 45,we obtain
tan-tan-tan- =
2
2
2
4 4 s - a > ( s - b ) ( s - 4.
P
s2
this, together with Herons formula and eq. (4)on page 44 yields formulas I
and 11.
Equation (11) on page 45 yields
CY
p y sz/s(s
as--0s-cos=
2
2
2
- a) (s - b) (s - c)
abc
t
abc
which, together with the relations

c
-
abc
siny = - that is, R =

=2R
2siny
4t
(see Figure 8) implies formula 111. Finally, formulas I1 and I11 imply formula IV.
B
Figure 8
Note 2. The inscribed and circumscribed circles of a given triangle.

Consider two circles, k and k, one containing the other, say k contains k;
see Figures 12 and 13. Let 0 and 0denote their centers, R and I their radii,
and d the distance between 0 and 0.
SOLUTIONS
F i i 12
51
F i i 13
To the theorem of Eulert used in the Second Solution, part b) of this

problem, we shall now add the converse and then prove the full theorem.
~ R E M A. neGessary and su..cient c o n d W f m the exiStenCeof atrkngk
with Circumcircle k and incircle 'k i s the epzlality
(1)
&o= E - 2Rr.
PROOF:a) Let A be any point on the circle k (see Figure 12). What is a
necessary and sufficient condition for the existence of points B and C such
that the triangle ABC satisfies the requirements of the theorem? If there
are such points B and C, they can be defined as the points of intersection of
the circle k with the tangents drawn from A to k'. Triangle ABC will have
k' as its incircle if BC is also tangent to .'k A necessary and sufficient condition for BC to be tangent to 'k is
(2)
<Ago'=
<:O'BC.
(In the diagram of Figure 12, this condition is not satisfied.)
From ( 2 ) we shall now obtain some equivalent necessary and sufficient

conditions.
t L.Eder was born in Basel, Switzerland, in 1707 and died in St. Petersburg (now
Leningrad) in 1783. At the time of his death, he was director of the Mathematics
Section of the Academy of Sciencesthere. His &tinpished researches enriched almost
all branches of mathematics. A significant portion of the mathematicalnotation in use
today is due to him; for example, he introduced the use of the letter r for the ratio of
circumferenceto diameter of a circle, and the letter e for the base of naturallogarithms,
i.e., c is the limit of the sequence
52
Draw the line from A through 0' until it intersects the circle k again, in
the point D. Then AO' bisects the angle BAC made by the tangents from
A to k'. That is
QO'AB = QCAO'.
Moreover QCAO' = QCBD because both these angles intercept the arc
CD of the circle k . Hence
0:O'AB = (c CBD.
(3)
Since angles O'AB and ABO' are interior angles of the triangle O'AB opposite to the exterior angle BO'D, we have
QO'AB = QBO'D
Furthermore, clearly
- QABO'.
< CBD = QO'BD - QO'BC.
If we substitute these values for angles O'AB and CBD into (3) we obtain
(4)
QBO'D
- QABO'=
QO'BD
Thus condition (2) holds if and only if
- QO'BC.
QBO'D = 4 O'BD.
(5)
These two angles are opposite the sides BD and O'D of triangle BO'D, so
that we can reformulate our result as follows:
Triangle ABC has the circle k' as incircle if and only if
BD = 0'0.
(In the diagram of Figure 13, this condition is satisfied.)
b) In order to show the equivalence of this necessary and sufficient condition with the condition ( l ) , we shall work with the ratio BDIO'D and
express it in terms of the quantities R,r and d.
Let E denote the point of tangency of the circle k' with the line A B (see
Figure 12) and let D' denote the point diametrically opposite to D on the
circle k. We claim that the triangles AO'E and D'DB are similar. This is a
consequence of the following two facts:
i) they are right triangles, the right angles being a t the vertices E and B,
respectively;
ii) the angle at A of the first triangle is equal to the angle a t D' of the
second because both angles are inscribed in the circle k and intercept its arc
BD.
From the similarity of triangles AO'E and D'DB it follows that
AO' = D'D
-.
EO'
BD'
that is,
AO'
2R
=r
BD
SOLUTIONS
or
53
AO BD = 2Rr.
(6)
Next we apply the following corollary of the theorem on inscribed angles
and their intercepted arcs: If two chords of a circle intersect, then the product
of the segments of one i s equal to the product of the segments of the other. Let A D
be one of the chords of the circle k and let the diameter of k through the point
0 be the other. Then, by this corollary,
(7)
AO*OD= ( R + d ) * ( R - d d ) .
If we divide equation (6) by (7), we obtain

BD
OD
-=
2Rr
(R+d)(R-
d)
Clearly, BD is equal to OD if and only if
(8)
2Rr= ( R + d ) ( R - d ) .
c) The results of parts a) and b) can be summarized as follows:

If A is an arbitrary fioint on the circle k, then the necessary and s u f i k t condition for the existence of @oints B and C such that triangle ABC has k as its
incircle is that equation ( 8 ) hold.
This establishes Eulers theorem [since equation (8) is the same as equation (l)] and its converse.?
Finally, observe that condition (8) does not depend on the choice of the
particular point A on the circle k. Therefore:
If for some W*ntA of the circle k we can construct a triangle AB C huving k
as circumcircle and k as incircle, then this construction is fiossible for any
point A on k. This principle has been formulated by Poncelett for arbitrary
conic sections and for polygons instead of triangles as follows: Aut semper,
aut nunquam (either always or never).
t See also the discussion of Eulers Theorem in Geometric Imqualities by N. D.

Kazarinoff, in this series.
$ The French general and mathematician J. V. Poncelet was born in Metz in 1788and
died in Paris in 1867. His main work was Trait6 des propi&% pojectives des jigures
(1822). He was a member of the Acudhie des Sciences of the Imtitlct.
The French Institut has five sections called academies; that concerned with mathematics and sciences is called Acudhie des Scimces and should not be confused with
the Acadhie Franpise whose members are commonly referred to as the 40 immortals. The Acudhie FranGaise is the poetry and literature section of the
Imtit~t;however, distinguished thinkers in other fields are sometimes honored by
election to the Acadkmie Fraqaise. The famous mathematican H. Poincar6 (1854des Sciences and the Acodknie
1912), for instance, was a member of both the Ac&ie
Fraqaise.
54
1897j3. Let ABCD be a rectangle and let M , N and P, Q be the points

of intersection of some line e with the sides AB, CD and AD, BC, respectively (or their extensions). Given the points M , N , P, Q and the
length p of the side AB, construct the rectangle. Under what conditions
can this problem be solved, and how many solutions does it have?
First Solution. Construct a right triangle PSQ with hypotenuse

PQ and with P S = p ; see Figure 14. Draw l i e s parallel to P S through
the points M and N,draw lines perpendicular to P S through the points
P and Q. Clearly, these four lines determine a rectangle which satisfies
the requirements of the problem.
Figure 14
The problem has a solution only if the triangle PSQ can actually be
constructed, that is if
p < PQ;
and if this condition is satisfied, triangle PSQ may be constructed on
either side of the line e. Hence we get two solutions; each is a reflection
of the other in the line e.
Second Solution. Before solving the problem, consider the following
generalization: Let ABCD be a parallelogram whose sides AB, CD, AD,
BC (or their extensions) intersect a line e in the points M , N, P, Q,
respectively; see Figure 15. Given the points M , N , P, Q, the length p
of side AB, and the angles of the parallelogram, construct the
parallelogram.
The first step is to construct the point of intersection S of the side BC
with a line through P parallel to AB. Once this point S is found, the
previous construction can be carried out easily.
We shall determine the point S from the following two conditions:
Since ABSP is a parallelogram, its opposite sides are equal; thus
PS = AB = p,
so that the point S lies on the circle with center at P and radius p.
SOLUTIONS
P\
Figure 15
Moreover, AB is parallel to P S so that angle PSQ is equal either to

the angle 0 at B, or to its supplement. We claim that, if the segments
MN and PQ are oriented in the same sense (i.e. if either N is to the right
of M and Q to the right of P , or if N is to the left of M andQ to the left
of P ) , then
QPSQ = 8,
and if MN and PQ are oppositely oriented, then
9:PSQ
= 180" - 8;
S for which
0 (or 180" - p) consists of arcs k and k', one above PQ, the
see Figures 15 and 16. In either case, the locus of points
9:PSQ
other below PQ. (The arcs can be constructed by the method explained
in the Note to the First Solution of Problem 1895/2, and the radius of
the circle of which k (or k') is an arc has length PQ/(2 sin 0) ;see iii
of that same Note, page 32.) Thus, the desired point S is a point of
intersection of the circle with center P and radius p with the arc k or R'.
Figure 16
Now let us count the number of solutions. Observe that with any
parallelogram ABCD which solves the problem, we get a second solution
56
by reflecting ABCD in the line e. We shall agree not to consider two solutions different if they are just such mirror images of each other. To
enumerate all the different solutions, it therefore suffices to count all the
points of intersection on the arc k, since those on arc k' can be obtained
by reflection.
For p < 90" we have the following possibilities:
1) If p > d [where d = PQ/(sin p) is twice the radius of the arc K]
then the circle with center P and radius p does not intersect the arc k,
and our problem has no solution.
2) If p = d = PQ/(sin
then the circle with center P and radius p
is tangent to k and the problem has exactly one solution (and, of course,
its mirror image).
3) If PQ < p < d = PQ/(sin p), then the circle with center P and
radius p intersects the arc k in two points, and our problem has two
solutions (and, of course, their mirror images).
4) If p < PQ, then the circle with center P and radius p intersects K
in one point, and the problem has one solution (and its mirror image) ;
this is the situation pictured in Figure 15.
For 2 90",there are only two possibilities:
1) If p 2 PQ, there is no solution.
2) If p < PQ, there is exactly one solution (and its reflection).
In the special case ,L3 = 90",ABCD is a rectangle and the arcs k and
k' are semicircles with diameter PQ.
Once we have determined the point S, we may construct the parallelogram as follows: Draw a line through Q and S ; the side BC will lie on
this line. Draw a line parallel to QS through P; A D will lie on this line.
Finally, draw lines parallel to PS through M and N . These, together with
the pair of parallel lines previously constructed, will form the desired
parallelogram.
a),
1898 Competition
1898/1.
by 3.
Determine all positive integers n for which 2"
+ 1 is divisible
Solution. We start with the familiar identity
an-bn=
(a-b)(,p-l+p-2b+
...
+&-2+bn-l).
If we set a = 2 and b = -1, we obtain

2"
- (-1)"
= [2
- (-1)]A
3A,
where A , the second factor on the right of the identity, is an integer
57
SOLUTIONS
since we substituted integers for a and b. Hence
+ 1 = 2" - (-1)n + 1 + (-1)n = 3A + 1 + (-1)n

which shows that 2" + 1 is divisible by 3 when n is odd, but gives a
2"
remainder of 2 when n is even.
Note 1. The notion of congruence. In order to formulate the essential

part of the above solution, let us introduce a concept suggested by Gauss,
the concept of congruence.
Definition: Let a, b, and m be three integers. We say that a i s congruent
to b modulo m and write
a= b (mod m)
if the difference a - b is divisible by m, that is, if there is an integer k such
that
a - b = km;
m is called the modulus. Note that the integers a, b, m, k need not be positive;
they may be negative, and a, b, and k may even be equal to zero. We may
assume without loss of generality that m is positive, since the fact that an
integer is divisible by m does not depend on the sign of m.
Examples:
22= 1, 24= 1, 26= 1,
(mod 3)
states that the numbers
22 - I = 3, 24-
I = is, 26- I =
are divisible by 3. The relation
a, ...
a = 0 (mod m)
says that a is divisible by m.

Let m = 3 and let us arrange the integers in increasing order by writing
three consecutive numbers in one column, the next three in the next column,
etc. Then we obtain the array
-9 -6 -3 0 3 6 9 .a.
...
***
-8
-5
-2
1 4 7 10
***
***
-7
-4
-1
2 5 8 11
* * a .
I t is easy to see that any two integers in the same row are congruent, and
any two integers from different rows are not congruent.
Similarly, for an arbitrary modulus m, we can arrange the integers in an
array consisting of m rows. This shows that, given any integer a, there is
exactly one integer r among the numbers
0, 1, 2,
***,
m - 1,
such that r = a (mod m) We shall call r the remainder of a with respect to m.
58
The following statements and theorems show that the concept of congruence resembles, in many respects, the concept of equality. A congruence is
reflexive: for every integer a, a = a (mod m) ;
symmetric: if a = b (mod m), then b = a (mod m) ;
transitive: i f a = b(modm) a n d b = t ( m o d m ) , t h e n a r c ( m o d m ) .
THEOREM.
If a = a' (mod m) and b E b' (mod m), then
a+ b = a'
4- b' (modm),
a- b = a'-
b' (modm), ab= a'b' (modm).
We shall prove the last statement in this theorem, the assertion about
products, merely by using the definition of congruence. The expressions
a = a' (mod m) and b
b' (mod m)
say that there exist integers k and 1 such that

therefore
a = a'
4-km,
b = b' 4- lm;
ab - a'b' = m(a'1 4- b'k 4- klm),
so that the difference ab
- a'b'
is divisible by m,that is
ab = a'b' (mod m) .
The following is a consequence of the above theorem: Ifa = b (mod m),

then an= b" (mod m).
Note 2. The last result may be used to solve our problem. The congruence
implies that
Zn= (-1)"
Hence for odd n
2=-1(mod3)
(mod3) or Zn+ I = (-I)*+

2.+
1 (mod3).
1= 0 (mod 3),
+ 1 is divisible by 3; and for even n,

2n + 1 = 2 (mod 3),
that is, 2" + 1 is not divisible by 3.
that is, 2"
1898/2. Prove the following theorem: If two triangles have a common

angle, then the sum of the sines of the angles will be larger in that triangle
where the difference of the remaining two angles is smaller.
On the basis of this theorem, determine the shape of that triangle for
which the sum of the sines of its angles is a maximum.
Solution. a) Denote the angles of the triangles by a,0, y and a', fl',
SOLUTIONS
59
y', respectively and let a = a

'. We iirst ask under what conditions the
inequality
sin a sin8 sin y < sin d
sin8' sin y'
or (equivalently, since a = a')
sin8
(1)
+ sin y < sin@'+ sin y'
holds. This inequality can be written in the form

(2)
B+Y
2sin-
8-7
8'+
cos -< 2sin-cos-.
2
Since (Y = a', we have /3
+ y = 8' +
Y'
8' - Y'
2
2
y' < 180, so that
Dividing both sides of (2) by the positive number sin 3 (8
8-Y
8'-
cos -C C O s - .
2
+y)
we get
Y'
This holds if and only if the absolute value of 8' - y' is less than that of
8 y ; that is, if and only if the difference between the angles 8' and y'
is less than the differencebetween the angles 8 and y.
b) According to a) : If not all three angles of a triangle A are equal,

say 8 # y, then it is possible to h d a triangle A' of different shape, such
that the sum of the sines of the angles in A' is larger than the corresponding s u m for A. To construct A', take A, leave its angle a unchanged,
and replace 8 and y by their arithmetic mean $ ( p y ) . The angles of
A' are a' = a, 8' = y' =
Y).
&(a+
Thus, the sum of the sines of the angles in a non-equilateral triangle
can always be increased in this manner. Therefore, only the equilateral
triangle may maximize this sum.
Note. The existence of a maximum. Our previous considerations show

that if there is a maximum among the sums of the sines of the angles in a
triangle, then this maximum is attained only in an equilateral triangle. However, the existence of such a maximum is far from obvious. For, the sum of
the sines in question may take on infinitely many values, and among infinitely
many numbers there may not be a largest one.
The fact that a maximum exists in our case requires proof; we must show
that the sum of the sines of the angles in an equilateral triangle A, is greater
than the corresponding sum in any other triangle A.
1) If one angle of A is 60, then A, and A both have an angle of 60".
60
The difference of the other two angles of Ax is zero, therefore of smaller

absolute value than the corresponding difference for A.
2 ) If A has no 60' angle, and if a 5 5 y, then clearly a < 60' < y, so
that
60'-a
and y-60'
are positive numbers.
Draw a triangle A' with angles
a' = 60',
Then
0'
= P, hence y' = a
+ y - 60'.
- = (a+ - 60') - 60' = (7 - 60') - (60' - a)

80 that I y' - a
' I is the positiveone of the two numbers
- (y - 60').
(y - 60') - (60'- a) and (60'7'
(Y'
(Y)
In either case
17'- a']'< (y
- 60')
+ (600 - a) = y - a = Iy - 4 .
Now, since a' = 60, we know from 1) that the sum of the sines of the angles
is smaller for A' than for Ax. Moreover 0' = 0 and, as we just saw, the difference of the other two angles of A' is smaller in absolute value than the corresponding difference for A. It follows from a), therefore, that the sum of the
sinesis smaller for the angles of A than for those of A' ;hence, it is also smaller
than for those of A,.
1898/3. Let A , B, C, D be four given points on a straight line e. Construct
a square such that two of its parallel sides (or their extensions) go through
A and B respectively, and the other two sides (or their extensions) go
through C and D, respectively.
Solution. We shall denote the length of a line segment X Y also by
X Y . Let PQRS (Figure 17) be a solution. If we rotate it about its
center, a 90' turn carries CD into C'D'. Of course C'D' I CD
and C'D' = CD. But then there is a point B' with the following
property: The segment BB' is perpendicular to the line e through C
and D, has length CD, and B' is on the side PS through A .
Thus the construction is this: Through B we draw the perpendicular
to the line through C and D and measure off BB' = CD. The sides of
the square PQRS will lie on the straight line AB', on the line through B
parallel to AB', and on the lines through C and D perpendicular to AB'.
Since there are two segments BB' equal and perpendicular to CD,
there are two solutions symmetrical with respect to the line e through C
and D.Figure 17 shows only one of these squares.
Jzutijication of the Sohtim. Througb, B draw the segment B L per-
SOLUTIONS
61
pendicular to AB', and through C the segment CN perpendicular to

D S ; then in the right triangles BLB' and CND, BB' = CD and
(C LBB' = (CNCD because the sides are pair-wise perpendicular.
Consequently, these triangles are congruent. Therefore B L = CN, i.e.
two adjacent sides of the constructed rectangle are equal.
D'
Figure 17
Remark. This construction is correct word for word whatever the position of the given points in the plane may be; see Figure 18.
62
1899 Competition
1899/1. The points Ao, A I , Az, Aa, Ad divide a Unit circle (circle of
radius 1) into five equal parts. Prove that the chords AoAI,A d z satisfy
(AOAI*AoA2)2= 5.
First Solution. In the Unit circle, the length of a chord is twice the
sine of half the subtended central angle. Thus (see Figures 19, 20)
A d 1 = 2 sin 36" = 4 sin 18" cos 18",
and
AoAz = 2 sin 72' = 2 cos 18",

AoAI.AoA2 = 8 sin 18'
COS'
18".
But 2 sin 18" is the length of a side of a regular 10-gon inscribed in the
unit circle; this length, as we know, is $( l/s 1). Hence
&-1
sin 18" =
C O S ~18" =
1 - sinz18" = 1 -
and
8 sin 18" cosZ 18" =
Therefore, in fact,
Figure 19
6 - 21/5
16
5-1
d- d.
4
(AoAl*AOA2)2=
Al
'
(&)2
= 5.
A,
Figure 20
63
SOLUTIONS
Note. Regular polygons. In our problem AoAl is a side of the 5-gon in

Figure 19 (the sides intersect only at the vertices), AoA2 is a side of the 5-gon
in Figure 20. Both figures show regular polygons because, by definition,
every figure constructed as follows is a regular polygon: We start at some
point A0 of the circumference of a circle and measure off a chord a so chosen
that, after measuring it off n times consecutively (following the circle in the
same sense), we come again to Ao and all n vertices are distinct.
For a given n we get all possible inscribed regular n-gons if we choose a as
the chord subtending the arc
taking for k,one after the other, all positive integers less than n/2 and prime
to n. If k is not prime to n, the construction described above leads us back
to A0 too soon; for example, n = 10, k = 4 gives a regular polygon with only
5 instead of 10 vertices.
For k = 1, we get a regular polygon whose sides meet only at their end
points.
For k > 1, we get a regular polygon whose sides intersect also at other
points.
If n is an odd prime p, k may take on all integral values from 1to b ( p - 1) ;
so in this case there are +(p 1) different inscribed regular p-gons.
If n is a composite number, there are fewer than #(n - 1) different regular
n-gons;thus, if n = 24,there are only four.
Second Solution. In the following solution our knowledge of regular

polygons is not used.
The lengths of the segments AoAI, A0.42 are connected in a simple
way with the cosines of those angles cp for which Scp is an odd multiple of
90, that is, those for which
cosscp = 0.
We may restrict our attention to pvalues from 0 to 180" since, for values
greater than MOO, we can always find an angle between 0 and 180" with
the same cosine. With this restriction, the cosines are:
cos (**No)
= cos 18" =
CQS
(3.90") = cos 54' =
cos (8-90") = cos 90" =
sin 72"
aAoA2,
sin36" =
$AoAi,
0,
cos (3.90")= cos 126" = -sin36'
= -4AoA1,
cos (f.90") = cos 162' = -sin 72" = -&AoA2.
We shall rewrite the equation cos Scp = 0 as a polynomial equation in
64
cos p (which we shall denote by x) ; the above expressions will be roots

of this equation, and the value of ( A o A I - A o Acan
~ ) ~then be found easily.
To this end we shall express
sin (n 1)p
cosw
and
sin p
recursively for n = 0, 1, 2, 3, 4, 5 in terms of x = cosq. w e use

sin (n 1) p/sin p instead of sin (n 1) p in order to obtain expressions
free of square roots.]
V,(x) we h d for cos np, sin (n l)p/sin p,
The expressions Tn(x),
respectively, in terms of x = cos p are obtained as follows:
sin (0
coso = To(x)= 1,
= Uo(x)= 1,
sin p
cosp = Tl(X)= x,
sin 2p 2 cosp = Ul(X) = 2%.

Sin p
In general, we shall use the recursion identity

cos (n
+ 1)p = c o s p c o s ~- sinpsinlzrp
=
that is,
sin np
cospcoslzrp - (1 - cos2p) -,
Sin p
and the recursion identity
sin(n+2)p
that is
Sin p
1
-[sin
sin Q
p cos
(n
+ l ) p + cos p sin + l)p],
Un+1(x) = Tn+l(X)
(2)
(fi
+ xVn(4.
Alternate application of (1) and (2) gives
- (1 - X")Uo(X) = x*x - (1 - x2) = 2x2 - 1

Uz(x)= T2(4 + xV,(x) = 2x2 - 1 + x-2x = 4x2 - 1
T&) = x(2x2 - 1) - (1 - x2)*2x= 4x8 - 3%
U~(X)
= 4x8 - 3%+ 4 4 9 - 1) = 8x8 - 4%
T~(x)
=~ ( 49 3 ~) (1 - 9)(W - 1) = 8x4 - 8x2 1
U4(x) = 8x' - 8x2 + 1 + 48x8 - 4 ~ =
) 16x4 - 12x2 + 1
T4x)
xT1(x)
SOLUTIONS
= ~ ( 1 6 ~ 420x2
65
+ 5) = x[(~x)' - S(22)' + 51.
In our notation, Tb(x)= cosSq, where cp is an angle such that

cos Scp = 0. Therefore, the roots of Ts(x)= 0 are just the five cosines
3Ad1, -4AoA1,
0, 3AoA2, -3AoAz
evaluated earlier. If from Tb(x), we take out the factor 2 corresponding
to the root 0, and if we set
( 2 4 2 = u,
we get the quadratic equation
ua - 5u
+ 5 = 0;
its roots are just ( A O A I )and

~ ( A o A ~ Since
) ~ . the product of the roots is
the constant term, it follows that
(AoA1*AoAe)*= 5.
Note 1. Generalization of the theorem proved above. Chebyahev
polynomials. The theorem just proved is a particularly simple instance of
the following:
If n = p , where p is a pfime and k a positive integer, and if the different
regular n-gons inscribable in the unit circle have side lengths a1, a,
at, then
9,
(a1
a .**
at)* = p.
Figure 22
Figure 21
For example, for n = 2*, the sides of the regular inscribed 8-gons (Figures
21,22) have lengths
45O
a = 2sin-
and
135'
2
a- 2sin-=
45O
2ws--.
2
66
The product u1u2 is
49
45O
4sin-m2
and
(UlU#
= 2sin45O = ~ 2 ,
= 2.
A proof of the general theorem involves a closer examination of the

equations
UI(X)
T Z ( X ) = 0,
= 0,
TZ(X)
= 0,
UZ(X)
= 0,
.**.
The polynomials m ( x ) and Vn(X) which express
cosnq and
sin ( n 1)cp
sin cp
as functions of x = cow are called Chebyshey po1ynomials.t
Note 2. A geometric application of complex numbers. The following

holds for regular polygons with any number of sides:
If Ao, A I ,
An-1 are the Vertices of a regular n-gm inscribed in the unit
circle, then
,4021 AoAz * * * A o A k i = n.
0 ,
We give a simple proof for those acquainted with complex numbers.$

We represent the regular n-gon in the complex plane, drawing the unit
circle about the origin and placing the vertex Ao at the point 1 on the real
axis (see Figure 23). If we denote the nearest vertex (going from A0 in the
counter clockwise direction) by the complex number e, then the vertices of
the n-gon correspond to the complex numbers
1,
3,
0 - 0 ,
-I;
each has absolute value 1, and their amplitudes are
0,
360
n ,
360"
2-,
*--
360
(n- 1)-9
n
respectively.
t P. L. Chebyshev, a Russian mathematician,was born in 1821 near Moscow; he

died in 1894 in St. Petersburg, where he was a professor at the university. He proved
the following oft-quoted theorem of number theory: I j x 2 1, t h there is u prime
#forwhich x < p s 2 l c . Forexmple,if x = 1, then 1 < 2 5 2 ; if % = 10, then
10 < 11,13,17,19
< 20.
3 For some basic concepts concerningcomplex numbers, see for example K. Knopp,
Bments of the Theory of Functions, New York, Dover Publications,Inc., 1952.
67
SOLUTIONS
Figure 23
The lengths of the segments from Ao to the other vertices correspond to

the absolute values of the complex numbers 1 - E, 1 - e',
1 - F1,
so that the theorem, in terms of complex numbers, states
0 ,
I(1- e)(1- 8)***(1(1)

T o prove this, we note that
E= 1
and, in general, fos integers k,
(8)"= (")k
Thus 1, E, 3, - - * ,
I = n.
-I)
= 1.
are the zeros of the polynomial
P- 1 =
l)(et.-'+
(2-
zn-2+
- 0 -
+e+
1).
Since the first factor on the right vanishes only at I= 1, the numbers
c,
*--,
2,
-I
are the zeros of the polynomial

f(2)
= eel+
2-2+
**-
+z+
1.
Therefore, f ( e ) factors into

f(2)
(2-
E)(2-
If we substitute 1for e, we obtain
f(1) = (1 - E ) ( 1 -
2)*--(e-
-I).
3)-*.(1--I)
12.
This is more than (1) claims: not only the absolute value of the product,
but the produe itself is n.
68

1899/2. Let x1 and x2 be the roots of the equation
9 - (a
+ d ) x + ad - bc = 0.
Show that xi and x i are the roots of

y2
- (a8 + d8 + 3abc + 3bcd)y + (ad - bc)8 = 0.
Solution. By known relations between the roots and coefficients of

second degree equations, the fact that XI, x2 are roots of
implies that
9- ( a + d ) x + a d - b c = O
+ d = x1 + xz, .
ad
- bc = x1xz.
We use these to express the coefficients of the second equation in terms

of ~1 and ~ 2 :
a8
+ d8 + 3abc + 3bcd = a8 + d8 + 3(a + d)bc

= ( a + d)8 - 3(a + d) (ad - bc)
= (xi
= xi
and
X2)8
- 3(S
+ S)xi%
%;
(ad - bc)8 = .:xi.
Therefore, the equation
r - (a8 + d8 + 3abc + 3bcd)y + (ad - bc)8 = 0
is equivalent to
and this factors into
+ xi)y + .:xi
r - (4
(y
- xi) (y - xi)
= 0,
= 0.
Thus its roots are indeed x: and xi.
Note. On reducing irrational equations. Denote a

d by p and
ad - bc by q ; then the problem answers the following question:
If the cube root of a number y satisfies
3c2+ p x
q = 0,
does y also satisfy an equation in integral powers of y, and if so, how is this
equation found? In short, can
be reduced to an equation in which no fractional powers of y appear, and if
SOLUTIONS
69
so, how? In the above problem the desired equation was given and merely
had to be verified. However, one can derive the equation and this derivation
will show how, in general, one can eliminate radicals from an equati0n.t
First, multiply equation
( 1)
by
fi,obtaining
(rn2+plS/;;+a=0
P(fi)2+S4G+Y=0.
(2)
Next, multiply this result by
1"Jj;:
(3)
Eliminate ( 13/;;)2 from equations (1) and (2), then from (1) and (3). This
gives
(Y - afi)lS/;;+ p y
Then, eliminate fifrom these two to get
- q2 = 0.
which is the Same as
If we substitute the values of p and p into the last equation, we obtain the
polynomial of the original problem.
In this procedure i t was inessential that the radical sign stood over the
unknown y alone; one could just as well have had a Kth root,
of a more
complicated expression in which the unknown is combined with constants
by means of the operations
-, X ,+.Nor need K be 3. To handle the more
general case, we multiply the given equation repeatedly by the radical to
be eliminated until we have the right number of equations in order to eliminate every power of that radical. This number is finite, since only the first
k - 1powers of the radical can 0ccur.S Then, all we have to do is to eliminate
successive powers of the radical and thus eliminate it entirely.
g,
+,
t It is to be understood that we deal here with polynomial equations Q ( x ) = 0

having rational coefficients, i.e., with algebraic equations, but that the unknown may
appear in irrational form; the same question might be formulated also as follows:
If fil,
&arerootsof Q ( x ) = O,constructanequationQ'(z) = Owithrootsnandrs.
$ Any power greater than k - 1 of the radical can be reduced to a power less than or
equal to k - 1 by means of the relation
For example,
(l"TYY>k= Y.
(qn"= Y($?)?
70
Should there be several radicals, possibly with different exponents, the

process is to be repeated for each, until no radical sign stands over an expression involving the unknown.
This plausibility argument shows: A n equation containing the unknown in
irrational form is reducible to one in which only integral powers of the unknown
occur, no matter how many radicals the original equation contains and to
what powers. We shall not consider the details of the proof.
Though a method of constructing the desired equation has been outlined,
i t is to be noted that, as the number of radicals and their powers increase,
the calculations become quite difficult.
1899/3. Prove that, for any natural number
2903" - 803" - 464"
1z,
the expression
+ 261"
is divisible by 1897.
Solution. By virtue of the well-known identity
an - b" = ( a - b) (an-1
+ a"-% +
ah"-2
bn-1)
the difference of the lzth powers of two integers is divisible by the difference of the integers themselves.
The number A equals
2903" - 464" - (803" - 261"),
and 2903" - 464" is divisible by
2903 - 464 = 2439 = 9 271,

while 803" - 261" is divisible by
803
- 261 = 542 = 2 * 271,
so that A is divisible by 271; that is A

But A is also equal to
271B, where B is an integer.
2903" - 803" - (464" - 261"),
where 2903"
- 803" is divisible by
2903 - 803 = 2100 = 7 300,
and 464" - 261" is divisible by
464 - 261 = 203 = 7 * 29,
so that A is also divisible by 7.
SOLUTIONS
71
Since 271 is not divisible by the prime 7, A = 271B is divisible by 7

only if B is [see part a) of the Note to the Second Solution of 1894/1].
Hence, B = 7C,where C is an integer, so that A = 271 7C = 1897C is
indeed divisible by 1897.
1900 Competition
1900/1. Let a, b, c, d, be fixed integers with d not divisible by 5.
Assume that m is an integer for which
am8
+ bm2 + cm + d
is divisible by 5. Prove that there exists an integer n for which
dn8
is also divisible by 5.
+ cn2+ bn + a
Solution. The integer m is not divisible by 5. For, otherwise,
am8
+ bm2 + cm + d = m(am2+ bm + c) + d
would be divisible by 5 only if d were divisible by 5, contrary to the

assumption.
The number m, therefore, is of the form
m=5k+r
where k is an integer and r a positive integer less than 5.
For the sake of brevity, let us set
A=am8+bm2+cm+d,
dn8
+ cn2 + bn + a.
If we eliminate d from these expressions, we get
+ bn(m2n2- 1) + cn2(mn- 1)
= (mn - l)[a(rnrc+ mn + 1) + bn(Pnn + 1) + cnz].
Now, if m = 5R + r is a number such that A is divisible by 5, and if we
An8 - B
a(m8n8- 1)
can choose n so that the right side of the last identity is divisible by 5,
then B will be divisible by 5. We shall achieve this by choosing, for every

m not divisibleby 5, an integer n so that the factor (mn - 1) on the right
is divisible by 5. In particular, since m = 5R r, mn = 5kn rn, so
that mn - 1 is divisible by 5 whenever rtz is 1 plus a multiple of 5. So,
72
if Y = 1, choose n = 1; if Y = 2, choose n = 3; if Y = 3, choose n = 2;

and if Y = 4, choose n = 4. In this way, for every m,we have found an
n which makes mn - 1 divisible by 5, and hence, for which B is divisible
by 5.
1900/2. Construct a triangle ABC, given the length c of its side AB,
the radius Y of its inscribed circle, and the radius r, of its ex-circle tangent
to the side A B and to the extensions of BC and C A .
Solution. a) Let D be the point of tangency of the inscribed circle

k to the side BC;let Dbe the point of tangency of the other circle k to
the extension of BC. (See Figure 24.)
Figure 24
We know [cf. 1896/3, Note 2, (1) and (2)] that, if a, b, c are the
lengths of the sides and 2s the perimeter of the triangle,
CD = s - C,
so that
DD = s -
CD = S,
(S
- C)
C.
b) From the fact that the circles k and k (with centers 0 and 0
respectively) do not intersect, we shall derive a relation between Y , Y,
and c. To this end, draw a line through 0 parallel to BC, intersecting
OD in P; AOPO is a right triangle with sides OP = DD = c,
OP = rc - Y, and hypotenuse 002 Y, Y. By Pythagorastheorem,
that is
(00)Z= 8
I:
+ (rc -
Y)Z
1 (rc
+ r)2,
- 2YcY + P 2 < + 2YcY + P,
SOLUTIONS
73
c) If the given data G, Y , and r, obey condition (1), and if r, > r, then
AABC may be constructed as foIIows: Draw a segment DD' of length
G, and draw perpendicular segments OD, O'D' of lengths Y , Y, at D, D',
respectively. Draw the circle k of radius Y about 0, and the circle k' of
radius r, about 0'. Since Y, > Y , the line through 0 and 0' intersects the
line through D and D'. Their point of intersection is the vertex C of the
desired AABC, since C is the only point on D'D from which two common
outer tangents can be drawn to both circles.
The circles k and k' do not intersect, because condition (1) implies
( O o y = 8 + (Yc - r)2 =
(r,
+ +
r)2
G2
- &Y, 2
(rc
Y)?
Next, draw one of the inner tangents common to both circles. It will
intersect the outer tangents from C in the desired vertices A and B of
AABC because, by the reasoning in a), A B = DD' = G. (The other
inner tangent yields another solution congruent to the first. If k and k'
are tangent, there is only one solution.)
1900/3. A cliff is 300 meters high. Consider two free-falling raindrops
such that the second one leaves the top of the cliff when the first one has
already fallen .001 millimeters. What is the distance between the drops
at the moment the first hits the ground? (Compute the answer to within
.1 mm. Neglect air resistance, etc.)
Solution. Let u be the distance the first drop has fallen at the instant
the second drop leaves the cliff, and let r be the time needed for this
descent of u. Let s be the total height of the cliff and t the time needed
for a raindrop to fall this entire distance. Let s' be the distance the second
drop falls from the time it leaves the cliff to the time the first hits the
ground, that is, in the interval t - r. We wish to compute s s'.
Then, by the law governing the free fall of bodies under the influence
of gravity, we have
u =2gr
l2
Hence,
s = +gt2, and s' = $g(t - T)~.
t This much of the solution answers also 1927/3.
74
so that
and
s - sf = 2
- *.
In our problem, u = .001 mm., s = 300 OOO mm., so that

s
- s' = 2
a - .001- 34.6 mm.
1901 Competition
1901/1.
Prove that, for any positive integer n,

In
+ 2" + 3" + 4n
is divisible by 5 if and only if n is not divisible by 4.

First Solution. Each of the numbers
14'1,
24=16=5*3+1,
+
+
34=81=5*16+1,
44=256=5*51+1
is of the form 512

1, where 12 is an integer. Moreover, if A stands for
any of the numbers 1, 2, 3 or 4, and if 1 is a positiye integer, then A4' is
of &e form 512 1 because (cf. identity used in 1898/1)
A4'- 1 = (A4)'- 1 = ( A * - 1)[(A4)"+
(A4)'--2+
.*-
+ 11,
where the fist factor on the right is divisible by 5 and the second is an
integer.
Any positive integer may be written in the form 41 r where 1 is a
positive integer or zero, and the remainder r is one of the numbers 0, 1,
2, or 3. In particular, our exponent n may be written in this form so that
s = 1" + 2" + 3" + 4"

=1
+ 24149 + 34'49 + 44*
+ 24'27 + 3413' + 44'4'

= 1+
+ 1)2'+ (512s + 1)3'+ + 1 ) 4
= 5(k# + 1
2
3 + k&) + (1 + 2' + 3' + 4')
=i
(5128
(5121
= 5%
+ R,
where m is an integer, and
=1
+ 2' + 3' + 4'.
SOLUTIONS
75
Thus, S is divisible by 5 if and only if R is. If n is divisible by 4, then

= 0 and R = 4; in this case S is not divisible by 5. But if n is not
divisible by 4, then r = 1 , 2 or 3, and the cOrreSpOndingvalues of R are
l0,30 or 100; in this case S is divisible by 5.
Second Solution. Observe (cf. Note 1 to 1898/1) that
24 = 16 = 1 (mod S),
l4= 1 zs 1 (mods),
34 = 81 = 1 (mod 5),
Let n = 4k
or 4,
Thus
Y,
44 = 256 = 1 (mod 5).
where k is an integer and Y is 0, 1, 2 or 3. If u is 1 , 2 , 3
a4 = 1 (mods), hence ua = 1 (mods).

an = uycaC E ur (mod 5) ,
and by the theorem on congruences (cf. Note 1 to 1898/1)
S = 1"
It follows that
+ 2" + 3" + 4" = lr+ 2' + 3' + 4 (mod 5).

S = 4=4(mod5);
S = 10=O(mod5);
S = 30=O(mod5);
if r = 0 ,
if 1 = 1 ,
if Y - 2 ,
if Y = 3,
S = 100 = 0 (mods);
so that S is indeed divisible by 5 if and only if n is not divisible by 4.

Note. Fermat's theorem. Our solution is based on the fact that
14=
3 ~4 s
1 ~(mod 5).
This congruence is a special case of the following theorem due to Fermat:t

If a is an integer not divisible by the prime 9, then
UP-'
= 1 (mod 9).
t The French mathematicianP. Fermat was born in 1601 and died in 1665,a member
of the Toulouse parliament. He proved many beautiful number theoretical results,
among them the one stated here. Following the custom of his time, he published his
results without proof. Later mathematicians endeavored to prove his theorems and
in the course of their efforts, contributed much to number theory. Fermat also played
a part in the founding of analytic geometry.
Fermat's Theorem, stated here, should not be confused with the famous conjecture
known as "Fermat's Last Theorem", which says that no non-zero integral values of
2; y, e exist that satisfy the equation P
p = @ i f n is an integer greater than 2;
see 1909/1,Note.
76
To prove this, consider the set of all permutations of m things, p at a

time, with repetitions allowed. Classify these mp permutations by putting
into one class all those that can be obtained from each other by cyclic permutations (cf. 1895/1, Note 1). The number of elements in any one class
is a divisor of p (as we showed in Note 1 to 1895/1), and hence is either
1 or p. We shall show that there are precisely m classes with one element
each, and so the remaining permutations fall into classes of p elements each.
An arrangement consisting of a p-fold repetition of the same object is in a
class by itself because a cyclic permutation (which replaces every object
by its left neighbor, and the first by the last) just yields the same element
back again. Since there are m objects, there are m different p-fold repetitions
of one object; hence there are m classes each containing one such permutation.
Each of the remaining mp - m permutations contains at least two distinct objects, and a cyclic permutation of such an arrangement yields a
new permutation. Consequently, these mp - m permutations must fall
into classes containing p elements each.
Thus mp - m is p times the number of these classes, that is,
mp-m=O(modp).
It follows that
m(mp-
1)
= 0 (modp).
This proves Fermats theorem for any positive integer m, since the last
congruence implies that, if m is not divisible by p , then mp - 1 is. (Cf.
Note to Second Solution of 1894/1.)
To prove the theorem for any integer a, choose a positive integer m congruent to a (mod p ) . It is easy to show that Fermats theorem for m implies
i t for a.i
1901/2.
If
u = cot 2230,
1
sin 2230
prove that u satisfies a quadratic and v a quartic (4th degree) equation

with integral coefficients and with leading coefficients 1.
There are many ways of proving Fermats theorem. Two proofs by Euler were the
f i s t to appear in print. An earlier proof by Leibniz appeared only in 1863 in his collected works. Gauss found, independently, a proof similar to that of Leibniz. The
proof given here is based on those due to Gauss and Leibniz.
G. W. Leibniz, philosopher, mathematician, and diplomat, was born in Leipzig in
1646 and died in Hannover in 1716. He invented the differentid and integral calculus
at the same time as Newton did.
SOLUTIONS
77
Solution. We first construct an angle of 22'30', that is one-fourth of
a right angle. This may be done as follows (see Figure 25) : Let ABC be
an isosceles right triangle with C A = BC, extend the side C A to a point
D such that A D = AB; draw the segment BD. The angle at D of the
isosceles triangle DAB is half the angle CAB, which in turn is 45'. So,
q A D B = 22'30'.
QBc
If we choose AC as the unit length, then
DA=AB=1/AP+BP=aY
and
DC=DA+AC=a+l,
DB = ~ / D P BC? = 1/(a I ) * + 1 =
The values of u and v are:
DC
u = cot 22'30' = - -a+&
BC
V =
so that
These give
DB
--=
sin 22'30'
BC
1
(U - 1)2 = 2, v2 = 4
+2a,
m.
dwm,
(9
- 4)2 = 8.
~ 2 - 2 ~ - 1 = 0and. 0"-88v2+8=0.
Note. Algebraic and transcendental numbers. Algebraic integers.
If a number a satisfies an algebraic equation
x^+urxn-'+
***+,=
0
with rational coefficients UI, a, ,an, then a is called an algebraic number.
Every rational number Y is algebraic since it satisfies the algebraic equation
with rational coefficients x - I = 0. However, not every algebraic number
is rational; for example, d2 is an algebraic number because it satisfies the
equation x2 - 2 = 0,yet it is irrational.
78
Not every number is algebraic. Those real or complex numbers which are
not algebraic are called transcendental numbers.? Among the well-known
numbers, e (the base of natural logarithms) and a are examples of transcendental numbers. The first proof of the transcendence of e is due to Hermite$
( 1873) ;that of the transcendence of ?r is due to Lindemanns (1882). ( I t is a
consequence of the transcendence of a that it is impossible to construct,
with straightedge and compass alone, a line segment equal in length to the
circumference of a given circle, or a square having the area of a given circle.)
Among algebraic numbers, we distinguish algebraic integers: A number a
is said to be an algebraic integsr if it satisfies an algebraic equation with integral coefficients and leading coefficient 1. For example,
a1 =
-1
++5
2
a2 =
-1
-d
2
are algebraic integers in spite of their fractional form, because they are roots
of the quadratic equation
X2+x-l=0
with integral coefficients and leading coefficient 1. I t can be shown that the
sum, difference, product and quotient (with a non-zero divisor) of two algebraic numbers are algebraic numbers, and that the sum, difference and
product of two algebraic integers are algebraic integers.
Our problem was to prove that the cotangent and the reciprocal of the
sine of 90/4 are algebraic integers.
1901/3. Let a and b be two natural numbers whose greatest common
divisor (g.c.d.) is d:
(a, b] = d.fi
Prove that exactly d of the numbers

are divisible by b.
a, 2a, 3a,
***,
(6 - l ) a , ba
t For simple expositions of the two classical proofs of the existence of transcendental
numbers, see I. Niven, Numbers: Rational and Irrational in this sexies.
$ Hermite was born in 1822 and died in 1901 in Pans, where he was a professor at
the Ewle Polytechnique and at the University. He was president of the Acadhie des
Scimces. As the most outstanding mathematician of his time, he was a member of
many scientiiic societies.
F. Lmdemann was born in 1852 and died in Hannover in 1939; he was a professor
at the University of Munich.
a The symbol (x, y ) means g.c.d. of x and yjsee Note 2 for a precise definition of
g.c.d.
SOLUTIONS
Solution. We have
79
b=&;
a=&,
hence [see Note 3, themem (i)]

(T,
s) =
1.t
, (6 - 1)a, ba by 6, we obtain the quotients
If we divide a, 2a,
2T
-Y -
...
( 6 - l)r
(ds)r
S
Since Y and s are relatively prime, the only integers here are the quotients
in which the coefficient of I in the numerator is a multiple of s; see Note
3, theorem (ii). Since b = ds, this happens d times as the coefficients go
through the values 1,2,
ds.
a,
Note 1. The divisors of a positive integer. It is easy to find the divisors

of a positive integer if we can write it as a power product of primes (cf. Note
to 1896/1). If
a = vp,
wp,
where v and w are not divisible by the prime p , then the product ab contains
fi to the power a @.(This becomes evident when the multiplication is
Camed out.)
It follows that a positive number a is divisible by b if and only if b contains
no prime to an exponent higher than the corresponding exponent in the
prime power product for a.
For example, the number a = F p , where p is an odd prime, has the following divimrs:
1, 2, 22,
P,
2P, 22p,
-*-,
2-1,
..., 2 9 ,
T,
2P.
Note 2. The greatest common divisor. If the integers a, b, -, m are

not all zero, they have a finite number of common divisors. The largest of
these is always @ d i v e and is called their greatest common divisor; it is often
, m)
denoted by (a, b,
Since zero is divisible by every number, the greatest common divisor of
any set of integers is equal to that of the non-zero integers among them.
Moreover, negative numbers may be replaced by their absolute values.
Hence, for any set of integers, finding the g.c.d. is reduced to finding that of
a set of positive integers.
From Note 1 we know that if a, b, .*-, m are positive integers and if d is
their g.c.d., then the @wer of a given prime in the prime power product for d is
--
t If the g.c.d.
of two numbers is 1, they are called relatively prime; see Note 3.
80
the smallest that occurs i n the prime power productsfor a, b,

,and m for that
prime.
For exampre
(26
324 7 9 1116, 2 9 310 138) = 2 6 310.
. .
. . .
This shows also that a number k divides each of the numbers a, b,

m only i f k divides their g.c.d.
-**,
and
Note 3. Relatively prime numbers. If two or more integers have 1 as

their g.c.d., i.e., if they have no cemmon divisors other than 1, they are
called relatively prime.
For example, if p and p are distinct primes, then and are relatively prime
for all @airs r, s of integers.
The following two theorems were used in the solution:
(i) If a, b,
integers
- 0
0 ,
m are integers, not all zero, and
if d is their g.c.d.,
then the
are relatively prime.

Suppose, to the contrary,that
Then
(a, b,
***,
m) = d,
a = ad, b = bd,
..*,
with d
> 1.
m = md
> d, so that d would not be the g.c.d. of a, b, ,m.

If b is a non-zero integer, ( a , b) = 1, and if the product ak of a and an
are all divisible by dd
(ii)
integer k i s dievisible by b, then k i s divisible by b.
For, if k were not divisible by b, then the prime power product for I b I
would contain a t least one prime p raised to a positive power 0, while the
power product for 1 k 1 would contain p to a power (Y < /3 (possibly a = 0).
Since ( a , b) = 1, p cannot divide a ; therefore p has the exponent a < /3 in
the prime power product for ka , and hence ka is not divisible by b. This
contradicts the hypothesis. Therefore k is divisible by b. (If a is a prime, this
theorem is identical with theorem a) in the Note to the Second Solution of
1894/1.)
I I
The following theorem is also noteworthy:

If the integer A is divisible by a and b and if ( a , b) = 1, then A i s divisible
by the product ab.
For, if A is divisible by a, there is an integer k such that A = ak; and by
theorem (ii), since ( a , b) = 1, A = ak is divisible by b only if k is, i.e., if
k = bk where k is an integer. Hence A = abk so that ab divides A.
SOLUTIONS
81
1902 Competition
1902/1. Prove
that any quadratic expression
Q(x) = A d + B x + C
a) can be put uniquely into the form
Q(x) = k
x(x - 1)
1.2
+ Ix + m,
where k, I , m depend on the coefficients A , B , C, and

b) Q ( x ) takes on integral values for every integer x if and only if k, 1,
m are integers.
Solution. a) Since Z = Z - x
d = 2
+ x = x(x - 1) + x, we may write
x(x
- 1)
+ X
and substitute this expression in Q. We obtain
Q(x) = 2 A
x(x- 1)
2
and this has the desired form
x(x- 1)
Q(x) = k
with
k=2A,
-I- ( A
1.2
+B)x+
C,
+ Ix + m,
1= A+B,
m = C.
b) (i) If Q ( x ) is an integer whenever x is, then, in particular, the

values Q(O), Q ( l ) ,Q(2) obtained by setting x = 0, 1 , 2 are integers. But
Q(0) = m,
so m is an integer,
Q(1) =
so 1 = Q ( l ) - m is an integer,
Q(2) = k
+ m,
+ 21 + m, so
k = Q ( 2 ) - 21 - m is an integer.
(ii) Conversely, if k, I , m are integers, then for every integral value of
x, Q ( x ) is an integer. For, if x is an integer, then either x or x - 1 is an

even number so that x(x - 1 ) / 2 is an integer; so, if k, 1, m are integers,
Q(integer) is the sum of three integers, hence itself an integer.
a2
Note 1. Polynomials that take on integral values for integral values

of the variable. The following generalization of statement a) in the problem is easy to prove by induction:
Every nth degree polymmia2
F(x) = ao+ uix+
wz+ * * *
can be brought into the form
where
&,
( p - 1 )
t)=
x(x-
1.2
+ an9
1)(x-
1.2.3
2)
l
...
and the coe5cients bi are uniquely determined by the coe&cients ai. Moreover, (f
the ai are integers, then the bi are integers.
PROOF.
The theorem is clearly true for first degree polynomials; part a ) of
the solution to the problem shows that it is also true for quadratic polynomials. Now, let us assume that it is true for polynomials up to degree
k - 1, and show that it then follows for polynomials of degree k. Consider
A(%)
= a0
+ alx + + aL1&l+
= Pk-l(X)
***
a d
By multiplying out, we can verify that

X(X
- 1) ( X - 2)**.(~- k + 1) = xk + GL~(x)
where G h l ( x ) is a polynomial of degree k

Thus
and
xk=
X(X-
I)(x-
- 1 with
2)**.(~- k + 1)
integral coefficients.
- G~i(x),
The difference of two polynomials of degree k - 1 with integral coefficients

is a polynomial of at most degree k - 1with integral coefficients; therefore,
by the induction hypothesis, the firsttwo terms on the right may be put into
the form
But then Pk is of the desired form
83
SOLUTIONS
where bk = k !& Moreover, if the a, are integers, then so are the bi.
Statement b) of the problem is generalized by the following theorem:

The polynomial F(x) takes on integral values for every integer x if and only
@ the m e e t s bo, b1,
,b, i n ( 1) are integers.
PROOF. The proof follows the pattern of the proof given above for the
a n = 2.
(i') Assume that the coefficients bi are integers; if we can show that the
expressions
are integers when x and k are integers, k 2 0, then F(x) will be a sum of
products of integers, and hence an integer.
We prove that
(3
X(X-
1 ) ( ~ -2 ) * * * ( ~ k- + 2 ) ( ~ - k + 1)
1 * 2 * 3 * * * ( k -l ) k
is an integer whenever x is as follows:
If x is a positive integer or zero,then
is the number of possible combinations (without repetitions) of x things, R

at a time,t and hence an integer; if x < R, this number of combinationsis 0.
If x is a negative integer, say x = -y , then
(-3
-y(-y-
= (-1)k
l)***(-y- R+2)(-y1 2**.(R - 1)k
b + k -
k + 1)
1)(Y+k-2)***(~+1)~
1 2***(K l ) k
Here y 4-k - 1is a positive integer, so that
is indeed an integer.
t Cf.Note 2, b) to 1895/1, page 26.
84
Next, we must show that, if F ( x ) is an integer for every integer x, then

the coefficients bi are integers.
(ii) As in the case n = 2, observe that bo = F(0) is an integer,
F(1) = bo
F ( 2 ) = bo
+ bl,
+ 2bl+
- bo is an integer,
62 = F ( 2 ) - bo - 2b1 is an integer
so that 61 = F( 1 )
so that
b2,
and, in general, for any integer r I n,
since
(L)
but
0 for r
(;)
< k;
= 1,
so this equation can be written as
We are given that F(r) is an integer, have shown that
is an integer whenever k is, and have already shown that bo, bl,
b-1
are integers. Therefore, b, is a sum of products of integers, and hence an
integer. We use this argument for r = 0, 1, 2, .-*, n and so show that all bi
in (1) are integers.
a * * ,
Remark. Observe that in the above proof, we deduced that the b, are integers
by using only the integral values F( 0) , F( 1) , *
F(n). The values of
F ( n 1), F(n 2),
did not matter. But once we have integers bi, the
fact that F(x) is an integer for all integers x can be deduced from (i). Thus
we have the theorem:
If the polynomial F ( x ) of degree n takes on integral values for x = 0,1, -,
n,then it takes on integral values for all integers x.
+ --.
a,
Note 2. The binomial theorem and the binomial series. The expressions
occur in two familiar situations for positive integers x, k : first as the number
85
SOLUTIONS
of combinations of x things, k at a time (see Note 2 to 1895/1), and second

as coefficients in the binomial expansion of the expression ( a b)". To show
that for a positive integral exponent x,
(a
+ bIZ
we merely analyze what it means to multiply the expression a b by itself

x times: Each term in the result will consist of a product made up of one term,
either a or b, from each of the x factors; these products must therefore be of
the form
a*kbk,
0I
kI
x.
For a fixed value of k , we get as many products aekP as there are ways of
picking k b's from x factors. Hence, the coefficient of the product aPkP is
the number of combinations of x things, R at a time. (It is clear that this will
be equal to the number of combinations of x things, x - k at a time; for, the
number of ways of picking out k things from x must be the same as the number of ways of holding on to x - k out of x things.) This proves the identity (1).
If we set a = b = 1, we obtain
2.=l+(:)+()I)f...+(
)+O;
n- 1
we arrived at the same identity in Note 2, c) to 1895/1.
When x is not a Dositive integer, we no longer have these simple combinatorial interpretations of
However, the expressions
called binomial coeficients, still retain their significance for binomials in the
following sense: If z is a number such that - 1 < z < 1and x any real exponent (not necessarily a positive integer), then (1 2)" is expressible as the
infinite series
86
HUNGARIAN PROBLEM
BOOK
It is called the binomial series and Newtoni was the first to consider it for
arbitrary exponents. (The proof that this infiniteseries represents the value
(1 a)z belongs to the cal~ulus.)In the special case x = -1, this series
becomes the infinite geometric progression
1- a + $ -
9+
whose sum, for - 1 < a < 1, is indeed

(1
z) =
--*
1
l+s
1902/2. Let S be a given sphere with center 0 and radius Y. Let P be

m y point outside the sphere S, and let Sbe the sphere with center P
and radius PO. Denote by F the area of the surface of the part of S
that lies inside S. Prove that F is independent of the particular point P
chosen.
Solution. Let T be the point on S antipodal to 0; that is, let the

other endpoint of the diameter through 0 be T.Consider a plane through
0, T and a point Y common to both spheres, see Figure 26; let k and k
be the great circles in which this plane cuts the spheres S and S, respectively. Then rZ and k have a second point of intersection, denoted by N .
The lines MN and OT intersect in a point Q.
Figure 26
t Isaac Newton was born in 1642 and died in 1727.He was professor of mathematics
at Cambridge University. He was the first to state the law of universal gravitation and,
simultaneously with Leibniz, created the differential and integral calculus. He is
buried in Westminster Abbey; the inscription on the marble monument that gram
his tomb ends with the lines: Sibi grcrtdentur mwtales tale tantumque ezstitisse hzrmani
gnrcris decus. (Mortals rejoice at so great an ornament to the human race.)
SOLUTIONS
87
We claim that the surface area F of the spherical cap of S' inside S is
equal to the product of the circumference of k' and the height OQ of the
cap i.e., that
(1)
= 1 -OT
OQ,
and does not depend on the position of the center P of S'.

We s h d make formula (1) plausible iri Note 1, below.
Note 1. On the surface area of the spherical cap, spherical segment,
and sphere. a) These three surfaces are generated respectively by the
following portions of a semicircle: a terminal arc, an inner arc, the entire
semicircle (Figure 27a, b, c) . Each surface is obtained by rotating the arc in
question about the diameter of the circle. We can approximate the surface
area by dividing the arc into equal segments, connecting the points of division by line segments and rotating these along with the arc; see Figure 27a',
b', c'. Smaller subdivisions of the arcs give closer approximations. This is
intuitively clear. What we get is the sum of the surface areas of (truncated)
cones; 80 let us first find these areas.
b'
C'
Figure 27
b) Let f i be the radius of the base of a right circular cone and d its slant
height. If the cone is cut along a line from the vertex and laid out in the plane,
the lateral surface is a sector of the circle of radius d, whose bounding arc has
88
length 27rpl; see Figure 28. Let $1 be the area of the sector; then, since the
area of the sector is to the area G?T of the whole circle as the length 2lrp1 of
its arc is to &e circumference 2 ~ ofd the whole circle, we have
27Wl
-tl= - =
2lrd
d+r
This gives
PI
tl = 7rp1a
as the lateral surface area of the cone.
Figure 28
Figure 29
To calculate the lateral surface area of a truncated cone, the plane figure
we consider is a sector of a ring, i.e. the region bounded by two concentric
circles and two radii. If YI and ~2 stand for the radii of the concentric circles,
the area of the ring is
2
Yl7r
- Y i l r = T(Y1 + Y2) (Yl - Y Z ) ,
where YI - ~2 = d is the width of the ring (see Figure 29). The number YI+ YS
may be interpreted as double the radius Y of the central circle of the ring
(or double the average of YI and ~ 2 ) . Thus the area of the ring is 27rrd, that
is, the product of the circumference of the central circle and the width of the
ring. The area k of a sector of the ring is to the area of the whole ring as the
SOLUTIONS
89
arc h of the central circle of the sector is to the length of the whole circumference of the central circle; thus we have
t -o h
-=
2 w a 2w'
and hence h = hd.
The truncated cone has slant height d, and a circle of circumferenceh at slant
distance d / 2 from the base; see Figure 29. Designating the radius of this
middle circle by p and the lateral surface by t, we have
t = 29rpd.
(1)
If we take
p = - Pl
2
(which is the radius of the circle about a cone a t distance d/2 from the base),
this expressea once more the area of the cone derived in part a).
Figure 30
c) Let us return to the sphere. Let d be the length of one segment in the
polygonal line approximating a great circle (see Figure 30). Let p be the
distance from the center of this line segment to the diameter of the sphere
and let I be the distance from the same point to the center of the sphere;
90
finally let p be the length of the projection of the line segment don the diameter. Then the area of the surface of revolution of the segment is given by (1).
The right triangles OMA and OMA (see Figure 30) are similar since
corresponding sides are painvise perpendicular. Hence
P
I
- = -d , i.e.
and so
pd = fil;
t = 27rp1.
In words: If we rotate a chord of a circle about a diameter it does not intersect,

the surface generated has area 27r times the distance of the chord to the center of
the circle times the projection of the chord on the diameter. If we inscribe a polygonal line in an arc of a circle, every chord of length d has equal distance from
the center, while the sum of the projections of the chords on the diameter
equals the height m of the surface of revolution. (In case the semicircle is
rotated, m is the length of the diameter.) Thus the area of the surfae generated by the p o l y g o ~ line
l is
2.lrlm.
As the subdivision is refined, I approaches the length of the radius R of the

sphere. Hence the surface area of spherical cap and spherical segment is
25rRm,
where R is the radius of the sphere, m the height of the surface in question.
For the whole sphere, m = 2R, so the sphere has area k R 2 .
Note 2. The geometry of Bolyai. The theorem just proved plays a
considerable role in the geometry of JAnos Bolyai, a system of geometry independent of the parallel axiom. ,Let us glance at the investigations of Bolyai.
a) The @rallel axiom. The history of the problem of parallel lines begins
with Euclids Elements, which takes its departure from the definitions of
some basic concepts, a few broad mathematical axioms, and, what is more
important, five postulates. These basic assumptions form the foundation of
Euclidean geometry. We shall now consider only the fifth postulate: If (in
the plane) two lines are cut by a transversal L so that on one s& of L the i n t e r h
angles add up to less than two right angles, then the two lines intersect on that
side of L. This is called the parallel postulate because it is the basis of the
theory of parallel lines.
Euclid calls two lines parallel if they are in the same plane but do not
intersect, no matter how far they are extended in either direction. He proves,
without using his fifth postulate, that if two lines in the plane are cut by a
transversal L so that the sum of the interior angles is two right angles, then
these lines are parallel (in the sense just stated). It follows that, for any line
I and through any point A not on I , there exists at least one line parallel to 1.
SOLUTIONS
91
Euclid also proves that there is only one parallel to I through A , but this
proof is based on his fifth postulate; in fact, the parallel postulate was designed for this purpose. While the other building blocks of Euclids geometry
are simple and straightforward, the fifth postulate sounds artificial and as if
invented as an expedient. This is why the Greeks already tried to deduce i t
from the other, accepted, basic tenets of geometry.
b) Jdms Bolyai. In the subsequent twenty-two centuries, many a brilliant

mind labored on the problem, none with greater zeal than Farkas Bolyai
(Jbnoss father).
Farkas Bolyai was born in Hungary in 1775 and died in 1856. He studied
at Nagy-Enyed and Kolozsvzlr, became a friend of Baron Simon KemCny,
and then went to the German universitiesof Jena and Gbttingen (1796-1799)
where he befriended Gauss, then a student there. F. Bolyai and Gauss corresponded (except for a long gap from 1816 to 1831) until Gauss death.t Only
J h o s Bolyais work revealed why his father was unable to prove the fifth
postulate.
Jzlnos Bolyai (1802-1860) was very young when he first heard of the importance of deriving the fifth postulate from the other postulates. He tried
to show its dependence on the others by denying the truth of the fifth postulate and then searching for a contradiction. First, he worked in the plane.
Only when he turned his attention to three-dimensional space did it dawn
upon him that no such contradiction need exist-that the fifth postulate
need not be a consequence of the others. Thus he was led to build a geometry
in which the fifth postulate is false while all the others remain true, and found
his system to be free of contradictions. From a logical point of view, this was
all that mattered.
The most beautiful discoveries of J h o s Bolyai date back to 1823, but his
finished paper appeared only in 1832. Bolyais father sent this work of 28
pages to Gauss who, on March 3, sent the saddening reply that he (Gauss)
had already obtained these results, but had withheld them for fear of an
uproar. This and the later news that the Russian mathematician Lobachevskyz had made the same discovery was a great blow to the young scientist. His fathers advice, in 1825, to publish at once, proved to be sound. But
Gauss judgment was also correct; for the work of both Bolyai and Lobachevsky remained unreadable in their time. Yet their bold ideas were vindicated; today, the geometry of Bolyai-Lobachevsky is esteemed as one of the
greatest accomplishmentsof the mind of man.
t The Correspondence of Karl F&d&cL Gauss and Farkas Bolyai, ed. by F.Schmidt
and P. Staeckel, Magyar Tud. Akademia.
$ N. I. Lobachevsky (1793-1856) was professor in Kazan. He published his gee
metrical work in various forms; the shortest one is a booklet published in Berlin in
1840. The Bolyais saw only this one (via Gauss, in 1848).
92
It has since been recognized not only that one may construct still a different geometry by using the other negation of the fifth postulate (i.e. by assuming that there exists no line parallel to a given line t through a point
outside I ) , but that yet other geometries, free of contradictions, can be constructed by dropping other postulates.
Which of the many geometries corresponds to reality? This is not a mathematical problem. The investigation of real space (and time) is an interesting
and knotty problem in the realm of the natural sciences. Whatever the
answer may be, the fact remains that the problem could not even have been
posed as long as the human mind could not think of spaces other than the
Euclidean space.
1902/3. The area T and an angle -y of a triangle are given. Determine
the lengths of the sides a and b so that the side c, opposite the angle 7,
is as short as possible.
First Solution. The law of cosines gives
+ b2 - 2abcosy
= ( a - b)2 + 2ab(l -
8 = a2
COS+~).
We also know that
T = iabsiny,
or
4T
2ab = -.
Sin y'
hence,
8=
(a
- b)2 + 4T 1-Sincos-y y = ( a -
b )2 +4 T t an -.
2
Since T and y are fixed, the second term on the right is a constant. The
fist term is zero when a = b and positive otherwise. Therefore 8,and
hence c, is a minimum when the triangle is isosceles, that is, when a = b.
In this case,
4T
2ab = 2a2 =
sin7 '
and
SOLUTlONS
93
Second Solution. We shd show geometrically that c is shortst

when the triangle is isosceles ( a = b) Let AoBoC be an isosceles triangle
with area T and angle y; see Figure 31. Let A1B& be a second triangle
With the same area, with 4 A1CB1 equal to y, and with side CAI longer
than side CB1. Sice the small triangles AoA1B1 and AoBoBI have equal
areas and the side AoBl in common, they have equal altitudes; that is,
AoAlBoBl is a trapezoid. Moreover, QAoAlBo < +BlBoAI because
+AoAiBo < QCAoBo = +&BOA0 < qBiBoAi.
To show that AoBo < AIB1, it now suflices to prove the following
W.Let g be one of the parallel sides of a non-iSosceles traQeoid.
Then the h g e r diagonal goes through that end Qmnt of g at which the i n w
angk is smaller.
P~OOF:
In the trapezoid PQRS (see Figure 32), let PQ and RS be
parallel and assume that
+SPQ > 4 P Q R .
R
Figure 32
S'
94
Let PQSS be symmetric with respect to the perpendicular bisector

of PQ; then S is on the line through S and R and beyond R, and
SRSP = QQSR. Since QPRS is an exterior angle of ASPR,
SPRS
> QPSR,
and S P S R = %PSQ
+ SKQSR.
Hence
SPRS
> QQSR = QRSP.
Since, in a triangle, the longer side lies opposite the larger angle, it
follows that in APRS
PS
> PR.
Since PS = QS by symmetry, we have proved that
QS
> PR.
This establishes the lemma. When applied to the trapezoid AoAIB&, it

establishes the fact that the side A&o=c belonging to an isosceles
triangle is shortest.
Third Solution. Consider all triangles with a given area T and a
common angle y at the vertex C. If one of these has a shorter base c than
the others, shrinkt all the others until their bases also have lengths c.
The resulting figures will have areas smaller than T. We have transformed our problem into that of finding, among all triangles with a
given base G and a given vertex angle y, the one with the largest area.
The vertices C of all such triangles lie on an arc of a circle with radius
c/2 sin y and chord G (see Note to 1895/2, First Solution). The triangle
with the longest altitude to c will have the greatest area; but the altitude
to G is longest when it is erected at the midpoint of the chord c, so the
solution is an isosceles triangle.
It follows that, among a31 triangles of area T and angle y at C, the side
opposite C is shortest for the isosceles triangle.
1903 Competition
1903/1. Let n = 2P1(2P - l), and let 2 p

1 be a prime number.
Prove that the sum of all positive divisors of n (not including n itself) is
exactly n.
t Shrinking a plane figure means to decrease its size while its shape is retained;
i.e., at each stage of Shrinking,the @re is similar to the original.
SOLUTIONS
Solution. If we set q = 2p
(d. 1901/3, Note 1):
1, 2,
- 1,
22,
4, 24, Z2q,
n = 2p-'q, the divisors of n are
.**)
2p-2, 2p-1,
*.*,
P2q.
We s u m the first geometric progression:

1+ 2
+ 22 +
...+
2P-2
95
2p - 1
= 2p - 1 = q;
2-1
the s u m of the other geometric series is
q(2p-1 - 1) = 2p-lq
- q = n - q.
Therefore, the desired sum is
q+n-q=n.
Note. Perfect numbers. A positive number n is called a e f f e c t number
if the sum of all its positive divisors less than n is exactly n. Perfect numbers
were treated already in Euclid (Elements, IX, no. 36). The smallest perfect
number is
2(22 1) = 6 = 1 2 3.
+ +
Our theorem yields only even perfect numbers. Deeper analysis reveals
that it yields every even perfect number.? The questions:
(a) Are there infinitely many even perfect numbers?
(b) Are there any odd perfect numbers at all?
are still unanswered.
For a given pair of values x and y satisfying x = sin a,

there can be four different values of e = sin (a
a) Set up a relation between x, y and z not involving trigonometric
functions or radicals.
b) Find those pairs of values ( x , y ) for which z = sin (a
takes on
fewer than four distinct values.
+ a).
1903/2.
y = sin
a,
+ a)
Solution. a) If sina = x, sin 6 = y, then

COSCY
= z k ~ G - 2 , cosa =
*dry,
As of April 1962, only 20 of Euclid's perfect numbers have been computed. The
first 18 may be found in Recreational Mathematics,August 1961, pp. 56-59, and the
19th and 20th appear in the April 1962 issue of the same magazine, pp. 29-31.
96
and
= sin (a
+ 8) = sin a cos8 + cos a sin@
may take on the following four values:
x d m + yd-)
= -XdW - y d
21 =
-21
22
=X
=-
-22
-y
xdFfl+
d n )
y d n .
The equation having these values as roots is

(2
or
(2)
(2
where
2:
and
2:
21) ( 2
- 2:) - 2:)
(e2
= 2[x2(12122
= x2(1
21) (2
= z
n ,
- 22) (2 +
- +
(2:
22)
2:)z
= 0,
+ at:
= 0,
r) + yy1 - x2)l = 2(a? - 2239 + u)
- r) - r(1-
9) = x2
- u.
If we substitute these values in (2) we obtain the relation

(3)
s4 - 2(xe - 2x29
y2)2
+ (a? - ry = 0
between x, y and e, and this expression is free of trigonometric functions

and radicals.
SOLUTIONS
97
b) In general, eq. (3) has the four distinct roots given in (1). [The
graph of the sine function in Fig. 33 illustrates how each value of x
between -1 and 1 leads to two angles a between 0 and 'K such that
a! = arcsin x, and similarly for y and @, and how different values of
e = sin (a
@)result, depending on which a! we combine with 0.3
Equations (1) tell us that we obtain fewer than four distinct roots in
the following cases:
1.
2%=
zz, or el = -22, which arise if

y d G z =
or
xdv
= 0,
that is, if y = 0,or 9 = 1, or if x = 0, or y2 = 1; in other words, if either

x or y takes on any one of the values, 0,1, or 1.
2. el =
-21,
or
22
-22,
which arise if
x d m = -yd=
or x d w = yd-,
that is, if 2"( 1 - 9)= y2( I - a?) or, equivalently, if
x2 = ys or
x = fy.
1903/3. Let A, B, C, D be the vertices of a rhombus; let k1 be the

circle through B, C, and D;let kz be the circle through A, C, and D; let
ka be the circle through A, B, and D; and let k4 be the circle through A,
B, and C. Prove that the tangents to k~ and k3 at B form the same angle
as the tangents to kz and k4 at A.
Figure 34
98
Solution. The theorem is true for any quadrilateral ABCD. (Figure

34 shows a convex quadrilateral and Figure 35 shows a non-convex
quadrilateral. The proof we present below holds for all.)
Figure 35
Definition. The angle between two intersecting curves is defined to be

the angle between the tangents to the curves at their point of intersection.
Thus, the theorem to be proved may be phrased: The angle between
b and Ka at B is equal to the angle between kz'and b a t A .
Observe first that the angles between any two intersecting circles are
equal at both intersections. To see this, join the centers by a line; the
whole figure is symmetric with respect to this line and symmetric angles
are equal. It follows that the angles marked by the same letters in Figure
34 are equal. The same is true for Figure 35.
99
SOLUTIONS
Moreover, in each case,

r+x+s=
180'
atA;
s+y+t=
180'
atB;
t+x+u=
180'
atc;
rc+y+r=
180'
atD.
Therefore
(r+x+s)
+ (t+x+u)
- (s+y+t)
or
2(x
- ( u + y + r ) =o,
- y ) = 0.
Hence, x = y, as we set out to prove.
1904 Competition
1904/1. Prove that, if a pentagon (five-sided polygon) inscribed in a
circle has equal angles, then its sides are equal.
Solution. In the pentagon AOAlA2A8AI of Figure 36, .the vertices

Ao, A, A2 and -48, A2, A1 determine congruent triangles because:
1.
AlA, = A d , ,
2. QAd1A2 = Q A d A
by assumption, and
3. &41AoA2= QA2Ad1, because they are inscribed angles

subtending the same arc.
A4
It follows from the congruence of these triangles that
AoAl= AoAs.
Similarly,by using other pairs of congruent triangles, it can be shown
100
that
AoAl=
A2A3
= A&
AiAz =
AJ4.
This theorem is true for polygons with an odd number (2n 1) of sides,
and the same proof applies. For polygons with an even number (2n) of
sides, the theorem is false; the simplest counter example is an inscribed
rectangle.
1904/2. If a is a natural number, show that the number of positive
integral solutions of the indeterminate equation
g 3. 2
( 1)
+ 3x8 + + nx, = a
***
is equal to the number of non-negative integral solutions of
[By a solzction of equation (1) ,we mean a set of numbers f a , xz,

which satisfiesequation (1)
.]
Solution. The set
if the set
*,
xn)
{zl,
a,* * * , zn] satisfies equation (1) if and only
(y1=x1-1,
y2=X2-l1,
yn=xn-1Ij
*.-,
satisfies the equation
+ + + + + +
(3) (y1+ 1) 2(y2 1) 3(y3 1)

***
4% 1) = a,
obtained from (1) by substituting y~ 1 for xi (i = 1, 2, m a - , n ) .
Equation (3) is equivalent to
y1+ 2y2
+ + nyn = a - (1 + 2 + + n)
-**
=a-
+ 1)
- 0 .
4%
2
Moreover, the xs are positive integers if and only if the ys are nonnegative integers.
Thus, for every set of values {XI,s, - - * , xn), there is exactly one set
of values lyl, y2,
yn1. Hence, (1) and (2) have the same number of
solutions of the specified types.
0 ,
1904/3. Let A182 and B1B2 be the diagonals of a rectangle, and let 0
be its center. Find and construct the set of all points P that satisfy
simultaneously the four inequalities
AiP > OP, A2P
> OP,
BiP
> OP,
B . 8 > OP
SOLUTIONS
101
Solution. Let P be any point in the plane and let P be its projection
on the line through A1 and A2; see Fig. 37. Then A1P is greater than,
equal to, less than OP according as A1P is greater than, equal to,less
than OP. Hence the locus of points satisfying A I P > OP is the half
plane containing 0 and bounded by the perpendicular bisector of the
segment A10. Similarly, the locus of points P satisfying A2P > OP is
the half plane containing 0 and bounded by the perpendicular bisector
of A2O; the locus of points P satisfying, respectively, B1P > OP and
BeP > OP consists of the half planes containing 0 and bounded by the
perpendicular bisectors of B10 and B20.
Figure 37
The locus of points satisfying all four inequalities simultaneously is

the common part of all these loci, i.e., the interior of the parallelogram
formed by the perpendicular bisectors of the segments AlO, A20, B10,
B20. This parallelogram is always a rhombus because its diagonals bisect the angles at which AlAz and B1Bz intersect, and hence are
perpendicular.
1905 Competition
1905/1. For given positive integers n and
d i c i e n t conditions for the system of equations
x+py=n,
p , find necessary and
x+y=p*
to have a solution (x, y , 8) of positive integers. Prove also that there is

at most one such solution.
102
Solution. We observe first that the second equation cannot be

satisfied by the positive integers x, y, e unless p > 1, that is p - 1 > 0.
Next, we solve the second equation for y and substitute its value in
the iirst equation, so that
x
+ py = x + p(p" - x ) = x - px + p"+' = x ( l - p ) + p
or, since p
- 1 # 0,
X =
(1)
+I=
n-1
- p e + l - l --
P+l-n
p-1
p-1
p-1'
From the second given equation, we find
pel
y=pc-x=p-
- - p"C1-
P- 1
p"
- pet1 + n
P-1
that is
(2)
n-p.
n-1
8-1
y=-=---*
p-1
p"-1
p-1
since, for every positive integer z,

pd1-
P-1
= p"
+ p-'+
and
p-i - p-'+
--
p-2
P-1
...+ 1
+ ... + 1
are integers, relations (1) and (2) tell us that x and y are integers if
and only if n - 1 is a multiple of p - 1. Moreover, x is positive if and
only if
pel
and y is positive if and only if
> n,
> p;
thus x and y are positive if and only if
p"t' > n > p,

(3)
so that n must be a number between two consecutive powers of p, e
being the value of the smaller exponent. We summarize the necessary
SOLUTIONS
103
and s a c i e n t conditions for the given system to have a solution in positive integers:
(4
P > 1,
- 1 (and hence n 1 p ) ,
(c) n is not an integral power of p.
(b) n - 1is a multiple of p
If these conditions are satisfied, then the inequalities (3) determine z

uniquely and once the value of e is determined, eqs. (1) and (2) determine x and y uniquely. Therefore, if (a), (b), (c) are satisfied, there is
a unique solution (x, y, 2;).
1905/2. Divide the unit square into 9 equal squares by means of two
pairs of lines parallel to the sides; see Figure 38. Now remove the central
square. Treat the remaining 8 squares the same way, and repeat the
process n times.
a) How many squares of side length 1/3" remain?
b) What is the sum of the areas of the removed squares as n becomes
infinite?
Figure 38
Solution. a) After the first step 8 squares are left, each with side
length 3. After two steps
8.9-8=8(9-
1) ' 8 '
squares are left, each with side length 1/32. After three steps
82 9
- 82= 8'(9 - 1) = 8'
104
squares are left, each with side length 1/38. After n steps, 8" squares are
left and each has side length 1/3".
b) The s u m of the areas of the squares that are left is
therefore, the s u m of the areas of the removed squares is

1-
8'.
As tz increases, (8/9)" becomes smaller and smaller and approaches 0

since 8/9 is less than 1. Therefore, the sum of the areas of the removed
squares approaches 1 as n approaches iniinity.
1905/3. Let C1 be any point on side AB of a triangle ABC (see Figure
39), and draw CIC. Let A1 be the intersection of BC extended and the
line through A parallel to CCl;similarly let B1 be the intersection of AC

extended and the line through B parallel to CC1. Prove that
1
1
+-=-
AAi
BBi
CCi'
Solution. Since AA1, BB1, CC1 are parallel, ACAC1
ACBC1
AA1BA. It follows that
A&AB and
Figure 39
SOLUTIGYS
Adding these relations, we find that

BIB
ACi CiB - AB
-=
AB
AB
CCi CC1
-+-=
A1A
105
1,
and dividing by CC1, we obtain the desired resultt
t If a and b are two positive numbers, then the number c, defined by

(i.e. the number whose reciprocal is the average of the reciprocals of the given numbers), is called the harmonic mean of a and b.
The theorem just proved shows that, if a and b are lengthsof parallel sides of a trapezoid, then their harmonic mean is the length of the segment through the intersection
of the diagonals and parallel to the parallel sides.
1
+-=-BBI
CC;
Classification? of Problems
1894/1,
1899/3,
1901/3,
Number theory:
1896/1,
1900/1,
1902/1,
Combiiatorics:
1895/1.
Quadratic equations:
1896/2,
Quadratic functions:
1902/1.
Equations of higher degree:
1903/2.
Diophanthe equations:
1894/1,
1904/2,
Number sequences:
1905/1,
1905/2.
InequaEties:
1896/1.
1898/1,
-
1901/1,
1903/1.
1899/2.
1905/1.
Plane geometry (without trigonometry)
Proofs:
1903/3,
Computations:
1894/3.
Loci:
1904/1,
1895/2,
1904/3.
comtructions:
1894/2,
1897/3,
1895/2,
1898/3,
Extreme values:
1902/3.
Inequalities:
1897/2,
1905/3.
1896/3,
1900/2.
1904/3.
t Since the problems and specially their solutions ofteninvolve several mathematid
disciplines, this dasdiation is necessrtrily arbitrary and somewhat incomplete.
107
HUNGARIAN P R O B L E M BOOK
108
Angle measurement:
1899/1,
1901/2,
1903/2.
Trigonometry:
1894/3,
1897/1,
1899/1.
1895/3,
1897/2,
1896/3,
1898/2,
Solid geometry:
1902/2.
From physics:
1900/3.
List of Explanatory Notes

Divisibility. Classification of the integers
An important property of prime numbers
Permutations with repetitions.
Combinations
The locus of points P at which a given line segment XY subtends a
given angle
Brocard points
The law of tangents
Factorization into a product of powers of primes
On the orthic triangle of an obtuse triangle
On circles tangent to the sides of a triangle
Connection between the angles and the sides of a triangle
Relation between sums and products of trigonometric functions
The meaning of some trigonometric products
The inscribed and circumscribed circles of a given triangle
The notion of congruence
The existence of a maximum
Regular polygons
Chebyshev polynomials
A geometric application of complex numbers
On reducing irrational equations
Fermats theorem
Algebraic and transcendental numbers. Algebraic integers
The divisors of a positive integer
The greatest common divisor
Relatively prime numbers
Polynomials that take on integral values for integral values of the
variable
The binomial theorem and the binomial series
On the surface area of the spherical cap, spherical segment, and
sphere
The geometry of Bolyai
Perfect numbers
109
19
20
24
26
30
33
34
37
43
43
45
46
49
50
57
59
63
65
66
68
75
77
79
79
80
82
84
87
90
95
List of Biographical Notes

Euclid
Gauss, K. F.
Brocard
Euler, L.
Poncelet, J. V.
Chebyshev, P. L.
Fermat, P.
Leibniz, G. W.
Hermite
Lindemann, F.
Newton, I.
Bolyai, J.
Bolyai, F.
Lobachevsky, N. I.
19
26
33
51
53
66
75
76
78
78
86
91
91
91
110
List of Winners
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
P a Pap
Mih&ly Seidner
Emil Xesz
AladAr Visnya
BemAt Fazekas
Tivadar K h m h
Od6n Kornis
Ir6neusz Juvantz
Gyula P6ka
D6nes K6nig
Alfred Haar
Marcel Xesz
Gyula Ujj
IgnAc Pilczer
Gy6G Zempl6n
Lip& Fej&
GAbor Gr6ffits
Odzjn Spiczer
K&m& Szmodics
Em6 Barany6
Hildeghd Szmodics
B6la Horvay
Istv6n Fuchs
Constantin Neubauer
111

Hungarian Problem Book I (Number 11) (Bk. 1) (E. Rapaport)

Uploaded by

Copyright:

Available Formats

Hungarian Problem Book I (Number 11) (Bk. 1) (E. Rapaport)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Hungarian Problem Book I (Number 11) (Bk. 1) (E. Rapaport)

Uploaded by

Copyright:

Available Formats

HUNGARIAN PROBLEM BOOK I

Jzsef Krschk (18641933) was born and educated