This action might not be possible to undo. Are you sure you want to continue?
A Friendly Mathematics
Competition
35 Years of Teamwork in Indiana
© 2003 by
The Mathematical Association of America (Incorporated)
Library of Congress Catalog Card Number 2002107971
ISBN 0883858088
Printed in the United States of America
Current Printing (last digit):
109 8 7 6 5 4 3 2 1
A Friendly Mathematics
Competition
35 Years of Teamwork in Indiana
Edited by
Rick Gillman
Valparaiso University
Published and Distributed by
The Mathematical Association of America
MAA PROBLEM BOOKS SERIES
Problem Books is a series of the Mathematical Association of America
consisting of collections of problems and solutions from annual mathemat
ical competitions; compilations of problems (including unsolved problems)
specific to particular branches of mathematics; books on the art and practice
of problem solving, etc.
Committee on Publications
Gerald Alexanderson, Chair
Roger Nelsen Editor
Irl Bivens Clayton Dodge
Richard Gibbs George Gilbert
Art Grainger Gerald Heuer
Elgin Johnston Kiran Kedlaya
Loren Larson Margaret Robinson
A Friendly Mathematics Competition: 35 Years of Teamwork in Indiana,
edited by Rick Gillman
The Inquisitive Problem Solver, Paul Vaderlind, Richard K. Guy, and Loren
L. Larson
Mathematical Olympiads 19981999: Problems and Solutions From Around
the World, edited by Titu Andreescu and Zuming Feng
Mathematical Olympiads 19992000: Problems and Solutions From Around
the World, edited by Titu Andreescu and Zuming Feng
The William Lowell Putnam Mathematical Competition 19852000: Prob
lems, Solutions, and Commentary, Kiran S. Kedlaya, Bjorn Poonen,
Ravi Vakil
USA and International Mathematical Olympiads 2000, edited by Titu
Andreescu and Zuming Feng
USA and International Mathematical Olympiads 2001, edited by Titu
Andreescu and Zuming Feng
MAA Service Center
P. O. Box 91112
Washington, DC 200901112
18003311622 fax: 13012069789
www.maa.org
Preface
A college level mathematics competition has been held in Indiana for
more than 35 years as of the writing of this book. Orginally known as the
Indiana Small College Mathematics Competition, its official title became
the Indiana College Mathematics Competition. However, many people refer
to it by its popular nickname "The Friendly Exam" earned because of the
collegial nature of the competition and the modest level of competitiveness
between the participants and the institutions involved. It is really more
about getting together to do mathematics and to challenge oneself than it
is about winning a competition.
This book shares the efforts of the many students and faculty who have
participated in the Indiana College Mathematics Competition over its first
35 years.
I would like to specifically identify the individual faculty members who
have composed the exams. Unfortunately, this is the one piece of the history
that has been lost over the years. Their attentive work at identifying or
creating good questions has been a major reason why the competition has
been a success for so long.
Over the years, exam authors have borrowed many classic problems from
other sources. These have been included to provide a complete record of
the exam, but have not generally been attributed to an original source.
I'd like to take this opportunity to thank all of the people who have
made this book possible by helping in one way or another. Among all of
these people, the following individuals provided noteworthy help: James
Lee, Ken Luther, Steve Shonefeld, Pat Sullivan, Eric Tkaczyk, and the
Purdue University Math Club.
I would also like to thank Pete Edson for having the idea for the
competition, and extend a very special thank you to Paul Mielke on
vi Preface
behalf of myself and my colleagues in Indiana. Without Paul's efforts
to develop the contest, to support it over the years, and to lead the section
in building the endowment Indiana and the MAA would not have this
wonderful example of a "friendly" competition that does so much to build
the mathematical community in Indiana.
Contents
Preface v
The Indiana College Mathematics Competition: A Short History 1
An Update of the History of the ICMC 5
Exams 7
Exam # 11966 7
Exam #21967 8
Exam #31968 9
Exam #41969 10
Exam #51970 II
Exam #61971 12
Exam #71972 14
Exam #81973 15
Exam #91974 16
Exam #101975 . 16
Exam #111976 . 17
Exam #121977 . 18
Exam #l31978 . 19
Exam #141979 . 20
Exam #151980 . 22
Exam #161981 . 23
Exam #171982 . 24
Exwn #181983 . 25
Exam #191984 . 26
Exam #201985 . 28
vii
viii Contents
Exam #211986 . 29
Exam #221987 . 30
Exam #231988 . 31
Exam #241989 . 32
Exam #251990 . 34
Exam #261991 . 36
Exam #271992 . 37
Exam #281993 . 38
Exam #291994 . 40
Exam #301995 . 40
Exam #311996 . 41
Exam #321997 . 43
Exam #331998 . 44
Exam #341999 . 46
Exam #352000 . 48
Solutions 51
Exam #11966 51
Exam #21967 55
Exam #31968 61
Exam #41969 64
Exam #51970 67
Exam #61971 69
Exam #71972 73
Exam #81973 75
Exam #91974 78
Exam #101975 . 81
Exam #111976 . 86
Exam #121977 . 88
Exam #131978 . 90
Exam #141979 . 95
Exam #151980 . 98
Exam #161981 . 103
Exam #171982 . 106
Exam #181983 . 109
Exam #191984 . 112
Exam #201985 . 117
Exam #211986 . 120
Exam #221987 .. 122
Contents ix
Exam #231988 . 126
Exam #241989 . 130
Exam #251990 . 133
Exam #261991 . 136
Exam #271992 . 139
Exam #281993 . 144
Exam #291994 . 148
Exam #301995 . 150
Exam #311996 . 156
Exam #321997 . 160
Exam #331998 . 166
Exam #341999 . 171
Exam #352000 . 176
Index by Problem Type 181
This page intentionally is no longer blank
The Indiana College Mathematics
Competition: A Short History
Paul T. Mielke
On April 27, 1965, Peter Edson, a trustee of Wabash College, sent our
dean a memorandum that included a newspaper clipping about a unique
mathematics competition that was held among high schools in New Jersey.
In this competition, each school entered a team that worked as a team on a
set mathematics examination. Edson wondered if anything of that sort was
done at the college level. I answered that I knew of no such competition
but that I would be willing to suggest the idea to my Indiana colleagues.
Thus, on May 18, 1965, I wrote a letter to heads of departments in the
small colleges of Indiana relaying Edson's suggestion. Response to my
letter was immediate and favorable, so I wrote a second letter, which read
in part:
The Mathematics Department of Wabash College is pleased to invite
its counterparts in the other colleges of Indiana to a friendly mathe
matics competition to be held on Saturday afternoon, March 12, 1966,
at Wabash College, Crawfordsville. The purpose of this competition
is to stimulate friendship and mathematical interest among members
of the various departments and their undergraduate students. Program
for the day, which we are calling "Pete Edson Day" in honor of the
Wabash College trustee who conceived the idea for the contest, is as
follows:
11 :30 A.M.  12:30 P.M. Registration, Baxter Hall Lounge,
Wabash College
12:30 P.M.  Luncheon, South Mezzanine, Campus Center.
Participants will be guests of the College.
2:00 P.M.  4:00 P.M.  The competition, a written examination
for undergraduates
2 The Indiana College Mathematics Competition: A Short History
2:00 P.M.  4:00 P.M.  Informal seminar for members of the
various departments
4:00 P.M.  Coffee hour at which test papers will be gathered,
Baxter Lounge
The test, prepared by Professor George F. Springer of Indiana
University, will be similar to, though (one hopes) not as difficult as,
the Putnam test. There will be a significant difference, however. Each
school is invited to bring a team of three undergraduates who will
work on the test as a team, consulting with one another as needed.
Tests will be sent to Professor Springer for grading and announcement
of results. A suitable prize will be awarded by Wabash College to
members of the winning team. Ranking of other teams will not be
revealed by Professor Springer.
This first contest was held as scheduled, with eight schools partici
pating: Anderson College, DePauw University, Earlham College, Franklin
College, Marian College, Rose Polytechnic Institute (now RoseHulman),
Valparaiso University, and Wabash College. In the spring of 1967 the
second competition was held at Marian College with 15 participants:
Anderson, Butler, DePauw, Earlham, Evansville, Franklin, Manchester,
Marian, RoseHulman, St. Joseph's, St. MaryoftheWoods, St. Mary's
(of Notre Dame), Taylor, Valparaiso, and Wabash.
Since its beginning, the contest has been called familiarly, "The Friendly
Math Competition", from the phrase used in the above letter. It has
been held every year, a total of 24 times as of 1989, with a minimum
participation of eight schools in that first year, a maximum participation of
nineteen in 1972, and an average participation of over thirteen schools. The
nineteen participants in the 1972 competition were Butler, DePauw, Earl
ham, Evansville, Franklin, Huntington, IVPUI Indianapolis, IV Northwest,
IV Southeast, Manchester, Marian, Oakland City, Purdue Calumet, Purdue
Ft. Wayne, RoseHulman, St. MaryoftheWoods, Taylor, Valparaiso, and
Wabash.
Throughout its history, the essential nature of the contest as a team
competition has remained the same, with a team consisting of at most
three members. The tests have usually been set and graded by a single
person, most often a professor from one of the large state universities.
Duration of the tests has been maintained at two hours. Before each test,
a team member from each school draws a number which is given to an
impartial judge, who attaches the number to the school's name in a memo.
The Indiana College Mathematics Competition: A Short History 3
Teams are instructed to identify their test papers with only their numbers,
so that only these are seen by the person who sets and grades the test.
Upon completing hislher work, the grader announces the results by team
number to the judge who holds the identification key. This person in turn
reports the results to the contestants.
Some other ground rules developed naturally. There was early agitation
for publicly announcing the ranking of the first three teams. This idea was
accepted, but it led to another problem. A school may enter any number
of contestants in the Putnam competition, but it must name its team of
three for the competition beforehand. RoseHulman has always been a
worthy adversary in this Indiana competition, and a decision had to be made
with regard to its wish to enter several teams. A consensus was reached
that mUltiple teams of three could be entered, and that a school need not
designate "its team." It was then decided that public announcement of
the top three places in the competition would refer to schools rather than
teams, so that if two teams from a given school scored second and third,
for instance, the school would be awarded second place in the competition,
and third place would be awarded to the next highest school.
Until 1987 the competition was restricted to the small colleges of
Indiana, including branches of IU and Purdue, but in 1987 it was opened
to all colleges, including the large universities.
The contests have always been managed by the host schools, though in its
fall meeting of 1977, the Indiana Section of the Mathematical Association
of America agreed to make them a part of its spring meeting, and this
practice has been followed since 1978. Maintaining the "Edson Trophy"
and awarding of prizes was assumed by Wabash College, which had a
small fund for that purpose, to which Mr. Edson contributed while he was
alive. Since this fund was exhausted, I have had the pleasure of supplying
both trophy and prizes.
Each member of the winning team is awarded a suitably inscribed book
prize, and the winners' names are inscribed on the Edson Trophy. The
books are traditionally selected from those offered by the MAA. The Edson
Trophy is a rectangular walnut plaque measuring 9 x 12 inches. To it is
affixed a permanent identifying plate with this inscription:
THE
PETER EDSON TROPHY
INDIANA COLLEGE
MATHEMATICS COMPETITION
4 An Update of the History of the ICMC
Below this plate is a smaller one containing the year and the name of the
most recent winning college and its team members. Each year, this smaller
plate is moved to the back of the plaque, and a new one replaces it. The
trophy "travels"; it resides each year at the winning college. The back of
the first trophy was filled by 1978, so it resides permanently at Wabash
College, which was the winner in that year. The current trophy is at Purdue
University, the 1989 winner. (Since writing the above, I have succeeded in
making individual trophies for each of the past winners, affixing all the old
winners' plates to them. The new trophies are made of sassafras, one of
Indiana's most beautiful hardwoods. The two old trophies have been retired
to Wabash College, where they record its six wins. Three special walnut
trophies have been made to hold the 12 winning plates for RoseHulman.
Henceforth, a nontravelling trophy will be awarded to each winner.)
It should finally be noted that the task here has been to write exami
nations that will challenge and separate the competitors, yet present some
problems that all can solve, so that there will be enjoyment for everyone
involved. As one might guess, the actual examinations varied in their
success according to these criteria.
An Update of the
History of the .CMC
I was introduced to the Indiana College Mathematics Competition
(ICMC) during the spring of 1987, during my first year on the faculty at
Valparaiso University. This was, as you read in Paul's comments, the first
year that the competition was opened up to the large state universities and
explains why I had not known of the competition while an undergraduate
student at Ball State University.
I have been an enthusiastic supporter of the competition ever since.
While the nickname "friendly" has not been used very much in recent
years, the competition still maintains the same friendly approach that it
had when it began. It is always very exciting and refreshing to be around
the students from the various schools as they wait for the exam to begin,
and to hear their conversations with each other and with their faculty
advisors after the exam. In many years, a lot of friendly banter and good
mathematical conversations occur among my VU students during the drive
back to campus from the meeting.
In 1992, I became the Indiana section's first Student Chapter Coordinator
and one of the responsibilities was to manage the ICMC. The first
competition that I organized was at St. Mary's College in Notre Dame,
Indiana, at a joint meeting of the Indiana, Michigan, and Illinois sections.
Thirty teams from the three states participated in the competition that year.
Subsequent Student Chapter Coordinators have helped implement changes
to the ICMC that have made it the central part of the section's efforts to
reach undergraduate students.
Since 1990, the average number of teams participating in the ICMC
has been 25, generally from twelve or thirteen colleges or universities.
Occasionally, a team from the Indiana Mathematics and Science Academy
also participates. The largest competition was at Ball State University in
5
6
An Update of the History of the ICMC
1998. That year, 29 teams from 22 institutions participated. Somewhat
surprisingly, the large state universities have not dominated the competition
since their inclusion, in either the number of teams participating or the
number of competitions won.
The most significant change in the competition came in the form of a
decision to move the competition from Saturday afternoon (on the second
day of a day and a half meeting) to Friday afternoon as a lead off event for
the meeting. Students can then stay for a banquet and after dinner speaker.
Rather than having a single person grade the exams, a team of faculty
graders do this work overnight. On Saturday, there is a session where
solutions to the exam are discussed and the winners of the competition are
announced late in the day during the section's business meeting.
The institution hosting the ICMC and the section meeting is invited to
have one of its faculty members write the exam and solutions. If they
decline this opportunity, the section finds an external source to do this
work. Each team competing pays a $5.00 registration fee. Originally this
fee was paid to the exam writer for composing and grading the exam. Now
these fees are paid to the writer for writing the exam and for leading the
problem solving session on Saturday.
This format has increased the number of students participating in the
mathematics sessions of the meeting, both as presenters and as members
of the audience. It has also established a new structure for the section
meeting. Sessions on professional and pedagogical issues are now held
on Friday afternoons during the competition itself, so that mathematical
presentations are the central focus when students are part of the audience.
The afterdinner speaker on Friday night is selected with the knowledge
that at least half of the audience of 100120 individuals are students.
The section has also been able to experiment with late night workshops
for students on Friday evenings, knowing that students will be staying
overnight for the Saturday portion of the meeting.
By 1996, the Indiana section had established a small endowment to
support the ICMC. This endowment will enable the section to continue to
hold the exam each year and present prizes to the winning team members
and the winning institution in the form of books and the Peter Edson
Trophy, respectively. Proceeds from the sale of this volume will go the
Indiana section to support undergraduate student activities in the section.
Rick Gillman
Spring, 2002
Exams
Exam #11966
As stated in the introduction, the first "friendly competition" was held at
Wabash College, located in Crawfordsville, a bit northwest of Indianapolis.
Eight schools participated in the competition that year. It was won by the
team from Wabash College consisting of James Clynch, Albert Hart, Jr.,
and Larry Haugh.
P19661. Show that the equation x
2
 y2 = a
3
always has integer
solutions for x and y whenever a is a positive integer.
P19662. Consider any five points P
l
, P
2
, P
3
, P
4
, Ps in the interior of
a square of side length 1 (one). Denote by d
ij
the distance between points
Pi and Pj . Prove that at least one of the distances d
ij
is less than V2/2.
P19663. Let p be a prime number and let al, a2, ... ,a
p
be integers not
necessarily arranged in consecutive order and with possible repetitions.
Establish the existence of integers m and n such that 1 ::::: m ::::: n ::::: p and
n
such that L aj is divisible by p.
j=m
P19664. Two functions of x are differentiable and not identically equal
to zero. Find an example of two such functions having the property that
the derivative of their quotient is the quotient of their derivatives.
P1966S. For two given positive integers n and k, how many different
sequences of positive integers al ::::: a2 ::::: a3 ::::: ... ::::: ak are there in
which ak ::::: n?
P19666. A sequence {x
n
} is defined by the following rule: Xn+l =
vi ax; + b with Xl = c. Show that this sequence converges whenever
7
s Exams
o < a < 1 and b > 0 regardless of the value of the real number c, and
determine the limit of the sequence.
P19667. By an interval we shall mean a set of points x on the real line
satisfYing a :s: x :s: b for a pair of real numbers a and b with a < b.
Suppose that we are given a collection of intervals h, h, ... , In which
cover an interval I; that is I ~ U ~ = l I
k
. Prove that we can select mutually
disjoint intervals from this collection which cover at least half of I.
P1966S. Let us assume that a given pair of people either know each
other or are strangers. If six people enter a room, show that there must be
either three people who know each other pairwise or three people who are
pairwise strangers.
Exam #21967
This competition was held at Marian College in Indianapolis. The winning
team, consisting of David Hafting, Albert Hart Jr., and Robert Spear, was
again from Wabash College.
P19671. A = {aij} is a symmetric (i.e., aij = aji) n x n matrix with n
odd, and each row of the matrix is a permutation of the integers 1,2, ... ,n.
Prove that the main diagonal is also a permutation of 1, 2, ... ,n.
P19672. Two parabolas have parallel axes. Prove that their common
chord bisects their common tangent.
P19673. Show that for all a ~ 0 and b ~ 1, ab :s: e
a
+ b(ln b  1).
P19674. For each positive integer n the binomial coefficients ( ~ ) , 0 :s:
r :s: n, are integers, some odd, some even. Show that for each n the number
of odd binomial coefficients is a power of 2.
P19675. Show that if lanl < 2 for 1 :s: n :s: N, then the equation
1 + alz + a2z2 + ... + aNz
N
= 0 has no root z such that Izl < 1/3.
P19676. Prove that if the set S of points in or on the boundary of
the unit square is partitioned into three disjoint sets A, B, and C, i.e.,
S = A U B U C and A n B = A n C = B n C = 0, then the least upper
bound of the diameters of A, B, and C is greater than or equal to V65 /8.
The diameter of a set is the least upper bound of the distances between
two points of the set.
Exam #31968 9
P19677. Given a > 0 and Xo > 0, show that there exists one and only
one sequence of positive numbers {Xo, Xl, X2, ... } such that
00
Xn= L xj,
j=n+l
for n = 0, 1,2, ....
P19678. Let T be a mapping of the Euclidean plane into itself which
preserves all rational distances. Prove that T preserves all distances.
Exam #31968
Held at Franklin College, located just south of Indianapolis, this competi
tion was won by a team from Earlham College. The team members were
William Roha, Thorn Sulanke, and William Wilson. It is unique because
it is the only competition that came with a warning.
WARNING: The statements below should be viewed as conjectures. At
least one cannot be done.
P19681. Let f be a realvalued function defined on the closed interval
[a, b]. Show that if the set of Riemann sums for f is bounded, then f is
bounded. By Riemann sum we mean a sum of the form
n
L f(ti)(Xi  xid
i=l
where a = Xo < Xl < ... < Xn = band Xil :s: ti :s: Xi for 1 :s: i :s: n.
P19682. Given four points which are the vertices of a convex quadrilat
eral in the plane and five points inside the quadrilateral such that no three
of the nine points are collinear, show that five of the nine points are the
vertices of a convex pentagon.
P19683. Let f : R
n
H R
n
be a differentiable function such that
f(tx) = tf(x) for X E Rn and t > O. Show that f is linear.
P19684. Find all integral solutions of the equation 2
X
 3
Y
= 1 or of
the equation 3
x
 2
Y
= 1.
P1968S. Let Un} be a sequence of realvalued functions defined on R.
Suppose that for each n, {xlfn(x) i= O} is bounded and that the sequence
10
converges unifonnly on R to the zero function. Show that
lim /00 in = O.
n+oo (X)
Exams
P196S6. Find two decreasing sequences {an} and {b
n
} of positive
numbers such that
00 00 00
Lan = 00 and Lbn = 00, but LCn < 00
n=l n=l n=l
where C
n
= min{ an, b
n
}.
P196S7. Let Zl, Z2, ... , Zn be complex numbers such that
Show that for some iI, i
2
, ... ,ib we have
1
Iz + z· + ... + z· I > 
1.1 t2 tk  7r.
P196SS. Let nand k be positive integers. Suppose line segments are
drawn joining each pair of n points and that each segment is painted blue
or green. Are there k points such that all the line segments with end points
among these k points are of the same color? Show that the answer is yes
if n is large enough. Can you guess how large n must be?
Exam #41969
This competition was held at RoseHulman Institute of Technology, located
in Terre Haute. The winning team from Valparaiso University consisted of
Gerald Anderson, Charles Frank, and Charles Spear.
P19691. Prove that sin x :::: x  (x
2
/ 7f) if 0 ::; x ::; 7f.
P19692. Suppose p, q, and r are positive integers no two of which have
a common factor larger than 1. Suppose P, Q, and R are positive integers
such that f + ~ + !f is an integer. Prove that each of f, ~ , and !f is an
integer.
P19693. Determine whether {i25 + 5V20 + {i25  5V20 is rational
or irrational.
P19694. A ball is thrown into the air. The only forces acting are gravity
(constant) and air resistance (proportional to the velocity). Which takes it
longer, to go up, or to come down?
Exam #51970 II
P1969S. An equilateral triangle is circumscribed about an arbitrary
triangle as shown in the figure below. Show without using calculus that
the maximum area it can have is
(b
2
+ c
2
 2bc cos ( A + i) ) .
I'
I '
I '
I
I ,
I ,
I
I ,
I ,
I ,
tJ!7\
P19696. Assume that f has a continuous second derivative, that a < b,
that f(a) = feb) = 0, and that 1f"(x)1 :::; M on a:::; x :::; b. Prove that
lib f(x)dxl:::;
Exam #51970
Held at Butler University, in Indianapolis, for the first of many occasions,
this competition was again won by a team from Valparaiso University. The
team consisted of Gerald Anderson, Jonathon Berke, and George Lueker.
A team from RoseHulman came in second.
P19701. Evaluate
1m ++ ... + .
1
. (1 1 1 )
n+oo n + 1 n + 2 2n
P19702. Let f(x) = xn + an_IX
n

1
+ ... + aIX + aD be a polynomial
with real coefficients. Prove that any root c satisfies
Icl :::; 1 + lanII + lan 2\ + ... + laol·
(Hint: consider Icl :::; 1 and Icl 2:: 1 separately.)
P19703. Squares ABEF, BCCH, CDJK, and DALM are placed
externally on the sides of a parallelogram ABCD, with X, Y, Z, and W
ilie respective centers of those squares. Prove that the lengili from X to
Z is the same as the length from Y to W, that the line segment from X
12 Exams
to Z is perpendicular to the line segment from Y to W, and finally that
XY ZW is a square.
P19704. On the border of a disk select an even number, e, of points.
Draw e/2 nonoverlapping curves in the disk whose ends are the e dots.
For instance, in the case e = 10 we may have the figure below. The curves
cut the disk into e/2 + 1 regions. Prove that the regions can be colored
with two colors in a way such that adjacent regions are colored differently.
P19705. Find the smallest natural number n that has the following
properties:
(a) Its base ten representation has a 6 as the last digit.
(b) If the last digit, 6, is erased and placed as the first digit in front of
the remaining digits, then the resulting number is four times as large
as the original number n.
Exam #61971
This competition was held at Earlham College, in Richmond. First place
was won by a team from RoseHulman Institute of Technologythe first
of their many winsconsisting of Tom Dehne, Bill Lipp, and Robert Klim.
There was a tie for second place between teams from Earlham College and
Goshen College.
There was a glossary of terms at the beginning of this exam, but this
has been lost over time.
P19711. Let S be a set and let P be an equivalence relation on S. That
is, for all x, y, z in S, xPx; xPy =} yPx; and xPy 1\ ypz =} xPz. For
each subset A of S, define A = {x E S::::I Yt A such that yPx}.
Exam #61971
(a) Prove that A c 5 and B c 5 implies
1. A c A,
11. A = A, and
Ill. Au B = Au B.
(b) Let 5 be the set of points in the Cartesian plane. Define
R: (Xl, Yl)R(X2, Y2) iff Yl  Y2 = 3(Xl  X2).
13
Granted that R is an equivalence relation on 5, describe or sketch A,
where A is the unit circle with center at (0,0). That is A = {(x, y) E
5 : X2 + y2 = I}.
P19712. Granted that the last three digits in the decimal representation
of 7
400
are 0, 0, 1, so that 7
400
= 10
3
k + 1 for some positive integer k,
find the last three digits in the decimal representation of 7
9999
.
P19713. Compute the derivative of I_el,/X' Then compute
1
1 I e
1
/
x
I
1 x2(1 e1/X)2 dx.
P19714. Let A be a countable subset of a Euclidean plane. Show that
there exists a line in the plane which contains no point of A.
P19715. For any real number x, let lxJ denote the greatest integer in x.
Thus, if n is the (unique) integer such that n :s: X < n + 1, then l X J = n.
Let X  l X J be called the fractional part of x.
(a) What is the limit of the fractional part of (2 + /2)k as k 7 oo?
(b) What is the behavior of the fractional part of (1 + .J2)k as k 7 oo?
P19716. A set of eleven bank robbers decided to store their loot from
a successful job in a safe (where else?) until the heat was off. Being
somewhat but not completely trustful of one another, they decided to be
able to open the safe when and only when a majority of the robbers was
present. Therefore, the safe was provided with a number of different locks
(the key to anyone not being able to open any other), and each robber was
given keys to some of the locks. How many locks were required, and how
many keys had to be given to anyone bank robber?
P19717. Is the following a Boolean algebra? Explain.
Given a nonempty set 5 together with a binary operation n and a unary
operation *. Suppose that
14
(a) for all A, B in S, AnB = BnA; and
(b) for all A, B, C in S, (A n B) n C = An (B n C); and
(c) for all A, B, C in S, An *B = C n *C {:} A n B = A.
Exam #71972
Exams
This year's competition was held at DePauw University, located in Green
castle, just west of Indianapolis. The winning team was from Purdue
UniversityCalumet, and it consisted of Reinhard Fritz, David Hasza, and
Lawrence Kus.
P19721. Suppose that A, B, C, and D are square matrices. A and Bare
not invertible. What conditions on A and B are necessary for the existence
of matrices X and Y such that AX + BY = C and BX + AY = D?
P19722. If A is a square matrix such that A
3
+ 4A2 + 3A + 2I = 0,
show that A is invertible (I denotes the identity matrix and 0 the matrix
all of whose entries are zero.)
P19723. Which is bigger: e
7r
or 7r
e
?
P19724. If I : ~ = l an converges, where an > 0 and an I 1 for all n,
does
converge? (Prove or give a counterexample.)
P19725. A cardshuffling machine always rearranges cards in the same
way relative to the order in which they were given to it. All of the hearts
arranged in order from ace to king were put into the machine, and then
the shuffled cards were put into the machine again to be shuffled again. If
the cards emerged in the order 10, 9, Q, 8, K, 3, 4, A, 5, J, 6, 2, 7, what
order were the cards in after the first shuffle?
P19726. Find the points of intersection of the curves whose equations
in polar form are
r = cos(O/2) and r = sin(O/2).
P19727. Can a group be a union of two proper subgroups?
Exam #81973 15
Exam #81973
This competition was held at St. Mary'softheWoods College, in Terre
Haute. The team consisting of Tom Seilke, Tom Stocks, and John VanDrie
from Wabash College won, with teams from Earlham College and Rose
Hulman taking second and third places respectively.
P19731. Evaluate lim x(e
1
/
X
 1).
x+oo
P19732. A rectangle is inscribed in a sector of a circle of radius 1 as
shown in the figure below. The central angle of the sector is a given angle
0, with 0 < 0 :s; 1f /2. Show that the maximum possible area for the
rectangle is
1  cos 0
2sinO'
P19733. Let n be a fixed positive integer greater than 1. Determine the
maximum value of IT:=1 ni, where ni are positive integers whose sum is
n. Note that IT is the product symbol and k is not fixed but can assume
any value from 1 to n.
P19734. Evaluate
1
1 nynl
lim dy.
n+oo 0 1 + y
P19735. Let A and B be ideals of a commutative ring R. The quotient,
A: B, of A by B is defmed as A : B = {r E R: rb E A for all bE B}.
Is A : B an ideal of R? Prove or disprove.
P19736. If n is a positive integer, show that an+!  n(a  1)  a is
divisible by (a  1)2.
P19737. Let A and B be square matrices. Prove that if AB IS non
singular, then both A and B are nonsingular.
16 Exams
P19738. Find all solutions in integers of x
2
+ y2 = Z2 with X, y, and z
in arithmetic progression.
P19739. A group of 5 men contains 3 Democrats and 2 Republicans.
Three men are chosen at random from the group. What is the probability
that both of the Republicans were selected?
Exam #91974
This competition was held at Butler University for the second time, where
the team of Stanley Lyness, Tom Sielke, and David Wilde from Wabash
College won the competition.
P19741. Let
P(x) = (1 + x) 1000 + x(l + x)999 + x
2
(1 + x)998 + ... + xlOOO.
(a) Find the coefficient of x
50
in P(x).
(b) What is the sum of all the coefficients of this polynomial?
P19742. Let S be a set with an associative multiplication, (x, y) t xy.
Suppose that for all x, y in S we have x
3
= x and x
2
y = yx
2
. Show that
the multiplication is commutative.
P19743. For each positive integer m, find two distinct pairs of positive
integers (nl, N
I
) and (n2, N2)(depending upon m) such that
(m
2
+ 1)( n; + 1) = Nl + 1 (i=1,2).
P19744. Determine whether at (0,0) f achieves a local maximum,
minimum, or neither, for the function f (x, y) = (y2  x) (2y2  x).
P19745. Suppose that y is a continuously differentiable function of x
which satisfies the condition y(O) = 1 and the inequality ~ +eXy+ 1 :s: o.
Show that y has a zero in the interval [0,3/4J.
P19746. Examine the validity of the following conjecture: The series
of positive terms I : ~ = o an diverges if and only if the series I : ~ = o a ~
diverges.
Exam #101975
A team from RoseRulman won this competition, which was held at
Wabash College. The winning team consisted of Robert E. Copus, Michael
J. Dominik, and Barry W. Carlin.
Exam #111976 17
P19751. Show that
(Xl
7r L a 7r
 < <
2 a
2
+ n
2
2·
n=l
P19752. A polygon having all its angles equal and an odd number of
vertices is inscribed in a circle. Prove that it must be regular.
P19753. Given m lines in the plane, with no two parallel and no three
concurrent; into how many components do they divide the plane? Prove
your assertion.
P19754. The expression alb means that a divides b. Suppose that
2ml(3m  1).
(a) Show that if m i= 1, then m is even, and
(b) Show that m = 1, 2, or 4.
P19755. Suppose that a > 0 and f is continuous for 0 ::; x ::; a. Define
g(x) = i
a
f(t) dt
x t
for 0 < x ::; a. Show that
loa g(x)dx = loa f(x) dx.
Exam #111976
This competition was held at DePauw University. The winning team of Jay
Ponder, Tom Sellke, and Matthew Wyneken was from Wabash College.
Teams from RoseHulman and Franklin College came in second and third,
respectively.
P19761. Determine all polynomials p( x) such that
p(x
2
 1) = Ip(x)12  1 and p(2) = 2.
P19762. Let n be a positive integer such that n + 1 is divisible by 12.
Prove that the sum of all of the divisors of n is divisible by 12.
P19763. Let
f(x) = al tanx + a2 tan ( ~ ) + a3 tan ( ~ ) + ... + an tan (;) )
IS Exams
where a1, a2, ... , an are real numbers and where n is a positive integer.
Given that If(x)l:S: Itanxl for x E {Jr/2,Jr/2}, prove that
I
a2 a3 an I
a1 + 2 + 3" + ... + ;;: :s: 1.
P19764. How many zeroes does the function f(x) = 3
x
1 2X2 have
on the real line? Prove that your answer is correct. Hint: You may need to
know that 1 < In 3 < 1.1.
P19765. Let A, B, and C be three noncollinear points in a rectangular
coordinate plane with coordinates (alla2),(b
1
,b
2
), and (C1,C2) respec
tively. Prove, using algebraic methods, that it is always possible to solve
for the coordinates of the center of the circle containing A, B, and C.
P19766. Suppose that f is a realvalued function of a real variable, and
that f(x + y) = f(x)f(y) for all x and y, f(1) I 0, and limx+o f(x)
exists. Prove that limx+o f (x) = 1.
Exam # 121977
This competition was held at RoseHulman and was also won by a team
from there. The team consisted of Rich Priem, Jan Slupesky, and Bob
Strickland.
P19771. Let f be a continuous function that maps the closed unit
interval J = [0, 1] into itself. Show that if f (f (x)) = x for all x III
J, then either f is strictly increasing on J or f is strictly decreasing.
P19772. Prove the following identities:
(a) cos
n
x = 2 ~ ~ ( ~ ) cos(n  2k)x
1 r 2n (2n)!
(b) :; io cos xdx = 22n (n!)2'
P19773. Let A = (aij) be an n x n matrix of real numbers such that
2:.7=1 aij = 1 for each i. Show that the matrix A  I is not invertible.
P19774. Find all of the isometries of the set R of real numbers, that is,
functions f from R into R with the property that the distance between any
pair of points is the same as the distance between their images under f.
Hint: Consider first the special case in which f(O) = o.
Exam #131978 19
P19775. Use the fact that lin diverges to obtain a proof that
there are infinitely many prime numbers P1,P2, .... Rint: For each integer
N there is an integer r such that
111
"7""c, + + ... + ,,,,,
... p';F ... pW ... p'j;
< (1 + + + ... + (1 + + + ... +
 P1 pi PI P2 P2
... (1 + + + ... +
PN P7v P'N
Exam # 131978
This competition, held at Earlham College, was again won by a team from
Wabash College. The team consisted of Kevin Fosso, Jay Ponder, and
Matthew Wyneken. Teams from RoseRulman and Manchester College
came in second and third, respectively. The exam came with instructions:
For problems 1 and 4 you need only show your computations, whereas
problems 2, 3, 5, and 6 require more justification.
P19781. The number 10 is a base for the positive integers because every
positive integer can be written uniquely as
where each d
i
is one of the digits 0,1,2,3,4,5,6, 7,8, or 9. The number
2 is a base for all integers using the digits 0 and 1. For example 1101
represents 3, since
1( _2)3 + 1( _2)2 + O( 2) + 1 = 3.
Find the representation in base 2 for the decimal number 2374.
P19782. Prove the following generalization of the Theorem of Pythago
ras: If ABCD is a 3dimensional tetrahedron such that each of the triangles
ABC, ABD, and ACD has a right angle at A, then
(areaABC)2 + (areaABD)2 + (areaACD)2 = (areaBCD)2.
P19783. Let k be a positive odd integer and let.S (n) = 2:7=1 jk. Show
that 2S(n) is divisible by n + 1.
20 Exams
P19784. Let aI, a2, a3, ... , an be real numbers, not necessarily distinct,
and let f(x) = ' E ~ = l l x  akl. For which value(s) of x is f(x) small
est? What is the minimum value of f(x)? You need not give a formal
justification for your answer.
P19785. Suppose that A, B, C, and D are four distinct points in the
plane. Find the point 0 in the plane so that the sum of the distances from
o to each of the four points is smallest, i.e., so that
10AI + lOBI + lOCI + 10DI
is smallest.
P19786. Design an experiment with a fair coin for which the probability
of success is 1/3. A fair coin is a coin for which the probability of tossing
heads and the probability of tossing tails are each 1/2. As an illustration,
we present the following experiment, for which the probability of success
is 1/4: Toss a fair coin twice. The experiment is a success if heads is
tossed both times.
Exam #141979
The team of Mike Hall, Tim Drabik, and Tony Mazzoni from RoseHulman
won this competition, which was again held at Butler University.
P19791. A piece of rope weighs four ounces per foot. It is passed over
a pulley, and on one end is suspended a weight, and on the other end a
monkey. The whole system is in equilibrium. The weight of the monkey
in pounds is equal to the age of the monkey's mother in years. The age of
the monkey's mother added to the age of the monkey is four years. The
monkey's mother is twice as old as the monkey was when the monkey's
mother was half as old as the monkey will be when the monkey is three
times as old as the monkey's mother was when the monkey's mother was
three times as old as the monkey. Each of the weight of the rope and the
weight at the end is half again as much as the difference in weight between
the weight and the weight plus the weight of the monkey. How long is the
rope? Assuming the pulley to be negligibly small, what fractional part of
the rope is on the same side of the pulley as the monkey when the system
is at rest?
P19792. Graph the relation sin x = sin y in the x, yplane.
Exam #141979 21
P19793. Snow starts to fall in the forenoon and falls at a constant rate
all day. At noon a snow plow starts to clear a highway_ The velocity of
the snow plow is such that it removes a constant volume of snow per unit
of time. The plow goes a mile during the first hour. It goes a half mile
during the second hour. What time did it start to snow? Give the time to
the nearest minute.
P19794. A hole six inches long is drilled through a sphere. (The hole is
six inches long after it is drilled, and the hole is through the center of the
sphere.) What is the volume of the part of the sphere that remains?
P19795. Note that 1 = 1
979
,2 = (_1)97 + V9, 3 = 1
9
+ 17  91,
4=1+V9+l7/9j.
Continue this through consecutive numbers k as far as you can. The
rules of the game are as follows:
(i) The only constants you may use are 1, 9, 7, 9, which must be used
exactly once and in that order.
(ii) Algebraic grouping symbols, operations, and functions may be used
as often as desired. Note that rule (i) prohibits use of 7r or e, as well
as use of higher order roots.
P19796. Find all points with integer x and y coordinates which are
strictly inside the polygon ABCDEFGH (described below) and also
strictly outside the circle centered at (4, 1) with radius 3/2. Here is a
description of the polygon:
B = (4, 21), D = (11, 1), E = (2, 2), F = (4,14)
Li (i = 1 to 8) are lines as follows:
Ll goes through Band (6, 2)
L2 goes through B and has slope 9/4
L3 goes through D and has slope 2/15
L4 goes through D and E
L5 goes through E and F
L6 goes through F and is perpendicular to x  16y = 4
L7 has equation x + 7y = 4
Ls has a yintercept of 6 and angle of inclination 45°
A is the intersection of Ll and Ls
C is the intersection of L2 and L3
G is the intersection of L6 and L7
H is the intersection of L7 and Ls.
22 Exams
Exam #151980
This year, for the first time, the competition was held at Valparaiso
University, located in Valparaiso. The winning team from RoseHulman
consisted of Michael Call, Randall Ekl, and Douglas Englehart. A team
from Butler University came in second.
P19801. Let {an} be a sequence of positive numbers. If there exists a
sequence {b
n
} of positive numbers and a constant a > 0 such that
show that the series
00
is convergent.
P19802. Consider a balance that is used to measure loads of integral
weights. The balance has two scales, a load scale and a weight scale. On
the weight scale one can place only certain measuring weights. On the load
scale one can place the load to be measured and any desired subset of the
measuring weights. Show that with four suitably chosen weights one can
measure the weight of any load whose weight is an integer between 1 and
40 pounds.
P19803. Suppose that f is a function that satisfies
f"(x) + f'(x)g(x)  f(x) = 0
for some continuous function g. Prove that if f vanishes at a and at some
b> a, then f is identically zero on [a, b].
P19804. Show that det(I + xyt) = 1 + xty for any vectors x and y in
Rn. Note that t denotes transpose.
P1980S. Find all positive functions f that satisfy
f(x) < 2(xy)2
f(y) 
for all x, y.
P19806. Show (a) that the integral
roo sin x dx
10 x
Exam #161981
is convergent. Then show (b) that
P19807. If
show that
Exam #161981
.!£ ( roo sin ax dX) 2 = O.
da 10 x
n (_l)k+l
Sn = L k '
k=l
n 1
S2n= Lk·
k=l n +
23
This competition, again won by a team from RoseHulman, was held at
Indiana UniversityPurdue University in Indianapolis (IUPUI). The team
consisted of Michael Call, Thomas Douglas, and Tim Drabik.
P19811. Express 3/8 as a base 7 fraction of form 0.aOala2a3···.
P19812. Before steers are introduced to a pasture, there is a given
amount of grass per acre, and the grass keeps growing at a constant rate.
If 12 steers take 16 weeks to deplete the grass on 10 acres, and if 18 steers
take 8 weeks to deplete the grass on 10 acres, how many steers does it
take to deplete the grass on 40 acres in 6 weeks?
P19813. A ball of radius 1 is in a comer touching all three walls. Find
the radius of the largest ball that can be fitted into the comer behind the
given ball.
P19814. The winning team of the World Series must win four games out
of seven. Assuming that teams are equally matched, find the probabilities
that the Series lasts
(a) exactly four games,
(b) exactly five games,
(c) exactly six games, and
(d) exactly seven games.
P19815. A man is standing atop a tall building. At a point 50 feet above
his eye atop a building 100 feet away, a rock is dropped. If the man watches
the rock fall, at what point in its descent is his head moving the fastest?
24 Exams
Assume that the buildings are arbitrarily high, that air friction is neglected,
and that the acceleration of gravity is 32 ftlsec/sec.
P19816. A destroyer is hunting a submarine in dense fog. The fog
lifts for a moment, disclosing the submarine on the surface three miles
away, upon which the submarine immediately descends. The speed of the
destroyer is twice that of the submarine, and it is known that the latter
will depart at once at full speed on a straight course of unknown direction.
The wily captain of the destroyer sails straight to the point 2/3 of the way
to the spot where the submarine was sighted and then sets out on a spiral
course that is bound to make him pass directly over the submarine. What
is the equation of this spiral? HINT: Use polar coordinates with the origin
at the point where the submarine was sighted.
Exam #171982
This was the first time that the competition was held at Ball State
University, located in Muncie. The winning team of Jeffery Baldwin,
Randall Ekl, and 1. Anthony Kirk was from RoseHulman.
P19821. Find a cubic equation whose roots are the reciprocals of the
roots of the equation x
3
+ ax
2
+ bx + c = 0, with c i= O.
P19822. A subset {al, a2, ... ,ad of the set {I, 2, ... , n} is said to be
separated if ai+l  ai ;::: 2 for i = 1,2, ... ,n  1. For example, {2, 5, 7}
is a separated subset of {I, 2, ... , 8}, but {3, 4, 8} is not. Show that the
number of separated subsets of {I, 2, ... , n}, each having k elements is
(nZ+l).
P19823. Given finitely many points in the plane situated so that any
three of them are the vertices of a triangle of area :::; 1. Show that all the
points can be enclosed in a rectangle of area :::; 4.
P19824. Is the function
f(x) = {e
1jX2
, x i= 0
0, x =0
differentiable at x = O? Prove your answer. Hint: You may use the fact that
lim yTe
Y
= °
Y+OO
for any r.
Exam # 1 81983 25
P19825. A realvalued function f of a real variable is said to satisfy
a Holder condition with exponent a if there is a constant c such that
If(x)  f(y)1 ::; clx  yin for all x, y. Wherever these functions are used,
a is restricted to be ::; 1. Can you explain why?
P19826. The probability that the square of a positive integer (in decimal
notation) ends with the digit 1 is 2/10 because out of every 10 numbers
those and only those ending with the digits 1 or 9 have squares ending
with 1. What is the probability that the cube of a positive integer chosen
at random ends with the digits 11? Prove your answer.
P19827. Find the volume of a torus (doughnut) of inner radius b whose
crosssection by a plane through the axis is a semicircle of radius a, with
its straight boundary parallel to the axis and curved boundary away from
the axis.
Exam # 181983
RoseHulman won the competition again this year. Its winning team con
sisted of Baron Gemmer, 1. Anthony Kirk, and Tom Moss. The competition
was held at the Indiana University, Bloomington, campus for the first time.
Teams from Butler University and Manchester College tied for second
place.
P19831. Find limn+oo \inTo
P19832. Suppose you repeatedly toss a fair coin until you get two heads
in a row. What is the probability that you stop on the 10th toss?
P19833. Consider an isosceles right triangle with legs of fixed length
a. Inscribe a rectangle and a circle inside the triangle as indicated in the
figure below. Find the dimensions of the rectangle (and the radius of the
circle) which make the total area of the rectangle and circle a maximum.
P19834. Show that every positive real number is a sum (possibly
infinite) of a subset of the numbers {I, 1/2, 1/3, 1/4, ... }.
26 Exams
P1983S. This problem is a repeat of problem #3 on EXAM #7.
P19836. Prove the "restricted" Fermat conjecture: For any integer n > 2,
xn + yn = zn has no solutions in positive integers if z < n.
P19837. Let N be one more than the product offour consecutive positive
integers. What can you say about N? Prove it.
P19838. Let Xl, X2,"" Xn be (distinct) real numbers. Define polyno
mials PI(X), P2(X), ... ,Pn(x) by
Pk(X) = II XXj ,
j#k xk  Xj
Prove that = 1 for all X in R.
xER.
P19839. Suppose that g : [0,1] + [0,1] is a continuous function and
that gm(x) = X (g composed with itself m times) for all X and for some
positive integer m. Prove that g2 (x) = x for all x.
Exam #191984
This competition was held at RoseHulman and a team from RoseHulman
consisting of Todd Fine, Byron Bishop, and Dan Johnson won. Teams
from Wabash College and Valparaiso University came in second and third,
respectively.
P19841. What is the prime factorization of 1,005,010,010,005,001 ?
P19842. In the graph below, how many paths that never go up connect
node A to node B? (Paths must follow edges indicated.)
A
B
Exam #191984 27
P19843. A very long hallway has 1000 doors numbered 1 to 1000; all
the doors are initially closed. One by one, 1000 people go down the hall:
the first person opens each door, the second person closes all doors with
even numbers, the third person closes door 3, opens door 6, closes door
9, opens door 12, etc. That is, the nth person changes all doors whose
numbers are divisible by n. After all 1000 people have gone down the
hall: which doors are open and which are closed?
P19844. In Subsylvania, there is no paper money and there are only two
kinds of coins, one worth 7 quanta, the other worth 12 quanta. Find the
largest price for which it is not possible for a Subsylvanian customer to
give exact change, or prove there is no such price. [Naturally, all prices
are whole numbers of quanta. Note that it is possible for a Subsylvanian
customer to buy an item costing 5 quanta (she gives the merchant a 12
quanta coin and receives a 7 quanta coin in exchange), but it is not possible
for her to give the merchant 5 quanta in exact change.]
P19845. Choose two points at random in the interval [0,1]. These two
points cut the interval into three segments. What is the probability that these
three segments can form the sides of a triangle? Note: "at random" means
the two points are chosen independently and with uniform probability
distribution. The probability of choosing a point in the interval [a, b] is
b a.
P19846. Let f be a continuous function on ° < x < 00 satisfYing
f(l) = 5 and f (_X_) = f(x) + 2 forO < x < 00.
x+1
(a) Find lim f(x).
x+oo
(b) Prove that lim f(x) = +00.
x+o+
(c) Find all such functions f. (Part credit for finding some f.)
P19847. A train is being made up of Boxcars, Flat cars, and Hopper
cars. Boxcars weigh 25 tons, and Flat cars and Hopper cars weigh 50
tons each. How many different trains of total weight 1000 tons (ignoring
caboose and engines) can be made up? Trains are considered identical if
and only if they have the same sorts of cars in the same order: e.g., H H B,
H BH and F BH are distinct trains of total weight 125 tons.
28 Exams
Exam #201985
ReId at Ball State University, this competition was again won by a team
from RoseRulman. The team consisted of Todd Fine, Erick Friedman,
and Dan Johnson. Teams from RoseRulman and Goshen College came in
second and third, respectively. Notice that the contest rules for assigning
second place were broken this year; it was the first of several in which
this occurred.
P19851. Let (Y, {3, 'Y, and 8 be the roots of X4 + bx
3
+ cx
2
+ dx + e = O.
Compute
((Y2 + 1)({32 + 1)("(2 + 1)(8
2
+ 1)
in terms of b, c, d, and e.
P19852. Prove: 3333
4444
+ 4444
3333
is divisible by 7. Make up some
more problems of this type.
PI9853. Consider the set U = {I, 2, 4,5,8,10,11,13,16,17,19, 20}.
In the figure below, each of the three pathsthe solid, the dotted, and
the dashedrepresents a subgroup of order 6 of U under a certain binary
operation. Three of the points of these paths have been labeled. Find all
the others .
.................. ................. .
: ".' :
:. " ".:
\ ".... j
\ .......•...... /
PI9854. Each side of a square is subdivided into 101 segments. Find
the number of different triangles that have their vertices at these 400 points
of subdivision.
P19855. Let al = 0 and a2 = 1, and for n 3,
an = (n  l)(anl + an2).
Find
(a) a formula for an and
(b) lim
n
+
oo
';" .
Exam #211986 29
P198S6. Given n + 1 integers between 1 and 2n inclusive, prove that
one of them must be a multiple of some other one.
P198S7. Let a and b be positive constants with b > 1. Given that
x + y = 2a and all values of x between 0 and 2a are equally likely,
find the probability that
Exam #211986
This competition was held at Butler University. The winning team consist
ing of Erich Friedman, John Hoffman, and Dan Tretter was from Rose
Hulman. The second place team was also from RoseHulman, and the third
place team was from Valparaiso University.
P19861. This is a repeat of problem #1 on EXAM #14.
P19862. Let {b
l
, b
2
, b
3
} be an orthonormal basis for C
3
and let ¢ E
L(C
3
, C3) be given by
¢(bd = 2b
2
, ¢(b
2
) = 2ib
l
, and ¢(b
3
) = (1 + i)b
3
•
Find the eigenvalues, if any, for ¢.
P19863. Evaluate ii.
P19864. N points are chosen on a circle so that when all segments are
drawn between all pairs of points, no three segments intersect at the same
point. In how many points do these segments intersect in the interior of
the circle?
P1986S. Find an equation with integral coefficients one of whose roots
is y'2 + ~ .
P19866. Let f : R2 + R be given by
f( )
 { ~ I X ~ 4' if (Xl, X2) i= (0,0);
XI,X2  Xl +X
2
0, if(Xl,X2) = (0,0).
(a) Is f continuous at (0, O)? JustifY your answer.
(b) Is f differentiable at (0, O)? JustifY your answer.
30 Exams
Exam #221987
This competition was held at Butler University. The team consisting of
Daniel W. Johnson, Brenton Young, and Daniel Kniep from RoseHulman
won the competition. A team from Indiana University, Bloomington, came
in second, and a team from Wabash College came in third.
P19871. A positive integer n is called composite if there are positive
integers P i= 1 and q i= 1 so that n = pq. Find a sequence of 10 consecutive
positive integers each of which is composite and less than 1,000,000.
P19872. John's job at the Acme Cannonball Factory is to stack the
cannonballs (which are 6inch diameter spheres) neatly into tetrahedral
piles. For example, using 4 cannonballs, John can make a tetrahedral
pile with 2 cannonballs on each edge: three balls forming the bottom
triangle and one in the center on the top. Find a formula for the number
of cannonballs in a tetrahedral stack whose base is an equilateral triangle
with n balls on each edge.
P19873. Experiments have determined that when a particular steel ball
is bounced on a hard surface, it bounces to half its original height. For
example, if it is dropped from a height of 6 feet, it will bounce to 3 feet.
Assuming that the ball obeys this law exactly, for what length of time will
the ball continue to bounce if it is dropped from a height of 16 feet (or will
it bounce forever)? [Recall from calculus that since the acceleration due
to gravity is 32 ft/sec/sec, an object failing to the ground from height h
(in feet) or bouncing from the ground to height h requires Vli/4 seconds
to do so.]
P19874. Ten seniors who share a house decide to exchange graduation
presents. They each put their name into a hat, mix the name cards
thoroughly, and draw a card out at random. What is the probability that
none of the ten draws his or her own name?
P19875. Euclidean fourspace is R4 with
IXYI = ((Y1  X1)2 + (Y2  X2)2 + (Y3 + X3)2 + (Y4 + X4)2) 1/2.
(a) Find five points in the unit ball of Euclidean fourspace that are as
far from each other as you can make them. That is, find points A, B,
C, D, and E such that
10AI :s; 1, 10BI:S; 1, 10CI:S; 1, 10DI:S; 1, and 10EI:S: 1,
Exam #231988 31
where 0 is the origin, such that
min{IABI, IAOI, IADI, IAEI, IBOI, IBDI, IBEI, IODI, IOEI, IDEI}
is as large as you can make it.
(b) If your answer is best possible, prove that it is.
P19876. Jane begins a journey at Quito, Ecuador (at 80° west longitude
on the equator) and flies steadily northwest.
(a) How far does she fly before reaching the North Pole?
(b) How many times does she cross the Greenwich Meridian (longitude
0°) in the journey?
(Note: You should suppose that the earth is a perfect sphere of radius
4000 miles, that there is no effect of wind, etc., that compass headings are
exactly accurate, and that Jane's plane has an unlimited amount of fuel
available.)
Exam #231988
This competition was held at Butler University. The contest was won by a
team of two students from Indiana University, Bloomington: Radu Tudorica
and Kevin Pilgrim. A team from IUPUI came in second, and a team from
Purdue University came in third.
P19881. Going at top speed, Grand Prix driver x leads his rival y by
a steady three miles. Only two miles from the finish, x runs out of fuel.
Thereafter x's deceleration is proportional to the square of his remaining
velocity, and, in the next mile, his speed exactly halves. Who wins and
why?
P19882. Let (G, *) be a group with the following cancellation rule:
x * a * y = b * a * c implies x * y = b * c for all x, y, a, b, and c in G.
Prove that G is Abelian, i.e., that * is commutative.
P19883. Define
n
n(x) = IT (x  i),
i=l,i#k
and let P(x) be a polynomial of minimum degree in which P(k) =
5T
k
(k), k = 1,2, ... , n. If sand t are both integers such that 1 :s; s :s; n
and 1:S; t:s; n, prove that fst P(x)dx = o.
32 Exams
P19884. let 5
n
be the sum of the squares of the first n positive odd
integers. What is the units digit of 512345? Prove your answer.
P19885. Find the shaded area of the figure below, where the interior
circular arcs have their centers on the outer circle. (This was a proposed
problem in the then current issue of The Journal of Recreational Mathe
matics. Students were encouraged to send in solutions they liked.)
P19886. Show, with proof, how to construct with unmarked straightedge
only, a perpendicular from the point P to the line L, as shown in the figure
below. (This is problem 13 in the 1988 issue of The Old Farmer's Almanac.
A solution to this plus several other problems could have won the solver
50 dollars.)
o
•
p
L
P19887. A fair coin is tossed ten times. Find the probability that two
tails do not appear in succession.
Exam #241989
This competition was held on the Indiana University, Bloomington, cam
pus. A team from Purdue University won the competition. The members
of this team were Alex Gottlieb, Aaron Weindling, and Nung Kwan Yip.
Teams from Earlham College and Wabash College came in second and
third, respectively.
P19891. Three students A,B, and C compete in a series of tests. For
coming first in a test, a student is awarded x points; for coming second,
Exam #241989 33
y points; for coming third, z points. Here x, y, and z are positive integers
with x > y > z. There were no ties in any of the tests. Altogether A
accumulated 20 points, B 10 points, and C 9 points. Student A came in
second in the algebra test. Who came in second in the geometry test?
P19892. Describe all sequences {x 1, X2, ... ,x
n
} of real numbers that
satisfy
2mn
IXm  Xnl:::; 2 2
m +n
for all positive integers m and n.
P19893. Show that the area of the shaded triangle formed by the side
trisectors of a triangle (as shown) is 1/7 of the area of the whole triangle.
P19894. Two coins are given. One is fair [Prob(heads) = 1/2] and the
other is biased with Prob(heads) = 2/3. One of the coins is tossed once,
resulting in heads. The other is tossed three times, resulting in two heads.
Which coin is more likely to be the biased one?
P1989S. Two identical pipes have elliptical crosssections with semi
axes a and b. The pipes intersect at right angles as shown below. Find the
volume of their intersection.
34 Exams
P19896. Find the determinant of the matrix
X1Yn )
X2Yn
1 + ~ n Y n .
P19897. What are the dimensions of Smith's ranch, described in the
following conversation?
Smith. Down in Todd County, which is a 19 x 19 miles square, I have
a ranch. It is rectangular, not square, the short side and the long side both
measuring a whole number of miles.
James. Hold on a minute! I happen to know the area of your ranch. Let
me see if I can figure out its dimensions. (He figures furiously.) I need
more information. Is the short side strictly less than half the long side?
(Smith answers his question.)
James. Now I know the dimensions.
Kathy. I, too, know the area of your ranch, and although I did not hear
your answer to James's question, I, too, can tell you the dimensions.
Bill. I did not know the area of your ranch, but having heard this entire
conversation, I can now figure it out.
Exam #251990
Beginning in 1990, significantly more information was collected about the
competitions. This year, the contest was held at Purdue University (located
in West Lafayette) for the first time, on March 31, 1990 in connection with
the spring meeting of the Indiana Section of the MAA. Twentyfour teams
from twelve colleges participated. The winning team from RoseHulman
consisted of Joel Atkins, John O'Bryan, and Kevin O'Bryant. They scored
51 out of 60 points on the following examination. The second place team
from Purdue University consisted of Mark Sepanski, Peter Sepanski, and
Nung Kwan Yip. The third place team from Indiana University consisted
of Urmi Bhatacharya, Lucia Demetrios, and Radu Tudorica.
P19901. Find all positive integers which are one more than the sum of
the squares of their base ten digits. For example, 35 = 1 + 3
2
+ 52.
P19902. A digraph in a word is an ordered pair of consecutive letters;
a word with n letters has n  1 digraphs. How many ways can the letters
Exam #251990 35
I, N, D, I, A, N, A be arranged so that no digraph is repeated? (Thus, the
arrangement A, N, D, I, I, N, A is counted, but not I, A, N, N, I, A, D,
which contains "I, A" twice.)
P19903. Find all real functions f such that, for all real x,
f(x + 2) = f(x) and f'(x) = f(x + 1)  2.
P19904. Let ABCDEFG denote a regular heptagon with side 1. By
connecting the vertices ACEGBDF A, in that order, we create a new reg
ular heptagon HIJKLM with side r. (See the figure below). Determine
r.
A
F c
P1990S. Evaluate the following limit.
P19906. A regular hexagon of side 1 is inscribed in the intersection of
two identical parabolas, oriented (in opposite directions) with their axes
parallel to the yaxis (see the figure below). Find the area of the (shaded)
region inside the parabolas and outside the hexagon.
36 Exams
Exam #261991
The contest was held at Anderson University, located in Anderson, on
March 23, in connection with the spring meeting of the Indiana Section
of the MAA. Twentythree teams from fourteen colleges participated.
The winning team from RoseHulman consisted of Jonathan Atkins, John
O'Bryan and Mark Roseberry. The second place team from Purdue Univer
sity consisted of Alex Gottlieb, Peter Sepanski and BoonLock Yeo. The
third place team from Goshen College consisted of Dave Cooper, Kathy
Steiner and Ken Wenger.
P19911. Find the area of that portion of the xyplane which is enclosed
by the curve with equation
12x  11 + 12x + 11 + ~ = 4.
P19912. Between I and 1,000,000 inclusive, which are more numerous,
those integers whose baseten representations contain a "1" or those that
do not? How many of each type are there?
P19913. P, Q and R are arbitrary points on sides BC, DA and CD
respectively, of the parallelogram ABCD, illustrated below. A is joined to
P, P to Q, Q to B, B to Rand R to A to fonn a starpentagon APQBR.
The regions inside the parallelogram and outside the starpentagon are
colored red (r). The pentagonal region bounded by the sides of the star
pentagon is colored blue (b). Show that the red area minus the blue area
is independent of the choice of the points P, Q and R.
R C
A ~                           ~ .
P19914. If a, band e are integers and if a+b+e is even, find an integer
n (in tenns of a, b and e) such that ab + n, be + nand ea + n are all
squares of integers.
Exam #271992 37
P19915. (a) Show that there exist infinitely many twosomes of consec
utive positive integers each of which can be written as the sum of two
nonzero perfect squares. For example, 17 = 12 + 4
2
, 18 = 3
2
+ 3
2
and
25 = 3
2
+ 4
2
, 26 = 12 + 52.
(b) Find three consecutive positive integers each of which can be written
as the sum of two nonzero perfect squares. Are there infinitely many
such trios?
(c) Do there exist four consecutive positive integers each of which can be
written as the sum of two nonzero perfect squares?
P19916. The harmonic mean of a set of positive numbers is the recip
rocal of the arithmetic mean (ordinary average) of the reciprocals of the
numbers. Find limn+oo(Hn/n), where Hn is the harmonic mean of the n
positive integers n + 1, n + 2, n + 3, ... , n + n.
P19917. The centroid of a set of n points {(Xi, Yi), i = 1,2,···, n}, in
n n
the xyplane is given by ( ~ 2: Xi, ~ 2: Yi). A lattice point in the xyplane
1 1
is a point whose coordinates are integers. Given nine lattice points in the
xyplane, show that some three of the points have a lattice point as centroid.
Exam #271992
The contest was held at The University of Indianapolis on April 11, in
connection with the spring meeting of the Indiana Section of the MAA.
There were 24 teams with a total of 67 participants from 13 colleges.
The winning team from RoseHulman consisted of Jonathan Atkins, Tony
Hinrichs, and Kevin o 'Bryant. The second place team from IndianaPurdue
University at Fort Wayne consisted of Jeff Gerardot, Aaron Pesetski, and
Mark Pragan. The third place team from Purdue University consisted of
Peter Sepanski and PokYin Yu.
P19921. A car rode over an ant on the pavement. The ant stuck to the
tire for one revolution and then was deposited back onto the pavement.
Assuming that the radius of the tire is one foot, find the length of the
curve traveled by the ant between its death and its final resting place.
P19922. At a movie theater, n patrons have lined up to buy tickets.
The ticket seller calls a patron viewable if he (she) is taller than all the
people in front of him (her) in line; otherwise he (she) is hidden. Given
that no two patrons are precisely the same height, find the average number
of viewable patrons among all possible permutations of the patrons.
38 Exams
P19923. A collection of n gossips each knows a unique tidbit of scandal
not known to any of the others. They communicate by mailing letters. Of
course each gossip will share all of the scandal he (she) knows at that time
whenever he (she) sends a letter. Find, with proof, the minimum number
of letters that can suffice to share all of the scandal.
P19924. Identify all finite groups G of order n having at least 2
n

2
proper subgroups.
P19925. Evaluatei: e
x2
dx.
P19926. For A = ( _ ~ 9 0 _56 )
(a) find the eigenvalues of A.
(b) Find four different integral matrices B so that B2 = A. (Hint: How
do the eigenvalues of B relate to the eigenvalues of A 7)
(c) For an arbitrary 2 x 2 matrix A, what is the maximum number of
integral matrices B that can satisfy B2 = A7
P19927. Evaluate
P19928. We all know that the lengths 3, 4, and 5 form the sides of a right
triangle. Notice that 3 and 4 are consecutive integers. Determine all other
integral right triangles whose legs are consecutive integers.(The hypotenuse
must also be an integer, but need not be consecutive as 5 happens to be.)
Exam #281993
The contest was held at Saint Mary's College, located at the northern edge
of South Bend, on April 24, in connection with the joint spring meeting of
the Indiana Section of the MAA with the Illinois and Michigan Sections.
There were 30 teams, with representation from all three states. The winning
team from RoseHulman consisted of Jon Atkins, Tony Hinrichs, and
Nick Tallyn. The second place from from IndianaPurdue University, Fort
Wayne, consisted of Joel Holcombe, Brian Johnson, and Aaron Pesetski.
The third place team from Calvin College, Michigan, consisted of Alan
Baljeu, Milce Bolt, and Tim Hotlebeek.
Exam #281993 39
P19931. Let P(x) be a real cubic polynomial for which P'(x) has
distinct real zeros. Prove that
pili ( X ) (P" ( x ) ) 2
P'(x) < 2 P'(x)
for all x for which P' (x) I O.
P19932. Let f and 9 be mappings from the set A to itself for which
f(g(f(a))) = g(a) and g(f(f(a))) = f(a) for all a in A. Prove that
f = g.
P19933. Let A be a square matrix of rank 1 and trace 1. Prove that
A2 = A.
P19934. For n > 1, a permutation al,a2, ... ,a
n
of {l,2, ... ,n} is
"orderly" if, for each i = 1,2, ... , (n  1), there is a j > i for which
laj  ai I = 1. (For example, for n = 4, the permutation 1,4,2,3 is orderly
whereas 3,1,4,2 is not.) How many permutations of {I, 2, ... ,n} are
orderly? HINT: What are the possible values for al ?)
P19935. Let al = 1 and a,+l = val + a2 + ... + ai, for i > O.
Determine lim (an).
n+oo n
P19936. Prove that 2
n
+
l
is a factor of 1(V3 + 1)2nl for all positive
integers n. (Here I x l denotes the smallest integer not less than x.)
P19937. Find all integers A, B, C, D, E (A :::; B :::; C :::; D :::; E)
which, when added in pairs, yield only the sums 401, 546, 691, and 836.
P1993S. In the figure below, ABCD is a rectangle. Find the area of the
parallelogram abed.
5
5
A
6 2
40 Exams
Exam #291994
The contest was held at Depauw University, on March 19, in conjunction
with the spring meeting of the Indiana Section of the MAA. Twenty teams
from eleven colleges participated. The winning team from Wabash College
consisted of Faisal Ahmed, Avijit Sarkar, and Faisal Syud. The second
place team from RoseHulman consisted of Jonathan Atkins, Nick Tallyn,
and James Moore. The third place team from Purdue University consisted
of Ilya Gluhovsky, Brian Singer, and William Stier. (These problems were
used with the permission of Crux Mathematicorum, © 1975, 1976, 1977.)
P19941. Calculate
l
x
2 2
lim e
t
x dt.
x+= 0
P19942. Show that for positive integers n, vn < 1 + y'27n.
P19943. If x > y > 0, show that
P19944. The sides of a triangle have lengths 4, 5, and 6. Show that one
of its angles is twice another.
P19945. Find all sets of five positive integers whose sum equals their
product.
P19946. Show that if one of the coordinates of the center of a circle
is irrational, then the circle contains at most two points both of whose
coordinates are rational.
Exam #301995
The contest was held at TriState University, located in the far northeastern
community of Angola on March 31, in conjunction with the spring meeting
of the Indiana Section of the MAA. Twentythree teams from eleven
colleges participated. The winning team from IU, Bloomington, consisted
of Daniel Bliss, Matt Laue, and Seth Patinkin. The second place team from
Wabash College consisted of Eham Ahmed Chowdhury, Matthew Gladden,
and Shirish Ranjit. The third place team from RoseHulman consisted of
Jamie Kawabata, Rick Mohr, and Nick Tallyn.
Exam #311996
P199S1. Let Zl and Z2 be complex numbers. Prove that
IZl + Jzi  + IZl  Jz?  = IZl + z21 + IZI  z21·
P199S2. Find a closed formula for the sum
k=O
P199S3. Let p be an odd prime. Prove that the integer part of
(V5 + 2)P  2
P
+
1
is divisible by 20p.
P199S4. Prove that
4(x
2
+ x + 1)3  27x
2
(x + I? = (x  1?(2x + 1)2(x + 2)2.
P199SS. Let n be any integer greater than 1. Prove that
n(n+I)
1 2 3 n (2n + 1) 2 
1 ·2 ·3 ···n < 3
41
P199S6. ABC is an equilateral triangle with each side of length a. L
is a line parallel to BC and b units below BC. A solid S is generated by
rotating the triangle ABC about the line L; find the volume of S.
P199S7. The vertices of a triangle ABC have coordinates (a cos ai,
a sin ad, i = 1,2,3. Find the coordinates of the orthocenter of triangle
ABC.
P199SS. Find the coefficient of x2n in the expansion of in closed
form.
Exam #311996
The contest was held at Butler University on March 29, in conjunction
with the spring meeting of the Indiana Section of the MAA. Twentyfive
teams from eleven colleges participated. The winning team from Rose
Hulman consisted of Rick Mohr, Nick Tallyn, and James Moore. Indiana
University, Bloomington, Purdue University, and Wabash College tied for
second place. The IU team consisted of Seth Padnkin and Richard Swartz.
The Purdue University team consisted of Dan Crosby, Lukito Muliadi,
and Tze Chao, Ng. The Wabash College team consisted of Alexander
Radnovich, Andrew Jones, and David Whittaker.
42 Exams
P19961. The San Francisco 4gers derive their name from the last two
digits of the year of the California Gold Rush, 1849. Note that both
numbers are perfect squares (49 = 7
2
and 1849 = 43
2
). How many other
fourdigit squares have this property, that their last two digits also form
squares (count 00, 01, 04, and 09 as squares)?
P19962. Find the largest possible area of a pentagon with five sides of
length 1 and a right interior angle.
P19963. A hundred armed rats enter a bar containing a hundred fat cats.
After the cats are lined up against the wall, the first rat marches down the
line taking a dollar from every cat. On his return, he passes the $100
that he collected to the second rat who proceeds to give a dollar to every
second cat. On his return, he passes the remaining $50 to the third rat who
proceeds to take a dollar from every third cat. This continues with the rats
alternately giving and taking until the 100th rat gives a dollar to the lOOth
cat. At this point the rats and the money they have taken disappear in a
cloud of smoke.
(a) How many cats profited? lost money? broke even from this?
(b) Which cat(s) profited the most and how much did they get?
(c) How much money did the rats leave with?
P19964. A square with sides parallel to the coordinate axes is inscribed
in the region
{(x,y): x,y > 0, y ~ 3x  x
3
}.
If its area is written as ijA + ifIj, where A and B are integers, find
A+B.
P19965. In triangle ABC, LA = 90
c
irc and AB = AC = 2. If curve
I joins points of AB and AC to bisect the area of triangle ABC, find the
length of the shortest possible such curve.
P19966. Two couples each agree to have exactly n children. They plan
to eventually pair off all their children in marriage so that sons and
daughters of the first couple marry daughters and sons of the second couple.
Assuming that sons and daughters are equally likely to be born, find the
a priori probability p that such an arrangement is mathematically possible.
Using Wallis' product, Sterling's formula, or any other wellknown result,
determine limn>CXJ (Pn)( yin).
Exam #321997 43
P19967. Call two datasets "statistically equivalent" if they have the same
number of elements, the same mean, and the same standard deviation.
Find all datasets of positive integers (in any order) which are statistically
equivalent to {I, 9, 9, 6}.
P19968. Find all continuous functions f(x) whose graph G (of y =
f(x)) has the following property: For each chord C of G, if C's projection
onto the xaxis has length d
2
, then C's midpoint lies d units above G.
Exam #321997
The contest was held at Franklin College, on March 14, in conjunction
with the spring meeting of the Indiana Section of the MAA. Twenty
five teams from thirteen universities participated. The winning team from
Wabash College consisted of lun Ma, Robert Dirks, and Abishai Daniel.
Teams from RoseHulman and Purdue University placed second and third,
respectively. The RoseHulman team consisted of Kyle Lacey, Chris Prince,
and Tyson Patterson. The Purdue team consisted of Dan Crosby, TzeChao
Ng, and Elad Harel.
P19971. In Indiana, license plates consist of2 digits followed by a letter
and then 4 more digits. Find the probability of getting a license plate in
which the last four digits are nondecreasing.
P19972. It is well known that 1 + 2 + 3 + 4 + ... + n = n(n + 1)/2
for any positive integer n. Find a similar expression for the sum
1 + (1 + 2) + (1 + 2 + 3) + ... + (1 + 2 + 3 + ... + n).
P19973. Find the smallest positive number A so that
21 < A
Iz4  5z
2
+ 61 
for every complex number z on the circle Izl = 3. Prove that A is the
smallest such number.
P19974. Find all solutions (x, y) of the equation x
Y
numbers x, y > O.
P19975. Given the following set of axioms:
(1) Every line contains exactly four points.
(2) No pair of points may be on more than one line.
yX for real
44
(3) Each point must be on at least four lines.
(4) Not all the points are on one line.
(5) There is at least one line.
Exams
(6) If a point is not on a given line, then the point must be on exactly
one line that has no points in common with the given line
(a) What is the fewest number of points and lines necessary to satisfy
these six axioms?
(b) If n were substituted for "four" in (1) and (3), what is the fewest
number of points and lines necessary to satisfy these six axioms?
P19976. The usual way of measuring the length of a vector x
(Xl, X2,"" xn) in ndimensional Euclidean space is the Euclidean norm
But there are other norms which can be used, such as
n
IlxliI = L IXil
i=l
( a) We know that the set of points {xER2 : II X 112 = I} is the unit circle.
Find each of the following sets of points:
{X E R2 : Ilxlll = I} and {x E R2 : Ilxll= = I}.
(b) The triangle inequality Ilx + yll :s:: Ilxll + Ilyll holds for all norms.
For the 2norm, equality only holds if X and yare colinear. Determine
when equality holds for the inorm and the (Xlnorm.
P19977. A farmer has a trough 4 feet long with semicircular cross
sections that are 2 feet in diameter. The top of the trough is parallel to the
ground. Initially, the trough is full of water. If the farmer tilts the trough
(along the side) at an angle ex with the horizontal, express the volume (in
terms of ex) of the remaining water.
Exam #331998
This was the largest ICMC to date. Twentynine teams from 22 universities
and schools participated in this ICMC, held at Ball State University.
The winning team from Wabash College consisted of Robert Dirks,
Exam #331998 45
Abishai Daniel, and Jun Ma. Teams from RoseRulman and the University
of Evansville placed second and third, respectively. The RoseRulman
team consisted of Kyle Lacey, Randy Motchan, Matt Lepinski, and the
Evansville team consisted of Siddartha Naidu, Robert Linne, Riten Sonpal.
P19981. Let C be a circle with diameter AB. Let P be any point
of segment AB. Let C, D, E be distinct points of C all on the same
side of line AB such that DPl..AB and L.CPD = L.DPE. Show that
(PD)2 = PC· PE.
P19982. Show that an integer n with final (decimal) digit u is divisible
by 7 if and only if nlou  2u is divisible by 7.
P19983. Prove that an odd integer n > 1 is composite iff there exist
nonnegative integers p and q such that n = p2  q2 with p  q > 1.
P19984. Let fa = 0, il = 1, and fk = fkl + fk2 be the Fibonacci
numbers.
(a) Show that the points F = {Uk, fk+d} lie on two hyperbolas: y2 
xy  x
2
= ±1.
(b) Show that the only points on the hyperbolas y2  xy  x
2
= ±1 with
nonnegative integer coefficients are points in F.
P1998S. Let fa, il, h be three nonnegative increasing realvalued func
tions defined on the nonnegative real numbers with fi(O) = 0 and
fa (x) :s: il(x) :s: h(x). For any nonnegative x, let R(x) denote the
rectangle whose vertices are (0,0), (x, 0), (0, il(x)), and (x, h(x)) Then,
il is said to bisect fa and h in area if for every nonnegative x, the area of
the portion of the R( x) lying between the curves y = fa (x) and y = il (x)
has the same area as the portion of R( x) lying between the curves
y = il (x) and y = h ( x ). If fa ( x) = xf3 and il ( x) = ax
f3
(a > 1,,8 > 0),
and il bisects fa and h in area, then what is h(x)?
P19986. Let Xn = {1,2, ... ,n}. A permutation ofn objects is a 11
function, cp, from Xn onto itself. cp is called a derangement if cp( x) i= x
for all x in X
n
.
(a) Find a closedform expression for Dn the number of derangements of
n objects.
(b) Show that Dn is equal to the nearest integer to ~ .
(c) Find the radius of convergence of the power series 2 : ~ = 1 D;.(.
46 Exams
P19987. Let us say that two nonsingular 2 x 2 matrices (with complex
entries) A and B are equivalent if there exists a complex number w such
that A = wB. Let [A] denote the equivalence class of A, that is, the set of
all matrices equivalent to A. Note that multiplication of such equivalence
classes is a welldefined operation.
(a) Find as simple a method as possible for calculating the square root(s)
of an equivalence class of matrices other than Here is an
important example to consider: the square root(s) of are
and
since
5) = (14 35) = 7 (2
4 7 21 1
and
(b) How many distinct square roots can an equivalence class that is not
equal to have? Realize that [A] = [A] so that the usual sign
ambiguity doesn't exist here.
(c) How many distinct square roots does have?
Exam #341999
This year the competition was held on the Indiana University in Bloom
ington. The team of Thomas Horine, James Lee, and Keith Henderson of
Purdue University won the contest. Teams from RoseHulman and Wabash
College placed second and third, respectively. The RoseHulman team
consisted of Matt Lepinski, Dennis Lin, Randy Motchan, and the Wabash
team consisted of Robert Dirks, Barry Weliver, Jun Ma.
P19991. In a large urn there are 1999 orange balls and 2000 yellow balls.
Next to the urn is a large pile of yellow balls. The following procedure is
performed repeatedly.
Two balls are chosen at random from the urn:
(i) If both are yellow, one is put back, the other thrown away;
(ii) If both are orange, they are both thrown away and a yellow ball from
the pile is put into the urn;
Exam #341999 47
(iii) if they are of different colors, the orange one is put back into the urn
and the yellow one is thrown away.
What is the color of the last ball in the urn?
P19992. Let P be any point inside an equilateral triangle T. Show that
the sum of the three distances from P to the sides of T is constant.
P19993. If x and yare positive numbers, show that
CiY + (XiY)2 J('f)2 + V
X2
i
y2
~     ~ ~  <       ~     
2 2
and find all cases of equality.
P19994. The statement
1 n (k) r
1
:;;, 2:.f :;;, + 10 f(x)dx
k=l
as n + 00
is familiar to all students of calculus. Several textbooks claim that the
convergence is monotone in n. Show that this claim is false even for
continuous increasing functions.
P1999S. What is the probability that a dart, hitting a square board at
random, lands nearer the center than the edge?
P19996. Does any row of Pascal's triangle contain three consecutive
entries that are in the ratio 1: 2 : 3? Recall that the nth row (n = 0, 1, 2, ... )
consists of the entries
(
n) n!
k  k!(n  k)!
(k=0,1,2, ... ,n).
P19997. Does the series
f COS(I;(k))
k=l
converge or diverge?
P1999S. Let A be a square matrix and suppose that there exist positive
integers m and n such that Am = I and An I I. Calculate
48 Exams
Exam #352000
This year's winning team was from Purdue University and consisted of
James Lee, Thomas Horine, and Keith Henderson. The second place team
from Wabash College consisted of Roberts Dirks, Zhenyu Yang, and Chris
Duefel. The team consisting of Matt Lepinski, Dennis Lin, and Lucas
Beverlin from RoseHulman came in third. The contest was held at Earlham
College.
P20001. Four suspects, one of whom was known to have committed a
murder, made the following statements when questioned by police. If only
one of them is telling the truth, who did it?
Arby: Becky did it.
Becky: Ducky did it.
Cindy: I didn't do it.
Ducky: Becky is lying.
P20002. Call a number N fortunate if it can be written with four equal
digits in some base b E Z+.
(a) Clearly 2222 is fortunate; why is 2000 fortunate?
(b) Find the greatest fortunate number less than 2000.
P20003. Determine whether
converges or diverges.
P20004. Prove that a group G of order 15 must be cyclic.
P2000S. Let the Fibonacci sequence Fk be defined as Fo = 0, Fl = 1,
and Fk+2 = Fk+l + Fk for k ~ O. It is known (and you may assume) that
Fk is the closest integer to Tk / V5 where T = (1 + V5) /2. Show that
P20006. In a triangle one angle is twice another. What is the largest
possible ratio of the area of the triangle to that of its circumcircle? You
Exam #352000 49
may express your answer in terms of
_J4+VfO
g  12'
P20007. Prove or disprove the statement: an irrational power of an
irrational number is irrational. You may assume that V2 is irrational but
be sure to prove all other results used.
P2000S. Find all functions f (x) which satisfY the equation
f(x) + f"(x) = f(3)(x) + f(5)(x)
and have
lim f(x) = f(O) = O.
x+oo
This page intentionally is no longer blank
Solutions
Exam #11966
519661
We will solve the more general problem of determining which integers A
yield integer solutions x and y to the equation x
2
 y2 = A. Since the
LHS factors, we have the system of equations
x  y = r and x + y = s,
with the condition that A = rs. The system has solutions
r+s
x=
2
rs
and y = 2'
Thus x and yare integers whenever rand s have the same parity. This
happens for all odd values of A (r = A, s = 1) and when A is a multiple
of 4.
For our particular problem, let a
3
= A. If a is odd, then so is A, and
if a is even, then A is a multiple of 4, so solutions exist for all positive
integer values of a.
Look under Diophantine Equations in the Index for similar problems.
519662
Divide the unit square into four smaller squares each of side length 1/2
(by drawing two lines which bisect parallel sides of the original square).
Since there are five points and only four smaller squares, two points must
lie in the same square. It is clear that any two such points can be separated
by a distance of less than V2/2 unless the two points are at opposite ends
51
52 Solutions
of the diagonal of the smaller square, which cannot happen since they are
both interior to the original square.
Look under Geometry in the Index for similar problems.
519663
Consider the numbers
There are p + 1 numbers in this list so two of them are congruent mod p.
The difference between those two is the required sum divisible by p. (Note
that p being prime is irrelevant.)
Look under Number Theory in the Index for similar problems.
519664
Solution 1: Suppose the functions are f(x) and g(x), then we must have
(
f(X))'
g(x)
f'(x)
g'(x) .
Applying the quotient rule, you get
g(x) f'(x)  f(x) g'(x)
(g(x) )2
f'(x)
g'(x) .
Multiply this equation by (g (x)) 2 g' ( x) to simplify the equation and we
get
g(x) g'(x) f'(x)  (g'(x))2 f(x) = (g(X))2 f'(x).
Ifwe know one of the functions, say g(x), then we have a first order linear
differential equation for the other function, that we can solve. Choosing
g(x) = x so g'(x) = 1, we find that f(x) must satisfy
x f'(x)  f(x) = x
2
f'(x).
Rewriting this equation, we get
f'(x) = ~ ~ = x(1 ~ x) f(x),
or
df dx
f x(1x)"
Integrating, we have
In If(x)1 = In Ixl In Ix 11 + C,
Exam #11966 53
so f(x) must be
x
f(x) = C x _ l'
An example of two functions f (x) and g( x) having the property that the
derivatives of their quotient is the quotient of their derivatives is given by
f(x) = x/(x  1) and g(x) = x.
Solution 2: Begin the same as in Solution 1, but let g(x) = exp(kx),
so that g' (x) = ke
kx
and the differential equation simplifies to
(k  1)f'(x) = f(x).
Solving this yields
f(x) = Cek"',.
Then, for any k =J 0, 1, the functions f and g satisfY the necessary
requirement.
Look under Differentiation or RealValued Functions in the Index for
similar problems.
519665
Solution 1: (In what follows, "sequence" refers to an ascending sequence
of positive integers as in the problem.) Let S(n, k) be the number of
different sequences of length k in which every number from 1 to n occurs
at least once. Then S (n, k) is simply the number of ways to partition a k
element sequence into n nonempty subsets. Hence, S (n, k) = i). Let
T (n, k, r) be the number of different sequences of length k in which only
r (out of n) distinct numbers occur, then we have
T(n,k,r) =
Finally, we see that the solution is
where the latter equality is an invocation of the Vandermonde convolution.
Solution 2: Make n + k  1 blanks and fill in k of them with x' s. For
any such arrangement of x's, define
ai = 1 + the number of blank to the left of the ith x
54 Solutions
for each i from 1 to k. This gives a 1to1 correspondence between the
sequences we are trying to count and the ways of putting k x's in some
of the n + k  1 blanks. But the number of ways of doing the latter is
obviously
Look under Enumeration in the Index for similar problems.
S 19666
Solution 1: Let f(x) = v'ax
2
+ b. We want to show that f(x) is a
contraction, i.e. :30 < C < 1 s.t. If(x)  f(y)1 :; Clx  yl for all
x, y E R. In this case, the sequence Xl = C, Xn+l = f(xn) converges to
the unique fixed point of f.
If(x)  f(y)1 = Ivax
2
+ b  vay2 + bl
= valvx
2
+ b/a  Vy2 + b/al
:; val
x
 YI·
To see that lv'x
2
+ c  Vy2 + cl :; Ix  yl for any c 2: 0, Multiply and
divide by the conjugate to get
Iv' x
2
+ c + V y2 + cl'
Since
h/x
2
+ c+ Vy2 + cl 2: Ixl + Iyl 2: Ix + YI,
the inequality follows.
Since 0 < va < 1, f is a contraction. The limit of the sequence
{x
n
} is simply the fixed point of f. Solving x = v' ax
2
+ b yields
limntoo Xn = Vb/(l  a).
Solution 2: After computing the first several terms of the sequence,
we see that
xn+1 = va
n
c
2
+ b(l + a + a
2
+ ... + a
n

l
)
Exam #21967 55
Since 0 < a < 1, we may take the limit giving
V
10 M
lim (xn+d = O· c
2
+ b() = .
ntcx; 1  a 1  a
Look under Limit Evaluation in the Index for similar problems.
519667
Assume that the set {h} is minimal in the sense that none of the intervals is
a proper subset of the union of some others. Then we can order the intervals
by the value of their leftendpoint, so that Ii begins to the left of IiH for
all i. Then we have I
j
and 1)+2 are disjoint for each j = 1, ... , n  2;
otherwise Ij+1 c 1j U 1)+2, or 1)+2 C 1jH' either of which violate the
minimality assumption.
Thus, the set of intervals with odd subscripts consists of mutually
disjoint sets, as does the set of intervals with even subscripts. At least
one of these two sets covers 1/2 of I, since their union covers all of I.
519668
Choose a person P. If P is a friend of k other people, then P is a stranger
of 5  k other people, and one of k or 5  k must be at least 3. Without
loss of generality, assume that P is a friend of at least 3 other people:
Q1, Q2, Q3. If Q1, Q2, Q3 are all pairwise strangers, then we are done.
Otherwise, Q i and Q j are friends (for some i, j). In this case, P, Q i, Qj
forms a triple of pairwise friends.
Look under Enumeration in the Index for similar problems.
Exam #21 967
519671
In a symmetric matrix, all nondiagonal entries are paired with another;
the two must be the same. There are n( n  1) such entries. If each row is
a permutation of the set {I, 2, ... , n}, then there must be n copies of each
number in the matrix and, since n is odd, at most n  1 copies of each
number can paired off. The remaining copy of each number must be on
the diagonal of the matrix, implying that the diagonal is also a permutation
of the set {1,2, ... ,n}.
Look under Matrix Algebra in the Index for similar problems.
56 Solutions
519672
Without loss of generality, assume that the axes of the parabolas lie parallel
to the xaxis; let (Xl, Yl) and (X2' Y2) be the points of contact of the
common tangent with the parabolas. Then the equations for the parabolas
can be given by
and
y2 + 2B
2
x + 2C
2
y + D2 = o.
Using implicit differentiation, we have the tangent line to parabola
through (Xi, Yi) is given by
YYi + Bi(X + Xi) + Ci(y + Yi) + Di = o.
Since both of the points (Xl, Yl) and (X2' Y2) lie on each of these lines,
we have
and
Subtracting equation (2) from equation (1), we obtain
2(Bl  B
2
) (Xl; X2 ) + 2(C
l
_ C
2
) (Yl ; Y2) + (Dl  D
2
) = 0,
that is, the midpoint lies on the line
which is the common chord of the parabolas. (The common chord is the
line segment connecting the two points of intersection of the parabolas.
That this chord exists is implicitly assumed here.)
Look under Analytic Geometry in the Index for similar problems.
519673
Let f (a, b) = e
a
+ b(log b  1)  ab, for a ;::: 0 and b ;::: 1. If f ( a, b) ;::: 0,
then the result follows. Note that if b = e
a
, then f(a, b) = O.
Further, since
we have
of
ob = 10gb  a,
of =O·fb= a
ob 1 e ,
of 0 ·f b a
ob < 1 < e ,
Exam #21967
and
of O'f b a
ob > 1 > e .
57
Therefore, fixing ao greater than or equal to zero, consider the function
f(ao, b), for 1 :::; b:::; 00. From the facts about of job above, f(ao, b) has
an absolute minimum at b = e
ao
. Further, since f (ao, e
ao
) = 0, we have
f(ao, b) 2: 0, for 1 :::; b:::; 00. Since ao was arbitrary, the result follows.
Look under RealValued Functions in the Index for similar problems.
519674
Suppose the proposition false and let n be the smallest positive integer
such that the number of odd coefficients in the expansion of (x + y)n is
not a power of 2. By inspection n is not 1, nor is n equal to 2.
Case 1. Let n = 2k, with k a positive integer. If m is odd, then
(
n) = nm+1( n )
m m m1
is an even integer since n  m + 1 is even. Thus all the odd coefficients
must occur for even values of m. Let m = 2q. Then (;:,) is congruent
(mod 2) to
2x4x .. ·xn
2 x 4 x ... x m x 2 x 4 x ... x (n  m)
by removing all odd factors and divisors. (That is, if x and yare congruent
(mod 2) then so are hx and ky for any odd integers hand k, and
conversely.) But the latter fraction is evidently thus (;:,) = is
odd if, and only if, is odd and so the number of odd coefficients in
(x + y) n is the same as the number of odd coeficients in (x + y) k, which
contradicts the assumption that n was the least positive integer for which
the proposition fails.
Case 2. Let n = 2k + 1, with k a positive integer. Since (;:J =
and, since nand n  m are not congruent (mod 2), the number of
odd coefficients is twice the number of odd coefficients obtained by
considering only even values of m in the symbols (;:,). But as before
we find
1
) == (mod 2). Thus the number of odd coefficients in
(x + y)n, is just twice the number of odd coefficients in (x + y)k, again
giving a contradiction. Thus the proposition is true.
Look under Enumeration in the Index for similar problems.
58 Solutions
519675
Suppose that z is a root and Izi < 1/3. Then
implies that
or
lalz + a2z
2
+ ... + aNzNI = l.
But using the triangle inequality, the requirement that Ian I < 2, and
Izi < 1/3, we obtain
a contradiction.
Look under Complex Numbers or Polynomials in the Index for similar
problems.
519676
In the illustration below, the distance from A to F, and the distance from
B to G are each V65/8. The distances AG, AH, AT, BH, and BJ are
all larger. If the square is divided as illustrated, each piece has a diameter
of V65/8.
N M
o
Q
R
A .! B
8
3
8
L K J
H
G
F
c D E
Exam #21967 59
Now suppose that the square is divided into three arbitrary pieces. One
of the three pieces must contain at least six of the sixteen boundary points
symmetrically arranged in the illustration. We will now show that this piece
has a diameter of at least V65/8.
Suppose the set contains a comer point, WLOG point A. If the set
contains any of points F, G, H, J, K, L, or M, the set has diameter at
least V65/8. If the set contains none of those points, but contains, WLOG,
comer E then the set must also contain one of N, P, Q, or R, and again
the diameter is at least V65/8. Finally, if the set contains none of E, F,
G, H, J, K, L, M, or N, then the set must contain one or both of D and
P, and one or both of C and Q. Again the diameter is at least V65/8.
Suppose the set contains none of the comer points, but does contain a
nonmidpoint of one of the sides, WLOG, say B. If the set also contains
G, H, K, or L, the set has diameter at least V65/8. If the set contains
none of those points, but does contain M, then it must contain one of C,
D, or F. Again the set has diameter at least V65/8. If the set doesn't
contain M either, it may contain P. In this case, it must also contain D or
F, and hence has a diameter of at least V65/8. This leaves the possibility
of the set consisting of B, C, D, F, Q, R, in which case it has a diameter
of at least V65/8.
This leaves the possibility of a set consisting solely of the midpoints,
but there are only four, so this can't happen.
Look under Geometry in the Index for similar problems.
519677
First note that for a > 0, y > 0, the equation x + x
a
= y has a unique
solution with x > 0. This is because x + x
a
is a continuous, monotonic
increasing function of x with limxtoo x + x
a
= 00 and is equal to zero
at x = 0. Thus it assumes every positive real value exactly once for
x E (0,00). Now, if the sequence {x
n
} satisfies the conditions given in
the problem, then
00 00
xn = L xj = Xn + 1 a + L xj
j=n+l j=n+2
= X ~ + l + Xn +1·
So Xn+1 must be the unique solution of this equation. If follows that the
sequence {x
n
} is unique. Further, if we define Xn inductively by setting
60 Solutions
Xn+l equal to the solution of x + x
a
= x
n
, then it is clear that
xn = + Xn+l
= + + Xn+1 = ...
co
= L xj.
j=n+l
Look under Infinite Series in the Index for similar problems.
519678
Suppose that p, q E R
2
and let m be the midpoint of p and q. Let 5 be
the line segment of length 1 which is perpendicular to pq and has m as
its midpoint. By symmetry, for every S E 5, Ip  sl = Iq  sI Now, for
each n E N, let Sn E 5 be such that Ip  snl E Q and 1m  snl < lin.
Such a choice is possible because Q is dense in Rand Ip  xl is a
continuous function of x. So as x varies along 5 in a of
m, Ip  xl assumes rational values. Now, as n 00, Sn m, hence
Ip  snl + Iq  snl Ip  ql, and we see that
IT(p)  T(q)1 ::::; IT(p)  T(sn)1 + IT(q)  T(sn)1
= Ip  snl + Iq  snl·
Letting n 00, we have IT(p)  T(q)1 ::::; Ip  ql·
Now, suppose c > 0. We can choose q' E R2 such that Iq  q'l < c and
Ip  q'l E Q. As a consequence of the previous argument, we have that
IT(q)  T(q')1 ::::; Iq  q'l· Also, we can invoke the triangle inequality to
see that IIT(P) T(q)IIT(p) T(q')11 ::::; IT(q') T(q)1 ::::; Iqq'l < c.
IIT(p)  T(q)1  Ip  qll
::::; IIT(p)  T(q)Ilp  q'll + lip  q'Ilp  qll
= IIT(p)  T(q)IIT(p)  T(q')11 + lip  q'Ilp  qll
::::; IIT(p)  T(q)IIT(p)  T(q')11 + Iq  q'l
< 2c.
Letting c 0, we see that IT(p)  T(q)1 = Ip  ql for all p, q E R2.
Look under Geometry in the Index for similar problems.
Exam #31968 61
Exam #31 96 S
519681
Since the set of Riemann sums is bounded, there exists an M such that
n
M;::: I L:!(ti)(Xi XiI)I for all ti,xi,Xil.
i=1
Thus we have
implying
that is, that ! is bounded.
Look under Riemann Sums in the Index for similar problems.
519682
Assume the claim is false for quadrilateral ABC D and five interior points.
If P and Q are two of the five interior points, then the line through P and
Q cannot intersect ABC D in adjacent sides. To see this, suppose, without
loss of generality, that the line intersects AB and BC. Then the points A,
C, D, P, and Q form a convex pentagon. Therefore, exactly one of the
rays PQ or QP intersects either AB or BC. Let S AB be those rays PQ
intersecting AB and S BC be those rays PQ intersecting BC, where P
and Q range over the five interior points. Thus S AB U S BC contains ten
rays, one for each pair of points P, Q. If rays PQ and QR are both in,
say, SAB, then either the points P, Q, R, B, and C or the points P, Q, R,
A, and D form a convex pentagon, depending on which side of line P R
the point Q lies. It follows furthermore that if PQ and PR or if QP and
RJ3 are both in, say, SAB, then QR or RQ must be in SBC. Without loss
of generality, assume S AB contains at least five rays. Because no point can
be both the initial and second point of such rays, it follows that some point
P is the initial or second point in at least three of these rays. Let the other
points of these rays be Q, R, and S. The previous discussion showed that
the rays QR or RQ, QS or stJ, and Ri; or sR must be in S BC. However,
this possibility is also precluded by the previous discussion, proving the
claim.
Look under Geometry in the Index for similar problems.
62 Solutions
519683
Fix t and define L : R
n
It R
n
by L(x) = tx. Then L is differentiable
on Rn, with L'(x) = tI, where I is the identity matrix. Now let F(x) =
f(L(x)), so that by the chain rule we have F'(x) = I'(L(x))L'(x) =
tf(tx). By the definition of f, F(x) = tf(x), so that F'(x) = tf'(x)
also. Therefore l' (tx) = l' (x) for any t > O. Therefore l' (x) is constant
and f ( x) must be linear.
Look under Multivariate Calculus in the Index for similar problems.
519684
The only integral solutions to the first equation are (1,0) and (2,1). The
only solutions to the second are (2,3) and (1,1). This is a special case
of Catalan's Conjecture and a general exposition of the progress that has
been made on the conjecture can be found in MAA FOCUS, vol 21 #5.
Look under Diophantine Equations in the Index for similar problems.
519685
This conjecture is false. Let
fn(x) = {nIx, on[1,e
n
],
0, otherwise.
Since 0 ::::; fn(x) ::::; ~ , for all x, fn 7 0 uniformly on R. Now
I: fn(x) dx = 1
so that limn+oo J ~ o o f n = l.
Look under Infinite Series, Limit Evaluation or Riemann Sums in the
Index for similar problems.
519686
Consider any two sequences {c
n
} and {d
n
} each tending monotonically
to zero, with d
n
> en, and L: C
n
converging while L: d
n
diverges.
Since J;i d
n
diverges, for each n there exists an integer ¢( n) such
that L:j=: d
j
2: e where e is any fixed positive number. Define the
sequence nk recursively by nI = 1 and dnk+l < cq,(nk)' It is clear that
nk < ¢(nk) < nkH'
Exam #31968 63
Now a solution to the problem can be given by the two series with terms
{an} = {Cnl , Cnl +1,"" c</>(nt},dn2 , d n2+1 , ... , d</>(n2)' Cn
3
,···,
C</>(n3)' d n4 ,·· .},
{bn } = {d
n"
d n, +1, ... , d</>(n,), C
n2
, C
n2
+1, ... , C</>(n2)' d
n3
, ... ,
d</>(n3)' Cn4 ,·· .}.
Clearly, each of the series z= an and z= b
n
diverge since they con
tain infinitely many stretches of terms adding up to more than e, and
z= min(a
n
, b
n
) converges by comparison with z= Cn·
Look under Infinite Series in the Index for similar problems.
519687
f(z, B) = max{lzl cos(B  arg z), O}. Then
1 J7I" Izl J7I"/2 Izl
 f(z, B) dB =  cosBdB = .
27f 71" 27f 71"/2 7f
Therefore,
J7r f(z·,B)dB = t hl
J 7f 7f
71" j=1 j=1
Because the maximum of a function is at least its average, it follows that
there exists B such that
1 n
7f j=1
L I Zj I cos( B  arg Zj)
cos(Oarg Zj
L IZjlcos(argzjB)
cos(Oarg Zj
=Re
z·e
iO
<
J 
cos(Oarg Zj cos(Oarg Zj
L Zj.
cos(Oarg ZJ
Look under Complex Numbers in the Index for similar problems.
64 Solutions
519688
This problem asks you to look for the Ramsey number n = r(k, k). We
will give an existence result for the more general problem of finding r (s, t),
that is, for what n can we find a set of s blue arcs or a set of t arcs. In
particular, we show that
r(s, t) :::; r(s  t, t) + r(s, t  1).
Working inductively on k = s + t, we have the result holding for k = 4
by inspection. Now suppose the result holds for all values smaller than k.
Let n = r(s  1, t) + r(s, t  1) be the number of points that we have,
and let each arc be colored blue or red. Let v be any point. There are either
r(s 1, t) blue arcs or r(s, t 1) red arcs at v. If there are r(s 1, t) blue
arcs, then the set of points at the other end of these arcs contain s  1 blue
arcs or t red arcs among them, by our induction hypothesis. In the first
case these blue arcs, plus those from v to these points, form a collection
of s blue arcs, as required. In the second case, we already have t points
all of whose arcs are red.
A similar argument holds if there are r (s, t  1) red arcs at v.
Look under Enumeration in the Index for similar problems.
Exam #41969
519691
We will show that
x
2
f(x) = sin x  x +  2: °
Jr
on [0, Jrj. By observation of f' (x) = cos xI + 2:, we find that 0, ~ ,
and Jr are critical points of f(x). By noting that
f
"( ) . 2
x =  SIll X +
Jr
has exactly two zeros on [0, Jrj, we can apply Rolle's Theorem to verifY
that 0, ~ , and Jr are the only critical points of f (x). (If f' had another zero,
f" would have a third zero.) Since f(O) = f(Jr) = ° and f ( ~ ) = 1  ~ ,
we can see that the minimum value of f(x) on [0, Jrj is 0.
Look under RealValued Functions in the Index for similar problems.
Exam #41969
519692
Suppose that
P + Q + R =A.
p q r
Multiply through by pqr yielding Pqr + Qpr + Rpq = Apqr. Now,
Pqr = Apqr  Qpr  Rpq = p(Aqr  Qr Rq),
65
hence p divides Pqr. But since (p, q) = 1 and (p, r) = 1, it must be that
p divides P. In the same manner, q divides Q and r divides R.
Look under Number Theory in the Index for similar problems.
519693
Let w = {h5 + 5J20 and w
8 = W + w. We have
V 25  5J20. We wish to determine
8
3
= w
3
+ w
3
+ 3ww(w + w) = 50 + 3·5·8.
To solve for 8 we note that 0 = 8
3
 15  50 = (8  5)(8
2
+ 58 + 10) has
one real solution, 8 = 5.
Look under Number Theory in the Index for similar problems.
519694
Label acceleration, a, downward at all points along the trajectory of height
h, and 9 the acceleration due solely to gravity. Let v be the velocity of the
ball at any time, in the general direction of travel (downward or upward)
and Vo be the upward velocity with which the ball is launched.
At launch, a = 9 + av, so v continually decreases until the highest
point of ascent, where v = 0 (by Rolle's theorem), so a = g. Thus, going
upward, the ball will always experience an acceleration downward greater
than g. The ascent will take the same amount of time as an object travelling
on a straight line with initial velocity v = 0 and initial acceleration a = g,
with a continually increasing through the distance h, to a final value of
a = 9 + avo. Since a is continually increasing, the amount of time to
traverse this distance, t
up
, is less than t, the amount of time for an object
with uniform acceleration 9 to traverse this distance.
Going downward, the acceleration begins with a = 9 and continues
with a = 9  av (since the air resistance is now upward, but the velocity
is downward). So the descent will take the same amount of time as
66 Solutions
an object travelling on a straight line with initial acceleration a = g,
with a continually decreasing through the distance h to a final value of
a = 9  aVj. Again, tdown, the time required to traverse this distance,
is greater than t, the time an object under uniform acceleration 9 would
require. Thus, tup < t < tdown.
Look under Differential Equations in the Index for similar problems.
519695
Let B denote the angle just above A. Using the law of sines to compute
the left side of the equilateral triangle, we find its area to be
v'3 [bsin(27f/3B) + CSin(B+A7f/3)]2
4 v'3/2 v'3/2
= ~ [bsin(27f/3  B)  csin ((27f/3  B)  (A + 7f/3)) r
= ~ [(b sin(27f /3  B)  c cos(A + 7f /3)) sin(27f /3  B)
+csin(A + 7f/3) cos(27f/3  B)r.
Now, if w is such that cosw = x/ Jx
2
+ y2 and sinw = y/ Jx
2
+ y2,
we may write x sin a + y cos a as J x
2
+ y2 sin( a + w), from which we
see that its maximum value is J x
2
+ y2. It follows that the maximum
area of the equilateral triangle is
v; [(b  ccos(A + 7f /3)) 2 + (csin(A + 7f /3)) 2]
v'3
= """4 (b
2
+ c
2
 2bccos(A + 7f/3)).
The claimed upper bound has been established. However, it need not
be the maximum value. To conform to the geometric setup, B must satisfY
° < B < 27f/3 and ° < 7f  (B + A) < 27f/3, which is equivalent to
7f/3  A < B < 7f  A. For instance, for b = 10, c = 1, and A = 7f  .1,
the maximum occurs for B outside this range.
Look under Geometry in the Index for similar problems.
Exam #51970 67
519696
From the Trapezoidal Rule we have
I
b
f(x) dx = b  a (f(a) + 2 ~ f(a + kh) + f(b)) _ (b  ~ 3 f"(c)
a 2n ~ 12n
k=l
for some c E [a, b]. If we set n = 1, the result follows immediately. A
complete proof of the Trapezoidal Rule itself can be found many places.
Look under RealValued Functions in the Index for similar problems.
Exam #51970
519701
We have
. 1 1 1. n 1
hm ( + + ... +) = hm ~  
n+oo n + 1 n + 2 2n n+oo ~ n + k
k=l
n (1 1)
 lim .
 n+ooL n Is. + 1
k=l n
1
1 1
= dx = In(2).
o 1 + x
Look under Limit Evaluation in the Index for similar problems.
519702
If lei :::; 1, then we are done. If Icl > 1, then we have
n n1 n2
C = an1C  an2c  ...  a1C  ao.
Divide by c
n

1
to obtain
Take absolute values and use the triangle inequality to see that
Since Icl > 1, we have the desired inequality.
Look under Polynomials in the Index for similar problems.
68 Solutions
519703
Solution 1: Without loss of generality, we may assume that the coordi
nates of the vertices of the parallelogram are A: (0,0), B: (1,0), C: (a, b),
D: (a + 1, b + 1). Then the coordinates of X, Y, Z, Ware
Y:(1+a;b, b;a),
w:(a;b, a;b).
Then XZ = Ja
2
+ (b+ 1)2 = YW.
We also have XZ = (a, b + 1) and YW = (b + 1, a). Thus, XZ·
YW = 0 and XZ .l YW.
Since the diagonals of a parallelogram bisect each other, the length of
each side is the hypotenuse of an isosceles right triangle with legs of length
J a
2
+ (b + 1)2 /2 and XY ZW must be a rhombus.
To see that the rhombus is a square, let E denote the intersection of X Z
with YW. Then mLW XY = mLW X E + mLY X E = a right angle.
Similarly, the other angles of XY ZW are right angles, and XY ZW is a
square.
Solution 2: Let YW and X Z meet at P. Then P is a center of
symmetry for the figure, so a 180° rotation maps the figure to itself.
Hence WXYZ is a parallelogram. Next, ZC = ZD and CY = DW.
Also LCZD = 90°, LZCB = LZDW = 90° + the acute angle of the
given parallelogram. Hence triangles ZCY and ZDW are congruent and
similarly oriented, so a 90° rotation carries one triangle into the other. Thus
ZY = ZW and LY ZW = 90
0
• Thus W XY Z is a square.
Look under Geometry in the Index for similar problems.
519704
The problem is trivial for e = 2, so we proceed by induction. Suppose
the result is true for all sets of points of order e, where e = k, k even.
Consider a set of points and arcs of order e = k + 2. Pick any arc A, say
from PI to P
2
and remove it together with its endpoints. By our induction
hypothesis, we can color the remaining regions with two colors.
Because of the nonintersection constraint on the arcs, PI and P
2
must
either be consecutive points around the circle, or be situated such that arc
A divides the remaining arcs into two sets, as illustrated below. In the first
Exam #61971 69
case, color the region created by arc A opposite of the color of the region
it is contained in. In the second case, color one of the two regions created
by arc A the other color (from what is was before) and switch colors for
all of the other regions on that side of arc A.
Look under Geometry in the Index for similar problems.
519705
Let d be the number of digits in n, then it is necessary that 6· 10
d

1
+
(n  6)/10 = 4n (from condition (b)). Multiplying through by 10 yields
6·lO
d
+(n6) = 40n (note here that n will also satisfY (a)) or equivalently
6.10
d
62(10
d
1)
n=
39 13
Hence it is necessary that 13 divide 10d  1. So we need
d1
10d  1 = 9 L 10
k
== 0 (mod 13).
k=O
Since gcd(9, 13) = 1, this is equivalent to
d1 d1
L 10
k
== L( _3)k == 0 (mod 13).
k=O k=O
Now we may simply add alternating powers of 3 until we reach zero:
(3)0 = 1,1 + (_3)1 = 2, 2 + (_3)2 = 7, 7+ (3)3 = 20 = 7,
7 + (3)4 = 74 = 9, 9 + (3)5 = 9  243 = 234 = 18 ·13 = O. So
d  1 = 5 =} d = 6. Thus if n = 2(10
6
 1)/13 = 153, 846 satisfies (a)
and (b) then it is the smallest such number. Indeed, since d is the correct
number of digits, (a) and (b) are true by construction.
Exam #6197 1
519711
For part (a) we have
(i) x E A =} xES =} xRx (with x E A) =} x E A, hence A c A.
 
(ii) By (1), A cA. Now, x E A =} :3y E A S.t. xPy. y E A=} :3z E A
S.t. yPz. xPy ~ y P z =} xpz (with z E A), so x EA. It follows that
A c A, hence A = A.
70 Solutions
(iii) x E Au B =} ::ly E A u B s.t. xPy. If yEA, then x E A
and if y E B, then x E B. Therefore AU B c Au B. Also,
x E Au B =} (x E A) V (x E B). So one of A or B contains a
y s.t. xPy. It follows that x E Au B and hence AU B = Au B.
(b) A = ((u,v) E S : v  y = 3(u  x) for some (x,y) E A} =
{(u, v) E S : v  3u = y  3x 1\ x
2
+ y2 = I}. First, we determine the
values that y  3x can take when x
2
+ y2 = 1. Thus we look at the minima
and maxima of the two functions VI  x
2
 3x and VI  x
2
 3x on
[1,1]. Since these functions are continuous, they will assume every value
between their respective minimum and maximum.
~ ( ± ~  3x) = =f x  3 = 0 =} 9(1  x
2
)
dx Vl x
2
= x
2
=} X = ±_3_.
v'lO
Checking the values x = ±1, ± Jk in ±Vl  x
2
 3x yields a
minimum and maximum of  v'lO and v'lO, respectively. Hence y  3x
assumes every value on [v'lO, v'lO]. Now we can write
A = {(u,v) E S: v  3u E [v'lO, v'lO]}
= {(u,v) E S: 3u  v'lO::; v::; 3u+ v'lO}.
It follows that A is the strip between the two parallel lines v = 3u  v'lO
and v = 3u + v'lO.
Look under Algebraic Structures in the Index for similar problems.
519712
We want to compute 7
9999
(mod 1000).
7
9999
== 7
1
(7
400
)25 == 7
1
(10
3
k + 1)25 == 7
1
• 1 == 143 (mod 1000).
So the last three digits are 1, 4, 3.
Look under Number Theory in the Index for similar problems.
519713
The derivative of (1  e ,,1 )1 is
1
ex
Exam #61971 71
Note, however that this is valid only for x =I 0; worse yet, (1  e ~ l )1
has a jump discontinuity at x = 0.
(b) (1e ~ l )1 is not an anti derivative of the integrand on the interval
[1, 1], but the function 9 is, where
{
(1  e ~ l )1 for x > 0,
g(x)= 1 for x =0,
1 + (1  e ~ l )1 for x < 0.
Hence the integral is g(1)  g( 1) = e':l'
A more direct approach would be to express the integral as the sum
of two integrals, one on the interval [1,0] and the other on the interval
[0,1].
Look under Differentiation or Integration in the Index for similar prob
lems.
519714
For each t E R, let L
t
be the line y(x) = t. Clearly the set B =
{Lt : t E R} is uncountable, Ltl n Lt2 = 0 for t1 =I= t2, and UtER Lt =
R
2
. It follows from the latter two statements that for any a E A, there is
exactly one line L
t
for which a E Lt. Let f : A ~ B be the function
which takes points in A to the line in B which contains them. Since A
is countable and B is uncountable, f cannot be smjective. It follows that
there exists an L
t
E B which does not contain a point of A.
Look under Enumeration in the Index for similar problems.
519715
For each positive integer k,
Lk/2J
(2 + Y'2)k + (2  Y'2)k = f; C:) 2
k
i+\
which is an integer; but ° < (2  J2)k < 1. Therefore, the fractional part
of of (2 + J2)k is 1  (2  V2)k, and its limit is 1.
(b) A similar calculation shows that the fractional part of
if k is odd,
if k is even.
72 Solutions
The first of these has limit 0 and the second the limit 1. Hence the fractional
part of of (1 + y'2)k has cluster points at 0 and 1.
Look under Limit Evaluation in the Index for similar problems.
519716
For each group of five robbers, there must be at least one lock to which
none has the key, and for any two groups, these locks must be different
(else the addition to one group of a new robber from the other would
provide a majority no member of which contains a key for such a lock as
we have associated with the first group). Therefore there are at least as
many locks as there are groups of five robbers, namely C;) = 462.
Now consider anyone robber. For each group of five among the
remaining 10 robbers, he must have a key to each lock to which they
collectively do not. Since each group has at least one such lock, he must
have at least as many keys as there are such groups, namely C50) = 252.
In order to show that no more locks or keys are required, we now
demonstrate that 462 locks and 252 keys are sufficient. In accordance with
any onetoone correspondence between a set of 462 different locks and the
collection of sixmember subsets of the robber band, we give keys for any
one lock to each member of its associated subset and to no other robber.
Each minority, being a subset of the complement of such a sixmember
subset, cannot open the lock associated with that subset. On the other
hand, consider any majority and any lock. Since only five robbers do not
have the key to that lock, at least one of the robbers in the given majority
does have the key.
Look under Enumeration in the Index for similar problems.
519717
We define three new items from the given operations as follows:
0=Xn*X, forallX,and
A C B iff An B = A.
With these definitions and the axioms given we can show, after a consider
able amount of simple but lengthy work, that we do indeed have a Boolean
Algebra.
Exam #71972 73
In fact, the converse is also true. If you start with the standard axioms
for a Boolean Algebra, you can derive (a), (b), and (c). Therefore, the
conditions of the problem give an alternate characterization of a Boolean
Algebra.
Look under Algebraic Structures in the Index for similar problems.
Exam #71972
519721
The following calculation shows that it is sufficient that A + B and A  B
be invertible. Supposing the existence of X and Y such that the desired
conditions hold, the equations
AX+BY=C,
BX+AY=D
yield (A + B)(X + Y) = C + D. Subtracting the second from the first
yields (A  B)(X  Y) = C  D. If A + B and A  B are invertible,
then
X  Y = (A  B)I(C  D),
X + Y = (A+B)\C+D),
and from these equations X and Y can be found. The only conditions are
that A, B, C, and D are the same size.
Look under Matrix Algebra in the Index for similar problems.
519722
From the given equation
I = A  2A 
22'
so AI exists and is equal to  2A 
519723
Let f(x) = lnx/x. Then f'(x) = (1  Inx)/x
2
and f'(x) = 0 when
lnx = 1, or when x = e. Since f'(x) changes sign from + to  as x
passes through e, f has a relative maximum at x = e. Further, it is an
74 Solutions
absolute maximum since f(x) t 0 as x t 00, and f(x) t 00 as
x t O. Thus f(7r) < f(e), so In7r/7r < Ine/e whence 7r
e
< e
7r
•
Look under Matrix Algebra or RealValued Functions in the Index for
similar problems.
519724
There exists N such that for all n > N, sufficiently large, 0 < an < 1/2.
Thus for n > N, a
n
/(l  an) < 2a
n
. So
00 N 00
0 < " ~ < " ~ + 2 " an
L....1a
n
L....1a
n
L....
n=1 n=1 n=N+I
and the series converges.
Look under Infinite Series in the Index for similar problems.
519725
Let P denote the permutation made by the machine. We are given
p2 = (A, 10, J, 6, 3, Q, 2, 9, 5, K, 7,4, 8)
in cycle notation. Since p2 is a cyclic permutation, P is also. Therefore
pI3 is the identity permutation. Thus
p = (p2)7 = (A, 9, 10,5, J, K, 6, 7, 3,4, Q, 8,2).
Look under Permutations in the Index for similar problems.
519726
The curves are identical, since (r, B) and (r, 37r  B) are different
coordinates for the same point. If r = cos( B /2), ilien
37rB 3 .3. .
r = cos 2 = cos 27rcos(B/2) + sm 27rsmB/2 =  smB/2.
519727
Suppose G = H U K with Hand K proper subgroups. Since H is proper,
there exists h E H such that h rt K. Similarly, there exists k E K such
that k rt H. Let 9 = hk. If 9 E H, then k = hIg E H, which is
impossible; if 9 E K, then h = gk
I
E K, which is also impossible. Thus
G cannot be H U K.
Look under Group Theory in the Index for similar problems.
Exam #81973
Exam #81973
519731
Using L'Hopital's rule,
e
1
/
x
 1
lim x(e
1
/
x
 1) = lim
x+oo x+oo 1/ x
. e
1
/
X
( 1/x
2
)
= hm = lim e
1
/
x
= 1.
x+oo 1/ x
2
x+oo
Look under Limit Evaluation in the Index for similar problems.
519732
75
Let a be the angle shown in in the figure below, with a variable and ()
fixed.
o
Let A be the area of rectangle ABCD. Since cot () = ~ ~ and BC =
OC sin a = sin a, We have
AB =OBOA
= OC cos a  AD cot ()
= cos a  BC cot ()
= cos a  sin a cot () .
Thus the area, A, of the rectangle is given by
A=BC·AB
= sin a cos a  sin
2
a cot ()
= ~ sin 2a  sin
2
a cot (),
76
which we differentiate to obtain
dA = cos 200  sin 200 cot O.
dr
Solutions
Setting this equal to 0, we find a = 0/2. Substituting into our area function,
we get
A= IcosO.
2sinO
Look under MaxIMin Problems in the Index for similar problems.
S 19733
Observe that the maximum occurs when no ni = 1. Further, (m2)·2 > m
if, and only if, m > 4. Thus, each ni > 4 can be replaced by (ni  2) + 2 in
the partition and the associated product will be increased. If some ni = 4,
replace it by 2 + 2, leaving the product unchanged. If three 2's occur
replace them by 3 + 3, increasing the product. Thus the maximum occurs
when each ni is a 2 or a 3, with at most two 2's occurring. Indeed, the
maximum is:
n  1 . 2i/2 where
33 )
S 19734
{
i = 4 if n = 1 (mod 3),
i = 2 if n = 2 (mod 3),
i = 0 if n = 0 (mod 3).
Integration by parts with u = 1/(1 + y) and dv = yn
1
dy gives
1
1 nyn1 yn 1 11 yn
dyI + dy
o 1 + y  1 + y 0 0 (1 + y)2 .
Since
the integral on the righthand side of the last equation approaches zero as
n approaches infinity. The desired limit is thus
lim (LI1) = 1/2.
n+oo 1 + Y 0
Look under Integration or Limit Evaluation in the Index for similar
problems.
Exam #81973 77
519735
The quotient A: B is an ideal of R. If a E A : B, then Vb E B,
ab E A=} (ab) E A =} (a)b E A,
hence a E A: B. Ifx,y E A: B, then
Vb E B, xb, yb E A =} xb + yb E A =} (x + y)b E A,
hence x + yEA: B. Finally, if a E A : B, r E R, then Vb E B,
ab E A =} r(ab) E A =} (ra)b E A so that ra E A : B (in particular,
° E A : B). It follows that A : B is an ideal.
519736
Let f(a) = a
n
+
1
 n(a 1)  a. Then
d
daf(a) = (n + l)a
n
 (n + 1).
Since f(l) = f'(1) = 0, f(a) is divisible by (a  1)2.
Look under Polynomials in the Index for similar problems.
519737
We will prove the contrapositive: If A or B is singular, then the product
AB is singular. Let B be singular. Then for some x =1= 0, we have Bx = O.
It follows that
A(Bx) = (AB)x = 0
implying that AB is singular. Now let A be singular. Then for some x =1= 0,
Ax = O. If B is nonsingular, there is a nontrivial solution to By = x.
Hence
Ax = A(By) = (AB)y = 0,
from which we see that AB is singular.
Look under Matrix Algebra in the Index for similar problems.
519738
Let x = a  d, y = d, and z = a + d. Then
(ad)2+d
2
= (a+d)2,
or
78 Solutions
or
a
2
= 4ad.
Thus a = 0 or a = 4d, and all solutions are (x, y, z) = (d, 0, d) or
(3d, 4d, 5d), where d is any integer.
Look under Number Theory in the Index for similar problems.
519739
There are ( ~ ) = 10 equiprobable ways to choose 3 men from the group.
Exactly 3 of those choices result in 2 Republicans being chosen (2
Republicans and the choice of lout of 3 Democrats). It follows that the
probability of both Republicans being selected is 3/10.
Look under Probability in the Index for similar problems.
Exam #91974
519741
(a) For a polynomial p(x), let [x
7n
Jp(x) denote the coefficient of x
7n
in
p(x). First, by the binomial theorem,
From this, we see that
It follows that
[X7nJ(XlOOOk(X + l)k) = [x
7n
(lOOOk»J(x + l)k
= (m ( 1 ~ 0 0  k)) = Coo: m).
Now, the polynomial given in the problem is equal to
1000
L xlOOOk(x + 1)k,
k=O
Exam #91974
so
Therefore the coefficient of x
50
is C ~ g l ) .
(b) From part (a), we see that the sum of all the coefficients is
1000 1000 ( )
fo[X
m
] = fo 1 ~ 1 = 2
1001
_l.
Look under Polynomials in the Index for similar problems.
519742
We have
yx = (yx)3 = (yx)(YX)2 = (yx?(yx)
= ((yx)2y)X = y(yx)2x = y(yx)(yx)x
= xy2yx
2
= xyx
2
= x
3
y = xy.
Look under Group Theory in the Index for similar problems.
519743
If m > 1, set
Then we have
(m
2
+ l)(ni + 1) = (m
2
+ 1) ((m  1)2 + 1)
= (m
2
+ 1)(m
2
 2m + 2)
= (m
4
 2m
3
+ 3m
2
 2m + 1) + 1
=(m
2
m+1)2+1
= Nl + 1,
79
so
(m
2
+ 1 ) ( n ~ + 1) = (m
2
+ 1)((m + 1)2 + 1)
= (m
2
+ 1)(m
2
+ 2m + 2)
= (m
4
+ 2m
3
+ 3m
2
+ 2m + 1) + 1
=(m2+m+1)2+1
= N ~ + 1.
Solutions
If m = 1, then n1 = 2,N1 = 3,n2
solutions.
17 are the smallest
S 19744
The point (0,0) is neither a minimum nor a maximum. We have f(O, 0) = 0;
let x = 8 and y = ~ where 0 < E < 8/2. Then f(x,y) < O. But
f(x,O) > 0 for all x.
Look under Multivariate Calculus in the Index for similar problems.
S 19745
Assume that y is never zero on [0,3/4]. We note that ~ < 1, so
y(3/4) > 0 means that y(x) > ~  x on [0,3/4]. Therefore
dy x x (3 )
 < 1 e y < 1 e   x .
dx   4
We see that
y(3/4) = 1 +  dx < 1 1 + eX   x dx = 2  e
3
/
4
.
1
3/4 dy 1
3
/
4
(3)
o dx 0 4
Since In(2) R::: .7, 2  e
3
/
4
< O.
(Or, 2
4
/
3
= 16
1
/
3
< 17.576
1
/
3
= 2.6 < e.)
Thus, y(3/4) < 0, contradicting our assumption. Hence, y must have a
zero on [0,3/4].
Look under RealValued Functions in the Index for similar problems.
S 19746
This conjecture is clearly false. Let an = 1/(n + 1), then
Exam #101975
(the harmonic series) diverges while
Look under Infinite Series in the Index for similar problems.
Exam #101975
519751
The result is obvious if a = 0, so consider a > O. Also let
a
f(x)= 2 2'
a +x
We see, using a lower Riemann Sum, that
00 a 1
00
a 7r
0<'" < dx=
a
2
+ n
2
a
2
+ x
2
2
n=l 0
81
and the result follows. If a < 0, the infinite sum is an odd function of a,
so the result follows in this case also.
Look under Infinite Series in the Index for similar problems.
519752
First we show, by way of contradiction, that no two of the vertices can
be joined by a diameter through the center of the circle. Let V be the
set of vertices, with [VI = n, odd. Suppose that Vo and are two
vertices joined by a chord through the center of the circle. This chord
partitions V  {vo, into two disjoints sets, V
l
and V
2
, of orders a and
b respectively. We have a + b = n  2, and without loss of generality, a is
odd and b is even.
Label the vertices of V
l
as vi, vi, ... , VI as we move around the circle
from Vo to Label the vertices of 112 as ... , as we again move
around the circle the other way from Vo to
Now each of the at
l
triplets
(
1 2) (2 3 4) (al a ')
VO'Vl'V
l
' Vl,Vl,V
l
, ... , V
l
,Vl,VO,
has the same angle, say fJ. Also note that the angles
L(vo, 0, vi), L(vi, 0, vi), ... , 0,
82 Solutions
(where 0 is the center of the circle) are all equal to 2n  2{3, and they
sum to n. Therefore a simple calculation gives us (a + 1) ( n  (3) = n.
Now consider the triplets
( V
I 2) (2 3 4) (b2 bI b)
vo, 2, v2 , V2' v
2
, v2 , ... , v2 , v
2
, v2 .
As before, each of these form an angle of measure (3 . The angles
...
each have measure 2(n  (3) and angle 0, vb) = a < 2(n  (3).
Therefore we have
(2)(n  (3) + a = n,
(a + l)(n  (3) = b(n  (3) + a,
and
a = (n  (3) (a  b + 1).
Now 0 < a < n  {3 implies that 0 < a  b + 1 < 2 and therefore, since
a and b are integers, a  b = 0, a contradiction.
Therefore, a chord starting a vertex Vo going through the center 0 of
the circle and ending at the point w on the circle divides the other vertices
into two sets of equal order. Thus
If nZ1 is even, we can proceed as before to obtain
Inl Inl
(2)(n  {3 + ad = (2)(n  (3 + (2)
2 2 2 2
nl
as illustrated below. This implies that the triangles 0, Wi, v;: and
nl
0, Vi, v:;:' are congruent. Letting the radius of the circle be r, this gives
us the result that an arbitrary side of the polygon has length 2r sin a, where
nl
a = n  (n  (3).
2
Exam # 101975 83
nI
I
VI
2
w
If n;l is odd, we let
nl
=/'1,
nl
0, W) = /'2,
nl 1 nl
L(V22 ,0,V
1
2
)=(/JI,
and can proceed as before to obtain
1 (n 1 ) 1 (n 1 )
2 2 1 (2)(7f,6+a1 )=2 2 1 (2) (7f,6+a2)
as illustrated below. This implies that the triangles O,W
'
,V;;l_l and
nl_l
0, Wi, v
2
2 are congruent. From this we can see that /'1 = /'2, so that
nl nl
triangles 0, Wi and 0, Wi are congruent. Again letting the
84 Solutions
radius of the circle be r, this gives us the result that an arbitrary side of
the polygon has length 2r sin ,,(, where "( = ¢  c¥.
w
Look under Geometry in the Index for similar problems.
519753
Let T m be the number of components yielded by m lines in the plane.
Clearly we have To = 1 (the whole plane). Also, if there are m  1 such
lines forming T ml components, then adding an mth line according to the
rules given will yield m more components. To see this, note that the new
line will intersect each of the existing lines at a unique point. If there are
m  1 lines, then the new line will split m previous components yielding
m new components. It follows that T m = T ml + m. We see that the T m
are just partial sums so that
m m m(m+l)
T m = To + I> = 1 + I> = 1 + 2 .
k=l k=l
Look under Enumeration in the Index for similar problems.
Exam #101975 85
519754
(a) Ifm is odd, then 3
m
== (l)m == 1 (mod 4), so 3
m
1 == 2 (mod 4)
and is not divisible by 2
m
for odd m larger than 1.
(b) We note that 3
1
 1 = 2 is divisible by 2
1
,3
2
 1 = 8 = 4·2 is
divisible by 2
2
, and 3
4
 1 = 80 = 16 . 5 is divisible by 24.
By Euler's generalization of Fermat's little theorem 3<P(2"') == 1 (mod 2
m
),
where ¢(2=) is the Euler totient function. Thus, ¢(2m) = 2=1. Therefore
the order of 3 modulo 2
m
, ord
2
", (3), is a power of 2. Hence, 2=13=  1
only if m is a mUltiple of ord
2
", (3).
We now prove by induction that for m 2: 4, 3
2
"'3 == 1 + 2=1 (mod 2= ).
When m = 4, we have 3
21
= 9 == 1 + 2
3
(mod 2
4
). Assume that the
result is true for m = k, i.e., that 3
2k

3
== 1 + 2
k

1
(mod 2k).
Now consider the case m = k + 1. We recall that, by our induction
hypothesis,
Squaring, we see that
3
2k

2
== 1 + 2
2k

2
+ j
2
2
2k
+ 2· 2
k

1
+ 2j2
k
+ 2j2
2k

1
(mod 2k+l).
Since k 2: 4, this is 3
2k

2
== 1 + ·2
k
(mod 2
k
+
1
). Thus,
3
2
"'3 == 1 + 2=1 (mod 2=)
and the claim is proved.
Next, we will prove that when m 2: 3, ord
2
",(3) = 2
m

2
• Since the
order of 3 modulo 2
m
is a power of two, we need only check exponents
that are powers of 2. Since
3
2
"'3 == 1 + 2
m

1
(mod 2=), ord(3) > 2=3.
2'"
Squaring 3
2
"'3, we see that
3
2
"'2 == 1 + 2· 2
m

1
+ 2
2
=2 (mod 2=).
As m > 2, we have 3
2
"'2 == 1 (mod 2=). Hence ord2m(3) = 2m2.
When m > 4, m < 2=2. (This is also easy to see by induction. It's
clearthatm = 5 < 8 = 2
3
andm+1 < m+m = 2m < 2·2=2 = 2
m

1
.)
Therefore, m ::; 4 and the only values for which 2
m
l(3= 1) are m = 1,
2, or 4.
Look under Number Theory in the Index for similar problems.
86
519755
By reversing the limits of integration, we have
r g(t)dt = r l
a
f(t) dtdx
J
o
J
o
x t
i
a
it f(t)
= dxdt
~ O t
= r f(t) . tdt = r f(t) dt.
J
o
t J
o
Solutions
Look under Multivariate Calculus in the Index for similar problems.
Exam # I 1 I 976
519761
By inspection, p( x) = x satisfies the conditions. Suppose that there
is another solution, and let its degree be n. From p(2) = 2 follows
p(22  1) = (p(2))2  1 or p(3) = 3. Similarly, p(8) = 8 and p(r) = r
implies p(r2 1) = r2 1. Thus n+ 1 numbers al, i = 1,2, ... n+ 1 can
be found such that p( ai) = ai. Thus the polynomial p( x)  x, of degree
n has n + 1 roots and this is impossible.
Look under Polynomials in the Index for similar problems.
519762
We are given that n == 11 (mod 12), whence n == 5 (mod 6). This
implies that the primepower decomposition of n contains a prime p,
p == 5 (mod 6) raised to an odd power. (If not, it would follow that
n == 1 (mod 6).) So we can write n = peN, (p, N) = 1, e odd. Then
a(n) = a(pe)a(N), and
a(pe) = 1 + p + p2 + ... + pel + pe
== 1 1 + 1··· + 1 1 == 0 (mod 6).
Hence 6Ia(n). If n == 11 (mod 12), then n == 3 (mod 4), and the
primepower decomposition of n contains a prime q, q == 3 (mod 4),
raised to an odd power. A calculation similar to the above shows that
4Ia(n), which gives the result.
Look under Number Theory in the Index for similar problems.
Exam #111976 87
519763
The quantity whose absolute value is to be shown to be less than 1 is
f' (0). Since
If' (0) I = I lim f ( h) I = lim If ( h) I < lim I tan hi = 1
h+O h h+O Ihl  h+O Ihl '
we are done.
Look under Differentiation or RealValued Functions in the Index for
similar problems.
519764
f"' > 0 for all x, so f" has at most one zero, f' has at most two, and f
at most three. Since f(O) = f(l) = 0, and f'(1)31n3  4 < 3.3  4 < 0,
and f (x) 7 00 as x 7 00, f has another zero to the right of 1. f has
exactly three zeros.
Look under RealValued Functions in the Index for similar problems.
519765
Suppose that the equation of the circle is
Then
x
2
+ y2 + rx + sy + t = O.
ral + sa2 + t = ai  a3,
rh + sb
2
+ t = bi  b ~ ,
rCI + SC2 + t = ci  c ~ .
The determinant of the coefficients is (b
l
al)(c2b2)(CI bl)(b2a2),
and since the points are not collinear
bl  al =I CI  bl .
b
2
 a2 C2  b
2
Thus the determinant is not zero, and the system has a unique solution.
Look under Analytic Geometry in the Index for similar problems.
S 19766
Since for all i E R, x = 2y for some y, we have f(x) = f(2y) =
f(y)2 ~ O. Further,f(l) = f(O + 1) = f(O)f(l), and since f(l) =I 0,
ss Solutions
f(O) = 1. Now let L = limx+o f(x). Then
1 = f(O) = f(x  x) = f(x)f( x) t L2.
Since L cannot equal 1 since this would imply f (x) ::::; 0 for some x, we
have L = 1.
Look under Limit Evaluation in the Index for similar problems.
Exam #121977
519771
f(a) = f(b) =} f(J(a)) = f(J(b)) =} a = b;
therefore f is injective on J. Since f is continuous, it assumes all
intermediate values. It follows that if f were not strictly monotone then it
could not be injective.
Look under RealValued Functions in the Index for similar problems.
519772
(a) Since, for n = 1, we have cos x = H cos x + cos( x)), we proceed
by induction. We therefore assume that
cos
m
x = 2 ~ f (;) cos(m  2k)x
k=O
and consider
= 2 ~ f (;) cos(m  2k)x cos x
k=O
= 2 ~ + l f [ (;) cos (( m + 1)  2k) x + cos (( m  1)  2k) x]
k=O
since
b
cos(a + b) + cos(a  b)
cos a cos = 2 .
The result now follows after some tedious, but routine manipulation of the
summation, including shifting an index of summation and adding binomial
coefficients.
Exam #121977 89
(b) We have
1 r 1 1 2n (2 ) r
;10 cos2nxdx=;22nL : 10 cos((2n2k)x)dx
k=O
= !_1 ( ~ (2n) sin(2n 2k)x[1f + (2n)7r)
7r 22n 6 k 2n  2k ° n
k=O,k#n
1 1 ((2n)!) (2n)!
= ; 22n (n!)2 7r = 22n(n!)2'
Look under Integration or Trigonometry in the Index for similar prob
lems.
519773
MUltiply the matrix A by the vector 1 = (1,1, ... ,1) to obtain Al = 1.
Thus (A  I)x = 0 has a nonzero solution and A  I is therefore not
invertible.
Look under Matrix Algebra in the Index for similar problems.
519774
Suppose that f(O) = t. Then
If(t)  tl = If(t)  f(O)1 = It I
and
If( t)  tl = If( t)  f(O)1 = Itl·
Also, If(t)  f( t)1 = 2It1 so that one of f(t) = 0 or f( t) = 0 must
hold. It follows that f (x) = 0 for some x in R. Now for any y in R,
If(y)1 = If(y)  f(x)1 = Iy  xl
so that any isometry must be of the form f (y) = ± (y  x) for some x in
R. It is also easy to see that anything of that form must be an isometry.
(These are translations followed by reflections.)
Look under RealValued Functions in the Index for similar problems.
90 Solutions
519775
Assume that there are only N primes. Then for each k, the kth partial sum
of the harmonic series can be written
k 1 1 1 1
L; = pa1pa2 •. • paN + bl b
2
bN + ... + pC=1p=C2...
p
=CN
n=1 I 2 N PI P2 ... PN I 2 N
since each integer n can be written as a product of our N primes. From
the hint, we therefore have
N 1
<II
 Ip··
i=1 2
This implies that the harmonic sequence is increasing and bounded above,
and therefore converging; a contradiction. (The hint is verified by picking
r 2: max{ai,b
i
, ... ,ei}, a finite set.)
Look under Number Theory in the Index for similar problems.
Exam #131978
519781
To find the decimal expansion of an integer M, one divides by 10
successively, recording the remainders:
M = lOM
I
+d
o
,
MI = 10M2 + dl ,
where the d
i
are 0, 1, ... ,8, or 9 to obtain the expansion
Exam #131978 91
To write M in the base 2, one proceeds by an analogous algorithm,
bearing in mind that the remainders must be 0 or + 1. So
2374(10) = 101111001110(_2)·
Look under Number Theory in the Index for similar problems.
519782
To solve this problem, one must get at the area of a nonrightangled
triangle. One technique is to use the cross product. Recall that Ilv x wll
gives the area of the parallelogram with sides v and w. To set this up, we
may assume that A is the origin in 3space, and that B, C, and D lie on
the axes, say at (x, 0, 0), (0, y, 0), and (0,0, z), respectively. Then
xy xz yz
area(ABC) = 2' area(ABD) = 2' area(ACD) = 2'
and the area of (BCD) is
1 1
:2IICD x CBII = :2
V
(x
y
)2 + (xz)2 + (yz)2.
Look under Analytic Geometry in the Index for similar problems.
519783
We recall that when k is odd, a
k
+ b
k
factors as
a
k
+ b
k
= (a + b)(a
k

1
_ a
k

2
b + a
k

3
b
2
+ ... + b
k

1
).
Now S(n) = 'L.?=l jk, and also S(n) = 'L.?=l (n+ 1 j)k. Adding these
two expressions, we get
n
2S(n) = 2)(n + 1 j)k + II
j=l
Look under Finite Sums in the Index for similar problems.
92 Solutions
519784
A first step in discovering the solution of this problem might be to graph
the function for small values of n, as illustrated. The x for which f (x)
is smallest is the median of aI, a2,' .. , an. That is, if n is odd, f(x) is
smallest at the middle a and if n is even, f (x) is smallest in the interval
between the middle a's.
n = 1 n=2
n=3 n=4
In view of this, the proof will be easier to write down if we reindex
the given numbers so that a1 ::; a2 ::; ... ::; an. We divide the proof into
two cases.
If n = 2m + 1, then
2m+1
f(am+d = L lam+!  akl
K=l
m 2rn+1
= L(arn+1  ak) + L (ak  am+!)
k=l k=rn+2
m 2rn+1
= marn+1  L ak + L ak mam+1
k=l k=rn+2
2rn+1 rn
L ak Lak.
k=rn+2 k=l
Exam #131978
Now suppose that az :::; x :::; aZ+1 where I :::; m. Then
2m+l Z 2m+l
f(x) = L IXakl=Lxak= L ak
x
k=l k=l k=l+l
2m+l Z
= L ak  L ak  (2m  21 + l)x.
k=l+l k=l
So that
[
2m+l t 1
f(x)  f(am+d = k ~ l ak  ~ ak  (2m  21 + l)x
m
= 2 L (ak  x) + (am +1  x) 2: O.
k=l+l
93
In an analogous way, we see f(x) 2: f(am+d if x:::; 0, x 2: an, or if
az :::; x :::; at+1 and I 2: m + 1.
If n = 2m, and am:::; x :::; a
m
+1, then
2m m 2m 2m m
f(x)=Llxakl=Lxak+ L akX= L akLak'
k=l k=l k=m+l k=m+l kl
Now if az :::; x :::; aZ+l, where I 2: m + 1, then
2m Z
f(x) = L akLak+(212n)x
k=Z+l k=l
and
Analogously, we can show
2m m
f(x) 2: L ak  Lakl
k=m+l k=l
if x :::; ao, x 2: an, or if az :::; x :::; aZ+l for I :::; m  1.
Look under RealValued Functions in the Index for similar problems.
94 Solutions
519785
We distinguish three cases. If A, B, C, and D lie on a line, this is problem
19784 restated; any point on the middle segment solves the problem.
If A, B, C, and D are the vertices of a convex quadrilateral, suppose
AC and BD are the diagonals and that they intersect at O. Let 0' be any
other point on the plane. Since AOC is a straight line and 0 is between
A and C,
IAOI + 1001 = IACI ~ IAO'I + 10'CI·
Similarly,
IBOI + 10DI = IBDI ~ IBO'I + 10' DI·
Thus,
IAOI + lOCI + IBOI + 10DI ~ IAO'I + 10'01 + IBO'I + 10'DI·
If A, B, C, and D are not the vertices of a convex quadrilateral, then
one of the points, say D, is contained in the triangle whose vertices are
the other three points, and 0 = D. Let 0' be any point in the plane other
than D. For one pair of the points A, B, and C, the triangle formed by
this pair and 0' contains the triangle formed by this pair and D, so let's
say AO'C contains ADC, so that
IAO'I + 10'01 2 IADI + IDCI·
By the triangle inequality, we have
10' BI + 10' DI 2IBDI,
so we then have
IAO'I + 10'CI + IBO'I + 10' DI 2 IADI + IBDI + ICDI·
Look under Geometry in the Index for similar problems.
519786
The first step in this solution is to realize that no experiment that always
finishes after N tosses of the coin can have probability of success 1/3.
(That is, there are 2N possible outcomes of flipping a coin N times, and
k/2
N
i= 1/3 for any integer k. Thus the solution should be an experiment
which a patient experimenter would expect to finish in a finite amount of
time with probability 1/3. One such experiment is:
Exam #141979 95
Toss a coin until (a) the coin has been tossed an even number of times
and (b) at least one head has been tossed. The experiment is a success
if two heads have been tossed.
Thus the experiment consists of two tosses, or four tosses, or six tosses,
or .... The experiment is over after two tosses unless both are tails;
after four tosses unless all four are tails; and so forth. The experiment
will terminate unless the experimenter tosses tails forever, an event whose
probability is zero.
The probability that the event is a success is given by
Pr ( success)
00
= L Pr(success given that the experiment requires exactly 2n tosses)
n=l
x Pr( experiment requires exactly 2n tosses)
00 1 3 1
= "'"""'  .  = .
63 4
n
3
n=l
Look under Probability in the Index for similar problems.
Exam #141979
519791
Let Mo (resp., Bo) be the age of the monkey's mother (resp., the monkey)
at the present time; let Ml (resp., Bd be the age of the monkey's mother
(resp., the monkey) at the time t
1
; let M2 (resp., B
2
) be the age of the
monkey's mother (resp., the monkey) at the time t2; and let M3 (resp.,
B3) be the age of the monkey's mother (resp., the monkey) at the time t3.
Let W w be the weight of the weight in ounces, W m be the weight of the
monkey in ounces, L be the length of the rope, Lm be the length of the
rope on the side of the monkey, and Lw be the length of the rope on the
side of the weight. Then we have
(1) Mi  Bi = k, a fixed difference in ages,
(2) Mo + Bo = 4,
(3) Mo = 2Bl,
(4) Ml = ~ B l '
96
(5) B2 = 3M
3
,
(6) M3 = 3B3·
Solutions
Since B3 = M3  k, equation (6) gives M3 = (3/2)k. So by (5)
B2 = (9/2)k and by (4) Ml = (9/4)k. Therefore
Bl = Ml  k = (9/4)k  k = (5/4)k
and by (3) Mo = (5/2)k. But by (1) and (2), Mo + Mo  k = 4, so k = 1
and Mo = 5/2.
Thus Wm = (5/2) . 16 = 40 OZ, WW = (3/2) ·40 = 60 OZ, and
L = 6
4
0 = 15 ft. Finally, p = 3/4 of the rope is on the side of the monkey
when the system is at rest since we have
60 + 40(1  p) = 40 + 40p.
519792
If sin x = sin y, then
O· . 2 x+y. xy
= SIn x  SIn y = COS 2 sm 2·
Therefore
x+y 7r xy
 =  + n7r or  = m7r
2 2 2'
where m and n are integers. The graph thus consists of the two families of
lines y = x + 7r + 2n7r and y = x + 2m7r, where m and n are integers.
Look under Trigonometry in the Index for similar problems.
519793
Let V(t) be the volume of snow removed at time t, d(t) be the depth
of the snow, a = ~ ~ be the rate at which snow is removed, b = ~ ~ be
the rate at which snow falls, and S(t) be the location of the plow. Then
dV = kd(t)ds, so ~ ~ = kd(t). Since
dV dV ds ds a
dt ds dt' dt kd(t)"
Noting that d(t) = bt + c, if to is the time at which it starts snowing,
then to = c/b. Suppose t = 0 at noon. We have
r
1
ds r
1
abdt * 1
1 = S(I)  S(O) = Jo dt = Jo kbd(t) = k In (d(t)lo
= k* (In(c + b) In(c)) = k* In (1 + n = k* In ( 1 t ~ ) .
Exam #141979 97
(Letting k* = a/k.) Similarly,
3/2 = 8(2)  8(0) = k* In (1  .
Therefore multiplying corresponding sides of these two equations gives
us
3k* In(1  = 2k* (In 1  .
to to
Solving for to, we get to = Since it started snowing before noon,
to must be negative and therefore we choose to = 0.618 hours. We
convert this to minutes by multiplying by 60, to get 37 minutes before
noon. Therefore it started snowing at 11 :23 a.m.
Look under Differential Equations in the Index for similar problems.
519794
Let r be the radius of the original sphere. Drill the hole along the xaxis,
as illustrated in the figure below. Then
V = Jr 27rY' (2V
r2
 y2) dy
v'r29
_ _ (r2 _ y2)3/2Ir
 27r (3/2) v?=9
= 47r . 93/2 = 367r.
3
y
x
Look under Volumes in the Index for similar problems.
98
519795
The next several positive integers are:
5 = 1 . 9  7 + v'9,
6 = 1 + 9  7 + v'9,
7 = 1 . )9 + 7 + v'9,
8 = 1 9 + 7 + 9,
9 = (_1)9 + 7 + v'9,
10 = 1 + 9 + l7 /9 J.
This list can be extended to at least 38.
519796
The set of points consists of
Solutions
(6,1), (5,3), (4,5), (4,4), (4,1), (3,7),
(3, 6), (3,1), (3,2), (3,3), (3,4), (3,5), (2, 9),
(2, 8), (1, 16), (1, 10), (1, 2), (0, 13), (0, 2),
(1, 15), (1, 2), (2, 17), (2, 2), (3, 19), (3, 2).
Look under Analytic Geometry in the Index for similar problems.
Exam # 1 519 SO
519801
We have
hence
albl > albl  aN+lbN+l
N
= 2 ~ ) a n b n  an+lbn+d
n=l
n=l
Look under Inequalities in the Index for similar problems.
Exam #151980 99
519802
The problem is to find the weights W1 < W2 < W3 < W4 such that every
positive integer I ::::; 40 can be written in the form I = 2::;=1 XiWi, where
the Xi is either 1, 0, or 1. (A given weight can be placed on the load
pan, not used, or on the weight pan to achieve a balance.)
It seems reasonable to choose W1 = 1 and W1 + W2 + W3 + W4 = 40.
To measure a 2 pound load, we need W2 = 3 (I + W1 = W2). Then we can
measure the loads I = 1,2,3,4. As 5 + 3 + 1 = 9, we can also measure
1=5 if we take W3 = 9. Then W4 = 40  2::{ Wi = 27.
Now note that if
4
o = L ZiWi = Z1 + Z2
W
2 + Z3
W
3 + Z4W4,
1
with each IZi I ::::; 2, then 3 divides Z1 and hence Z1 = O. So 0
3(Z2 + 3z
3
+ 3
2
z
4
) and this gives Z2 = O. Similarly, Z3 = Z4 = O.
Therefore, the set
s = { ~ X ' W ' : 1 < X· < 1} ~ 'l, 1, _ 1,_
i=1
has exactly 3
4
= 81 members.
As I E S implies I E S, it follows that there are exactly 40 positive
integers in S and each of these is at most 2::i Wi = 40. So S coincides
with the set of integers 1,2, ... ,40.
Look under Number Theory in the Index for similar problems.
519803
Choose Xo in [a, bj such that f(xo) is the maximum value of f on [a, bj.
Then f'(xo) = 0 and we must also have f"(xo) ::::; O. Therefore
0= f(a) ::::; f(xo) = !" ::::; 0,
so f (xo) = O. That is f (x) ::::; 0 for all X in [a, bj. A similar argument
shows that f (x) ~ 0 on [a, bj and hence f vanishes identically on [a, bj.
Notice iliat this argument assumes that Xo is in the interior of [a, bj, but
the same conclusions can be drawn if Xo = a or b.
Look under Differential Equations in the Index for similar problems.
100 Solutions
S 19804
For
and yt = (Yl Y2 . .. Yn ) ,
the matrix
c
y
,
XIY2
x,Yn)
X2Yl X2Y2 X2Yn
A = xyt = :
. .
XnYl X n Y2 XnYn
(We can assume Y i= 0, as the result is trivially true otherwise.) A short
computation shows that if
then the vector Az = (y, z)x where (,.) denotes the usual inner product
in Rn. If z(l),ze
2
), ... ,ze
n
) is an orthonormal basis with ze
i
) = y/llyll,
then, with respect to this basis, the matrix I + A has the form
I+A=
(
(Z(1),y)(x,ze
1
») +1 0
(z(1),y)(x,Z(2») 1
o
(ze1),y)(x,z(n») 0
o
o
.:. 0)
: 0
o
1
so the determinant of I + A is the product of the diagonal entries, that is
det(I + A) = (ze
1
),y)(x,z(1») + 1
= (y/llyll,y)(x,y/llyll) + 1 = (x,y) + 1.
Another proof can be given based on the fact that the determinant is
unchanged on adding a multiple of one column to another. With this, one
reduces A to a lower triangular matrix and then det(I + A) is easy to
compute.
Look under Matrix Algebra in the Index for similar problems.
Exam #151980 101
519805
::; 2(xy)2 =} log2 f(x) Iog
z
f(y) ::; (x _ y)2.
If we fix x = xo, divide by y  xo, with y > xo, and then let y tend to
xo, we get
f'(xo) d
f( ) = d log2 f(x)1 _ ::; 0.
Xo x XXo
A similar argument with y ::; Xo gives f' (xo) If (xo) 2: 0. Hence l' (xo) =
° for all xo, so that f must be a constant function. Conversely, any constant
function satisfies the condition.
Look under RealValued Functions in the Index for similar problems.
519806
(a) The improper integral has two apparent singularities, at x = ° and
x = 00. But as the function
{
sinx
f(x) = ('
xolO
x=o
is continuous on [0,00], the integral 10
1
dx makes sense. So we need
only check that
j
oo sinx d _ I' jb sinx d
 x 1m  x
1 X b+oo 1 X
exists and is finite.
Now
j
b sin x dx = _ cos x I b _jb x dx
1 X XII x
cosb jb cosx
= cos 1     dx
b 1 x
2
and Icotbl ::; lib 0. If g(x) = cosxlx
2
, then
lb Ig(x)1 dx 50 lb :2 dx = 1
so that limb+oo I: Igl exists and is finite. This implies that 1imb+00 lIb g( x) dx
exists and is finite by an argument similar to the one used to show that an
absolutely convergent series is convergent.
Put g+ = + g), g = g). Then g+ 2: 0, g 2: 0, so that
I: g+ and I: g are increasing functions of b that are each bounded above
102 Solutions
by It Igl, hence limb+oo I: g+, limb+oo I: g both exist and are finite
so that
has a finite limit as b 7 00. It now follows that 10
00
S i ~ x dx is convergent.
(b) Fix a 2 O. Then
1
00 sin ax d l' lb sin ax d l' lQb sin Y d 1
00
sin x d
 x = 1m  x = 1m  y =  x.
o X b+oo 0 X b+oo 0 Y 0 x
Similarly, if a :::; 0,
1
00 sin ax dx = 1
00
sin x dx.
o x 0 x
Therefore, (10
00
SinxQx dx) 2 = (10
00
S i ~ x dx) 2 is constant for a i= O.
Hence its derivative with respect to a is 0 for any a i= O. The derivative
fails to exist at a = 0 because of the discontinuity at that point.
Look under Integration in the Index for similar problems.
519807
We have
= 1 + ( ~  ~ ) + ~ + ( ~  ~ ) + ... + ( ~  2)
2 2 3 4 4 2n 2n
2n 1 n 2 n 1 n 1 n 1
= L k  L 2k = L k + L n + k  L k
k=l k=l k=l k=l k=l
n 1
= L
n
+
k
'
k=l
Look under Finite Sums in the Index for similar problems.
Exam #161981
Exam #161981
5198 II
If
then
3 a1 a2
7·
7
=ao +7"+7
2
+ ...
103
so that ao must be the greatest integer in 28
1
, that is, ao = 2. Consequently,
21 16 a1 a2
=+++ ...
8 8 7 7
2
5 a1 a2
or  =  +  + ....
8 7 7
2
Since 7 . ~ = a1 + a! + ... , a1 must be the greatest integer in 3;, so
a1 = 4. Continuing in this manner, we find a2 = 2, a3 = 4, a4 = 2.
Apparently, (iho = (.24)r.
For an alternate solution, we have
310
810
from which we can obtain the result by long division in base 7.
Look under Number Theory in the Index for similar problems.
519812
Letting A denote the number of acres, Go denote the initial amount of
grass per acre, S the number of steers, t the number of weeks until the
grass is depleted, v the growth rate of the grass per acre per week, and r
the rate of consumption of the grass per week, we have the relationship
GoA + Avt = Srt.
The information given yields the following equations:
(1) Go· 10 + 10· v . 16 = 12· r . 16,
(2) Go . 10 + 10 . v . 8 = 18 . r . 8,
(3) Go· 40 + 40 . v . 6 = S . r . 6.
Subtracting equation (2) from (1) gives 40v = 24r, and subtracting
equation (3) from four times equation (1) gives 200v = 384r  3Sr. Solve
for v, substitute, and cancel the r, to find S = 88 steers.
Look under Systems of Equations in the Index for similar problems.
104 Solutions
519813
Draw a plane passing through the center C of the given sphere, the comer
0, and the point P of tangency of the given sphere with one side of the
comer. On this plane, we have the trace of the given sphere and the trace
of the sphere whose radius we seek, as illustrated below. Since the radius
of the given sphere is 1, we know that CO = v'3. If C' is the center of
the sphere of radius r that we seek, then C'O = rv'3. Thus
CO = CQ + QC' + C'O,
V3 = 1 + r + rV3,
V3  1 = r(1 + V3),
v'3 1
r = v'3=3 +1 .
Look under Geometry in the Index for similar problems.
519814
The probability of winning the world series in exactly n games is
1) . 2 __ 1 (n 1)
2 2
n

1
 2
n

1
3 '
since 1/2 is the probability of winning the nth game, is the
probability of winning exactly 3 of the previous n  1 games, and we
have 2 teams to choose from. We have Pr(n = 4) = Pr(n = 5) = i,
Pr( n = 6) = 1
5
6' Pr( n = 7) = 1
5
6' which sum to one.
Look under Probability in the Index for similar problems.
Exam #161981 105
51981·5
Let y be the position at time t, measured from the horizontal. Then
y = 50  16t
2
, 1;jf = 32t, 0 = tan 1 160' and by the chain rule,
dO dO dy
=
dt dy dt
100
y2 + 100 (32t).
Therefore
100 100(2y) dy
y2 + 1002 (32)  32t (y2 + 1002)2 dt = O.
Substituting for 1;jf, and then y, after some simplification we arrive at
3125 + 400t
2
 192t
4
= O. By the quadratic formula, y2 = 125/24, and
so y = 100/3 feet below the observer's eye level.
Look under MaxIMin Problems in the Index for similar problems.
519816
Suppose the path of the destroyer is given by
x = r cos 0,
y = rsinO,
where r = reO) is some function of O. If s denotes arclength along the
destroyer's path as illustrated below, then
Since dx = r sin 0 dO + cos 0 dr, and dy = r cos 0 dO + sin 0 dr,
Since the submarine is moving at a constant rate c, the distance r that it
travels in a straight line is given by r = ct. Equating the distance s + 2
that the destroyer has gone at the time t of interception with the distance
2ct that the submarine has gone, we find s + 2 = 2ct. Thus s + 2 = 2r,
so ds = 2dr. Consequently,
and thus 3(dr)2 = r2(dO)2. Taking square roots, separating the variables,
and integrating gives r = K exp(O/V3) for some constant K. Since r = 1
when 0 = 0, K = 1. .
106 Solutions
Look under Analytic Geometry in the Index for similar problems.
Exam #171982
519821
Let Xl, X2, X3 be the roots of x3 + ax2 + bx + c. It is known that the
coefficients of a polynomial are symmetric functions in the roots. Let
Then we have
Let a', b', c' be the corresponding coefficients of the cubic whose roots
are reciprocals of the original cubic. Then
, 111 1
c ==,
Xl X2 X3 C
Exam #171982 107
b' = _1_ + _1_ + _1_ = Xl + X2 + X3
a
,
XlX2 XlX3 X2X3 XlX2X3
C
and
a' = _ + + = _ XlX2 + XlX3 + X2
X
3 =
Xl X2 X3 XlX2X3 c
Therefore the desired cubic is
3 b 2 a 1
X + x + x +.
c c c
Look under Polynomials in the Index for similar problems.
519822
A separated subset of order k from a set of order n can be represented by a
sequence of k dashes and n  k slashes where no two dashes are adjacent.
For example, the separated subset {2, 5, 7} with n = 8 can be represented
as /////. Given a set ofnk slashes, and conditions of this problem, there
are n  k + 1 places to possibly place the k dashes (this includes before
the first slash and after the last slash). Thus there are different
separated subsets.
Look under EnUllleration in the Index for similar problems.
519823
Let the two points that are furthest apart be denoted A and B. Join them by
a line, wlog horizontal. Now draw the smallest rectangle containing all of
the points with top and bottom sides parallel to the line AB, vertical sides
through A and B. There will be points C and D on the top and bottom,
respectively, of the rectangle. We consider the quadrilateral formed by
ACBD. The area of triangle ACB is :::; 1 and the area of triangle
ABD is AB :::; 1, where hand h' are the altitudes of the triangles.
Thus the area of the rectangle is
(h + h')AB = hAB + h' AB :::; 2 + 2 = 4.
Look under Geometry in the Index for similar problems.
519824
f (x) is differentiable at x = 0 if
( ) 1/h
2
lim f(h)  f 0 = lim _e _
h h
lOS Solutions
exists. Set y = Ilh
2
, then
_1/h
2
lim _e __ = lim yl/2eY = o.
h+O h y+oo
Look under Differentiation in the Index for similar problems.
519825
If a :::; 1, then f is uniformly continous on its domain.
Look under RealValued Functions in the Index for similar problems.
519826
If a
3
ends in the digits 11, then a
3
== 11 (mod 100). This is equivalent
to the two conditions a
3
== 11 == 3 (mod 4) and a
3
== 11 (mod 25). By
an easy calculation, 3 is the only residue satisfying a
3
== 11 modulo 4.
Now, a
3
== 11 (mod 25) =} a
3
== 1 (mod 5). By another easy calculation,
a = 1 is the only residue modulo 5 which satisfies this equation. That
means we need only check the cubes of 1, 6, 11, 16, and 21 modulo 25.
1
3
= 1,6
3
= 36 . 6 = 11 . 6 = 16, 16
3
= 256· 16 = 6· 16 = 96 = 21,
21
3
= (4)3 = 64 = 14 = 11. So a = 21 is the only solution
modulo 25. By the Chinese remainder theorem, there is only one number
satisfying a == 3 (mod 4) and a == 21 (mod 25) (that number is 71). It
follows that the probability that a random cube ends with the digits 11 is
1/100.
Look under Number Theory or Probability in the Index for similar
problems.
519827
Using the disk method for finding volumes of revolution, we obtain the
integral
f
a ( 2) 2 47fa
3
7f ( va
2
 y2 + b)  b
2
dy = 7f2 • a . b + .
a 3
Look under Volumes in the Index for similar problems.
Exam # 181983 109
Exam #181983
519831
lim \YrJ> lim = lim v:;;ri = 00
Look under Limit Evaluation in the Index for similar problems.
519832
There are 2
10
possible outcomes when one flips a com 10 times. Let
Ti be the result of the ith flip. We want to compute the number of
possible outcomes which have Tg = TlO = H and no adjacent H's in
T
1
, T
2
, ..• , T
s
, Tg• Certainly then, Ts = T and we wish to know how
many sequences of length 7 have no adjacent H's. Let Sn be the number
of sequences of length n without 2 consecutive occurrences of H. Each
such sequence must either start with HT followed by a sequence of length
n  2 with no consecutive H's or start with T followed by a sequence
of length n  1 with no consecutive H's. Therefore Sn = Snl + Sn2.
Also, SI = 2, S2 = 3. We see then that Sn = Fn+2 where Fi denotes
the ith Fibonacci number. Hence S7 = Fg = 34. It follows that there
are 34 possible outcomes which satisfy our condition. The probability that
such an outcome occurs is then 34/1024. (We have solved a more general
version of the problem: if 10 is replaced by an arbitrary n E N, then the
probability that we stop after the nth flip is Fn_I/2n.)
Look under Fibonnacci Sequences or Probability in the Index for similar
problems.
519833
From the illustration below, we want to maximize
f(x) = (a  x)x + 7fr2.
Since we have the relationship
 x)(a  x) =  x) +  x) +  x)
2 2 22'
we see that
(ax)(ax)
r=
(a  x) + (a  x) + J2(a  x)
ax
2+J2·
110
Thus we want to maximize
7f(ax)2
f(x) = (a  x)x + (2 + V2)2'
where 0:::; x:::; 1. Letting A = 7f/(2 + V2)2, we can write
f(x) = x
2
(A  1) + x(a  2aA) + Aa
2
.
Differentiating gives us
f'(x) = 2x(A  1) + a(1 2A),
Solutions
which has a critical point at x = a(1  2A)/2(1  A). Further this is a
maximum since A :::; 1 gives us f"(x) = 2(A  1) :::; O. The dimensions
of the rectangle and circle easily follow.
a
a
Look under Geometry or MaxIMin Problems in the Index for similar
problems.
519834
Solution 1. Suppose that t E R is not a finite sum of elements from
S = {1, 1/2, 1/3, 1/4,·· .}. Define f(x) = m a x { ~ E S : ~ :::; x}. Since
zero is a limit point of S, f is well defined for x > O. Finally, define
to = f(t), ti+l = ti + f(t  ti). Since t is not a finite sum of elements
from S, we must have t  ti =1= 0 for all i. Now, since f only attains
positive values, the sequence {ti} is monotone increasing. Furthermore,
since f(t  ti) < t  ti, we have ti < t for all i so that {til is bounded
above. Every bounded monotone sequence converges, so ti + s for some
Exam # 1 81983 III
s E R. If s < t, then let = f(t  s). ti t S, so there exists II t, Itlr
which s  ti < 1... It follows immediately that ti+l > ti + > II
m  lit C
contradiction. Therefore t = s.
Solution 2. Case 1: assume that our positive number, p, satisfies
o < p < 1. Then p has a binary representation of the form
1 1 1
P = O.hb2 b3 ... = b1 "2 + h4" + b
38
+ ...
with each b
k
either zero or one. This gives a representation of p as a sum
(possibly infinite) of a subset of the numbers {1/2, 1/4, 1/8, ... }. Case
2: assume that our positive number M satisfies M 2:: 1. Since the series
+ i + + ... diverges, there is a finite sum S = such that
M = S + q with 0 < q < 1. Applying case 1 to this q gives the result.
Look under Infinite Series in the Index for similar problems.
S 19835
This is a repeat of problem # 3 on Exam # 7 (1972).
519836:
Assume that there is a solution with z < n. Without loss of generality we
have x :s; y, and therefore x :s; y < z < n. From
= (z  y)[znl + yzn2 + ... + ynl]
> (1)[x
n

1
+ x
n

1
+ ... + xnl]
= nx
n

1
.
But now xn > nx
n

1
implies x > n, a contradiction.
Look under Diophantine Equations in the Index for similar problems.
S19837
N is a perfect square.
N = (n  l)n(n + l)(n + 2) + 1 = n
4
+ 2n
3
 n
2
 2n + 1
=(n2+n1)2.
Look under Number Theory in the Index for similar problems.
112 Solutions
519838
First, note that Pk(Xk) = 1 (this is obvious from the definition) and
Pk(Xj) = 0 for j i= k (because of the x  Xj term in the numerator).
Let
n
P(X) = LPk(X)  1,
k=l
then P( x) is a polynomial of degree at most n  1 because each Pk (x) is
a polynomial of degree n  1. But notice that for each Xj,
Since the Xj are distinct, it follows that P(x) has n distinct zeros. Any
polynomial of degree at most n  1 with n zeros must be identically zero,
therefore P(x) = 0 which implies = 1.
Look under Polynomials in the Index for similar problems.
519839
g(X) = g(y) =} gm(x) = gm(y) =} X = y, therefore 9 is injective. Any
injective realvalued continuous function must be strictly monotone on its
domain. Suppose that 9 is increasing on [0,1], then for x E [0,1],
X> g(x) 1 g(x) > g2(X) =} ... =} gml(x) > gm(x),
hence x > gm (x) = x which is obviously false. Also,
x < g(x) =} g(x) < g2(X) =} ... 1 gml(x) < gm(x),
so that x < gm(x) = X. SO we must have g(x) = x. Now if 9 is decreasing
on [0,1]' then for x, y E [0,1], x < y =} g(x) > g(y) =} g2(X) < g2(y)
so that g2 is increasing on [0,1]. Using the above argument we see that
g2(x) = x must hold.
Look under RealValued Functions in the Index for similar problems.
Exam #191984
519841
By the binomial theorem 1,005,010,010,005,001 is equal to (1001)5
7
5
(11
5
) 13
5
.
Look under Number Theory in the Index for similar problems.
Exam #191984 113
519842
Every path from A to B goes along 9 edges, 5 slanting right, 4 slanting
left and any such path joins A to B. Thus there are (;) = 126 paths.
Look under Enumeration in the Index for similar problems.
519843
Door D is changed by the nth person if and only if n divides D. Doors
that have an even number of divisors are closed, those with an odd number
are open. If n divides D, then so does D / n, thus D has an even number of
divisors unless D is a perfect square. Thus, doors 1,4,9,16, ... are open,
the rest are closed.
Look under Number Theory in the Index for similar problems.
519844
This problem concerns the numerical semigroup
S = {ap + bql a, b are nonnegative integers}
generated by p = 7 and q = 12 and asks "What is the largest integer
not in S?" (In the general solution here, all variables will be integers and
p and q will denote relatively prime positive integers.) In this question
the emphasis is on the nonnegativity because, since p and q are relatively
prime, every integer can be expressed in the form ap + bq for some integers
a and b.
Two fairly easy observations, which we offer without proof, enable one
to do the stated problem with a little trial and error.
Lemma 1. If n is a positive integer such that for each k = 0,1,2, ... ,
p  1 there are nonnegative ak and b
k
with n + k = akP + bkq, then every
integer greater than n is in the semigroup S.
Lemma 2. Let m = ap + bq where a and b are integers. Then
m = a' p + b' q if and only if there is an integer k so that a' = a + kq and
b' = b kp.
Corollary. If m is an integer; then there are unique integers a and b
with ° :s; a :s; q  1 so that m = ap + bq.
Now since 65 = (11)7 + (1)12 and 65 = 7a + 12b with a and b
nonnegative implies a :s; 9, the uniqueness in the corollary shows that it
is not possible to give exact change of 65 quanta. Moreover,since 66 =
6(7) + 2(12), 67 = 1(7) + 5(12), 68 = 8(7) + 1(12), 69 = 3(7) + 4(12),
114 Solutions
70 = 10(7) + 0(12), 71 = 5(7) + 3(12), and 72 = 0(7) + 6(12), Lenuna
1 shows 65 is the largest such price. This is actually a special case of an
old (1884) theorem by Sylvester.
Theorem. If p and q are relatively prime positive integers, then
(p  1) (q  1)  1 cannot be written in the form ap + bq for nonnegative
integers a and b, but every larger integer can be. (Proof omitted.)
The following theorem, also by Sylvester in 1884, shows that there are
33 prices for which Subsylvanians cannot give exact change.
Theorem. If p and q are relatively prime positive integers and S is the
numerical semigroup generated by p and q, then there are (p 1) (q 1) /2
positive integers not in S. (Proof omitted.)
Look under Number Theory in the Index for similar problems.
519845
In order for three segments to form the sides of a triangle, it is necessary
and sufficient that the sum of the lengths of any two segments must be
greater than the length of the third. In our situation, this means that none
of the segments formed may have length greater than or equal to 1/2.
The easiest way to find the required probability is to represent the
two points chosen in the interval as an ordered pair from the unit square
and decide which pairs lead to triangles. In the figure below, the region
corresponding to possible triangles has been shaded.
~       ~ ~       ~             x
Because of the independence and uniformity assumptions, the probability
that the first point is chosen from the interval [a, bJ and the second from
the interval [c, dJ is just (b  a) (d  c). More generally, the probability that
the points x and y are chosen from [0, 1], so that (x, y) is in the subset A
Exam # 191984 115
of the unit square, is equal to the area of A. Thus the probability that a
triangle can be formed is the area of the shaded region in the figure, 1/4.
We may obtain the same result more analytically with the principles of
conditional probability. Recall that if Xl, X2, •.. , Xn form a partition of the
space of outcomes and z is any event, then
n
Pr(z) = LPr(zlxi) Pr(xi).
i=l
In our situation, letting X be the random variable for the choice of the
first point, then this becomes
Pr(a triangle) = l ~ o Pr(a triangle X = x) dx.
Now, when 0 :::; x :::; 1/2, in order for a triangle to be formed, the
other point must be in the interval (1/2,1/2 + x), which has probability
(x + 1/2)  x = x. When 1/2 < x :::; 1, in order for a triangle to be
formed, the second point must be in the interval (x 1/2,1/2), which has
probability 1/2  (x  1/2) = 1  x. Thus, the probability of having a
triangle is
1
1/2 11
xdx+ (1x)dx=1/4.
x=O x=1/2
(This might be interesting to verify with a computer simulation.)
Look under Probability in the Index for similar problems.
519846
(a) From the functional equation and the continuity of 1 at 1, we have
lim f(x) = lim 1 (x/[x + 1])  2 = 1(1)  2 = 3.
x + 00 x + ex)
(b) We note first of all that it is not sufficient to observe that 1 (1/ n) =
2n + 3 for n = 1,2,3, ... ; the function g(x) = (2/x) cos(2n/x) + 3
has the property that g(l/n) = 2n + 3 for n = 1,2,3, ... but certainly,
limx+o+g(x) =J +00.
Since 1 (x) is continuous on 0 < x < 00, it has a finite minimum value
on every bounded closed interval. Let c be the minimum value of 1 on the
interval [l/(n + 1), l/n]. Since l/(n + 1) equals (l/n)/[(l/n) + 1], the
functional equation shows that Cn+l = Cn +2, hence that Cn+1 = C1 + 2n.
116
Thus, if 0 < x::; 1/(1 + n), we have
f(x) 2: inf Cn = cn+l = Cl + 2n
m ~ n + l
which means limx+o+ f(x) = +00.
Solutions
(c) The given conditions relate the values of f on the intervals [1, (0),
[1/2,1), [1/3,1/2), ... to each other and specify the values at the end
points, but besides f(l) = 5 and limx+oo f(x) = 3, the function f is
an arbitrary continuous function on [1, (0). A succinct way of say this is:
if 9 is any continuous function on (00,00) such that limx+±oo g(x) = 0,
then f(x) = 3 + (2/x) + g(cot(7r/x)) satisfies the given conditions and
any solution can be written in this way for some g.
Look under Limit Evaluation in the Index for similar problems.
519847
It is helpful to consider 25 tons as one unit of weight; we want to count
the number of trains of total weight 40 units. Let f (n) denote the number
of trains of total weight n units: we want to find f (40).
It is easy to find f(n) when n is small: f(l) = 1 (the only train is B);
f(2) = 3 (BB, F, H); f(3) = 5 (BF, BH,BBB,FB,HB); f(4) = 11
(BFF, FF, HF, BBH, FH, HH, BFB, BHB,BBB,FBB,HBB);
etc. More importantly, we notice that an n unit train is either an n  1 unit
train with a boxcar stuck on the end or an n  2 unit train with a flat car or
hopper car stuck on the end. That is, f satisfies f(n) = f( n1)+2f(n2)
where f(l) = 1 and f(2) = 3. Thus f(5) = 11 + 10 = 21, f(6)
21 + 22 = 43, f(7) = 43 + 42 = 85, and so on until we arrive at
and
241
f(40) = 3'
our answer.
(It is easy to prove by induction, but perhaps hard to discover, that
f(n) = f(n  1) + 2f(n  2) where f(l) = 1 and f(2) = 3 implies:
22k  1
f(2k1)= 3
and
22k+l + 1
f(2k) = 3
for k = 1,2,3, .... )
Look under Enumeration in the Index for similar problems.
Exam #201985 117
Exam #201985
519851
We have
X4 + bx
3
+ CX
2
+ dx + e = (x  a)(x  (J)(x ,,)(x  6).
In the above equality, first put x = i, and then x = i, where i = A,
to get
1  c + e  (b  d)i = (i  a)(i  (J)(i ,,)(i  6),
and
1  c + e + (b  d)i = (i  a)( i  (J)( i ,,)( i  6).
Noting that (u+vi)( u vi) = u
2
+v
2
, multiplication of the corresponding
sides of the two equalities above yields
Look under Polynomials in the Index for similar problems.
519852
Since 3333 == 1 (mod 7) and 4444 == 1 (mod 7), we have
3333
4444
+4444
3333
== 14444 + (_1)3333 (mod 7) == 11
== 0 (mod 7).
Some more problems of this type:
(i) 22221111 + 11112222 is divisible by 3.
(ii) 11114444 + 44441111 is divisible by 5.
(iii) 2222
5555
+ 5555
2222
is divisible by 7.
(iv) 8888833333 + 3333388888 is divisible by 11.
Look under Number Theory in the Index for similar problems.
519853
Clearly U is the multiplicative group of units modulo 21. Since 6 is a
squarefree integer and U is abelian, each subgroup to be found is cyclic.
Hence the subgroup along the solid path could be {I, 2, 4, 8,16,11}; the
liS Solutions
one along the dashed path could be {I, 5, 4,20,16, 17}; and the one along
the dotted path could be {I, 13,4,10,16, 19}.
Look under Group Theory in the Index for similar problems.
519854
There are 2 types of triangles that can be formed: either each vertex lies
on a different side of the square, or 2 vertices lie on one side and the third
on another side.
We find the number of triangles of the first type as follows: Select 3
of the 4 sides of the square (this can be done in (i) or 4 ways) and then
select 1 of the 100 points of subdivision on each of these 3 sides (this can
be done in 100
3
ways). Thus, there are 4.100
3
triangles of the first kind.
The number of triangles of the second type is found as follows: Select
1 side of the square (4 ways of doing this) and 2 of the division points on
this side ( C ~ O ) ways) and then select 1 of the 3 remaining sides (3 ways)
and 1 of the 100 points of subdivision (100 ways) on this side. The total
number of triangles of the type is 4 . 3 . 100 . e ~ O ) = 6 . 99 . 100
2
.
Hence the required number is 4.100
3
+ 6.99.100
2
= 9,940,000.
Look under Enumeration in the Index for similar problems.
519855
From an = (n  1)(a
n
l + a
n
2), we get
an  nanl =  [anl  (n  l)an2]'
Iterating this relation n  3 times, we have
Hence,
an anl
n! (n I)!
and
~ (a
r
_ arl ) = ~ (It = ~ (It.
L..J r! (r  I)! L..J r! L..J r!
r=2 r=2 r=O
Since al = 0, the sum on the left telescopes to %3. Thus
(
n (It)
an = n! L;:y
r=O
Exam #201985 119
and
lim an = lim ~ (It = e
1
.
n+oo n! n+oo ~ r!
r=O
(Observe that an is the number of derangements of a set of n objects .)
Look under Limit Evaluation or Sequences in the Index for similar
problems.
S 19856
We will prove this result by induction. The result is clearly true for n = 1.
Assume the result true for n = m. Let S be a set of m + 2 integers between
1 and 2m + 2.
Suppose S does not contain both 2m + 1 and 2m + 2. Let Sl be the
subset of S not containing 2m + 1 or 2m + 2. Then Sl contains m + 1
integers between 1 and 2m. By the induction hypothesis, the result is true
for Sl and therefore for S.(Of course if neither 2m + 1 nor 2m + 2 is in
S, then by the induction assumption, the result is already true for S ).
Suppose S contains both 2m + 1 and 2m + 2. If m + 1 is in S, then
2m + 2 is a multiple of S, and thus the result is true for S. Therefore,
assume m + 1 is not in S. Let Sl be the set obtained from S by throwing
out 2m + 1 and 2m + 2 and putting in m + 1. Then Sl has m + 1 integers
between 1 and 2m. By the induction hypothesis, some element b of Sl is
a multiple of some other element a of Sl. If a and b are both in S, then
the result is true for S. Now a cannot be m + 1. Because the only positive
integers m + 1 can divide other than itself are k(m + 1), where k is an
integer ~ 2. But Sl does not contain such integers. Hence a must be in
S. If b is m + 1, then since 2m + 2 is a multiple of band b is a multiple
of a, 2m + 2 is a mUltiple of a. Thus the result is true for S.
Look under Number Theory in the Index for similar problems.
S 19857
From x+y = 2a, we have y = 2ax. Substituting in xy > [(b
2
1)/b
2
Ja
2
,
we get b
2
x(2a  x) > (b
2
 l)a
2
,
i.e., b
2
x
2
 2ab
2
x + (b
2
 l)a
2
< 0,
i.e., (bx(bl)a)(bx(b+l)a) <0,
l.e.,
(bl)a (b+l)a
b <x< a .
110
Thus, the number of favorable cases is
(b + 1)a
b
(b  1)a
b
Solutions
2a
b·
The number of all possible cases is when 0 < x < 2a , and hence is 2a.
The required probability is therefore
2a/b
2a
1
b·
Look under Probability in the Index for similar problems.
Exam #211986
519861
See Solution #1 on Exam #14.
519862
Let B = {b
1
, b
2
, b
3
}. The matrix of ¢; with respect to B is given by
A = ( ~ ~ ~ ) .
o 0 1 +i
To find the eigenvalues we set
det(A  {3I) = 0 = [(1 + i)  {3] [{32  4i]
which implies that {3 = 1 + i, {3 = V2 + V2i or {3 = V2  V2i.
Look under Complex Numbers or Matrix Algebra in the Index for similar
problems.
519863
ii = (e
7ri
/
2
)i = e7r/2.
Look under Complex Numbers in the Index for similar problems.
519864
For every four distinct points of the N points there is one and only one
pair of segments that intersect and have the given four points as endpoints.
Exam #211986
Thus the number of intersections is given by:
(
n) = n(n  l)(n  2)(n  3) .
4 24
Look under Enumeration in the Index for similar problems.
519865
Let x = V2 + ( V'3). Then
x
2
= 2 + 2V2( ij3) + ( ij3)2,
x
3
= 2V2 + 3(2( ij3)) + 3V2( ij3)2 + 3.
Similarly
(ij3)2 = ij9 = x
2
 2  2V2ij3,
But V'3 = x  V2, so
Now
ij9 = x
2
 2  2V2(x  V2) = x
2
 2V2x + 2.
x
3
= 2V2 + 6(x  V2) + 3V2(x
2
 2V2x + 2) + 3
= 2V2 + 6x  6V2 + 3V2x2  12x + 6V2 + 3
= 2V2 + 3V2x2  6x + 3.
Thus x
3
+ 6x  3 = V2(3x2 + 2). Squaring both sides gives
x
6
+ 36x
2
+ 9 + 12x4  6x 36x = 18x
4
+ 24x2 + 8
or
Look under Polynomials in the Index for similar problems.
519866
121
(a) If the limit of f(xl, X2) exists as (Xl, X2) tends to (0,0), then the limit
exists along any path to (0,0). In particular the limit must exist along the
path defined by Xl = Then the limit of f(xl, X2) as (Xl, X2) tends to
(0,0) now becomes
lim X2) = lim = ! 1= o.
X2+0 X2+0 x
2
2
Thus f(xl, X2) is not continuous at (0,0).
122 Solutions
(b) Since !(Xl,X2) is not continuous at (0,0), !(Xl,X2) is not differ
entiable at (0,0).
Look under Multivariate Calculus in the Index for similar problems.
Exam #221987
519871
The wellknown proofthat there are arbitrarily long sequences of composite
integers suggests that the solution should be 11!+2, 11!+3, ... ,11!+11, but
a calculation shows that these numbers are much too large. An appropriate
idea is present though; we should replace 11! by a number divisible by
2,3, ... ,11. For example, the least common mUltiple of these integers is
2
3
. 3
2
. 5 . 7 . 11 = 27720, so the sequence 27722,27723, ... ,27731 is a
solution. There are smaller solutions, the smallest solution is the sequence
beginning with 114.
Look under Number Theory in the Index for similar problems.
519872
Each layer is a triangle of cannonballs, the kth having k balls on each
edge. Thus, each layer has
1 2 k
_k(k+l)_I(k2 k)
+ + ... +  2 "2 +
cannonballs. The total number of cannonballs in a stack is therefore
1
2((1
2
+ 1) + (22 + 2) + ... + (n
2
+n))
= + 22 + ... + n
2
) + (1 + 2 + ... + n))
2
n(n+l))
2 6 + 2
1
= "6n(n + 1)(n + 2).
Look under Enumeration in the Index for similar problems.
519873
If the ball obeys the law exactly, it must bounce an infinite number of
times, but it is not clear wheilier it bounces for an infinite time. We should
Exam #221987 123
find the infinite series that represents the total time, decide if it converges,
and if it converges compute the sum.
Since it is dropped from a height of 16 feet, it will require vT6/4 =
1 second to fall to the ground. It then bounces to a height of 8 feet,
requiring VS /4 second to rise and VS / 4 second to fall to the ground
again. Continuing, we see that the total time is the sum of the series
1 + 2(VS/4) + 2(V4/4) + 2(v'2/4) +"',
or
(
2) 00 00 ( 1 )j
1+ "4 v'2
Since this is a geometric series with ratio less than 1, the series converges
and its sum is
1 + v'2 __
1
c:::
11/v'2
v'2+1
v'21'
Thus, this model predicts the ball will bounce slightly more than 5.8
seconds. (In fact, this model is in fair agreement with experiment; the
constant 1/2 in this problem is called the coefficient of restitution of the
ball.)
Look under Infinite Series in the Index for similar problems.
519874:
Let Sj denote the event that the jth senior draws hi1> or her own name
from the hat. We wish to find the probability that none of these events
occurs, that is, we want
Pr(Sf n S2 n ... Sfo)·
We may expand this as
Pr(Sf n S2 n··· n Sfo)
= 1  Pr(S1 U S2 U ... U S1O)
= 1 [ L Pr(Sj,)  L Pr(Sj, n Sj2)
1:'OiI:'01O 1:'Oj,:'012:'01O
+
PreS n S· n S· )  ...
)1 )2 )3
124 Solutions
But Pr( Sj, = ;6, and Pr( Sjl n S12) = 1
8
6, and so forth. Since the number
of ways to choose k indices from lOis the above becomes
Pr(Sf n n··· Sfo)
= 1 [ CI0) C20) + ... 
= 1  [:! ;! + :! ... 
1 1 1 1
=  +    + ... +  .368.
I! 2! 3! 1O!
Look under Derangements or Probability in the Index for similar prob
lems.
519875
A geometric construction: We give a construction that is an induction on
the dimension. This construction and plausibility argument can be made
into a proof, but the analytic proof below is more concise. For me, the
idea comes from the inductive step from dimension 2 to dimension 3, but I
will begin here with dimension 1, where the logic begins. Clearly, the two
most distant points in the unit ball of dimension 1, i.e., the unit interval,
are the points + 1 and l.
The three most distant points in the unit ball in dimension 2, i.e., the
unit disk, are the vertices of any equilateral triangle with vertices on the
unit circle. Indeed, if A, B, and C are points of the disk that do not satisfY
this condition, then assuming A and B are on the circle and the length
of AC is least, then the triple A, B, C' is an improvement, where C' is
the more distant intersection of the perpendicular bisector of AB and the
circle. To find the coordinates of such points, we choose the north pole
(0, 1) as the first point and note that the center of gravity of the three
points must be the origin. Thus, the other two points, being symmetrically
placed, have coordinates (?, 1/2). Since they are on the circle, and at
opposite ends of the segment (the Idimensional case), the coordinates
must be (±J3/2, 1/2).
The four most distant points in the unit ball in dimension 3 are four
equidistant points on the sphere, (As before, if not, assuming A, B, and
C are on the sphere and and the length of AD is least, then the 4tuple
A, B, C, D' is an improvement, where D' is the more distant intersection
of the sphere and the line perpendicular to the disk determined by A, B,
Exam #221987 125
and C and through the center.) Again assuming the north pole (0,0,1)
to be one of the points, we find that the other three points must have
coordinates (7,7, 1/3), Moreover, these three points solve the problem
for the disk they determine, that is, they are a scaled version of the
solution of the 2dimensional problem. To be on the sphere, since the third
coordinate is 1/3, the scale factor must be J879 so the three points are
(0, J879, 1/3) and (±y'273, y'279, 1/3).
The solution to the stated problem is five equidistant points on the unit
sphere in dimension 4. Assuming the north pole is one of the points, the
others have coordinates (7,7,7, 1/4). The first three coordinates must
be a scaled version of the solution to the 3dimensional problem, and the
scale factor must be VI5/16. Thus,the solution to the problem is
(0,0,0,1); (0,0, VI5/16, 1/4); (0, V576, V5/48, 1/4);
(±/578, V5/24, V5/48, 1/4)
and the mutual distance is V512.
An analytic proof: Letting Vl, ... ,V5 be the five vectors above, we see
that, for all i, j,
Now, if Xl,'" ,X5 are any vectors in the unit ball of R
4
, then
But
so
i#j
::::; 40  2 LXiXj.
if.j
 LXiXj::::; 5.
i#j
Combining this with the earlier inequality, we find
S ::::; 40 + 10 = 50.
Since the sum for S contains 20 terms, either all of the terms are 2.5 or
there is at least one of the terms that is less than 2.5. That is, the minimum
126 Solutions
distance between points is no more than V5fi and the above solution is
optimal.
Look under Geometry in the Index for similar problems.
S 19876
If Jane were flying straight north, she would fly a quarter of the way
around the earth before reaching the north pole, that is, she would fly
2l7r(4000)/4 = 20007r miles. Instead, her path is northwest, so she flies
further. To be precise, at each instant, her path is along the hypotenuse of
a right triangle with northsouth and eastwest legs and, since the surface
of a sphere is locally like the Euclidean plane, the distance along the
hypotenuse is J2 times the distance along the northsouth leg. Since this
is true at every instant of her journey, the total distance she flies before
reaching the north pole is 2000J27r miles.
To see how many times she crosses the Greenwich Meridian, we look at
the same infmitesimal triangle and relate the latitude and longitude angles.
In the usual spherical coordinates with the origin at the center of the earth,
longitude corresponds to the angle (J and (north) latitude corresponds to
7r /2  VJ. Thus, we want the total change in () when VJ changes from 7r /2
to 0, that is, we want
The northsouth leg of the triangle has length 4000dVJ whereas the eastwest
leg of the triangle has length 4000 sin VJd(J since the parallel corresponding
to VJ is a circle of radius 4000 sin VJ. Thus, the total change in () is then
11
° _.1 d(VJ) I = 00.
1r/2 SlllVJ
which means Jane crosses the Greenwich Meridian infinitely often.
Look under Integration in the Index for similar problems.
Exam #231988
S19881
Let x(t) denote distance between x's car and the finish line, and y(t)
denote the distance between y's car and the finish line. Suppose x(O) = 2
Exam #231988
and y(O) = 5. For t > 0,
d
2
x _ (dX)2 dx _ ~ () _ In(at+b)
d
2  ad=} d  b =} x t  + c.
t t t at+ a
But
Inb In(b)
x(O) = 2 =} c = 2 +  =} x(t) = 2 + Hat.
a a
Let it > 0 be such that x(tt) = 1 and, hence,
We have
dxl 1 dxl
dt tl = 2 dt o·
In( b )
1=2+ ~
a
1 11
and atl +b = 2(T)'
implying that a = In 2.
127
Let tx be the time required for driver x to reach the finish line. Then
o = x (tx) and a = In 2 together imply tx = 1 ~ b 2 ' Let ty be the time
required for driver y to reach the finish line. Then
dyl
dt all t
dxl
dt t=O
1
b
and y(O) = 5 together imply y(t) = 5  t, so that ty = 5b.
Finally,
since
3 3b
In 2 >  =} ty = 5b >  = t
x
,
5 In2
0> dxl
dt t=O
1
b
implies b > O. Driver x finishes first.
Look under Differential Equations in the Index for similar problems.
519882
Let x and y denote any two members of the group. Then
x * xI * Y = e * y = y = y * e = y * xI * X =} X * Y = Y * x,
by the given cancellation rule.
Look under Group Theory in the Index for similar problems.
128 Solutions
519883
Note that
So if sand t are as prescribed,
519884
The unit's digit of 512345 = 5. To verify this it suffices to show, using
induction on n, that 5
5n
= 5n + 10k
n
where k
n
is a positive integer for
each n. 55 = 12 + 3
2
+ 52 + 7
2
+ 9
2
= 165 = 5 + 10(16). Assuming the
result true for n gives
55(n+1) = 55n+5
= 5
5n
+ [2(5n + 1)  1]2 + ... + [2(5n + 5)  1]2
= 5
5n
+ (1On + 1)2 + (IOn + 3)2 + ... + (IOn + 9)2
= 5
5n
+ 10jn + (12 + 3
2
+ ... + 9
2
)
= 5n + 10kn + 10jn + 165
= 5(n + 1) + 10kn+1)
where kn+1 = kn + jn + 16.
Look under Number Theory or Sequences in the Index for similar
problems.
519885
The desired area is 8 times that of the shaded region 5 in shown below. If
the circle has radius a, then a Cartesian equation for the curve bounding 5
on the right is (x + a)2 + y2 = 2a
2
. Changing to polar coordinates gives
the polar equation r = f(O) = a[yll + cos
2
0  cosO]. So the area of 5 is
{'i f(O)2 dO = a
2
[1 v'3 + ~ l
Jo 2 2 2 3
Exam #231988 129
after much work. Hence, the desired area is
2a
2
[1  v'3 + iJ .
Look under Geometry or Integration in the Index for similar problems.
519886
First draw the two straight lines BAP and CDP, where Band Dare
opposite ends of the diameter lying within L. Extend straight lines AD
and BC downward until they intersect in a point E. P E is then a straight
line perpendicular to L.
p
B ~     '      ~     ~ ~   ~ L
To prove this, notice that line segments EA and PC are altitudes of
6BP E. Since these segments are concurrent with the line segment through
BD, that line segment must also be an altitude of 6BPE. Thus PE is
perpendicular to line L.
Look under Analytic Geometry or Geometry in the Index for similar
problems.
130 Solutions
519887
There are 2
10
= 1024 possible lOflip sequences. Let Hn denote the
number of nflips sequences not having two or more tails in a row but
ending with heads and let Tn denote the number of nflips sequences not
having two or more tails in a row but ending with a tail. Clearly H2 = 2
and T2 = HI = 1. For n 2:: 3, Tn = Hn l and Hn = Hn
l
+ Tn l =
Hn l + Hn 2. So H3 = H2 + HI = 2 + 1 = 3, H4 = 3 + 2 = 5, ... ,
Hg = 55, and HlO = 89. We need HlO + TlO = 89 + 55 = 144 and the
desired probability is / o ~ ~ = ;4'
Look under Probability in the Index for similar problems.
Exam #241989
519891
It is easy to show that there are three tests and that x + y + z = 13. From
C's total we deduce that 3z ~ 9, so that z = 1, 2, or 3. The only solution
satisfYing all the constraints of the problem is z = 1, y = 4,x = 8. It
follows that A scored 4 + 8 + 8, B scored 8+ I + 1, and C scored 1 + 4 + 4.
Thus C came in second on the Geometry test.
Look under Logic in the Index for similar problems.
519892
Any constant sequence satisfies the hypothesis of the problem. Conversely,
if we fix m and let n tend to infinity, we see that limntoo Xn = X
m
• By
uniqueness of limits, we deduce that Xl = X2 = ... , so that {Xl, X2, ••• }
must be a constant sequence.
Look under Sequences in the Index for similar problems.
519893:
If we label the figure as shown below and compare areas, we deduce that
2a = d,
2(a+h) =d+g+j,
2(a + h + e + b) = d + g + j + c + f + i.
Exam #241989
These equations may be reduced to
2a= d,
2h = 9 + j,
2(e+b)=c+f+i.
In a similar fashion we have
2b = e,
2i = h + j,
2(f + cO = a + d + g,
and
2c= f,
2g = i + j,
2(d + a) = b + e + h.
From these nine equations we deduce that
6j = a + b + c + d + e + f + 9 + h + i,
and this is equivalent to the assertion of the problem.
Look under Geometry in the Index for similar problems.
131
132 Solutions
519894
Assuming independence, if the fair coin is the one flipped only once, then
the probability of obtaining the results noted is
2

9
On the other hand, if the biased coin is the one flipped only once, then the
probability of obtaining the results noted is
~ ' G ) ( ~ ) 2 ( ~ ) 1
1

4
Hence it is more likely that the biased coin was the one flipped only once.
Look under Probability in the Index for similar problems.
519895
Let the y and zaxes be the axes of the cylinders. If we slice the
intersection by planes parallel to the yzplane, the crosssectional areas
are all squares (by symmetry). Hence the volume is
[aa 4b
2
(1 _ ::) dx = 16;b
2
•
Look under Volumes in the Index for similar problems.
519896
The characteristic equation is
Hence the eigenvalues are ,\ = 1 (multiplicity n  1) and
,\ = 1 + XlYl + ... + XnYn'
The determinant, being the product of the eigenvalues, is
1 + XIYl + ... + XnYn'
Look under Integration or Matrix Algebra in the Index for similar
problems.
Exam #251990 133
519897
The dimensions are 4 x 18. A complete solution requires lengthy discussion
to eliminate many potential solutions. A key ingredient is the fact that the
number 72 has three "admissible" factorizations: 4 x 18, 6 x 12, and 8 x 9
with answers YES, NO, NO to the question "is the short side strictly
less than half the long side?" Several other possible areas admit triple
factorizations, e.g., 48 = 3 x 16, 4 x 12 and 6 x 8, but here the answers
are always two YESes and one NO.
Note: The students were told at the beginning of the contest to assume
that the sides of Smith's ranch are parallel to the side of Todd county.
Look under Logic in the Index for similar problems.
Exam #251990
19901
Write n = 10
k
ak + 10
k

l
ak_l + ... + lOal + ao, where ai E {O, ... ,9}
and ak 2: 1. We assume
which implies
(10
k
 ak)ak + ... + (10  adal = a6  ao + 1. (*)
The righthand side of (*) is at most 9
2
 9 + 1 = 73. Since ak 2: 1,
the lefthand side of (*) is at least 10
k
 9. Therefore k :::; 1. Clearly,
there are no 1 digit solutions. It is easy to check that the only solutions to
(10  al)al = a6  ao + 1 are ao = 5, al = 3 or 7, so 35 or 75 are the
only solutions.
Look under Number Theory in the Index for similar problems.
19902
This is an inclusionexclusion problem, where the universe U is the set of
arrangements of I, N, D, I, A, N, A and the subsets AI' ... ' A6 are those
arrangements in which each possible digraph lA, AI, IN, NI, AN, NA, is
repeated. If, say, IA is repeated, then neither IN, nor NA can occur at all,
and AI could occur at most once. However AN or NA might occur twice
134 Solutions
if the tuples IA or NIA were repeated. In other words,
Continuing in this way, we find that 6 of the pairs Ai n Aj are possible
and none of the 20 tuples Ai n Aj n Ak may occur simultaneously. Thus
the total number of possibilities (by symmetry) is
assuming Ai n Aj is possible.
Note that lUI = 7!/[(2!)3 . I!] = 630. If IA is repeated, then we
wish to write (IA), (IA), N, D, N in some order which can be done in
5!/ [(2!)2 . I!] = 30 ways. If IA and AN are repeated, then we need to
arrange (IAN),(IAN), D, which can be done in 3!/(2! ·1!) = 3 ways. Thus
the desired answer is 630  6 . 30 + 6 . 3 = 468.
Look under Enumeration in the Index for similar problems.
519903
Since f' (x) = f (x + 1)  2, l' is differentiable and
f"(x) = f'(x + 1) = f(x + 2)  2 = f(x)  2.
It follows that f(x) = 2 + cle
x
+ C2ex, and, since f(x + 2) = f(x),
Cle
2
= Cl and C2e2 = C2 so Cl = C2 = o.
519904
A variety of approaches work. This one refers to the figure below. Since
the exterior angle to a regular heptagon has measure 2;, LGAB = 5;.
Since LGAF subtends an arc of J,f on the circle, its measure is J,f =
LGAK = LLAB, and LKAL = 3;, so LAKL = LALK = 2;. The
law of sines applied to r and to sand l gives
so
r s
and
s 1
sin 371"  sin 271"
7 7
sin]I  sin 571" '
7 7
sin ]I sin 371"
r = 7 7.
sin 271" sin 571"
7 7
This may be reduced to (among others):
sin ]I sin 371"
7 7
sin2 271"
7
7r 1
2cos 
7 cos J,f'
cos 271"
__ 7_
cos]I .
7
Exam #251990 135
Other proofs use the law of cosines, and dropping a perpendicular from
K to AG, etc.
Look under Geometry in the Index for similar problems.
519905
Construct the Riemann sum
2 t t 1 n n
2
1 1 1
n i=1 j=1 yn
2
+ ni + j = 8 n
2
J1 + i/n + j/n
2
'
which in the limit, gives us
[1 [1 v'1
dXdY
= [\2J1+x+y)l
y
1
=odx
Jo Jo + x + y Jo
= 10
1
(2v'2 + x  2v'1 + x) dx
= i(3
3
/
2
_ 2.2
3
/
2
+ 13/2) = i(3V3  4V2 + 1)
3 3
(;::::0 .719).
Look under Limit Evaluation or Riemann Sums in the Index for similar
problems.
519906
Insert a coordinate system centered at the center of the hexagon, so the
vertices of the hexagon are (0, ±1), (±4, The parabolas must have
the form y = ax
2
+ {3, and the points show that they are y = 1 
and y = 1 + which intersect at (±1,0). A routine calculation
136 Solutions
shows that the area between the parabolas is i V6 and that of the hexagon
is 6· "'1, so the total area of the shaded region is
8V2  9 8V6  9V3
2V3 6
Look under Analytic Geometry in the Index for similar problems.
Exam #261991
519911
The curve, clearly symmetric with respect to both the xaxis and the yaxis,
lies on or between the lines x = ±l. On [1, 1/2], Y = ±V3(x+ 1). On
[1/2,1/2]' Y = V3/2. On [1/2, 1], Y = V3(lx). The curve is a regular
hexagon of edge length 1 with vertices (±1/2, V3/2), (±1/2, V3/2) and
(±1,0). The enclosed area (that of six equilateral triangles of unit edge
length) is 6 x 1/2 x 1 x 1 x V3/2 = 3V3/2.
Look under Analytic Geometry in the Index for similar problems.
S 19912
The number of kdigit numbers without any 1 's is 8 X 9
k

l
, so between
1 and 1,000,000 inclusive there are 8(1 + 9 + 9
2
+ ... + 9
5
) = 9
6
 1 =
531,440 numbers without 1 's and 468,560 numbers with 1 'so
Alternate Solution The number of ways of filling six spaces with
digits 0, 2, 3, 4, 5, 6, 7, 8, 9 is 9
6
. But 000000 is not between 1 and
1,000,000, so there are 9
6
1 = 531,440 numbers without 1 's and 468,560
with l's.
Look under Enumeration in the Index for similar problems.
519913
Label the areas of the eleven regions inside the parallelogram as illustrated
below.
Let K be the area of the parallelogram ABGD; then
(1) area ABR = K/2 = r3 + W2 + W3 + W5 + b,
(2) area ARD + area BRG = K/2 = r2 + Wl + rl + r4 + W4 + r5,
Exam #261991 137
r          ~                   ~ c
A ~ ~                         ~ ~
(3) area QAB = area QAF, so T3 = Wl + b + W4'
(4) From (1) and (2), T3 +W2+W3 +W5 +b = T2 +Wl +Tl +T4 +W4 +T5.
(5) Substituting T3 = Wl + b + W4 from (3) in the left side of (4) and
Wl + W4 = T3  b (from (3)) in the right side, z= Wi + 2b = z= Ti  b.
Using K as the sum of the reds + sum of the "whites" + blue, we get
red area less the blue area is K/2 regardless of the choices of points F,
Q, and R.
Look under Enumeration or Geometry in the Index for similar problems.
519914
The sum a + b + c is even, so the sum a + b + c is even since parity is
not affected by a sign change in a summand. The identity
a
2
+ b
2
+ c
2
 2ab  2bc  2ca = a
2
+ b
2
+ c
2
 2ab + 2bc  2ca  4bc
suggests taking
Indeed,
a
2
+ b
2
+ c
2
 2ab  2bc  2ca
n= 
4
(
a+b+c)2 b
=n+ c
2 '
(
ab+c)2 b
=n+ a
2 '
(
a+bc)2
2 = n+ abo
and
Look under Number Theory in the Index for similar problems.
138 Solutions
519915
Remark: Ifn = a
2
+b
2
and m = c
2
+rf2, then nm = (acbd)2+(ad+bc)2.
Thus, the product of two sums of squares is also a sum of squares.
(a) Take a = c and b = d, then (a
2
 b
2
)2 + (2ab)2 = (a
2
+ b
2
)2. It
follows that if n is a sum of two squares, then n
2
is a sum of two squares.
The first example 12 + 4
2
= 17, 3
2
+ 3
2
= 18 leads to infinitely many
more, namely 8
2
+ 15
2
= 172 = 289, 12 + 172 = 290, and so on. From the
example 3
2
+4
2
= 52 = 25, 12 +5
2
= 26, one can generate infinitely many
more by taking (3k)2 + (4k)2 = (5k)2 = 25k
2
and 12 + (5k)2 = 1 + 25k
2
.
Any Pythagorean triple will serve as a starting point.
(b) If n  1, n, n + 1 are consecutive sums of two squares, then
n
2
1 = (n l)(n + 1), n
2
, n
2
+ 1 are consecutive sums of two squares.
For starters, take 72 = 6
2
+ 6
2
,73 = 3
2
+ 8
2
,74 = 52 + 7
2
.
(c) In any set of four consecutive positive integers one must be
congruent to 3 modulo 4. This integer cannot be a sum of squares since
a
2
+ b
2
== 0, 1 or 2 mod 4 for any a and b. Therefore, it is not
possible to have four consecutive positive integers which are sums of two
squares.
Look under Number Theory in the Index for similar problems.
519916
We have
lm= lmn  ++ ... + 1
. n 1· ( 1) (1 1 1 )
n+oo Hn n+oo n n + 1 n + 2 n + n
= lim (.!.) ( 1 + 1 + ... + :_1;)
n+oo n 1 + l/n 1 + 2/n 1 + n/n
 lim (.!.) t 1
 n+oo n k=l 1 + k/n
[1 1
= J
o
1 +x dx
= In(l + = In2.
Therefore,
lim Hn = _1_.
n+oo n In2
Look under Limit Evaluation in the Index for similar problems.
Exam #271992 139
519917
By reducing the coordinates of each of the nine given points modulo 3,
each can be identified with one of the nine points in the square array S:
{( 1, 1), (1,0), (1,1), (0, 1), (0,0), (0,1), (1, 1), (1,0), (1, I)}.
The existence of three lattice points with a lattice point as centroid
is equivalent to having a row, column, or diagonal (broken diagonals
included) of a 3 x 3 square array completely occupied. If only four or
fewer squares were occupied by the nine points, then one or more would
be occupied by at least three points which would have a lattice point as
centroid. Therefore, at least five of the squares in the 3 x 3 square array
must be occupied. To avoid three points in a row, two rows would contain
just two points each. It is now easy to show that it is impossible to place
the fifth without completing a row, column, or diagonal, (possibly broken).
Look under Analytic Geometry or Enumeration in the Index for similar
problems.
Exam #271992
519921
The curve is a cycloid which can be parameterized as x = t  sin t and
y = 1  cos t for ° ::; t ::; 27r. The arc length is given by
127r J(l cOS(t))2 + (sin(t))2 dt = 127r J2(1 cos(t)) dt
= 127r V4sin2(t/2)dt
= = 8.
Look under Integration in the Index for similar problems.
519922
Observe that the ith tallest person is viewable if and only ifhe is in front of
all the people 1,2,3, ... , iI. This happens with probability t. Therefore,
the answer is
n 1
L  ::.:::; In n + 0.5772157 ....
i=l n
Look under Probability in the Index for similar problems.
140 Solutions
519923
Select a head gossip Va. First have all n 1 other people send a letter with
their scandal to Va; then have Va distribute the assembled information via
n  1 letters to everyone else. We claim that this total of 2n  2 letters is
the minimum.
Let each letter be represented by an arc in a directed graph. No one can
know all the scandal until at least n  1 letters have been sent, for we
need a sub graph of arcs unilaterally connected with directed paths from
every vertex into whichever vertex becomes fully informed. Then each
subsequent letter can inform one more person, so at least n  1 more
letters are required. This gives a bound of 2n  2 which we have already
achieved above.
Look under Enumeration in the Index for similar problems.
519924
For n = 1, Zl has no proper subgroup, contrary to the stated condition. For
n = 2, Z2 works because it has 1 subgroup. For n = 3, Z3 fails because
it has 1, and not 2, proper subgroups. For n = 4, we need 4 subgroups.
The Klein 4group works, whereas Z4 only has 2. Thus we have found 2
groups that satisfy this condition, namely Z2 and K
4
.
We shall now show that no group of order n ::::: 5 has so many subgroups.
First, every subgroup contains the identity element, e, so the number of
subgroups of order i cannot exceed the number of subsets of order i  1
selected from G  { e }. There are ( 7 ~ ; ) of these, and the largest subgroup
order i is at most l ~ J. For any odd n = 2k + 1, this yields at most
k
8 (7 ~ n < 2
2k

1
= 2
n

2
subgroups. For any even n = 2k, we can find at most
k
L (7 ~ :) = 2
2k

1
t=l
= 2
n

2
subgoups. That is just barely enough, but only if every set of order l ~ J is
a subgroup. For n :s; 4, this is true of Z2 and K
4
.
But for n ::::: 6 this requires every subset of order 2 and 3 be a subgroup.
That is, {e, a} and {e, a, b} must both be subgroups. But the first forces
Exam #271992 141
the order of a to be 2 while the second requires the order of a to be 3, a
contradiction.
Look under Group Theory in the Index for similar problems.
_r2 (Xl
= He 10 = H.
Therefore I = Vi.
Look under Integration in the Index for similar problems.
519926
(a) The characteristic polynomial of A is
X2 13X +36 = (X 4)(X  9).
We find that the eigenvectors for A are and C2). This gives
p=(1 1)
3 2
and
pl = (2 1)
3 1 '
and therefore
(
1 1) (4 0) (2 1)
A = 3 2 0 9 3 1 .
(b) Clearly all four choices of signs in C = :3) yield C
2
=
Therefore, all four choices of B = PCpl give
142 Solutions
as desired. These are
(
5 1 ) ( 5 1) (13 5) ( 13 15 )
B = 6 0' 6 0 ' 30 12' 30 12 .
(c) If A has a repeated eigenvalue, for example if A = 2
n
I, we can
find at least 2(n + 1) different B's; namely, for any i with 0:::; i :::; n we
may have
(
0 2i)
B = ± 2ni 0 .
Thus, there is no maximum.
Look under Matrix Algebra in the Index for similar problems.
519927
Solutiou 1 Since y
x/VI  e
2x2
to get
x
2
is a continuous function, we may square
Now this has the form ~ so we may apply L'H6pital's Rule to get
lim 2x 2 = L2 = !.
x+o+ 4xe
2x
2
Therefore L = ± (1/ V2). But L 2: 0 since the fraction is positive on the
domain. Therefore L = + (1/ V2).
Solutiou 2 Since the form is g, we apply L'H6pital's Rule to get
VI  e
2x2
lim 2 2
x+o+ 2xe x
Since e
2x2
~ 1, this reduces to
VI  e
2x2
lim ,
x+o+ 2x
provided this limit exists. But this also equals L/2 provided L =1= o. We
find L = A so that, as above, L = ±(I/V2).
Look under Limit Evaluation in the Index for similar problems.
519928
We want integer solutions to a
2
+ (a + 1)2 = b
2
. This can be rewritten as
1 = 2b
2
 (2a+ I?
Exam #271992 143
Upon replacing c = 2a + 1, we have a Pell equation 1 = 2b
2
 c
2
• The
first solution is b
l
= Cl = 1. In general, given a solution (b
n
, cn), we can
generate another solution
because
Thus, we find the solutions (b
2
, C2) = (5,7), (b
3
, C3) = (29,41), (b
4
, C4) =
(169,239), ....
Can there be any other solutions? We can also write the recurrence in
reverse as (bn, cn) = (3b
n
+
1
 2C
n
+l, 4b
n
+
1
+ 3c
n
+d. Suppose that
(b, c) is the smallest solution not on our indexed list.
Upon applying the reverse recurrence we find that (3b  2c, 4b +
3c) is another smaller solution that also cannot be in our list. But this
contradicts our selection of (b, c). Consequently. the list includes every
possible solution.
Can we find an explicit solution of the simultaneous recurrence? Ap
plying the recurrences repeatedly we find
bn+1 = 3bn + 2cn
= 3bn + 2( 4bn  1 + 3Cn l)
= 3bn + 8b
n

1
+ 6( 4bn + 3c
n
)
= 21b
n
+ 8bn 1 + 18cn·
Now nine times the first equation minus the last yields
Similarly, we can show that Cn satisfies the same recurrence. Namely
Cn+! = 6c
n
 CnI' The roots of x
2
 6x + 1 are associated with this
recurrence, namely r = 3 ± VS.
We find that
b
n
= (2
4
J2) (3+vs)n+ (2+4J2) (3vs)n,
C
n
= ( J2
2
1 ) (3 + vs)n  ( J2
2
+ 1) (3  vs)n,
144 Solutions
and recall
an = en;l = ( \1"24
1
) (3 + V8)n _ ( \1"2
4
+ 1) (3 _ V8)n.
Look under Diophantine Equations in the Index for similar problems.
Exam#2S1993
519931
Let P(x) = ax
3
+ bx
2
+ ex + d. Then P'(x) = 3ax
2
+ 2bx + e;
P" (x) = 6ax + 2b; and pili (x) = 6a. The following inequalities are
equivalent:
pili ( X ) ! . (P" ( x ) ) 2
P'(x) < 2 P'(x) ,
2P'(x)P"'(x) < (p"(x)) 2,
2(3ax2 + 2bx + e)(6a) < (6ax + 2b)2,
36a
2
x
2
+ 24abx + 12ae < 36a
2
x
2
+ 24abx + 4b
2
,
12ae < 4b
2
,
(2b)2  4(3a)e > O.
Since P' (x) has distinct zeros, the last inequality holds and the result
follows.
Look under Polynomials in the Index for similar problems.
519932
For simplicity of notation, we write f and 9 for f(a) and g(a) and omit
parentheses.
f = gff = (fg!)ff = f(gff)f = fff
= ff(gff) = f(fg!)f = fgf
=g.
Look under Group Theory in the Index for similar problems.
Exam #281993 145
519933
Since A has rank 1, it has at least one nonzero row, r, and each of its
rows is a multiple of r. Therefore there is a 1 x n row vector X for which
A = XT . r. Now, for any p x q and q x p matrices Band C, the trace of
B . C equals the trace of C . B. Therefore we have
1 = trace of A = trace of (r . XT) = r . XT.
(The last equality holds since (r . XT) is a 1 x 1 matrix.) Thus,
A2 = (XT . r) . (XT . r) = XT . (r· XT) . r = XT . r = A.
Look under Matrix Algebra in the Index for similar problems.
519934
Let O(n) be the number of orderly permutations of {I, 2, ... , n}. We will
prove by induction that O(n) = 2
n

1
. Since both 12 and 21 are orderly,
0(2) = 2
2

1
= 2. Now assume that O(k) = 2
k

1
and note that any orderly
permutation of {I, 2, ... , k} must begin with 1 or k. (Let j be the first
number. Wherever the 1 is, 2 must follow it; 3 must follow the 2; and so
on and inductively, j must follow the 1.) For any orderly permutation, P,
of {I, 2, ... , n} we create two orderly permutations of {I, 2, ... , (k + I)}
as follows:
1) place (k + 1) at the begining of P, or
2) increase each number in P by 1 and place 1 at the beginning.
(For example, with k = 4 and P = 1423, we create 51423 and 12534.)
This process thus creates 2k distinct orderly permutations of {I, 2, ... ,
(k + I)}. Since, for any orderly permutation of {I, 2, ... , (k + I)}, we
can reverse the process and obtain the orderly permutation of {I, 2, ... ,k}
which created it, we see that 0 (k + 1) = 2k, completing our induction.
Look under Permutations in the Index for similar problems.
519935
Solution 1 Since a;+l = a; + an, one can show by induction that
n n
 'n<a <
2 YH_n_
2
,
so
an 1
 ~  .
n 2
146
Solution 2 Since a;+l = a; + an, one has
an+ 1  an = va; + an  an
a; + an  a;
va; + an + an
1
Solutions
Next since al = 1, the an's are increasing and the calculations above
show that
so an 7 00. Next, using an 700, the same calculations show that
This, in turn, shows
an 1
 7.
n 2
Look under Limit Evaluation in the Index for similar problems.
519936
First note that there are integers An and En for which
and further that
Thus, since (A
n
/2
n
) is an integer, An is divisible by 2n. Note that
is divisible by 2n+l. Finally, since 0 < (J3 . 1)2n < 1, it follows that
Look under Number Theory in the Index for similar problems.
Exam #281993 147
519937
If there are only two distinct integers, X < Y, among A, B, C, D,
E then there would be at most three distinct sums of pairs of them:
X + X < X + Y < Y + Y. Also, if there are at least four distinct
integers, W < X < Y < Z among A, B, C, D, E then there would be at
least five distinct sums of pairs of them: W + X < W + Y < X + Y <
X + Z < Y + Z. Therefore there are exactly three distinct integers among
A, B, C, D, E and we have the following six cases:
1) A = B = C < D < E,
3) A = B < C < D = E,
5) A < B = C < D = E,
2) A = B < C = D < E,
4) A < B = C = D < E,
6) A < B < C = D = E.
The first three cases are impossible since the smallest sum 401, is odd.
In case (4), we must have
A + B = 401, 2B = 546, A + E = 691, and B + E = 836.
(Why?) This yields
A = 128, B = C = D = 273, E = 563.
Similarly, in case (5) we have
A + B = 401, 2B = A + D = 546, B + D = 791, and 2D = 836,
which yields
A = 128, B = C = 273, D = E = 418.
In case (6) we have
A + B = 401, A + C = 546, B + C = 691, and 2C = 836,
from which we obtain
A = 128, B = 273, C = D = E = 418.
In summary, there are three collections of integers satisfYing the given
conditions:
A = 128, B = C = D = 273, E = 563;
A = 128, B = C = 273, D = E = 418;
and A = 128, B = 273, C = D = E = 418.
Look under Logic or Systems of Equations in the Index for similar
problems.
148 Solutions
519938
Let the area of abed be one 'unit'. We will determine the area of ABCD
in units. Since = = 5, and = = 4, we see that LAdD
contains 5x4 = 10 units Similarly since Ba = £ = 6 and Aa = £ = 3
2 . , ba 1 ad 2 '
LAaB contains = 9 units. Noting that 6AdD = 6CbB and
6AaB = 6CeD, we see that ABCD contains 2 x 10 + 2 x 9 + 1 = 39
units.
Finally, since area of ABC D = 48, we see that the area of abed is
Note: In general it can be shown that if 1, 5, 2, and 6 are replaced by
w, x, y, and z, then the area of abed is
wy(w + x)(y + z)
xz + (w + x)(y + z)·
Look under Geometry in the Index for similar problems.
Exam #291994
519941
Using L'H6pital's Rule,
fX et2dt x2
1
· Jo l' e
1m 2 = 1m
x+oo eX / x x+oo x·2xe
x
eX
x2
x
2
1
= lim
x+oo 2X2  1 2"'
Look under Limit Evaluation in the Index for similar problems.
519942
We have
n(n 1) 2 (2)n/2
(1 + J2/n)n = 1 + + .  + ... + 
2 n n
(
2)n/2
... +:;;: >n.
Now take nth roots.
Look under Inequalities in the Index for similar problems.
Exam #291994 149
519943
The inequality is
x
2
X Y
In<.
y2 y x
Let f(p) = Inp2  p + pl. Then l' (p) = _p2(p  1)2 < 0, for p > 1.
Since f(l) = 0, it follows that f(p) < 0 when p > 1. This implies the
inequality.
Look under Inequalities in the Index for similar problems.
519944
If C is the largest and A the smallest angle of the triangle, then the
law of cosines gives cos C = 1/8 and cos A = 3/4. Then cos 2A =
2 cos
2
A  1 = 1/8 and it follows that, since A and C are between 0 and
7f, 2A = C.
Look under Geometry in the Index for similar problems.
519945
Let the integers in nondecreasing order be a, b, c, d, and e. Then abede =
a + b + e + d + e ::; 5e so abed::; 5.
There are five cases:
abed = 1 implies e = 4 + e, impossible,
abed = 2 implies 2e = 5 + e, so e = 5,
abed = 3 implies 3e = 6 + e, so e = 3,
abed = 4 implies 4e = 6 + e or 4e = 7 + e; the first gives e = 2 and the
second is impossible, and
abed = 5 implies 5e = 8 + e, but since d = 5, e = 2 is impossible.
The three solutions are thus {1, 1, 1,2, 5}; {1, 1, 1, 3, 3}; and {1, 1,2,2, 2}.
Look under Number Theory in the Index for similar problems.
519946
Suppose that the circle contains three distinct rational points. Then, sub
stituting their coordinates in x
2
+ y2 + Ax + By + C = 0 determines
rational values for A, B, and C. But then the coordinates of the center,
(A/2,  B /2), are both rational, a contradiction.
Look under Analytic Geometry in the Index for similar problems.
150 Solutions
Exam #301995
S 19951
Let u = Z1 + /zr  z ~ , and v = Z1  /zr  z ~ . Then u+v = 2zl. and
u  v = 2/zr  z ~ . By the Parallelogram Identity,
1
lul
2
+ Ivl2 = 2(1u + vl
2
+ lu  v1
2
)
= 21z112 + 21zr  z ~ l ,
from above. But, luvl = Izr  (zr  z ~ ) 1 = IZ212. Hence
(lui + IV\)2 = lul
2
+ 21uvl + Ivl
2
= 21z112 + 21z212 + 21zr  z ~ l .
Again, by the Parallelogram Identity,
21z112 + 21z212 = IZ1 + z21
2
+ IZ1  z212.
Noting that Izr  z ~ 1 = IZ1 + z21·lz1  z21, we see that
(lui + IV\)2 = (IZ1 + z21 + IZ1  z21)2.
Since both bases are nonnegative, this yields lui + Ivl = IZ1 + z21 + IZ1  z21
which establishes the required result.
Look under Complex Numbers in the Index for similar problems.
S 19952
Let w be a complex cube root of unity, with w i= 1. Then w is a root
of x
3
 1 = (x  1)(x
2
+ X + 1) = 0, Since w i= 1, we must have
w
2
+ w + 1 = 0. Let 8 be the given sum,
8 _ ~ ( 3n )
1  6 3k+ 1 '
k=O
8
2
= I: (3k
3
: 2)'
k=O
Use w
3
= 1 and the Binomial Theorem to get
(1 + 1)3n = 8 + 8
1
+ 8
2
,
(1 + w)3n = 8 + w8
1
+ w
2
8
2
,
(1 + w
2
)3n = 8 + w
2
8
1
+ w8
2
.
Exam #301995 151
Use 1+w+w
2
= 0, so that (1+w)3n = (_w
2
)3n = (_1)
3n
w
6n
= (_1)n,
since w
3
= l, and(1 + w
2
)3n = (_w)3n = (_1)
3n
w
3n
= (_1)n, and add
the three results above to get
Hence
2
3n
+ 2. (1)n
S = :3,.....''.
Look under Finite Sums in the Index for similar problems.
519953
Let (y'5 + 2)P  2
P
+
1
= N + f, where N is an integer and ° < f < 1.
We need to show 20p divides N. Let f' = (y'5  2)p. Then, since
° < y'5  2 < 1, we have ° < f' < 1. Hence 1 <  f' < 0, and
1 < f  f' < 1. But
N + f  !' = (y'5 + 2)P  2
P
+
1
 (y'5  2)P
= 2(p . 1 . 2 . 5(pl)/2 + P . 3· 2
3
. 5(p3)/2
+ ... + p . p  2 . 2
P

2
. 5 + 2
P
)  2
P
+
1
is an integer, since p is odd. Hence f  f' is an integer, which must be
= 0, since 1 < f  f' < 1. Thus
N = 4 . 5(p . 1· 5(p3)/2 + p . 3 . 22 . 5(p5)/2 + ... + pp  22
P

1
.
But p is an odd prime implies p divides pr for every r such that 1 :::; r :::;
p  1. This shows that 20p divides N.
Look under Number Theory in the Index for similar problems.
519954
Let
Then
!,(x) = 12(x
2
+ X + 1)2(2x + 1)  27(2x(x + 1) + 2x
2
(x + 1)).
We can easily verify that
° = f (1) = f (  ~ ) = f (  2) = !' (1) = !' (~ ) = f' (  2).
152 Solutions
Since f (x) is a sextic, we conclude that 1,  ~ , and 2 are double zeroes
of f (x) and that f (x) has no other zeros. As the leading coefficient of
f(x) is 4, we now see that
f(x) = (x  1)2(2x + 1)2(x + 2)2.
This establishes the result.
Look under Polynomials in the Index for similar problems.
519955
Let N = 11 .2
2
.3
3
... nn. Then there are
n(n + 1)
1+2+3+"'+n= 2
factors in N, their geometric mean is N2/n(n+l), and their arithmetic mean
equals
1 + (2 + 2) + (3 + 3 + 3) + ... + (n + n + ... + n)
n(n+l)
2
2(1 + 22 + 3
2
+ ... + n
2
)
n(n + 1)
2n(n + 1)(2n + 1)
6n(n + 1)
2n+ 1
3
But the geometric mean of k unequal positive nUlllbers is always less than
their arithmetic mean. Hence, N
2
/
n
(n+l) < 2n + 1/3, that is,
(
2n+ 1)n(n+l)/2
N< 
3
Look under Inequalities or Integration in the Index for similar problems.
519956
Let G be the centroid of the triangle. Then G is
b+! (aV3) = b + ~ = c
3 2 2V3
units above L.
1 aV3 a
2
V3 .
The area of the triangle ="2a ' 2 = 4 square umts.
Exam #301995
By Pappus's Theorem, the required volume
V = (27rc) a
2
J3
4
= 7ra
2
J3 ( b + ~ )
2 2J3
7ra
2
J3 7ra
3
=+
'f 4
or = 7r: (2J3b + a) cubic units.
Look under Volumes in the Index for similar problems.
S19957
153
Solution 1 Let G be the centroid of the triangle. Then G has coordinates
 "" acoso' "" asino' .
(
1 3 1 3 )
3 ~ " 3 ~ t
i=l i=l
Clearly, by looking at the coordinates of the vertices, we see that the
circumcenter of the triangle is the origin, 0. Let H be the orthocenter of the
triangle. Since, by geometry, 0, G, H are collinear and OG : OH = 1 : 3,
we see that the coordinates of Hare
(t, a cos Oi, t, a sin Oi) .
Solution 2 (Solution 2 does not use the property of the Euler line
quoted and used in Solution 1. In fact, it proves that 0, G, Hare
collinear and that OG : OH = 1 : 3.) Let A, B, C be respectively
(a cos 0i, a sin Oi), i = 1, 2, 3, and H be the orthocenter. Observe that the
origin, 0, is the circumcenter of triangle ABC. If L is the middle point
of BC, then OL.lBC. But AH.lBC also. Hence AHIIOL, i.e., AH is
parallel to the vector
II
I (02+03) (0
2
0
3
). (0
2
+0
3
) (0
2
0
3
))
\ cos 2 cos 2 ' sm 2 cos 2
II
I (02+03) . (0
2
+0
3
))
\ cos 2 ,sm 2 .
154
Hence, the parametric equations of AH are
{
X = aCOSDl + tcOS(<>2t<»)
, t E R.
Y = a sin Dl + t sin( <>2t<>3 )
Similarly, the parametric equations of BH are
{
X = aCOSD2 + UCOS(<>3t<»)
, t E R.
Y = a sin D2 +usin(<>3t<>3)
These two lines meet at H(x, y), where
= aCOSDl + aCOSD2
2 . (ct
1
 D2) . (Dl + D2)
= asm 2 sm 2
and
. (D2 + D3) . (ct
3
+ Dl )
tsm 2  usm 2
= a sin Dl + asinD2
2 . (ct
1
 D2) (Dl + D2)
=  a sm 2 cos 2 '
Solving for t we find,
D3  D2
t = 2acos 2 .
Thus,
Solutions
x = a cos ctl + 2a cos ( D3 ; D2 ) cos ( ct3 ; D2 ) cos ( D3 ; D2 )
= a cos ctl + a cos D2 + a cos D3.
Since y is obtained from x by replacing cos with sin, we see that y =
a sin Dl + a sin D2 + a sin D3. Thus the coordinates of H are
Look under Geometry in the Index for similar problems.
Exam #301995
519958
We have
(1+x)n k) k}
(1 x)3 = k x "2 + 1)(k + 2)x .
The coefficient of x2n is
(7) (2n)(2n + 1)
+ ... +
= t  k + 1)(2n  k + 2)
k=O
= ta {(2n + 1)  k}{ (2n + 1)  (k  1)}
= ta 1)2  2n: 1 ta 1)
+ ta 1)
= +4n+ 1 +2n+ 1) ta (2n+ 1) ta
+ ta 1)
1 1
= "2 (4n
2
+ 6n + 2)2n  (2n + 1)(n· 2
n

1
) + "2 n(n  1)2
n

2
= 2
n

3
{ 4(4n
2
+ 6n + 2)  4n(2n + 1) + n
2
 n}
= 2
n

3
(9n
2
+ 19n + 8).
155
Note: We have used the following results, obtained by putting x = 1
in the binomial expansion of (1 + x)n and in the results obtained by
differentiating with respect to x once and twice.
156 Solutions
Of course, we have also assumed that Ixl < 1, so that
1 ~ k 1 ~ kl
1  x = L.. x, (1 _ x )2 = L.. kx ,
k=O k=O
and
1 1 L:
oo
k2
( )3 =  k(k1)x .
1 x 2
k=O
Exam #311996
519961
The problem is equivalent to finding digits (integers from 0 to 9) a, b,
c such that n = lOa + b ;::: v1000 and n
2
 c
2
is divisible by 100.
Factoring, we see that (lOa + b + c)(10a + b  c) is divisible by 100. If
5 divides both factors, then 5 divides their sum 20a + 2b and therefore
divides b. Otherwise, one of the factors must be divisible by 25, and in
fact by 50 since the two factors differ by the even number 2c. Thus one
of a + (b + c)/10), a + (b  c)/10 is a mUltiple of 5. Since band c are
digits, (b ± c)/10) must be 0 or 1 and a = 0,4,5,9. But for this problem,
lOa + b ;::: 32, so either a = 4,5,9 or b = 0,5. There are 30 + 13  6 = 37
numbers of this form and thus 36 others besides 43
2
= 1849.
Look under Number Theory in the Index for similar problems.
519962
Let ABCDE denote the pentagon P with LA = 90°. Then the area oftri
angle EAB is 1/2 and quadrilateral T = EBCD has sides y'2, 1, 1, 1, l.
We may assume that P is convex. Note that
area(EBCD) = area(BCD) + area(DEB) = 1/2sinC + V2/2sinE.
Also, by the law of cosines,
so
BD2 = 2  2cosC = 3  2V2cosE
1 y'2 1
 cos C   cos E = .
224
Squaring and adding both equations,
2 11 V2
(area(EBCD)) = 16  2 cos(C + E).
Exam #311996 157
This is clearly maximized when C + E = n, (which is true if and only if
EBCD is a trapezoid) and the area of EBCD is
Jll yf2
16 + 2 .
Look under Geometry in the Index for similar problems.
519963
(a) Number the cats from 1 to 100, and consider the nth cat. For each
even divisor of n, he will get a dollar and for each odd divisor, he'll
lose a dollar. If n = 2
a
then the number of even divisors is
a (ai + 1) and the number of odd divisors is (ai + 1). Hence
the nth cat will get f (n) = (a  1) (ai + 1). The cat loses money if
a = 0, i.e., n is odd, so 50 cats lose money. If a = 1, the cat breaks even,
and there are 25 of these. Finally, the remaining 25 cats profit.
(b) To maximize f(n), note that if a < 2, f(n) ::::; a and if2
a
b ::::; 100,
for a = 6, 5,4, 3, 2, then b ::::; 1, b ::::; 3, b ::::; 6, b ::::; 12, b ::::; 25. So
calculating f(2
6
) = 5, f(2
5
·3) = 8, f(2
3
.3
2
) = 6, and f(22 ·3·5) = 4
gives the 96th cat to gain the most with $8.
(c) The total amount the rats left with is
100
2) J = 68.
k=l
Look under Number Theory in the Index for similar problems.
519964
Let the two x coordinates of the square's vertices be a and b with a> b.
Then a
3
3a = b
3
3b = ab so a = b
3
2b and b = 4aa
3
. Equivalently,
a
3
= 4ab and b
3
= a+2b. Subtracting, (ab)(a
2
+ab+b
2
) = 3(ab)
so a
2
+ ab + b
2
= 3. Substituting a = b
3
 2b,
b
6
 4b
4
+ 4b
2
+ b
4
 2b
2
+ b
2
= 3,
b
6
 3b
4
+ 3b
2
= 3,
b
6
 3b
4
+ 3b
2
 1 = 2,
(b
2
 1)3 = 2,
b
2
= 1 +v2.
ISS
The area of the square is
(b
3
_ 3b)2 = b
2
(b
2
_ 3)2
so A + B = 108.
= (1 + ?'2)(?'2  2)2
= (1 + ?'2)( + 4  4?'2)
= 6  = 6 + {I108
= {l216 + {I108
Solutions
Look under Analytic Geometry in the Index for similar problems.
519965
Reflecting ABC about the yaxis and then the result about the xaxis, gives
a closed curve r bisecting the area of a square of side length VB. Thus r
encloses a region of area (VB)2 = 4. By the isoperimetric theorem, if a
closed curve r encloses a fixed area, the length of r is minimized when
r is a circle. Therefore r is a circle, say of radius r, with 7fr2 = 4, so
r = vf41ir and the length of 'Y is
1 1
'Y = (27fr) = (27f)  = Vi·
4 4 7f
Look under Geometry in the Index for similar problems.
519966
Such an event is mathematically possible if out of the 2n children, exactly
n are males. Thus, Pn = 2n e:). The expression n . P; can be written
as
2 = (1/2)4n(2n)2 = (2n)!(2n)!
nPn n n n (2nn!)4
1 . 2 . 3 ... 2n 1 . 2 . 3 ... 2n
= n,::: ,::::
(2·4·6···2n)2 (2·4·6··· 2n)2
1·3·5· .. (2n  l) 1 ·3·5· .. (2n  l)
= n ..,=''' ,::'::'
2 . 4 . 6 ... 2n 2 . 4 . 6 ... 2n
1 . 3 . 3 . 5 . 5· .. (2n  l) ... (2n  l)
2 . 2 . 4 . 4 . 6 ... 2n . 2n
1 3 3 5 5 2n  1 2n  1 n
22446 2n  2 2n 2n
Exam #311996 159
Thus,
. 2 13355 2
2 hm np =      ... = 
n+oo n 2 2 4 4 6 n
by Wallis' formula, and the answer is 1/ yI7f.
An alternate solution can be found using Stirling's Formula: Since
n!
(
2n) = 2
2n
n [( )ny'2nn ] 2 y7m
whence Vii . Pn fir·
Look under Limit Evaluation or Probability in the Index for similar
problems.
S 19967
If {Xl, X2, X3, X4} is statistically equivalent to {I, 9, 9, 6} then
1 4 1
4 LXi = 4(1 + 9 + 9 + 6)
i=l
so that Xi = 25. If (J" is the standard deviation, then
and we must have
4
LX: = 12 + 9
2
+ 9
2
+ 6
2
= 199.
i=l
Letting Yi = Xi  6,
4 4
LYi = 1 and LY; = 43.
i=l i=l
Since squares are congruent to either 0 or 1 mod 4, three of the Yi (say
Yl, Y2, Y3) must be odd and one (say Y4) must be even. As odd squares
are congruent to 1 mod 8, we have + 3 == 43 mod 8 whence Y4 must
be a multiple of 4. Since each iYii < 7, iY4i = 0 or 4. If iY4i = 4, then
we must have three odd squares which add to 27. This is possible in two
different ways: 27 = 25 + 1 + 1 = 9 + 9 + 9. Checking to see if any of
the signed sums of the elements from {5, 4, 1, I} or {4, 3, 3, 3} equals 1,
160 Solutions
we find 5  4 + 1  1 = 1, 4  5 + 1 + 1 = 1, and 4  3 + 3  3 = 1. The
Xi = Yi + 6 = {11, 2, 7, 5}, {10, 1, 7, 7}, and {10, 3, 9, 3}. If IY41 = 0,
we must have three odd squares which add to 43. This is possible only
one way: 43 = 9 + 9 + 25. Checking to see if any of the signed sums of
the elements from {O, 3, 3, 5} equals 1, we find 0 + 3 + 3  5 = 1. Then
Xi = Yi + 6 = {1, 9, 9, 6}.
Look under Diophantine Equations in the Index for similar problems.
519968
Let (a, f(a)) and (b,J(b)) be arbitrary points on the graph with b > a
and let c = at
b
• We are given that
or equivalently,
f(b)  f(c) f(c)  f(a)
_.....:b'=c'::_=c'=a=_ = 8.
b a
Take the limit as b + a (and a fortiori b + c, c + a) to see that f" (a) = 8.
Since a was arbitrary, we have Y = 4X2 + Ax + B.
Look under Differentiation in the Index for similar problems.
Exam #321997
519971
We will consider combinations of the last four digits only. If the four digits
are all different, then there are C40) = 210 combinations. If we have three
different digits, then there are 3 . C30) = 360 combinations (e.g., the digits
{1, 2, 3} generate three combinations, 1123, 1223, and 1233). If we have
two different digits, then there are 3 C20) = 135 combinations (e.g., the
digits {1, 2} generate three combinations 1112, 1122, and 1222). Finally,
if all the digits coincide then there are ten combinations. The total number
of combinations is 210 + 360 + 135 + 10 = 715 and the probability is
p = 715/10,000 = 0.0715
Look under Enumeration or Probability in the Index for similar problems.
Exam #321997
S 19972
We calculate sums directly,
By using two wellknown formulas the above is equal to
~ (n(n+ 1)(2n+ 1) n(n+ 1)).
2 6 + 2
161
SimplifYing we get our answer n(n + 1)(n + 2)/6. It is interesting to note
that these numbers appear as the first n numbers in the 3rd diagonal of
Pascal's triangle.
Look under Finite Sums in the Index for similar problems.
S 19973
Note first that z4  5z
2
+ 6 = (z2  3)(z2  2). So
\z4  5z
2
+ 6\ = \(z2  3)(z2  2)\ = \Z2  3\\z2  2\.
But we know that \Zl  Z2\ ~ \\Zl\\Z2\\ for all Zl and Z2 in C. So
\z2  3\ ~ \\z2\  3\ = \\Z\2  3\ = \9  3\ = 6
and
\Z2  2\ ~ \\Z2\_ 2\ = \\Z\2  2\ = \9  2\ = 7.
Thus, \z2  3\\Z2  2\ ~ (6)(7) = 42. Finally,
21 21 1
.,..;,;:::; <  = .
\z4  5z
2
+ 6\  42 2
So letting A = 1/2 we get an upper bound. Further, A is the smallest
upper bound since if Z = ±3, then equality holds.
Look under Complex Numbers in the Index for similar problems.
S 19974
If x
Y
= yX for x, y > 0, then taking the natural logarithm of each side we
obtain y In x = x In y, which implies
lnx Iny
x y
162 Solutions
Obviously all the above is true if x y. So suppose x =1= y. Then if
we graph the function v = In u, we see that the line through the points
(x,lnx) and (y,lny) goes through the origin and has equation
v=mu.
Hence In y = my and In x = mx. We also note that this is possible only
if x, y > 1 and m > O. Let y = tx where t > 0 and t =1= 1. Then
Simplifying we get
lntx = mtx,
In t + lnx = t(mx),
lnt + lnx = tlnx.
tl!(tl) = x.
Then y = tx = tt!(tl). So all pairs of the form (tl!(tl), tt!(tl)) satisfy
the original equation. (If we restrict t > 1 we get all solutions with x < y.)
Some pairs include:
519975
t = 2 =} (2,4),
t = 3 =} (J3,3J3),
t = 3/2 =} (9/4,27/8).
We claim that 16 points and 16 lines satisfy the six axioms. The proof
that this is the fewest number of points and lines will be deferred to the
solution of part (b) below. The six axioms will guide us in constructing
what is essentially a 4 x 4 array. Each row (and column) will contain four
points, thus giving us 16 points in our solution. The fact that we also have
16 lines will become apparent later. Let us begin the construction.
By Axiom 5, there is at least one line, call it rl. By Axiom 1, there
are exactly four points on this line, call them AI, BI, Cl, and D
1
. We
adopt the convention that a line in our system can be described by the four
points that it contains. So in particular, we write rl = AlBlClD
l
. Now
by Axiom 3 and Axiom 2, there is a line Cl that contains the point Al but
does not contain any of the points B
l
, C
l
, and D
l
. Let A
2
, A
3
, and A4 be
the other three points on line Cl. By Axiom 6 and Axiom 2, there are three
new lines r2 = A
2
B2C
2
D2, r3 = A3B3C3D3, and r4 = A
4
B4C4D
4
.
These lines together add a total of nine new points, namely, B
i
, Ci, and
Di for 1 :S i :S 3. We now have the 16 points in our system. By Axiom
6, we now add three new lines, C2 = BlB2B3B4, C3 = C
l
C
2
C
3
C
4
, and
Exam #321997 163
C4 = DID2D3D4. It is now easy to picture the 4 x 4 array: we have four
row lines rb r2, r3, and r4 and four column lines CI, C2, C3, and C4. Note
that this simple construction satisfies all of the axioms except Axiom 3.
To satisfy Axiom 3, we must add more lines (this is the tricky part).
We start by satisfying Axiom 3 for the point AI' The possible lines
through Al are:
(a) A
I
B
2
C
3
D
4
,
(d) A
I
C
2
D
3
B
4
,
(c) A
I
C
2
B
3
D
4
,
(f) A
I
D
2
B
3
C
4
·
Let us choose line (a). Then by Axiom 2, we must exclude lines (b), (c),
and (e). However, we can choose line (d) and that gives us four lines that
contain point AI. We now want to satisfy Axiom 3 for the point B
I
. As
above, Axiom 2 limits the possible lines that we can choose. The possible
lines through Blare:
We choose lines (1) and (3) to satisfy Axiom 3 for B
I
. Note that Axiom 3
is also satisfied for the points C
3
and D
3
. In a similar fashion we choose
the following four lines to complete our system:
(i) C
I
A
2
B
3
D
4
, (ii) C
I
D
2
A
3
B
4
,
This system of 16 points and 16 lines satisfies all of the axioms.
For each positive integer n, the problem statement associates a set of
six axioms to n. Say that a collection of points and lines satisfying these
axioms is an nconfiguration. If C is an nconfiguration and x, y are lines
in C, we say that x and y are parallel if they share no points in C. For each
line, the set consisting of the line and all those parallel to it is a family
of parallels. For n = 1, there is a configuration with two points and two
lines that satisfies the axioms, and no configuration with fewer points or
fewer lines will do. The rest of the solution to part (b) follows from the
following:
Theorem. Let n be a positive integer with n > 1. If an nconjiguration
exists, then there is one with n
2
points and n
2
lines, and no nconjiguration
with fewer points or fewer lines exists.
Proof Let n be an integer, n > 1, and let C be an nconfiguration. Let
x be any line of C and choose a point P on x. By Axiom 3, there exists
another line y that contains p. By Axiom 1, Y contains exactly n points,
say PI = P,P2,'" ,Pn' For each i,2 ::; i ::; n, there is by Axiom 6 a
unique line Xi containing Pi and parallel to x. Let Xl denote x.
164 Solutions
Using Axiom 6, it is easy to verify that "equal to or parallel to" defmes
an equivalence relation on the set of lines of C, so Xi is parallel to Xj
when i # j. Thus, the n lines Xl, ... , xn together contain n
2
points. We
wish to show that these are all the points of C. To this end, let Q be any
point of C that is not on Xl. Then there exists (by Axiom 6) a unique line
z through Q and parallel to Xl. Again by Axiom 6, there Camlot be two
lines through P and parallel to z. So since z is parallel to Xl, necessarily
z meets y, say at Pj' But then z must be Xj, so Q is on Xj, which is a
contradiction.
Therefore C has exactly n
2
points. It is then clear that every line has
exactly n lines in its family of parallels.
Let k be the number of families of parallels for C. Since each point
is on exactly one line from each family, Axiom 3 implies k ~ n. This
shows that C must have at least n
2
lines. Suppose that k > n. Choose any
family of parallel lines and discard it to obtain a collection B of n
2
points
and k  1 families of parallel lines. Since we discarded an entire family
of parallels, B satisfies Axiom 6. Each point of B is on at least n lines
of B, so Axiom 3 is satisfied. Since n > 1 and C is an nconfiguration,
B satisfies Axioms 1, 2, 4, and 5, and thus B is an nconfiguration. It is
clear that we may continue in this fashion and discard families of parallel
lines until we have a minimum of n families of parallel lines. The resulting
nconfiguration then has n
2
points and n
2
lines.
We have shown that any nconfiguration must have at least (exactly)
n
2
points and at least n
2
lines. We also showed that if an nconfiguration
exists, then there is one with exactly n
2
points and exactly n
2
lines. Thus
n
2
is the fewest number of points and lines for an nconfiguration. This
proves the Theorem.
Look under Synthetic Geometry in the Index for similar problems.
519976
Part (a) is a standard exercise in norms. The first set of points {x E R2 :
Ilxlh = 1} is the square in the plane with vertices (1,0), (0, 1), (1,0),
and (0, 1). The second set of points {x E R2 : Ilxlloo = 1} is the square
in the plane with vertices (1,1), (1,1), (1, 1), and (1, 1).
For part (b), we first consider the 1norm. If x = (Xl, X2) and y =
(Yl, Y2), then x + y = (Xl + Yl, X2 + Y2). Then
Exam #321997 165
holds if and only if
Equality holds in this second equation if and only if
Thus, Xl and YI must have the same sign and similarly for X2 and Y2.
Geometrically, this means that x and y must be in the same quadrant.
For the oonorm,
Ilx + Ylloo = Ilxll oo + IIYlloo
holds if and only if
max{IXI + YII, IX2 + Y21} = max{lxII, IX21} + max{IYII, IY21}·
Let IXi + Yil = max{lxI + YII, IX2 + Y21}. Then
IXi + Yil :::; IXil + IYil :::; max{l
x
II,X21} + max{IYII, IY21}
and equality holds if and only if each of the following holds:
(1) IXil = max{lxII, IX21},
(2) IYil = max{IYII, IY21},
(3) Xi and Yi have the same sign.
The solution is easier to view geometrically. The diagonals Y = X and
Y = x partition the plane into the following four parts:
(a) RI = {(Xl, X2) E R2 : X2 :::: IXII},
(b) R2 = {(Xl, X2) E R2 : X2 :::; IXII},
(c) R3 = {(Xl, X2) E R2 : Xl :::: IX21},
(d) R4 = {(Xl, X2) E R2 : Xl :::: IX21}.
Then
Ilx + ylloo = Ilxll oo + IIYlloo
holds if and only if X and Y lie in the same part. The proof easily follows
from the properties (l), (2), and (3) above.
Look under Analytic Geometry in the Index for similar problems.
519977
The volume of water can be found by multiplying the crosssectional area
of the remaining water with the length (4 feet) of the trough. Thus, it
166 Solutions
suffices to find the crosssectional area of the remaining water when the
trough is tilted through an angle a with the horizontal, as illustrated.
To find the crosssectional area of the remaining water (region W) we
subtract the areas of the sector on the left and the isosceles triangle in the
middle of the figure above. One can show that the angle subtended by the
arc in the sector is 2a. So the area of the sector is
1 2
Asector = 2(2a)r = a.
The isosceles triangle has congruent angles of measure a. Using trigonom
etry, we get the area of the triangle
Atriangle = = = cos a sin a.
We subtract these areas from the total area of the crosssection, which is
7r /2, to get 7r /2  cos a sin a for the crosssectional area of the remaining
water. Now mUltiplying by the length of the trough, we get the volume of
the remaining water 27r  4a  4 cos a sin a feet
3
.
Look under Geometry in the Index for similar problems.
Exam #331998
519981
Here is a Cartesian proof: put P at the origin with AB along the xaxis and
C, D, E in the upper halfplane. Let C have equation (x  a? + y2 = r2,
and let the lines P E and PC have equations y = bx and y = bx,
respectively. Note that the slopes of these two lines are opposite since D
must lie on the yaxis and LCP D = LDP E. The coordinates of E and C,
then, are (Xl, bxd and (X2' bX2), respectively, where Xl and X2 are the
positive and negative (resp.) roots of the equation (x  a)2 + b
2
x
2
= r2.
Furthermore, D = (0, vr2  a
2
). Now, we calculate distances: PE =
xIVl+b
2
, PC = x2Vl+b
2
. Since Xl and X2 are solutions of the
same quadratic equation, we know that their product is the ratio of the
Exam #331998
167
constant term of that equation to the leading term. Therefore,
PE. PO = XlX2(1 + b
2
) = (a
2
 r2)(1 + b
2
) = r2 _ a
2
= PD
2
(1 + b
2
)
A transformational proof: reflect through AB to obtain 0", D" and E",
then observe that P E" is a continuation of PC and PO" is a continuation
of PE. Apply "mean proportional" to the chords D D" and 0 E".
Look under Geometry in the Index for similar problems.
519982
Let k = n/7 and m = (nw
u
 2u)/7. Then, 70m = 7k  21u so that k
is an integer if and only if m is an integer .
Look under Number Theory in the Index for similar problems.
519983
One direction is obvious: if n = p2  q2, then n = (p + q) (p  q) and
both of these factors are greater than 1. In the other direction, if n = 8t
with 8 2: t > 1, then we set p = (8 + t)/2 and q = (8  t)/2. Since n is
odd, both 8 and t must be odd, so that both their sum and difference are
even. Thus, p and q are nonnegative integers, with n = p2  q2. To see
that p  q > 1 observe that p  q = t.
Look under Number Theory in the Index for similar problems.
519984
For part (a), use induction. First, observe that (0, 1) is on one of the
hyperbolas. Then, assume that (fk2, hl) is on one of the hyperbolas,
that is, iLl  ikdk2  R2 = ±1.
Then, we verify that
R  hhl  ILl
= (fkl + h_2)2  (fkl + ik2)!kl  /;1
= ILl + 2ikdk2 + IL2  iLl  ikdk2  iLl
= iLl + ikdk2 + iL2
==F1.
Thus, inductively, we see that all points of':S are on one of these hyperbolas
(and they alternate between the two).
168 Solutions
For part (b), we will repeatedly need the following solution of y in terms
of x for fIrst quadrant points (the choice of sign corresponds to the choice
of hyperbola):
x + V5x
2
± 4 1 + J5 ± 4/ x
2
y= =x .
2 2
Now, using this, we observe a few preliminary facts: if (a, b) is a
fIrstquadrant integer point on one of these hyperbolas and a 2:: 2, then
b 2:: a (since 5 ± 4/ a
2
2:: 4). Furthermore, if a 2:: 2, then b :s; 2a (since
5 ± 4/ a2 :s; 9). Also, if a < 2 and is a nonnegative integer, then we
have only a few possibilities to consider (a = 0 and a = 1) with both
sign choices), leading to three integer points: (0,1), (1, 1) and (1,2), all
of which are in F.
Next, assume that there is a point (a, b) which is an integer, fIrstquadrant
point on one of the two hyperbolas, but is not in F. Let (ao, b
o
) be such a
point with ao as small as possible. From above, we know that ao 2:: 2 and
that ao :s; b
o
:s; 2ao. Consider the point (b
o
 ao, ao): it is clear that this
is still a fIrstquadrant integer point and that b
o
 ao :s; ao. We now show
that this point is on one of the hyperbolas:
a6  (b
o
 ao)ao  (b
o
 ao)2 = a6  boao + a5  b5 + 2b
o
ao  a5
= a6 + boao  b6 = ±1
since (ao, b
o
) is on one of the hyperbolas. Furthermore, there is exactly one
point at which b
o
= 2ao, namely,(1,2) so that (b
o
 ao, ao) has a strictly
smaller fIrst coordinate than (ao, b
o
), implying by our assumption that it
must be in F. But, then (ao, b
o
) must have been in F as well, contradicting
our choice of (ao, b
o
).
Look under Analytic Geometry or Fibonnacci Sequences in the Index
for similar problems.
519985
Denote by Al(X) the area of the portion of R(x) lying between fo and h.
Denote by A
2
(y) the area of the portion of R(r;\y)) lying between h
and h Then,
Exam #331998 169
Furthermore,
r r t1/f3
A2(Y) = io r;1(t)  r;1(t)dt = io a
1
/f3  f:;1 dt.
Now, the condition that it bisect fo and h in area may be rephrased
as A
2
(y) = A1U;1(y)) or
13+1
l
y
t1/f3 _ 1 _ (a  1)y13
1/13 f2 dt  13+1 •
o a (,8 + 1)a 13
Integrating and solving a bit further, we find that
l
y
1 13+1 a,8 + 1  a
f
2
(t)dt=y13 13+1·
o (,8+1)a13
Solving for f:;1 and then inverting, we find that
x
f3
a
f3
+
1
,8f3
fz(x) = (a,8 + 1  a)f3·
Look under Integration in the Index for similar problems.
519986
(a) An inclusionexclusion argument gives
Dn =nln(n1)1+ (;)(n2)1 ....
Rewriting this a bit gives
This latter form is useful for part (b): Since
00 (_1)i
e=L·,,
z.
i=O
it is clearly sufficient for part (b), to show that
00 (1)in'
\n!/eDn \=\ L ., ·\<1/2.
z.
i=n+1
Since this latter series is an alternating series with terms decreasing in
absolute value, its limit has absolute value strictly less than the absolute
value of the first term which is 1/ (n + 1). So, for n > 0 we have part (b).
170 Solutions
For part (c), using the expression above, together with the ratio test
method for computing radius of convergence, yields the following expres
sion for the radius:
. Dn+In! . n!(n + I)!
hm = hm , ..
n+oo Dn(n + I)! n+oo (n + l)!n!
(
_l)n+I
= lim 1+ .=l.
n+oo (n + 1) "n (.1)'
L..,,=o ,!
Look under Derangements in the Index for similar problems.
519987
For part (a), one very simple method is to observe that the example
generalizes in the following way: if det A = 1, then(A±I)2 = (tr A±2)A
so that one simple method of coming up with two square roots of [A] is
to find a representative of the class, say A', which has determinant 1 (do
this by dividing A by the square root of its determinant, which is possible
since we're dealing with complex matrices), then adding or subtracting I.
An alternate form would be
v1AT = [A ± v'det AI].
Generally, this will produce two distinct valid square roots, but occasionally
one of the two will be singular. This occurs when =Fv'det A is an
eigenvalue of A which occurs only when A has a repeated eigenvalue.
In this case, one of the two square roots will be singular (and its square
will be the zero matrix).
(b) This provides us with one or two distinct square roots. Is this all?
The answer, for classes other than the class of the identity, is "yes." The
way to see this is to take a representative of the equivalence class which
has determinant 1 and put it in Jordan canonical form, say 0 = BAB
I
is in Jordan form. Any square roots [D] of [0] correspond to square roots
[B
1
DB] of [A]. So, we are reduced to examining determinant 1 Jordan
forms. There are two such, diagonal and nondiagonal. .
Consider the nondiagonal form first. Here 0 = and we are
looking for D = such that D2 = wO for some w.
We calculate that
(
a b) 2 ( a
2
+ be
e d  e(a + d)
b(a + d))
d
2
+be'
Exam #341999 171
For this to be equal to we need e(a + d) = 0 and b(a + d) =
a
2
+ be = d
2
+ be = W =1= O. The first equation (together with one of the
latter) implies that e = 0 and a + d =1= O. But, since clearly a
2
= d
2
, we
must have a = d. Thus we are reduced to 2ab = a
2
which gives us one
equivalence class, namely [( ) ] .
In the diagonal case, C = A 1) with A =1= AI (otherwise C is
equivalent to J). So, we have b( a + d) = e( a + d) = 0, a
2
+ be = W A and
d
2
+ be = W A 1. The former implies that either b = e = 0 or a + d = 0, but
if a = d, the latter two cannot hold, since this would force WA = WA 1.
So we must have b = e = 0, a
2
= WA and d
2
= WA 1. This gives us two
classes of square roots,
l/Jx)] and
which are equivalent to the two square roots computed by the simpler
method above.
For part (c), we apply the calculations for part (b) and obtain:
b(a + d) = e(a + d) = 0, a
2
+ be = d
2
+ be.
These are all satisfied whenever a + d = 0, giving infinitely many
equivalence classes of square roots of [1], in addition to the obvious fact
that [1] is its own square root. To show that there are infinitely many
nonequivalent square roots, consider the matrices (! where a, bare
arbitrary complex numbers.
Look under Matrix Algebra in the Index for similar problems.
Exam #341999
519991
The key observation is that the procedure never alters the parity (even/odd
count) of the orange balls. Each performance results in a loss of one ball
from the urn, so that, eventually, just one is left. Its color must be orange
since the original number of such balls is odd.
Look under Logic in the Index for similar problems.
172 Solutions
519992
Various correct proofs can be given, some using trigonometry, some via
analytic geometry. The simplest and most elegant proof, however, is to
consider area. We have
area of T = area of AB P + area of BC P + area of CAP,
where A, B, and C are the vertices of T. The standard area formula
( ~ b a s e x height) for triangles then shows that the constant of the problem
is the altitude of T.
Look under Geometry in the Index for similar problems.
519993
The inequality may be rewritten as
(i)
We square both sides, in attempting to eradicate the radicals, giving
5x
2
+ 2xy + y :s; 2x2 + 2V2x V x
2
+ y2 + x
2
+ y2. (ii)
Cancellation leads to the simpler inequality
(iii)
Squaring again gives
(iv)
or
(v)
Inequalities (i) to (v) are all equivalent. (v) is valid, with equality only
when y = x, so that the same is true of (i).
Look under Inequalities in the Index for similar problems.
519994
There is a grain of truth to the claim mentioned in the question. Iff is a
convex jUnction, increasing or not, then the Riemann sums (over uniform
partitions) do tend monotonically to the integral. This is true whether the
sums are evaluated at the righthand endpoints (as they are here), or at the
left, or whether they are inscribed or circumscribed. A counterexample to
Exam #341999 173
the claim is provided by the piecewise linear function
{
0, if 0:::; x :::; 1/2,
f(x)= 6x3, if1/2:::;x:::;2/3,
1, if 2/3 :::; x :::; 1.
It is clear that
f(1/2) + f(l) = 1/2
2 '
and
f(1/3) + f(2/3) + f(l) _ 2/3
3 .
Look under Riemann Sums in the Index for similar problems.
519995
We take the dartboard to be the square centered at the origin with comers at
(±1, ±1). The points closer to the center than the edge are those enclosed
by four parabolic arcs as illustrated.
The desired probability is the ratio of the shaded area to the total area
of the square. The symmetry of the problem enables us to compute this
ratio by restricting our attention to the triangle indicated below.
(1, 1)
(0,0)
p __ I (1,0)
174 Solutions
The parabolic arc is the locus of points equidistant from (0,0) and the
line l. We have
J x
2
+ y2 = 1  x,
so that the equation of the arc is x = (1 y2)/2.
P is the point (1/2,0), while Q is the intersection of the arc with the
line y = x. Letting Q = (q, q), we must have q = (1  q2)/2, so that
Q=(}21,}21).
The shaded area of the triangle is
4}25
6
The total area of the triangle is 1/2, so the probability is 41
5
.
Look under Probability in the Index for similar problems.
519996
Three consecutive entries of the nth row, say
stand in the ratio 1 : 2 : 3 if and only if
nk nkl
 = 2 and
k+l k+2
3

2
These equations, obtained by expanding the binomial coefficients as pre
scribed in the question, may be rephrased as n = 3k + 2 and 2n = 5k + 8.
There is one, and only one, solution: n = 14, k = 4.
519997
The series diverges. Its terms lie between those of the (divergent) harmonic
series, :Z::::;;:O=ll/k, and those of the (convergent) alternating harmonic
series, :Z::::;;:O=l ( 1)k / k. There is some cancellation in our series, courtesy
of the cosine, but this occurs too slowly for convergence because of the
logarithm.
We show that the partial sums of our series are not Cauchy: more
precisely, we prove
(2n+i)"
e L cos(1og k) >
k=e(2n i)" k
(n=I,2, ... ).
Exam #341999
To do this, we observe that
whenever
so that
whenever
1
cos x > M
 y2
1
cos (log k) ;::: V2
The series (*) contains at least
terms, all of which are greater than
1
Look under Infinite Series in the Index for similar problems.
519998
175
The standard formula for the sum of a finite geometric series has a matrix
analogue,
1  Am = 0 by hypothesis, so that
(1 + A + ... + A
m

1
)(1  A)x = 0
no matter what the vector x. Since 1  A i= 0, there exist nonzero vectors
x and y with
(1  A)x = y.
But then, by (**),
(1 +A+ ... + Am1)y = 0
so that 1 + A + ... + Aml is singular and det(1 + A + ... + Aml) = o.
Look under Matrix Algebra in the Index for similar problems.
176 Solutions
Exam #352000
520001
If Cindy didn't do it, the other three are lying. In particular, Becky must
be lying. This means that Ducky is telling the truth. Therefore Cindy is
the only possibility.
Look under Logic in the Index for similar problems.
520002
The solution to part (a) is 2000 = 5555
7
.
(b) 1885 = 1111
12
. Let the number be N = aaaab = aMb where
Then since 1 ::; a ::; b  1,
Since N < 2000, Mb < 2000, b::; 12 and since 111112 = 1885, we have
N :::::: 1885 whence b
4
> 1885 so that b :::::: 7. The values of Mb for b = 7,
8, 9, 10, and 11 are 400, 585, 820, 1111, 1464 and no multiples of these
lie between 1885 and 2000.
Look under Number Theory in the Index for similar problems.
520003
It converges. Let
Since n + vn + 1 > n, Sn+1 :::::: Sn and therefore Sn are increasing. We
show the Sn < 2 for all n and thus we have convergence. For x:::::: 6, we
have
x + ffx < 2x  2.
Thus we see that
Exam #352000
Sn < 1+ + Vn 2+ Vn1 + y2n 2
< ... +Vn2+V2n=4
1 + )2+ V3+'" + Vn 3+ y2n 6
< ...
<
< 1+)2+ V3+ V4+V9
< VI + V2 +..)3 + 3
< VI + v'5 < 2.
The actual value is 1.75793 ....
Look under Sequences in the Index for similar problems.
520004
177
The order of an element divides the order of G, and can therefore be 1, 3,
5, or 15. For these k, let nk be the number of elements of order k. Clearly
nl = 1. We'll show that n15 > 0 (an element x of order 15 exists) and
so G =< x > is cyclic. Suppose G has A subgroup(s) of order 3 and
B subgroup(s) of order 5. Since 3 and 5 are prime, by Sylow's theorem,
A == 1 (mod 3), B == 1 (mod 5), and both A and B divide 15. Thus
A = B = 1 and n3 = 2, n5 = 4. Since nl + n3 + n5 + n15 = 15,
n15 = 15  1  2  4 = 8 > O.
Look under Group Theory in the Index for similar problems.
178
520005
Since
  T)k
V5 V5
Solutions
is an integer for k 2: 1, and 11  TI < 1, we have
so
and
Fk = _l_
T
k   T)k
V5 V5
(l)kFk = __ 2
sinh
(V5)
k! J5e 2
k=O
) (li!k Fk ) = sinh2 (V;) = cosh Vs)
by the double angle formula cosh 2x = 2 sinh2 x + l.
Look under Fibonnacci Sequences or Infinite Series in the Index for
similar problems.
520006
The desired ratio is p = K / 7f R2, where K denotes the area of the
triangle. We use the law of sines, sinG = c/2R and the area formula
K = to write
(a/R) (b/R) sinG 2. A . B . G
p = 27f = ;;: sm sm sm
= sin2BsinBsin3B
7f
= cos Bsin
3
B(4cos
2
B1)
7f
=  X
2
)3/2(4x
2
 1)
7f
where x = cos BE (1/2,1). Thus we seek the maximum of
4
p =  f(x), f(x) = (1  x
2
)3/2(4x
3
 x) on (1/2,1).
7f
Exam #352000 179
f'(x) = (1 x
2
)3/2(12x
2
1) + (4x
3
 x)(3/2)(1 X
2
)1/2(_2x)
= (1 X
2
)1/2 [(1 x
2
)(12x
2
 1) + (4x
3
x)(
= (1  X
2
)1/2(16x
2
 1 24x4)
so
16x
2
 1  24x4 = 0 or
2 4+v'IO
x =
12 '
the other root being less than 1/2. Thus the answer is
ig( 1l)3/2(4g
2
 1).
7r
This can be written as
i (1 (4 1) . h +
7r 3 12 3 12 V 3 12
= 7v'IO  2 /22 4v'iO
1087r V +
(7v'IO  2)( V20 + J2)
1087r
34J2 +5V5
547r
Look under Geometry or MaxIMin Problems in the Index for similar
problems.
520007
Consider x = J2v'2. If x is rational, it is a counterexample. Otherwise,
y = xv'2 is a counterexample since y = 2 is rational.
Look under Number Theory in the Index for similar problems.
520008
If D f = 1', the characteristic polynomial factors as
(D2 + 1)(D  1)(D2 + D + 1)f = 0,
which yields five linearly independent solutions, which can be combined
under the Principle of superposition to yield the general solution
f(x) = LCieTiX,
ISO Solutions
where ri is a root of the characteristic polynomial (±i, 1, w, w), with w =
(1 + v3i)/2. Since limf(x) = 0, Re(ri) < 0 and f(x) = Ae
iw
+ Be
iw
.
Since f(O) = 0, B = A and
f(x) = AeX/2[eixv'3/2 _ e
ix
v'3/2]
= Ce
x
/
2
(sin V; x)
for some constant C.
Look under Differential Equations in the Index for similar problems.
Index by Problem Type
Algebraic Structures
19711, 19717
Analytic Geometry
19672, 19765, 19782, 19796, 19816, 19886, 19906, 19911, 19917,
19946, 19964, 19976, 19984
Arclength
19921
Binomial Coefficients
19996
Complex Numbers
19675, 19687, 19862, 19863, 19951, 19973
Derangements
19874, 19986
Differentiation
19664, 19713, 19763, 19824, 19968
Differential Equations
19694, 19793, 19803, 19881,20008
181
182 Index by Problem Type
Diophantine Equations
19661, 19684, 19836, 19928, 19967
Enumeration
19665, 19668, 19674, 19688, 19714, 19716, 19753, 19822, 19842,
19847, 19854, 19864, 19872, 19902, 19912, 19913, 19917, 19923,
19971
Field and Ring Theory
19735
Fibonnacci Sequences
19832, 19984, 20005
Finite Sums
19783, 19807, 19952, 19972
Fundamental Theorem of Calculus
19883
Geometry
19662, 19676, 19678, 19682, 19695, 19703, 19704, 19752, 19785,
19813,19823, 19833, 19875, 19885, 19886, 19893, 19904, 19913,
19938, 19944, 19957, 19962, 19965, 19977, 19981, 19992,20006
Group Theory
19727, 19742, 19853, 19882, 19924, 19932,20004
Inequalities
19801, 19942, 19943, 19955, 19993
Infinite Series
19677, 19685, 19686, 19724, 19746, 19751, 19834, 19873, 19997,
20005
Index by Problem Type 183
Integration
19713,19734,19772, 19806, 19876, 19885, 19896, 19921, 19925,
19955, 19985
limit Evaluation
19666,19685,19701, 19715, 19731, 19734, 19766, 19831, 19846,
19855, 19905, 19916, 19927, 19935, 19941, 19966
logic
19891, 19897, 19937, 19991,20001
Matrix Algebra
19671, 19721, 19722, 19737, 19773, 19804,19862, 19896, 19926,
19933,19987,19998
Max/Min Problems
19732, 19815, 19833, 20006
Miscellaneous
19667, 19743, 19791, 19795
Multivariate Calculus
19683, 19744, 19755, 19866
Number Theory
19663,19692,19693,19705,19712,19738, 19754, 19762, 19775,
19781, 19802, 19811, 19826, 19837, 19841, 19843, 19844, 19852,
19856,19871, 19884, 19901, 19914, 19915, 19936, 19945, 19953,
19961, 19963, 19982, 19983,20002,20007
Permutations
19725, 19934
Probability
19739, 19786, 19814, 19826, 19832, 19845, 19857, 19874, 19887,
19894, 19922, 19966, 19971, 19995
184
Polar Coordinates
19726
Polynomials
Index by Problem Type
19675, 19702, 19736, 19741, 19761, 19821, 19838, 19851, 19865,
19931, 19954
RealValued Functions
19664,19673, 19691, 19696, 19723, 19745, 19763, 19764, 19771,
19774, 19784, 19805, 19825, 19839
Riemann Sums
19681, 19685, 19905, 19994
Sequences
19855, 19884, 19892, 20003
Solving Equations
19974
Systems of Equations
19765, 19812, 19937
Statistics
19967
Synthetic Geometry
19975
Taylor Series
19958
Trigonometry
19772, 19792
Index by Problem Type 185
Volumes
19794, 19827, 19895, 19956
This page intentionally is no longer blank
A Friendly Mathematics Competition
35 Years of Teamwork in Indiana
© 2003 by The Mathematical Association of America (Incorporated) Library of Congress Catalog Card Number 2002107971 ISBN 0883858088 Printed in the United States of America Current Printing (last digit): 109 8 7 6 5 4 3 2 1 .
A Friendly Mathematics Competition 35 Years of Teamwork in Indiana Edited by Rick Gillman Valparaiso University Published and Distributed by The Mathematical Association of America .
Chair Roger Nelsen Editor Irl Bivens Clayton Dodge Richard Gibbs George Gilbert Art Grainger Gerald Heuer Elgin Johnston Kiran Kedlaya Loren Larson Margaret Robinson A Friendly Mathematics Competition: 35 Years of Teamwork in Indiana. compilations of problems (including unsolved problems) specific to particular branches of mathematics. and Loren L. Solutions. Richard K. edited by Titu Andreescu and Zuming Feng The William Lowell Putnam Mathematical Competition 19852000: Problems. Bjorn Poonen. edited by Rick Gillman The Inquisitive Problem Solver. Kedlaya. Ravi Vakil USA and International Mathematical Olympiads 2000.MAA PROBLEM BOOKS SERIES Problem Books is a series of the Mathematical Association of America consisting of collections of problems and solutions from annual mathematical competitions.maa. Box 91112 Washington. edited by Titu Andreescu and Zuming Feng MAA Service Center P. DC 200901112 18003311622 fax: 13012069789 www. etc. Larson Mathematical Olympiads 19981999: Problems and Solutions From Around the World. edited by Titu Andreescu and Zuming Feng Mathematical Olympiads 19992000: Problems and Solutions From Around the World. Guy. Kiran S. edited by Titu Andreescu and Zuming Feng USA and International Mathematical Olympiads 2001. O. and Commentary. Committee on Publications Gerald Alexanderson. books on the art and practice of problem solving.org . Paul Vaderlind.
Unfortunately. Steve Shonefeld. this is the one piece of the history that has been lost over the years. Ken Luther. Among all of these people. It is really more about getting together to do mathematics and to challenge oneself than it is about winning a competition.Preface A college level mathematics competition has been held in Indiana for more than 35 years as of the writing of this book. its official title became the Indiana College Mathematics Competition. the following individuals provided noteworthy help: James Lee. and the Purdue University Math Club. but have not generally been attributed to an original source. These have been included to provide a complete record of the exam. and extend a very special thank you to Paul Mielke on . However. Orginally known as the Indiana Small College Mathematics Competition. I would like to specifically identify the individual faculty members who have composed the exams. Pat Sullivan. Over the years. I would also like to thank Pete Edson for having the idea for the competition. exam authors have borrowed many classic problems from other sources. many people refer to it by its popular nickname "The Friendly Exam" earned because of the collegial nature of the competition and the modest level of competitiveness between the participants and the institutions involved. Their attentive work at identifying or creating good questions has been a major reason why the competition has been a success for so long. This book shares the efforts of the many students and faculty who have participated in the Indiana College Mathematics Competition over its first 35 years. Eric Tkaczyk. I'd like to take this opportunity to thank all of the people who have made this book possible by helping in one way or another.
and to lead the section in building the endowment Indiana and the MAA would not have this wonderful example of a "friendly" competition that does so much to build the mathematical community in Indiana.vi Preface behalf of myself and my colleagues in Indiana. to support it over the years. Without Paul's efforts to develop the contest. .
. . . . .Contents Preface The Indiana College Mathematics Competition: A Short History An Update of the History of the ICMC Exams Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exwn Exam Exam v 1 5 7 7 8 9 10 II 12 14 15 16 16 17 18 19 20 22 23 24 25 26 28 vii # 11966 #21967 #31968 #41969 #51970 #61971 #71972 #81973 #91974 #101975 #111976 #121977 #l31978 #141979 #151980 #161981 #171982 #181983 #191984 #201985 . . . . . . .
. . . . . . . . . . . . . . . . . .viii Contents Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam #211986 #221987 #231988 #241989 #251990 #261991 #271992 #281993 #291994 #301995 #311996 #321997 #331998 #341999 #352000 . . . . . .. . 29 30 31 32 34 36 37 38 40 40 41 43 44 46 48 51 51 55 61 64 67 69 73 75 78 81 86 88 90 95 98 103 106 109 112 117 120 122 Solutions Exam #11966 Exam #21967 Exam #31968 Exam #41969 Exam #51970 Exam #61971 Exam #71972 Exam #81973 Exam #91974 Exam #101975 Exam #111976 Exam #121977 Exam #131978 Exam #141979 Exam #151980 Exam #161981 Exam #171982 Exam #181983 Exam #191984 Exam #201985 Exam #211986 Exam #221987 . . . .
. 126 130 133 136 139 144 148 150 156 160 166 171 176 181 Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Exam Index by Problem Type . . . . . . . . . .Contents ix #231988 #241989 #251990 #261991 #271992 #281993 #291994 #301995 #311996 #321997 #331998 #341999 #352000 . . .
This page intentionally is no longer blank .
a written examination for undergraduates . 2:00 P. Thus.M. .12:30 P. Participants will be guests of the College.The competition.Luncheon. 1966. sent our dean a memorandum that included a newspaper clipping about a unique mathematics competition that was held among high schools in New Jersey. at Wabash College. Edson wondered if anything of that sort was done at the college level. is as follows: 11 :30 A.4:00 P.M. each school entered a team that worked as a team on a set mathematics examination. Campus Center.The Indiana College Mathematics Competition: A Short History Paul T. In this competition. Mielke On April 27. . Registration. Crawfordsville. 1965. . Program for the day. 1965. so I wrote a second letter. Wabash College 12:30 P. Baxter Hall Lounge. The purpose of this competition is to stimulate friendship and mathematical interest among members of the various departments and their undergraduate students. South Mezzanine. Response to my letter was immediate and favorable. I wrote a letter to heads of departments in the small colleges of Indiana relaying Edson's suggestion. . which we are calling "Pete Edson Day" in honor of the Wabash College trustee who conceived the idea for the contest. I answered that I knew of no such competition but that I would be willing to suggest the idea to my Indiana colleagues. Peter Edson.M.M. which read in part: The Mathematics Department of Wabash College is pleased to invite its counterparts in the other colleges of Indiana to a friendly mathematics competition to be held on Saturday afternoon. on May 18.M. a trustee of Wabash College. March 12.
M. There will be a significant difference. from the phrase used in the above letter. Evansville. Taylor. Valparaiso. Before each test. prepared by Professor George F. Each school is invited to bring a team of three undergraduates who will work on the test as a team. the Putnam test. Springer of Indiana University. Earlham. Earlham.2 The Indiana College Mathematics Competition: A Short History 2:00 P. Taylor. however. Marian. and an average participation of over thirteen schools.M. Marian College. IV Southeast. . MaryoftheWoods. RoseHulman. Earlham College. with eight schools participating: Anderson College. Huntington. . Tests will be sent to Professor Springer for grading and announcement of results. Duration of the tests has been maintained at two hours. IVPUI Indianapolis. St. Butler. It has been held every year.M. a total of 24 times as of 1989. DePauw. Ranking of other teams will not be revealed by Professor Springer. though (one hopes) not as difficult as. DePauw University. the contest has been called familiarly.Coffee hour at which test papers will be gathered. most often a professor from one of the large state universities. DePauw. The nineteen participants in the 1972 competition were Butler. IV Northwest. Franklin College. with a team consisting of at most three members. Marian. St. and Wabash. . consulting with one another as needed. with a minimum participation of eight schools in that first year. Throughout its history.Informal seminar for members of the various departments 4:00 P. Baxter Lounge The test. Oakland City. St. Wayne. Rose Polytechnic Institute (now RoseHulman). Since its beginning. . This first contest was held as scheduled. and Wabash. Evansville. MaryoftheWoods. Manchester. the essential nature of the contest as a team competition has remained the same. Valparaiso University. "The Friendly Math Competition". Manchester. and Wabash College. In the spring of 1967 the second competition was held at Marian College with 15 participants: Anderson. RoseHulman. Joseph's. a team member from each school draws a number which is given to an impartial judge. Valparaiso. Mary's (of Notre Dame). Purdue Calumet. a maximum participation of nineteen in 1972. The tests have usually been set and graded by a single person. Purdue Ft.4:00 P. A suitable prize will be awarded by Wabash College to members of the winning team. St. Franklin. Franklin. who attaches the number to the school's name in a memo. will be similar to.
and a decision had to be made with regard to its wish to enter several teams. the Indiana Section of the Mathematical Association of America agreed to make them a part of its spring meeting. There was early agitation for publicly announcing the ranking of the first three teams. so that only these are seen by the person who sets and grades the test. This person in turn reports the results to the contestants. the grader announces the results by team number to the judge who holds the identification key. A school may enter any number of contestants in the Putnam competition. Some other ground rules developed naturally. and that a school need not designate "its team. but it must name its team of three for the competition beforehand. I have had the pleasure of supplying both trophy and prizes. This idea was accepted. Since this fund was exhausted. The books are traditionally selected from those offered by the MAA. to which Mr. for instance. Edson contributed while he was alive. Until 1987 the competition was restricted to the small colleges of Indiana. A consensus was reached that mUltiple teams of three could be entered." It was then decided that public announcement of the top three places in the competition would refer to schools rather than teams. Maintaining the "Edson Trophy" and awarding of prizes was assumed by Wabash College. which had a small fund for that purpose. the school would be awarded second place in the competition. but it led to another problem. though in its fall meeting of 1977. The Edson Trophy is a rectangular walnut plaque measuring 9 x 12 inches. but in 1987 it was opened to all colleges. and third place would be awarded to the next highest school. The contests have always been managed by the host schools. and the winners' names are inscribed on the Edson Trophy. To it is affixed a permanent identifying plate with this inscription: THE PETER EDSON TROPHY INDIANA COLLEGE MATHEMATICS COMPETITION . Upon completing hislher work. RoseHulman has always been a worthy adversary in this Indiana competition. including the large universities. and this practice has been followed since 1978.The Indiana College Mathematics Competition: A Short History 3 Teams are instructed to identify their test papers with only their numbers. Each member of the winning team is awarded a suitably inscribed book prize. including branches of IU and Purdue. so that if two teams from a given school scored second and third.
Henceforth. As one might guess. affixing all the old winners' plates to them. I have succeeded in making individual trophies for each of the past winners. . Three special walnut trophies have been made to hold the 12 winning plates for RoseHulman. The current trophy is at Purdue University.) It should finally be noted that the task here has been to write examinations that will challenge and separate the competitors. The trophy "travels". Each year. one of Indiana's most beautiful hardwoods. the actual examinations varied in their success according to these criteria. The new trophies are made of sassafras. where they record its six wins.4 An Update of the History of the ICMC Below this plate is a smaller one containing the year and the name of the most recent winning college and its team members. so it resides permanently at Wabash College. this smaller plate is moved to the back of the plaque. the 1989 winner. and a new one replaces it. (Since writing the above. it resides each year at the winning college. The back of the first trophy was filled by 1978. which was the winner in that year. a nontravelling trophy will be awarded to each winner. The two old trophies have been retired to Wabash College. so that there will be enjoyment for everyone involved. yet present some problems that all can solve.
Indiana. generally from twelve or thirteen colleges or universities. Mary's College in Notre Dame. during my first year on the faculty at Valparaiso University. This was. the average number of teams participating in the ICMC has been 25. The first competition that I organized was at St. at a joint meeting of the Indiana. as you read in Paul's comments. It is always very exciting and refreshing to be around the students from the various schools as they wait for the exam to begin. I have been an enthusiastic supporter of the competition ever since. Michigan. Thirty teams from the three states participated in the competition that year. I became the Indiana section's first Student Chapter Coordinator and one of the responsibilities was to manage the ICMC. a team from the Indiana Mathematics and Science Academy also participates. the first year that the competition was opened up to the large state universities and explains why I had not known of the competition while an undergraduate student at Ball State University. While the nickname "friendly" has not been used very much in recent years. a lot of friendly banter and good mathematical conversations occur among my VU students during the drive back to campus from the meeting. In 1992. Occasionally. In many years.An Update of the History of the . Since 1990. and Illinois sections.CMC I was introduced to the Indiana College Mathematics Competition (ICMC) during the spring of 1987. Subsequent Student Chapter Coordinators have helped implement changes to the ICMC that have made it the central part of the section's efforts to reach undergraduate students. and to hear their conversations with each other and with their faculty advisors after the exam. the competition still maintains the same friendly approach that it had when it began. The largest competition was at Ball State University in 5 .
Originally this fee was paid to the exam writer for composing and grading the exam. Somewhat surprisingly. the section finds an external source to do this work. 2002 . Students can then stay for a banquet and after dinner speaker. Sessions on professional and pedagogical issues are now held on Friday afternoons during the competition itself. Each team competing pays a $5. By 1996. respectively. Proceeds from the sale of this volume will go the Indiana section to support undergraduate student activities in the section. Now these fees are paid to the writer for writing the exam and for leading the problem solving session on Saturday. both as presenters and as members of the audience. The afterdinner speaker on Friday night is selected with the knowledge that at least half of the audience of 100120 individuals are students. knowing that students will be staying overnight for the Saturday portion of the meeting. Rick Gillman Spring. It has also established a new structure for the section meeting. If they decline this opportunity. in either the number of teams participating or the number of competitions won. there is a session where solutions to the exam are discussed and the winners of the competition are announced late in the day during the section's business meeting. so that mathematical presentations are the central focus when students are part of the audience. the Indiana section had established a small endowment to support the ICMC. Rather than having a single person grade the exams. the large state universities have not dominated the competition since their inclusion.6 An Update of the History of the ICMC 1998. This endowment will enable the section to continue to hold the exam each year and present prizes to the winning team members and the winning institution in the form of books and the Peter Edson Trophy.00 registration fee. On Saturday. The most significant change in the competition came in the form of a decision to move the competition from Saturday afternoon (on the second day of a day and a half meeting) to Friday afternoon as a lead off event for the meeting. The institution hosting the ICMC and the section meeting is invited to have one of its faculty members write the exam and solutions. This format has increased the number of students participating in the mathematics sessions of the meeting. 29 teams from 22 institutions participated. That year. The section has also been able to experiment with late night workshops for students on Friday evenings. a team of faculty graders do this work overnight.
P3 .. For two given positive integers n and k. Consider any five points P l .y2 = a3 always has integer solutions for x and y whenever a is a positive integer. Denote by dij the distance between points Pi and Pj . + A sequence {x n } is defined by the following rule: Xn+l = b with Xl = c. P1966S. P2 . viax. how many different sequences of positive integers al ::::: a2 ::::: a3 ::::: . Prove that at least one of the distances dij is less than V2/2. P19662. Eight schools participated in the competition that year. Ps in the interior of a square of side length 1 (one).. ::::: ak are there in which ak ::::: n? P19666.ap be integers not necessarily arranged in consecutive order and with possible repetitions. Jr. .. Establish the existence of integers m and n such that 1 ::::: m ::::: n ::::: p and such that L j=m n aj is divisible by p. P19664. Two functions of x are differentiable and not identically equal to zero. P19663. It was won by the team from Wabash College consisting of James Clynch.. and Larry Haugh. Find an example of two such functions having the property that the derivative of their quotient is the quotient of their derivatives. a2.Exams Exam #11966 As stated in the introduction.. P19661. Albert Hart. Let p be a prime number and let al. Show that this sequence converges whenever 7 . P4 . a bit northwest of Indianapolis. . located in Crawfordsville. Show that the equation x 2 . the first "friendly competition" was held at Wabash College.
Show that for each n the number of odd binomial coefficients is a power of 2. A = {aij} is a symmetric (i. and C is greater than or equal to V65 /8. was again from Wabash College. . P19675. and C. In which cover an interval I. . and each row of the matrix is a permutation of the integers 1.. consisting of David Hafting.. Let us assume that a given pair of people either know each other or are strangers. aij = aji) n x n matrix with n odd. some odd. then the least upper bound of the diameters of A. i.2. . Albert Hart Jr. P19676. and determine the limit of the sequence.. . some even. Prove that if the set S of points in or on the boundary of the unit square is partitioned into three disjoint sets A. By an interval we shall mean a set of points x on the real line satisfYing a :s: x :s: b for a pair of real numbers a and b with a < b. Show that if lanl < 2 for 1 :s: n :s: N. 0 :s: r :s: n. P19674. P19673. 2.. P1966S. Suppose that we are given a collection of intervals h. + aNz N = 0 has no root z such that Izl < 1/3..e.. that is I ~ U~=l I k . then the equation 1 + alz + a2z2 + .. show that there must be either three people who know each other pairwise or three people who are pairwise strangers. Prove that we can select mutually disjoint intervals from this collection which cover at least half of I. For each positive integer n the binomial coefficients (~).n. Prove that the main diagonal is also a permutation of 1. Exam #21967 This competition was held at Marian College in Indianapolis. . Show that for all a ~ 0 and b ~ 1.. The winning team. S = A U B U C and A n B = A n C = B n C = 0. . B. Two parabolas have parallel axes.n. are integers. The diameter of a set is the least upper bound of the distances between two points of the set. h. P19671.s o< a < Exams 1 and b > 0 regardless of the value of the real number c. Prove that their common chord bisects their common tangent.. P19672. and Robert Spear.e. . B.. P19667.. ab :s: e a + b(ln b  1). If six people enter a room.
Exam #31968
9
P19677. Given a > 0 and Xo > 0, show that there exists one and only one sequence of positive numbers {Xo, Xl, X2, ... } such that
Xn=
for n
L
j=n+l
00
xj,
= 0, 1,2, ....
P19678. Let T be a mapping of the Euclidean plane into itself which preserves all rational distances. Prove that T preserves all distances.
Exam #31968
Held at Franklin College, located just south of Indianapolis, this competition was won by a team from Earlham College. The team members were William Roha, Thorn Sulanke, and William Wilson. It is unique because it is the only competition that came with a warning. WARNING: The statements below should be viewed as conjectures. At least one cannot be done.
P19681. Let f be a realvalued function defined on the closed interval [a, b]. Show that if the set of Riemann sums for f is bounded, then f is bounded. By Riemann sum we mean a sum of the form
L
i=l
n
f(ti)(Xi  xid
Xil
where a = Xo <
Xl
< ... < Xn = band
:s: ti :s: Xi
for 1
:s: i :s: n.
P19682. Given four points which are the vertices of a convex quadrilateral in the plane and five points inside the quadrilateral such that no three of the nine points are collinear, show that five of the nine points are the vertices of a convex pentagon. P19683.
Let
f : Rn
H
R n be a differentiable function such that
f(tx) = tf(x) for X E Rn and t > O. Show that f is linear.
P19684. Find all integral solutions of the equation 2 X the equation 3x  2Y = 1.

3Y = 1 or of
P1968S. Let Un} be a sequence of realvalued functions defined on R. Suppose that for each n, {xlfn(x) i= O} is bounded and that the sequence
10 converges unifonnly on R to the zero function. Show that
n+oo
Exams
lim
/00 in = O.
(X)
P196S6. Find two decreasing sequences {an} and {b n } of positive numbers such that
00
00
00
Lan =
n=l
00
and Lbn
n=l
= 00,
but LCn <
n=l
00
where C n
P196S7.
= min{ an, bn }.
Let
Zl, Z2, ... , Zn
be complex numbers such that
Show that for some iI, i 2, ... ,ib we have
Iz + z· + ... + z· I > 7r.
1.1 t2
1
tk

P196SS. Let nand k be positive integers. Suppose line segments are drawn joining each pair of n points and that each segment is painted blue or green. Are there k points such that all the line segments with end points among these k points are of the same color? Show that the answer is yes if n is large enough. Can you guess how large n must be?
Exam #41969
This competition was held at RoseHulman Institute of Technology, located in Terre Haute. The winning team from Valparaiso University consisted of Gerald Anderson, Charles Frank, and Charles Spear.
P19691.
Prove that sin x :::: x  (x 2 / 7f) if 0 ::; x ::; 7f.
P19692. Suppose p, q, and r are positive integers no two of which have a common factor larger than 1. Suppose P, Q, and R are positive integers such that + ~ + is an integer. Prove that each of ~, and is an integer.
f
!f
f,
!f
P19693. Determine whether {i25 or irrational.
+ 5V20 + {i25 
5V20 is rational
P19694. A ball is thrown into the air. The only forces acting are gravity (constant) and air resistance (proportional to the velocity). Which takes it longer, to go up, or to come down?
Exam #51970
II
P1969S. An equilateral triangle is circumscribed about an arbitrary triangle as shown in the figure below. Show without using calculus that the maximum area it can have is
~ (b 2 + c2 I
2bc cos ( A
I'
'
+
i) ).
I
I
' ,
,
I
I I
I
I I
,
, ,
tJ!7\
P19696. that f(a)
= feb) = 0,
Assume that f has a continuous second derivative, that a < b, and that 1f"(x)1 :::; M on a:::; x :::; b. Prove that
lib f(x)dxl:::;
~(ba)3.
Exam #51970
Held at Butler University, in Indianapolis, for the first of many occasions, this competition was again won by a team from Valparaiso University. The team consisted of Gerald Anderson, Jonathon Berke, and George Lueker. A team from RoseHulman came in second.
P19701.
Evaluate . 1 1m
n+oo
(1+ + ... +) . 1 1 ++
n
1
n
2
2n
P19702. Let f(x) = xn + an_IX n 1 + ... + aIX + aD be a polynomial with real coefficients. Prove that any root c satisfies
(Hint:
Icl :::; 1 + lanII + la n 2\ + ... + laol· consider Icl :::; 1 and Icl 2:: 1 separately.)
P19703. Squares ABEF, BCCH, CDJK, and DALM are placed externally on the sides of a parallelogram ABCD, with X, Y, Z, and W ilie respective centers of those squares. Prove that the lengili from X to Z is the same as the length from Y to W, that the line segment from X
The curves cut the disk into e/2 + 1 regions. y. . There was a tie for second place between teams from Earlham College and Goshen College. P19705. First place was won by a team from RoseHulman Institute of Technologythe first of their many winsconsisting of Tom Dehne. There was a glossary of terms at the beginning of this exam. Find the smallest natural number n that has the following properties: (a) Its base ten representation has a 6 as the last digit. is erased and placed as the first digit in front of the remaining digits. Let S be a set and let P be an equivalence relation on S. of points. then the resulting number is four times as large as the original number n. and Robert Klim. Exam #61971 This competition was held at Earlham College. On the border of a disk select an even number. and finally that XY ZW is a square. For each subset A of S. Bill Lipp. 6. in the case e = 10 we may have the figure below. z in S. P19711. Draw e/2 nonoverlapping curves in the disk whose ends are the e dots. P19704. xPy =} yPx. xPx. Prove that the regions can be colored with two colors in a way such that adjacent regions are colored differently. and xPy 1\ ypz =} xPz. define A = {x E S::::I Yt A such that yPx}.12 Exams to Z is perpendicular to the line segment from Y to W. That is. (b) If the last digit. for all x. but this has been lost over time. e. For instance. in Richmond.
P19713. Y2) iff Yl . (a) What is the limit of the fractional part of (2 + /2)k as k (b) What is the behavior of the fractional part of (1 7 oo? 7 + . Granted that the last three digits in the decimal representation of 7400 are 0. Let A be a countable subset of a Euclidean plane. y) E 5 : X2 + y2 = I}. (b) Let 5 be the set of points in the Cartesian plane.Y2 = 3(Xl . P19715. Being somewhat but not completely trustful of one another. c A. 0. 1.l X J be called the fractional part of x. Let X . c 5 and B c 5 implies A 11. P19712. For any real number x. where A is the unit circle with center at (0. then l X J = n. Define R: (Xl. so that 7400 = 103 k + 1 for some positive integer k. How many locks were required. A = A. they decided to be able to open the safe when and only when a majority of the robbers was present. and = Au B Au B. P19714. Compute the derivative of I_el./X' Then compute 1I 1 1 x2(1. That is A = {(x.e1/X)2 e. A set of eleven bank robbers decided to store their loot from a successful job in a safe (where else?) until the heat was off. describe or sketch A. Given a nonempty set 5 together with a binary operation operation *. Suppose that n and a unary . Therefore. the safe was provided with a number of different locks (the key to anyone not being able to open any other). Thus.J2)k as k oo? P19716.X2). Ill. let lxJ denote the greatest integer in x. Yl)R(X2. Is the following a Boolean algebra? Explain.Exam #61971 13 (a) Prove that A 1. and how many keys had to be given to anyone bank robber? P19717. and each robber was given keys to some of the locks. find the last three digits in the decimal representation of 79999 . Show that there exists a line in the plane which contains no point of A. if n is the (unique) integer such that n :s: X < n + 1. Granted that R is an equivalence relation on 5.0).1 / x I dx.
where an > 0 and an I. Q.) P19725. 2. located in Greencastle. converge? (Prove or give a counterexample. Exam #71972 This year's competition was held at DePauw University. 5. B. J. and D are square matrices. does Which is bigger: e7r or 7r e ? If I:~=l an converges. 7. If A is a square matrix such that A 3 + 4A2 + 3A + 2I = 0. B. K. AnB Exams = BnA. and S. Find the points of intersection of the curves whose equations in polar form are r P19727. What conditions on A and B are necessary for the existence of matrices X and Y such that AX + BY = C and BX + AY = D? P19722. Suppose that A. show that A is invertible (I denotes the identity matrix and 0 the matrix all of whose entries are zero. C in = An (B n C). Can a group be a union of two proper subgroups? . and it consisted of Reinhard Fritz. An *B = C n *C {:} A n B = A. B in S. The winning team was from Purdue UniversityCalumet.1 for all n. what order were the cards in after the first shuffle? P19726. 6. 9. just west of Indianapolis. 3. B. If the cards emerged in the order 10. and (b) for all A. A cardshuffling machine always rearranges cards in the same way relative to the order in which they were given to it. (A n B) n C (c) for all A. A and Bare not invertible.14 (a) for all A. All of the hearts arranged in order from ace to king were put into the machine. and then the shuffled cards were put into the machine again to be shuffled again.) P19723. P19721. C. 8. 4. David Hasza. and Lawrence Kus. C in S. A. P19724. = cos(O/2) and r = sin(O/2).
then both A and B are nonsingular. 1f /2.1)2. Evaluate lim n+oo 1 0 1 nynl dy. Is A : B an ideal of R? Prove or disprove. P19737. A rectangle is inscribed in a sector of a circle of radius 1 as shown in the figure below. show that an+! . Let A and B be ideals of a commutative ring R. of A by B is defmed as A : B = {r E R: rb E A for all bE B}.cos 0 2sinO' P19733. If n is a positive integer. Evaluate lim x(e 1 / X x+oo  1). P19732.1) . with teams from Earlham College and RoseHulman taking second and third places respectively. The team consisting of Tom Seilke.Exam #81973 15 Exam #81973 This competition was held at St. Note that IT is the product symbol and k is not fixed but can assume any value from 1 to n. IS non . P19736. Show that the maximum possible area for the rectangle is 1 . Prove that if AB singular. The central angle of the sector is a given angle 0. Determine the maximum value of IT:=1 ni. The quotient. and John VanDrie from Wabash College won. in Terre Haute. Mary'softheWoods College. P19731. Let A and B be square matrices. P19734.n(a . Tom Stocks.a is divisible by (a . Let n be a fixed positive integer greater than 1. 1+y P19735. with 0 < 0 :s. where ni are positive integers whose sum is n. A: B.
. Examine the validity of the following conjecture: The series of positive terms I:~=o an diverges if and only if the series I:~=o a~ diverges. Determine whether at (0. N I ) and (n2. Show that y has a zero in the interval [0. find two distinct pairs of positive integers (nl. . y in S we have x 3 = x and x 2y = yx 2. P19745. Exam #101975 A team from RoseRulman won this competition. y) t xy. y.3/4J.. The winning team consisted of Robert E. A group of 5 men contains 3 Democrats and 2 Republicans. or neither. Michael J. Suppose that y is a continuously differentiable function of x which satisfies the condition y(O) = 1 and the inequality ~ +eXy+ 1 o. for the function f (x. For each positive integer m. Three men are chosen at random from the group. + 1) = Nl + 1 (i=1. which was held at Wabash College.x). Let S be a set with an associative multiplication. Carlin. (x. P19744. What is the probability that both of the Republicans were selected? Exam #91974 This competition was held at Butler University for the second time.2). P19743. (b) What is the sum of all the coefficients of this polynomial? P19742.0) f achieves a local maximum. y) = (y2 . minimum. + xlOOO. and David Wilde from Wabash College won the competition. Suppose that for all x. Copus. and Barry W. Tom Sielke.16 P19738. and z P19739. Dominik. :s: P19746.x) (2y2 . Find all solutions in integers of x 2 + y2 in arithmetic progression. P19741. Show that the multiplication is commutative. where the team of Stanley Lyness. N2)(depending upon m) such that (m 2 + 1)(n. Exams = Z2 with X. (a) Find the coefficient of x 50 in P(x). Let P(x) = (1 + x) 1000 + x(l + x)999 + x 2(1 + x)998 + .
2.1) = Ip(x)12 . Determine all polynomials p( x) such that p(x 2 . P19761. Let n be a positive integer such that n + 1 is divisible by 12. Define a f(t) dt t loa g(x)dx = loa f(x) dx. Prove that it must be regular. P19763. Exam #111976 This competition was held at DePauw University. Teams from RoseHulman and Franklin College came in second and third. Given m lines in the plane. + an tan (. Suppose that 2ml(3m . P19762.Exam #111976 17 P19751. Show that 7r .1. Tom Sellke. Suppose that a > 0 and g(x) = for 0 < x ::. into how many components do they divide the plane? Prove your assertion. a..1). A polygon having all its angles equal and an odd number of vertices is inscribed in a circle. The expression alb means that a divides b.) ) . and Matthew Wyneken was from Wabash College. Prove that the sum of all of the divisors of n is divisible by 12. P19755.. The winning team of Jay Ponder. x ::. P19753. with no two parallel and no three concurrent. Let f(x) = al tanx + a2 tan (~) + a3 tan (~) + . Show that i x f is continuous for 0 ::. then m is even.1 and p(2) = 2.< 2 L (Xl n=l a2 a 7r + n 2 <2· P19752. (a) Show that if m i=. and (b) Show that m = 1. P19754. a. respectively. or 4.
1. P19771. that is..Jr/2}. then either f is strictly increasing on J or f is strictly decreasing.0.. B. f(1) I. Hint: You may need to know that 1 < In 3 < 1. Show that if f (f (x)) = x for all x III J. io cos xdx r P19773. and that f(x + y) = f(x)f(y) for all x and y. . Prove. P19772. Let A.2X2 have on the real line? Prove that your answer is correct. Prove that limx+o f (x) = 1.C2) respectively.. using algebraic methods.. Suppose that f is a realvalued function of a real variable. and C be three noncollinear points in a rectangular coordinate plane with coordinates (alla2). and (C1. Exam # 121977 This competition was held at RoseHulman and was also won by a team from there. B. Let A = (aij) be an n x n matrix of real numbers such that 2:. . prove that a2 a3 an Ia1 + 2 + 3" + .(b 1.b2). and limx+o f(x) exists. and C. P19774.. an are real numbers and where n is a positive integer.7=1 aij = 1 for each i. Jan Slupesky. Hint: Consider first the special case in which f(O) = o.IS Exams where a1. 1] into itself.: I :s: 1. + . that it is always possible to solve for the coordinates of the center of the circle containing A. Show that the matrix A  I is not invertible. The team consisted of Rich Priem. and Bob Strickland. P19766. functions f from R into R with the property that the distance between any pair of points is the same as the distance between their images under f. (a) cos n x Prove the following identities: = 2~ ~ (~) cos(n = 22n (n!)2' (2n)! 2k)x 1 2n (b) :. Given that If(x)l:S: Itanxl for x E {Jr/2. . How many zeroes does the function f(x) = 3x 1. a2. Find all of the isometries of the set R of real numbers. P19765. P19764. Let f be a continuous function that maps the closed unit interval J = [0.
.8. 5. whereas problems 2.5. Let k be a positive odd integer and let.. (1 + ~ + P7v + .S (n) that 2S(n) is divisible by n + 1. The team consisted of Kevin Fosso..6. was again won by a team from Wabash College. or 9. and Matthew Wyneken. + . P19782.4.1. + p~l p~2 .. + PI (1 + ~ + ~ + ..Exam #131978 19 P19775... and 6 require more justification. held at Earlham College. 7..P2.. ABD. 3. pW + .. + ~) ~) pi .  < (1 + ~ + ~ + . Prove the following generalization of the Theorem of Pythagoras: If ABCD is a 3dimensional tetrahedron such that each of the triangles ABC.. P19781. The number 2 is a base for all integers using the digits 0 and 1.. Teams from RoseRulman and Manchester College came in second and third.. For example 1101 represents 3.F "7""c...3. then (areaABC)2 + (areaABD)2 + (areaACD)2 = (areaBCD)2. and ACD has a right angle at A.2. Rint: For each integer N there is an integer r such that 1 1 1 p~l p~2 .. The number 10 is a base for the positive integers because every positive integer can be written uniquely as where each di is one of the digits 0.. = 2:7=1 jk.. Use the fact that 2:~=1 lin diverges to obtain a proof that there are infinitely many prime numbers P1. p'.p~l p~2 . The exam came with instructions: For problems 1 and 4 you need only show your computations.. . since 1( _2)3 + 1( _2)2 + O( 2) + 1 = 3. Jay Ponder.. Show P19783. p'j.. . + ~) ~ P1 P2 p~ P2 PN P'N Exam # 131978 This competition. Find the representation in base 2 for the decimal number 2374... respectively.
what fractional part of the rope is on the same side of the pulley as the monkey when the system is at rest? P19792. A fair coin is a coin for which the probability of tossing heads and the probability of tossing tails are each 1/2. Find the point 0 in the plane so that the sum of the distances from o to each of the four points is smallest.. The monkey's mother is twice as old as the monkey was when the monkey's mother was half as old as the monkey will be when the monkey is three times as old as the monkey's mother was when the monkey's mother was three times as old as the monkey. so that 10AI + lOBI + lOCI + 10DI is smallest. P19791. for which the probability of success is 1/4: Toss a fair coin twice. The age of the monkey's mother added to the age of the monkey is four years. . and let f(x) = 'E~=llx . Suppose that A. It is passed over a pulley. C. and on the other end a monkey. a2.akl. For which value(s) of x is f(x) smallest? What is the minimum value of f(x)? You need not give a formal justification for your answer. i. we present the following experiment. B. .20 Exams P19784. Design an experiment with a fair coin for which the probability of success is 1/3.. an be real numbers. and Tony Mazzoni from RoseHulman won this competition. Let aI.e. not necessarily distinct.. The experiment is a success if heads is tossed both times. The weight of the monkey in pounds is equal to the age of the monkey's mother in years. yplane. Graph the relation sin x = sin y in the x. which was again held at Butler University. P19786. How long is the rope? Assuming the pulley to be negligibly small. and D are four distinct points in the plane. The whole system is in equilibrium. Exam #141979 The team of Mike Hall. As an illustration. A piece of rope weighs four ounces per foot. and on one end is suspended a weight. P19785. . Tim Drabik. a3. Each of the weight of the rope and the weight at the end is half again as much as the difference in weight between the weight and the weight plus the weight of the monkey.
2 = (_1)97 + V9. 21). 9.14) Li (i = 1 to 8) are lines as follows: Ll goes through Band (6. Snow starts to fall in the forenoon and falls at a constant rate all day. Continue this through consecutive numbers k as far as you can.16y = 4 L7 has equation x + 7y = 4 Ls has a yintercept of 6 and angle of inclination 45° A is the intersection of Ll and Ls C is the intersection of L2 and L3 G is the intersection of L6 and L7 H is the intersection of L7 and Ls. operations. (ii) Algebraic grouping symbols. as well as use of higher order roots. A hole six inches long is drilled through a sphere. 3 = 19 + 17 . The rules of the game are as follows: (i) The only constants you may use are 1. 9. It goes a half mile during the second hour. and functions may be used as often as desired. . Find all points with integer x and y coordinates which are strictly inside the polygon ABCDEFGH (described below) and also strictly outside the circle centered at (4. which must be used exactly once and in that order. P19796. 7. F = (4. What time did it start to snow? Give the time to the nearest minute. (The hole is six inches long after it is drilled. and the hole is through the center of the sphere. Note that 1 = 1979 . Note that rule (i) prohibits use of 7r or e. D = (11. 1). 2). The plow goes a mile during the first hour. At noon a snow plow starts to clear a highway_ The velocity of the snow plow is such that it removes a constant volume of snow per unit of time. P19794.Exam #141979 21 P19793.) What is the volume of the part of the sphere that remains? P19795. 2) L2 goes through B and has slope 9/4 L3 goes through D and has slope 2/15 L4 goes through D and E L5 goes through E and F L6 goes through F and is perpendicular to x . Here is a description of the polygon: B = (4. E = (2. 1) with radius 3/2. 4=1+V9+l7/9j.91.
Suppose that f is a function that satisfies f"(x) + f'(x)g(x) . A team from Butler University came in second. P19806. the competition was held at Valparaiso University. Show that det(I + xyt) = 1 + xty for any vectors x and y in Rn. Randall Ekl. Consider a balance that is used to measure loads of integral weights. The winning team from RoseHulman consisted of Michael Call. 2(xy)2 Show (a) that the integral 10 roo sin x dx x . located in Valparaiso. The balance has two scales. for the first time.f(x) = 0 for some continuous function g. P1980S. P19804. P19802.22 Exams Exam #151980 This year. a load scale and a weight scale. P19803. On the load scale one can place the load to be measured and any desired subset of the measuring weights. P19801. b]. y. On the weight scale one can place only certain measuring weights. Prove that if f vanishes at a and at some b> a. Note that t denotes transpose. Show that with four suitably chosen weights one can measure the weight of any load whose weight is an integer between 1 and 40 pounds. and Douglas Englehart. If there exists a sequence {b n } of positive numbers and a constant a > 0 such that show that the series 00 is convergent. Find all positive functions f that satisfy f(x) < f(y) for all x. Let {an} be a sequence of positive numbers. then f is identically zero on [a.
find the probabilities that the Series lasts (a) exactly four games. Assuming that teams are equally matched. Find the radius of the largest ball that can be fitted into the comer behind the given ball.!£ ( roo sin ax dX) 2 = O. and Tim Drabik. (c) exactly six games. If 12 steers take 16 weeks to deplete the grass on 10 acres. Express 3/8 as a base 7 fraction of form 0. P19812. a rock is dropped. At a point 50 feet above his eye atop a building 100 feet away. Then show (b) that . and if 18 steers take 8 weeks to deplete the grass on 10 acres. was held at Indiana UniversityPurdue University in Indianapolis (IUPUI). The team consisted of Michael Call. and the grass keeps growing at a constant rate.Exam #161981 23 is convergent. P19815. If Sn show that =L n (_l)k+l k ' n k=l S2n= Lk· n+ k=l 1 Exam #161981 This competition. how many steers does it take to deplete the grass on 40 acres in 6 weeks? P19813. there is a given amount of grass per acre. Thomas Douglas. da 10 x P19807. If the man watches the rock fall. A man is standing atop a tall building. again won by a team from RoseHulman. and (d) exactly seven games. at what point in its descent is his head moving the fastest? . Before steers are introduced to a pasture. P19814. P19811.aOala2a3···. The winning team of the World Series must win four games out of seven. A ball of radius 1 is in a comer touching all three walls. (b) exactly five games.
2. . . upon which the submarine immediately descends. 1. n} is said to be separated if ai+l .::: 2 for i = 1. . .. . . .. For example. .. n}. .n . 2.24 Exams Assume that the buildings are arbitrarily high. but {3.. and it is known that the latter will depart at once at full speed on a straight course of unknown direction. 8}. Hint: You may use the fact that Y+OO lim yTe. P19823.. x i= 0. Exam #171982 This was the first time that the competition was held at Ball State University.. The winning team of Jeffery Baldwin. . {2.1. Randall Ekl. Show that the number of separated subsets of {I. 5. 8} is not. Given finitely many points in the plane situated so that any three of them are the vertices of a triangle of area :::. The fog lifts for a moment. A subset {al. and 1. disclosing the submarine on the surface three miles away.Y = ° for any r. Find a cubic equation whose roots are the reciprocals of the roots of the equation x 3 + ax 2 + bx + c = 0. 4. The speed of the destroyer is twice that of the submarine.2. and that the acceleration of gravity is 32 ftlsec/sec. The wily captain of the destroyer sails straight to the point 2/3 of the way to the spot where the submarine was sighted and then sets out on a spiral course that is bound to make him pass directly over the submarine. 7} is a separated subset of {I. 4. P19824. that air friction is neglected... A destroyer is hunting a submarine in dense fog.1jX2 ..ad of the set {I. a2. 0 x =0 O? Prove your answer. located in Muncie.ai . 2. . each having k elements is (nZ+l). What is the equation of this spiral? HINT: Use polar coordinates with the origin at the point where the submarine was sighted. Show that all the points can be enclosed in a rectangle of area :::. P19821. Anthony Kirk was from RoseHulman. Is the function f(x) differentiable at x = = {e. P19816.. with c i= O. P19822.
Anthony Kirk. and Tom Moss. . Inscribe a rectangle and a circle inside the triangle as indicated in the figure below. A realvalued function f of a real variable is said to satisfy a Holder condition with exponent a if there is a constant c such that If(x) . Consider an isosceles right triangle with legs of fixed length a. The competition was held at the Indiana University. Find limn+oo \inTo P19832. campus for the first time. P19831. 1. Find the dimensions of the rectangle (and the radius of the circle) which make the total area of the rectangle and circle a maximum. What is the probability that you stop on the 10th toss? P19833. P19827. clx .. with its straight boundary parallel to the axis and curved boundary away from the axis. . Bloomington.yin for all x. Exam # 181983 RoseHulman won the competition again this year. Suppose you repeatedly toss a fair coin until you get two heads in a row. Its winning team consisted of Baron Gemmer. 1/4. The probability that the square of a positive integer (in decimal notation) ends with the digit 1 is 2/10 because out of every 10 numbers those and only those ending with the digits 1 or 9 have squares ending with 1. y.. 1. What is the probability that the cube of a positive integer chosen at random ends with the digits 11? Prove your answer. Wherever these functions are used. Show that every positive real number is a sum (possibly infinite) of a subset of the numbers {I. a is restricted to be ::. P19834. Teams from Butler University and Manchester College tied for second place.f(y)1 ::. 1/3. }.Exam # 181983 25 P19825. 1/2. Can you explain why? P19826. Find the volume of a torus (doughnut) of inner radius b whose crosssection by a plane through the axis is a semicircle of radius a.
P19838.) A B .005. P19836.Xj II xER.. j#k xk . Suppose that g : [0. Prove the "restricted" Fermat conjecture: For any integer n > 2. . Byron Bishop. has no solutions in positive integers if z < n. + yn = zn P19837.001 ? P19842. Teams from Wabash College and Valparaiso University came in second and third. Prove that g2 (x) = x for all x.1] is a continuous function and that gm(x) = X (g composed with itself m times) for all X and for some positive integer m. respectively.Pn(x) by Pk(X) = XXj .010. xn This problem is a repeat of problem #3 on EXAM #7. P19839. In the graph below. and Dan Johnson won. What can you say about N? Prove it. how many paths that never go up connect node A to node B? (Paths must follow edges indicated. What is the prime factorization of 1. Exam #191984 This competition was held at RoseHulman and a team from RoseHulman consisting of Todd Fine.010. Define polynomials PI(X).26 Exams P1983S. Prove that I:~=IPk(X) = 1 for all X in R. Let Xl. P2(X).1] + [0.. X2. P19841.005."" Xn be (distinct) real numbers. . Let N be one more than the product offour consecutive positive integers.
opens door 6.Exam #191984 27 P19843. In Subsylvania. one worth 7 quanta. Choose two points at random in the interval [0. That is. P19846. A train is being made up of Boxcars. and Flat cars and Hopper cars weigh 50 tons each. x+oo (_X_) = f(x) + x+1 +00. . A very long hallway has 1000 doors numbered 1 to 1000. (Part credit for finding some f. Find the largest price for which it is not possible for a Subsylvanian customer to give exact change. and Hopper cars.g. How many different trains of total weight 1000 tons (ignoring caboose and engines) can be made up? Trains are considered identical if and only if they have the same sorts of cars in the same order: e. 1000 people go down the hall: the first person opens each door. forO < x < (b) Prove that lim f(x) = x+o+ (c) Find all such functions f. H H B. These two points cut the interval into three segments. Let f be a continuous function on < x < f ° 2 00 satisfYing 00. closes door 9. b] is b. One by one. Note that it is possible for a Subsylvanian customer to buy an item costing 5 quanta (she gives the merchant a 12 quanta coin and receives a 7 quanta coin in exchange). Boxcars weigh 25 tons. there is no paper money and there are only two kinds of coins. the nth person changes all doors whose numbers are divisible by n.1]. all prices are whole numbers of quanta. H BH and F BH are distinct trains of total weight 125 tons. but it is not possible for her to give the merchant 5 quanta in exact change.a. all the doors are initially closed. the other worth 12 quanta. f(l) = 5 and (a) Find lim f(x). Flat cars. [Naturally. etc. After all 1000 people have gone down the hall: which doors are open and which are closed? P19844. the third person closes door 3.] P19845..) P19847. the second person closes all doors with even numbers. The probability of choosing a point in the interval [a. or prove there is no such price. opens door 12. What is the probability that these three segments can form the sides of a triangle? Note: "at random" means the two points are chosen independently and with uniform probability distribution.
28
Exams
Exam #201985
ReId at Ball State University, this competition was again won by a team from RoseRulman. The team consisted of Todd Fine, Erick Friedman, and Dan Johnson. Teams from RoseRulman and Goshen College came in second and third, respectively. Notice that the contest rules for assigning second place were broken this year; it was the first of several in which this occurred.
P19851. Compute
Let
(Y,
{3, 'Y, and 8 be the roots of X4 + bx3 + cx2 + dx + e = O.
((Y2
+ 1)({32 + 1)("(2 + 1)(82 + 1) + 44443333
is divisible by 7. Make up some
in terms of b, c, d, and e.
P19852. Prove: 33334444 more problems of this type.
PI9853. Consider the set U = {I, 2, 4,5,8,10,11,13,16,17,19, 20}. In the figure below, each of the three pathsthe solid, the dotted, and the dashedrepresents a subgroup of order 6 of U under a certain binary operation. Three of the points of these paths have been labeled. Find all the others .
(:~:=~.::::.;~:U:::::.·:::::::::=)
: ~~
:.
.................. ~ ..................
" "....
".:
".' :
j
\
\ .......•....../
PI9854. Each side of a square is subdivided into 101 segments. Find the number of different triangles that have their vertices at these 400 points of subdivision. P19855.
Let
al =
0 and a2 = 1, and for n
~
3,
an = (n  l)(an l
Find (a) a formula for an and (b) limn + oo
';" .
+ an2).
Exam #211986
29
P198S6. Given n + 1 integers between 1 and 2n inclusive, prove that one of them must be a multiple of some other one. P198S7. Let a and b be positive constants with b > 1. Given that x + y = 2a and all values of x between 0 and 2a are equally likely, find the probability that
Exam #211986
This competition was held at Butler University. The winning team consisting of Erich Friedman, John Hoffman, and Dan Tretter was from RoseHulman. The second place team was also from RoseHulman, and the third place team was from Valparaiso University.
P19861.
This is a repeat of problem #1 on EXAM #14.
P19862. Let {b l , b2 , b3 } be an orthonormal basis for C 3 and let ¢ E L(C 3 , C3) be given by
¢(bd = 2b2 , ¢(b2 ) = 2ibl , and ¢(b3 ) = (1
Find the eigenvalues, if any, for ¢.
P19863.
+ i)b3 •
Evaluate ii.
P19864. N points are chosen on a circle so that when all segments are drawn between all pairs of points, no three segments intersect at the same point. In how many points do these segments intersect in the interior of the circle? P1986S. Find an equation with integral coefficients one of whose roots is y'2 +~. P19866.
Let
f : R2 + R be given by
if (Xl,
X2)
~IX~ 2 f( XI,X2 )  { Xl +X4' 0, (a) Is
i= (0,0);
=
(0,0).
if(Xl,X2)
f continuous at (0, O)? JustifY your answer. (b) Is f differentiable at (0, O)? JustifY your answer.
30
Exams
Exam #221987
This competition was held at Butler University. The team consisting of Daniel W. Johnson, Brenton Young, and Daniel Kniep from RoseHulman won the competition. A team from Indiana University, Bloomington, came in second, and a team from Wabash College came in third.
P19871. A positive integer n is called composite if there are positive integers P i= 1 and q i= 1 so that n = pq. Find a sequence of 10 consecutive positive integers each of which is composite and less than 1,000,000. P19872. John's job at the Acme Cannonball Factory is to stack the cannonballs (which are 6inch diameter spheres) neatly into tetrahedral piles. For example, using 4 cannonballs, John can make a tetrahedral pile with 2 cannonballs on each edge: three balls forming the bottom triangle and one in the center on the top. Find a formula for the number of cannonballs in a tetrahedral stack whose base is an equilateral triangle with n balls on each edge. P19873. Experiments have determined that when a particular steel ball is bounced on a hard surface, it bounces to half its original height. For example, if it is dropped from a height of 6 feet, it will bounce to 3 feet. Assuming that the ball obeys this law exactly, for what length of time will the ball continue to bounce if it is dropped from a height of 16 feet (or will it bounce forever)? [Recall from calculus that since the acceleration due to gravity is 32 ft/sec/sec, an object failing to the ground from height h (in feet) or bouncing from the ground to height h requires Vli/4 seconds to do so.] P19874. Ten seniors who share a house decide to exchange graduation presents. They each put their name into a hat, mix the name cards thoroughly, and draw a card out at random. What is the probability that none of the ten draws his or her own name? P19875.
Euclidean fourspace is R4 with
=
IXYI
((Y1  X1)2
+ (Y2 
X2)2
+ (Y3 + X3)2 + (Y4 + X4)2) 1/2.
(a) Find five points in the unit ball of Euclidean fourspace that are as far from each other as you can make them. That is, find points A, B, C, D, and E such that
10AI :s; 1, 10BI:S; 1, 10CI:S; 1, 10DI:S; 1,
and
10EI:S: 1,
P19876.) Exam #231988 This competition was held at Butler University. and. IOEI. IAOI. IAEI.i).. *) be a group with the following cancellation rule: x * a * y = b * a * c implies x * y = b * c for all x. Jane begins a journey at Quito. prove that fst P(x)dx = o.e. Define n(x) = IT i=l.. IODI. P19883. Going at top speed. such that min{IABI. P19881. t:s. y. . A team from IUPUI came in second. If sand t are both integers such that 1 :s.i#k n (x . IADI. prove that it is. n. Let (G. . and a team from Purdue University came in third. IDEI} is as large as you can make it. that compass headings are exactly accurate. and that Jane's plane has an unlimited amount of fuel available. s :s. IBEI. Prove that G is Abelian. Who wins and why? P19882.. The contest was won by a team of two students from Indiana University. IBDI. b. Only two miles from the finish. k = 1. Grand Prix driver x leads his rival y by a steady three miles. in the next mile. x runs out of fuel. . Ecuador (at 80° west longitude on the equator) and flies steadily northwest.2. a. Thereafter x's deceleration is proportional to the square of his remaining velocity. that there is no effect of wind. n and 1:S. IBOI. that * is commutative. etc. and c in G. Bloomington: Radu Tudorica and Kevin Pilgrim. and let P(x) be a polynomial of minimum degree in which P(k) = 5Tk (k). his speed exactly halves..Exam #231988 31 where 0 is the origin. i. (b) If your answer is best possible. (a) How far does she fly before reaching the North Pole? (b) How many times does she cross the Greenwich Meridian (longitude 0°) in the journey? (Note: You should suppose that the earth is a perfect sphere of radius 4000 miles. n.
P19885. Bloomington. . A team from Purdue University won the competition. with proof. A fair coin is tossed ten times. P19891. let 5 n be the sum of the squares of the first n positive odd integers. The members of this team were Alex Gottlieb.) P19886. For coming first in a test. Teams from Earlham College and Wabash College came in second and third. and Nung Kwan Yip. Find the shaded area of the figure below. a student is awarded x points. A solution to this plus several other problems could have won the solver 50 dollars. as shown in the figure below. Aaron Weindling. Students were encouraged to send in solutions they liked.B. Show. and C compete in a series of tests. for coming second. (This was a proposed problem in the then current issue of The Journal of Recreational Mathematics. What is the units digit of 512345? Prove your answer. a perpendicular from the point P to the line L. campus. Exam #241989 This competition was held on the Indiana University. (This is problem 13 in the 1988 issue of The Old Farmer's Almanac. where the interior circular arcs have their centers on the outer circle. respectively. Find the probability that two tails do not appear in succession. Three students A.32 Exams P19884. how to construct with unmarked straightedge only.) p • o L P19887.
2mn m 2 +n2 P19893. B 10 points. Who came in second in the geometry test? P19892. The pipes intersect at right angles as shown below. z points. Student A came in second in the algebra test. Show that the area of the shaded triangle formed by the sidetrisectors of a triangle (as shown) is 1/7 of the area of the whole triangle. satisfy Describe all sequences {x 1. One is fair [Prob(heads) = 1/2] and the other is biased with Prob(heads) = 2/3. There were no ties in any of the tests. for all positive integers m and n. .xn } of real numbers that IXm . . P19894. . Altogether A accumulated 20 points. for coming third. X2. Which coin is more likely to be the biased one? P1989S.. resulting in two heads. The other is tossed three times. One of the coins is tossed once. Two identical pipes have elliptical crosssections with semiaxes a and b. Here x. Two coins are given. y.Xnl:::. and z are positive integers with x > y > z. and C 9 points. Find the volume of their intersection.Exam #241989 33 y points. resulting in heads..
For example. How many ways can the letters . What are the dimensions of Smith's ranch. The third place team from Indiana University consisted of Urmi Bhatacharya. on March 31. P19897. not square. too. I did not know the area of your ranch. Lucia Demetrios. described in the following conversation? Smith. and although I did not hear your answer to James's question. Let me see if I can figure out its dimensions. Hold on a minute! I happen to know the area of your ranch. I. Now I know the dimensions. Is the short side strictly less than half the long side? (Smith answers his question. P19902. Peter Sepanski. significantly more information was collected about the competitions. Find all positive integers which are one more than the sum of the squares of their base ten digits. A digraph in a word is an ordered pair of consecutive letters. Down in Todd County. and Nung Kwan Yip. They scored 51 out of 60 points on the following examination. This year.34 P19896. Find the determinant of the matrix X1Yn X2Yn ) Exams 1+ ~nYn . P19901.) James. I have a ranch. know the area of your ranch. 35 = 1 + 32 + 52. 1990 in connection with the spring meeting of the Indiana Section of the MAA.1 digraphs. the contest was held at Purdue University (located in West Lafayette) for the first time. which is a 19 x 19 miles square. Kathy. I can now figure it out. Bill. Exam #251990 Beginning in 1990. but having heard this entire conversation. the short side and the long side both measuring a whole number of miles. a word with n letters has n . The second place team from Purdue University consisted of Mark Sepanski. can tell you the dimensions. (He figures furiously. I.) I need more information. James. The winning team from RoseHulman consisted of Joel Atkins. Twentyfour teams from twelve colleges participated. and Kevin O'Bryant. and Radu Tudorica. too. John O'Bryan. It is rectangular.
N. A F c P1990S. A regular hexagon of side 1 is inscribed in the intersection of two identical parabolas. for all real x. in that order. D. A. D. we create a new regular heptagon HIJKLM with side r. but not I. A" twice. A be arranged so that no digraph is repeated? (Thus. By connecting the vertices ACEGBDF A. which contains "I. Determine r. I. Evaluate the following limit. N.) P19903. f'(x) = f(x f(x + 2) = f(x) and + 1)  2. oriented (in opposite directions) with their axes parallel to the yaxis (see the figure below). I.Exam #251990 35 I. A is counted. N. N. P19906. P19904. Let ABCDEFG denote a regular heptagon with side 1. the arrangement A. A. A. . D. N. I. Find the area of the (shaded) region inside the parabolas and outside the hexagon. N. I. Find all real functions f such that. (See the figure below).
located in Anderson.11 + 12x + 11 + ~ = 4. P. Twentythree teams from fourteen colleges participated. P19912. b and e) such that ab + n. Q and R. find an integer n (in tenns of a. P to Q. . illustrated below. of the parallelogram ABCD. R C A~~. Q to B. those integers whose baseten representations contain a "1" or those that do not? How many of each type are there? P19913. P19914. The second place team from Purdue University consisted of Alex Gottlieb. Peter Sepanski and BoonLock Yeo. Kathy Steiner and Ken Wenger. The winning team from RoseHulman consisted of Jonathan Atkins. The third place team from Goshen College consisted of Dave Cooper. A is joined to P.000 inclusive. be + nand ea + n are all squares of integers. P19911. which are more numerous. band e are integers and if a+b+e is even. on March 23. The pentagonal region bounded by the sides of the starpentagon is colored blue (b). in connection with the spring meeting of the Indiana Section of the MAA. DA and CD respectively. B to Rand R to A to fonn a starpentagon APQBR. Show that the red area minus the blue area is independent of the choice of the points P. John O'Bryan and Mark Roseberry. Q and R are arbitrary points on sides BC. Find the area of that portion of the xyplane which is enclosed by the curve with equation 12x .000. Between I and 1.36 Exams Exam #261991 The contest was held at Anderson University. If a. The regions inside the parallelogram and outside the starpentagon are colored red (r).
and Kevin o 'Bryant. n patrons have lined up to buy tickets. At a movie theater. The ticket seller calls a patron viewable if he (she) is taller than all the people in front of him (her) in line. The second place team from IndianaPurdue University at Fort Wayne consisted of Jeff Gerardot. Yi). There were 24 teams with a total of 67 participants from 13 colleges.···. i n n 1 1 = 1. P19922.2. where Hn is the harmonic mean of the n positive integers n + 1. The centroid of a set of n points {(Xi. Aaron Pesetski. . Tony Hinrichs. A car rode over an ant on the pavement. . n + 2. Exam #271992 The contest was held at The University of Indianapolis on April 11. ~ 2: Yi). find the length of the curve traveled by the ant between its death and its final resting place. The ant stuck to the tire for one revolution and then was deposited back onto the pavement. 26 = 12 + 52. in connection with the spring meeting of the Indiana Section of the MAA. n}. A lattice point in the xyplane is a point whose coordinates are integers. 18 = 32 + 32 and 25 = 32 + 42. and Mark Pragan. P19917. show that some three of the points have a lattice point as centroid. The third place team from Purdue University consisted of Peter Sepanski and PokYin Yu. The winning team from RoseHulman consisted of Jonathan Atkins. n + n. n + 3. Given that no two patrons are precisely the same height. 17 = 12 + 4 2. Find limn+oo(Hn/n). (b) Find three consecutive positive integers each of which can be written as the sum of two nonzero perfect squares. find the average number of viewable patrons among all possible permutations of the patrons. P19921. otherwise he (she) is hidden. (a) Show that there exist infinitely many twosomes of consecutive positive integers each of which can be written as the sum of two nonzero perfect squares.. Given nine lattice points in the xyplane..Exam #271992 37 P19915. . Assuming that the radius of the tire is one foot. The harmonic mean of a set of positive numbers is the reciprocal of the arithmetic mean (ordinary average) of the reciprocals of the numbers. Are there infinitely many such trios? (c) Do there exist four consecutive positive integers each of which can be written as the sum of two nonzero perfect squares? P19916. For example. in the xyplane is given by (~ 2: Xi.
x2 dx.38 Exams P19923. There were 30 teams. The winning team from RoseHulman consisted of Jon Atkins. Brian Johnson. and Aaron Pesetski. For A = (_~90 _56 ) (a) find the eigenvalues of A. Evaluate P19928. the minimum number of letters that can suffice to share all of the scandal. Michigan. 2 P19926. Determine all other integral right triangles whose legs are consecutive integers. Tony Hinrichs. The third place team from Calvin College. located at the northern edge of South Bend. (b) Find four different integral matrices B so that B2 = A. P19924. . and Tim Hotlebeek. 4. Of course each gossip will share all of the scandal he (she) knows at that time whenever he (she) sends a letter.(The hypotenuse must also be an integer. with representation from all three states.) Exam #281993 The contest was held at Saint Mary's College. Find. Fort Wayne. Identify all finite groups G of order n having at least 2n proper subgroups. with proof. on April 24. (Hint: How do the eigenvalues of B relate to the eigenvalues of A 7) (c) For an arbitrary 2 x 2 matrix A. consisted of Joel Holcombe. Notice that 3 and 4 are consecutive integers. P19925. They communicate by mailing letters. and 5 form the sides of a right triangle. and Nick Tallyn. Milce Bolt. The second place from from IndianaPurdue University. A collection of n gossips each knows a unique tidbit of scandal not known to any of the others. We all know that the lengths 3. consisted of Alan Baljeu. but need not be consecutive as 5 happens to be. in connection with the joint spring meeting of the Indiana Section of the MAA with the Illinois and Michigan Sections. Evaluatei: e. what is the maximum number of integral matrices B that can satisfy B2 = A7 P19927.
for n = 4.. . (Here Ix l denotes the smallest integer not less than x. For n > 1. 546. P19936. B~'='"':. Find all integers A. D. Let = (P" ( P'(x) x) ) 2 I.4. Prove that A2 = A.. a permutation al. .. E) which.2. (For example. f(g(f(a))) f and 9 be mappings from the set A to itself for which g(a) and g(f(f(a))) = f(a) for all a in A. Prove that pili ( X ) ~ P'(x) < 2 for all x for which P' (x) P19932.) How many permutations of {I.. n+oo n + a2 + .2. P1993S. Find the area of the parallelogram abed.ai I = 1.. there is a j > i for which laj .n} are orderly? HINT: What are the possible values for al ?) P19935.2.. B :::.2 is not. for each i = 1. P19933.3 is orderly whereas 3.) P19937.:"""?IC 5 5 A 6 2 . . .1. In the figure below. D :::. for i > O. . yield only the sums 401. the permutation 1. . + ai. Let A be a square matrix of rank 1 and trace 1. Prove that f = g. (n . P19934.a2. and 836.. . 2.an of {l.n} is "orderly" if.O. B. Let al = 1 and a. E (A :::.4. C.1).Exam #281993 39 P19931.+l = val Determine lim (an).. when added in pairs. Let P(x) be a real cubic polynomial for which P'(x) has distinct real zeros. Prove that 2n + l is a factor of 1(V3 + 1)2nl for all positive integers n. C :::. 691. .. ABCD is a rectangle..
The second place team from RoseHulman consisted of Jonathan Atkins. Matt Laue. then the circle contains at most two points both of whose coordinates are rational. and William Stier. Avijit Sarkar. and Shirish Ranjit. Find all sets of five positive integers whose sum equals their P19946. Brian Singer. in conjunction with the spring meeting of the Indiana Section of the MAA. Twentythree teams from eleven colleges participated. The third place team from RoseHulman consisted of Jamie Kawabata. Rick Mohr. 1976. Show that one of its angles is twice another. and 6. The winning team from Wabash College consisted of Faisal Ahmed. P19944. product. in conjunction with the spring meeting of the Indiana Section of the MAA. Twenty teams from eleven colleges participated. P19943. If x > y > 0. . 1977. and Nick Tallyn. Show that for positive integers n. 0 P19942.) P19941. (These problems were used with the permission of Crux Mathematicorum. The third place team from Purdue University consisted of Ilya Gluhovsky. Nick Tallyn. and Faisal Syud. Matthew Gladden. Exam #301995 The contest was held at TriState University. and James Moore. Show that if one of the coordinates of the center of a circle is irrational. P19945. and Seth Patinkin. The sides of a triangle have lengths 4. Calculate lim x+= l x et 2 2 x dt. The second place team from Wabash College consisted of Eham Ahmed Chowdhury. consisted of Daniel Bliss. Bloomington. The winning team from IU. 5. located in the far northeastern community of Angola on March 31.40 Exams Exam #291994 The contest was held at Depauw University. © 1975. on March 19. show that vn < 1 + y'27n.
The winning team from RoseHulman consisted of Rick Mohr.2 P199S6.1?(2x + 1)2(x + 2)2. Lukito Muliadi. . in conjunction with the spring meeting of the Indiana Section of the MAA. Let Zl be complex numbers. Prove that IZl + P199S2. form. Let p be an odd prime. Indiana University. Bloomington. i = 1. A solid S is generated by rotating the triangle ABC about the line L. Find the coefficient of x2n in the expansion of g~:~: in closed Exam #311996 The contest was held at Butler University on March 29. Nick Tallyn. Prove that 1 ·2 ·3 ···n 1 2 3 n (2n + 1) < 3 n(n+I) . find the volume of S. The Wabash College team consisted of Alexander Radnovich.3.z~1 + IZl . Purdue University. and Wabash College tied for second place. The vertices of a triangle ABC have coordinates (a cos ai.Exam #311996 41 and Z2 P199S1. Prove that 4(x 2 + x P199SS. Find the coordinates of the orthocenter of triangle ABC. Jzi . ABC is an equilateral triangle with each side of length a.z~1 = IZl + z21 + IZI . and James Moore.z21· Find a closed formula for the sum k=O tG~)' 2P+ 1 P199S3. The Purdue University team consisted of Dan Crosby. Ng.2. P199S4. and Tze Chao. and David Whittaker. P199SS. L is a line parallel to BC and b units below BC. Let n be any integer greater than 1. Twentyfive teams from eleven colleges participated. The IU team consisted of Seth Padnkin and Richard Swartz. + 1)3  27x 2 (x + I? = (x . P199S7.Jz? . Andrew Jones. Prove that the integer part of (V5 + 2)P is divisible by 20p. a sin ad.
find the length of the shortest possible such curve. If its area is written as ijA + ifIj. he passes the remaining $50 to the third rat who proceeds to take a dollar from every third cat. Find the largest possible area of a pentagon with five sides of length 1 and a right interior angle. P19963. Sterling's formula. 04. At this point the rats and the money they have taken disappear in a cloud of smoke. Using Wallis' product. 01. find A+B.42 Exams P19961.y > 0. y ~ 3x .y): x. On his return. find the a priori probability p that such an arrangement is mathematically possible. he passes the $100 that he collected to the second rat who proceeds to give a dollar to every second cat. A square with sides parallel to the coordinate axes is inscribed in the region {(x. . The San Francisco 4gers derive their name from the last two digits of the year of the California Gold Rush. This continues with the rats alternately giving and taking until the 100th rat gives a dollar to the lOOth cat. and 09 as squares)? P19962. Assuming that sons and daughters are equally likely to be born.x 3 }. Two couples each agree to have exactly n children. P19965. On his return. LA = 90 c irc and AB = AC = 2. P19966. or any other wellknown result. Note that both numbers are perfect squares (49 = 72 and 1849 = 43 2 ). In triangle ABC. If curve I joins points of AB and AC to bisect the area of triangle ABC. 1849. determine limn>CXJ (Pn)( yin). A hundred armed rats enter a bar containing a hundred fat cats. that their last two digits also form squares (count 00. where A and B are integers. They plan to eventually pair off all their children in marriage so that sons and daughters of the first couple marry daughters and sons of the second couple. How many other fourdigit squares have this property. the first rat marches down the line taking a dollar from every cat. After the cats are lined up against the wall. (a) How many cats profited? lost money? broke even from this? (b) Which cat(s) profited the most and how much did they get? (c) How much money did the rats leave with? P19964.
if C's projection onto the xaxis has length d 2 . Find all solutions (x. and Tyson Patterson. P19968. Find all datasets of positive integers (in any order) which are statistically equivalent to {I. + (1 + 2 + 3 + . Exam #321997 The contest was held at Franklin College. 9. Robert Dirks. respectively. The Purdue team consisted of Dan Crosby. Find the probability of getting a license plate in which the last four digits are nondecreasing. P19971. The RoseHulman team consisted of Kyle Lacey. Find a similar expression for the sum + 1)/2 1 + (1 P19973. and the same standard deviation. Teams from RoseHulman and Purdue University placed second and third. yX for real Given the following set of axioms: (1) Every line contains exactly four points. + n). 6}. Izl = 3. y) of the equation x Y numbers x. It is well known that 1 + 2 + 3 + 4 + . Twentyfive teams from thirteen universities participated. the same mean. P19972. then C's midpoint lies d units above G. y > O.5z 2 Find the smallest positive number A so that + 61  <A for every complex number z on the circle smallest such number. Find all continuous functions f(x) whose graph G (of y = f(x)) has the following property: For each chord C of G. The winning team from Wabash College consisted of lun Ma... TzeChao Ng.. 9.. and Elad Harel. 21 Iz4 . (2) No pair of points may be on more than one line. . Prove that A is the P19974. license plates consist of2 digits followed by a letter and then 4 more digits. on March 14. + 2) + (1 + 2 + 3) + ..Exam #321997 43 P19967. In Indiana. Chris Prince. P19975.. Call two datasets "statistically equivalent" if they have the same number of elements. in conjunction with the spring meeting of the Indiana Section of the MAA. + n = n(n for any positive integer n. and Abishai Daniel.
{X E R2 : Ilxlll = I} and {x E R2 : Ilxll= = I}. what is the fewest number of points and lines necessary to satisfy these six axioms? P19976. X2. Exam #331998 This was the largest ICMC to date. The usual way of measuring the length of a vector x (Xl. (b) The triangle inequality Ilx + yll :s:: Ilxll + Ilyll holds for all norms. such as IlxliI = L i=l n IXil (a) We know that the set of points {xER2 : II X 112 Find each of the following sets of points: = I} is the unit circle. Initially."" xn) in ndimensional Euclidean space is the Euclidean norm But there are other norms which can be used. Determine when equality holds for the inorm and the (Xlnorm.44 (3) Each point must be on at least four lines. then the point must be on exactly one line that has no points in common with the given line (a) What is the fewest number of points and lines necessary to satisfy these six axioms? (b) If n were substituted for "four" in (1) and (3). . express the volume (in terms of ex) of the remaining water. If the farmer tilts the trough (along the side) at an angle ex with the horizontal. The top of the trough is parallel to the ground. Exams (6) If a point is not on a given line. Twentynine teams from 22 universities and schools participated in this ICMC. (5) There is at least one line. A farmer has a trough 4 feet long with semicircular cross sections that are 2 feet in diameter. held at Ball State University. equality only holds if X and yare colinear. the trough is full of water. (4) Not all the points are on one line. For the 2norm. P19977. The winning team from Wabash College consisted of Robert Dirks.
then what is h(x)? P19986. D. il is said to bisect fa and h in area if for every nonnegative x. For any nonnegative x. h be three nonnegative increasing realvalued functions defined on the nonnegative real numbers with fi(O) = 0 and fa (x) :s: il(x) :s: h(x). (x. Let P be any point of segment AB.. (c) Find the radius of convergence of the power series 2:~=1 D. Let Xn = {1. il(x)). Randy Motchan.CPD = L. . and fk = fkl + fk2 be the Fibonacci (a) Show that the points F xy .Exam #331998 45 Abishai Daniel. and Jun Ma. the area of the portion of the R( x) lying between the curves y = fa (x) and y = il (x) has the same area as the portion of R( x) lying between the curves y = il (x) and y = h (x ). Riten Sonpal.q > 1. P19984. Let fa. (0. il. 0). P19983.2u is divisible by 7.2. cp. and (x. h(x)) Then. numbers. .. Robert Linne.0). from Xn onto itself. il = 1.DPE.AB and L. P1998S. Show that an integer n with final (decimal) digit u is divisible by 7 if and only if nlou .n}. Matt Lepinski. cp is called a derangement if cp( x) i= x for all x in X n .q2 with p .xy .. Let fa = 0. P19982. Prove that an odd integer n > 1 is composite iff there exist nonnegative integers p and q such that n = p2 . (a) Find a closedform expression for Dn the number of derangements of n objects. fk+d} lie on two hyperbolas: y2  (b) Show that the only points on the hyperbolas y2 . let R(x) denote the rectangle whose vertices are (0. Let C be a circle with diameter AB. If fa (x) = xf3 and il (x) = ax f3 (a > 1. (b) Show that Dn is equal to the nearest integer to ~.x 2 = ±1 with nonnegative integer coefficients are points in F. The RoseRulman team consisted of Kyle Lacey..8 > 0). Teams from RoseRulman and the University of Evansville placed second and third. and il bisects fa and h in area.(. Let C. A permutation ofn objects is a 11 function.x 2 = ±1. . = {Uk. E be distinct points of C all on the same side of line AB such that DPl. P19981.. and the Evansville team consisted of Siddartha Naidu. Show that (PD)2 = PC· PE. respectively.
Let us say that two nonsingular 2 x 2 matrices (with complex entries) A and B are equivalent if there exists a complex number w such that A = wB. the other thrown away. that is.46 Exams P19987. Let [A] denote the equivalence class of A. one is put back. (c) How many distinct square roots does (~ ~) have? Exam #341999 This year the competition was held on the Indiana University in Bloomington. Dennis Lin. (ii) If both are orange. respectively. Barry Weliver. Randy Motchan. The following procedure is performed repeatedly. Two balls are chosen at random from the urn: (i) If both are yellow. (a) Find as simple a method as possible for calculating the square root(s) of an equivalence class of matrices other than (~ ~) Here is an important example to consider: the square root(s) of (~ ~) are (~ ~) since and (~ ~) = (~ ~) (~ 5) (14 35) 7(21 ~) 4 7 21 = and (~~)(~ ~)=(~ 195)=3(~ ~). and Keith Henderson of Purdue University won the contest. Note that multiplication of such equivalence classes is a welldefined operation. The RoseHulman team consisted of Matt Lepinski. James Lee. Jun Ma. they are both thrown away and a yellow ball from the pile is put into the urn. The team of Thomas Horine. and the Wabash team consisted of Robert Dirks. . P19991. Next to the urn is a large pile of yellow balls. In a large urn there are 1999 orange balls and 2000 yellow balls. (b) How many distinct square roots can an equivalence class that is not equal to (~ ~) have? Realize that [A] = [A] so that the usual sign ambiguity doesn't exist here. the set of all matrices equivalent to A. Teams from RoseHulman and Wabash College placed second and third.
) consists of the entries n! ( n) k .. 2.f k=l 1 n (k) + 10 f(x)dx r :. 1.(k)) converge or diverge? P1999S. P19993.. Several textbooks claim that the convergence is monotone in n. .n).k)! (k=0. hitting a square board at random. 2:. .Exam #341999 47 (iii) if they are of different colors. P1999S.. Let P be any point inside an equilateral triangle T.. What is the probability that a dart. Does any row of Pascal's triangle contain three consecutive entries that are in the ratio 1: 2 : 3? Recall that the nth row (n = 0. Let A be a square matrix and suppose that there exist positive integers m and n such that Am = I and An I I. Show that the sum of the three distances from P to the sides of T is constant. 1 as n + 00 is familiar to all students of calculus.. show that ~~~ CiY + (XiY)2 2 < ~ J('f)2 + VX2 i y2 2 and find all cases of equality. The statement :. Show that this claim is false even for continuous increasing functions. P19997..2. lands nearer the center than the edge? P19996. If x and yare positive numbers. What is the color of the last ball in the urn? P19992.. Does the series f k=l COS(I. Calculate .1. P19994.k!(n . the orange one is put back into the urn and the yellow one is thrown away. ..
I didn't do it. P20004. and Lucas Beverlin from RoseHulman came in third. What is the largest possible ratio of the area of the triangle to that of its circumcircle? You . Zhenyu Yang. Fl = 1. made the following statements when questioned by police. The team consisting of Matt Lepinski. Prove that a group G of order 15 must be cyclic. Thomas Horine. and Keith Henderson. The second place team from Wabash College consisted of Roberts Dirks. Four suspects. and Chris Duefel. Let the Fibonacci sequence Fk be defined as Fo = 0.48 Exams Exam #352000 This year's winning team was from Purdue University and consisted of James Lee. P20003. Determine whether converges or diverges. Ducky did it. P20001. Becky is lying. (a) Clearly 2222 is fortunate. In a triangle one angle is twice another. Call a number N fortunate if it can be written with four equal digits in some base b E Z+. It is known (and you may assume) that Fk is the closest integer to Tk / V5 where T = (1 + V5) /2. who did it? Arby: Becky: Cindy: Ducky: Becky did it. P20002. and Fk+2 = Fk+l + Fk for k ~ O. The contest was held at Earlham College. If only one of them is telling the truth. Show that P20006. one of whom was known to have committed a murder. why is 2000 fortunate? (b) Find the greatest fortunate number less than 2000. P2000S. Dennis Lin.
Exam #352000 49 may express your answer in terms of g  _J4+VfO 12' P20007. Prove or disprove the statement: an irrational power of an irrational number is irrational. You may assume that V2 is irrational but be sure to prove all other results used. P2000S. Find all functions f (x) which satisfY the equation f(3)(x) f(x) and have + f"(x) = x+oo + f(5)(x) lim f(x) = f(O) = O. .
This page intentionally is no longer blank .
Solutions Exam #11966 519661 We will solve the more general problem of determining which integers A yield integer solutions x and y to the equation x 2 . then A is a multiple of 4. Since the LHS factors. s = 1) and when A is a multiple of 4. let a 3 = A. Since there are five points and only four smaller squares. Look under Diophantine Equations in the Index for similar problems. The system has solutions x= r+s 2 and y = 2' Thus x and yare integers whenever rand s have the same parity.y = r and x + y = s. rs with the condition that A = rs. 519662 Divide the unit square into four smaller squares each of side length 1/2 (by drawing two lines which bisect parallel sides of the original square). This happens for all odd values of A (r = A. we have the system of equations x .y2 = A. If a is odd. two points must lie in the same square. then so is A. so solutions exist for all positive integer values of a. It is clear that any two such points can be separated by a distance of less than V2/2 unless the two points are at opposite ends 51 . For our particular problem. and if a is even.
52
Solutions
of the diagonal of the smaller square, which cannot happen since they are both interior to the original square. Look under Geometry in the Index for similar problems.
519663
Consider the numbers
There are p + 1 numbers in this list so two of them are congruent mod p. The difference between those two is the required sum divisible by p. (Note that p being prime is irrelevant.) Look under Number Theory in the Index for similar problems.
519664
Solution 1:
Suppose the functions are f(x) and g(x), then we must have
( f(X))' g(x)
Applying the quotient rule, you get
f'(x) g'(x) . f'(x) g'(x) .
g(x) f'(x)  f(x) g'(x) (g(x) )2
Multiply this equation by (g (x)) 2 g' (x) to simplify the equation and we get
g(x) g'(x) f'(x)  (g'(x))2 f(x) = (g(X))2 f'(x).
Ifwe know one of the functions, say g(x), then we have a first order linear differential equation for the other function, that we can solve. Choosing g(x) = x so g'(x) = 1, we find that f(x) must satisfy
x f'(x)  f(x) = x 2 f'(x).
Rewriting this equation, we get
f'(x) =
or
~~ =
x(1
~ x) f(x),
df
dx
f
Integrating, we have In If(x)1
x(1x)"
= In Ixl In Ix 11 + C,
Exam #11966
53
so f(x) must be
f(x)
=
x C x _ l'
An example of two functions f (x) and g( x) having the property that the derivatives of their quotient is the quotient of their derivatives is given by f(x) = x/(x  1) and g(x) = x.
Solution 2: Begin the same as in Solution 1, but let g(x) = exp(kx), so that g' (x) = ke kx and the differential equation simplifies to
(k  1)f'(x)
Solving this yields
=
f(x).
f(x) = Cek"',.
Then, for any k =J 0, 1, the functions requirement.
f and
g satisfY the necessary
Look under Differentiation or RealValued Functions in the Index for similar problems.
519665
Solution 1: (In what follows, "sequence" refers to an ascending sequence of positive integers as in the problem.) Let S(n, k) be the number of different sequences of length k in which every number from 1 to n occurs at least once. Then S (n, k) is simply the number of ways to partition a k element sequence into n nonempty subsets. Hence, S (n, k) = (~= i). Let T (n, k, r) be the number of different sequences of length k in which only r (out of n) distinct numbers occur, then we have
T(n,k,r)
=
(~)S(r,k)= (~)G=~)'
Finally, we see that the solution is
where the latter equality is an invocation of the Vandermonde convolution.
Solution 2: Make n + k  1 blanks and fill in k of them with x' s. For any such arrangement of x's, define ai = 1 + the number of blank to the left of the ith x
54
Solutions
for each i from 1 to k. This gives a 1to1 correspondence between the sequences we are trying to count and the ways of putting k x's in some of the n + k  1 blanks. But the number of ways of doing the latter is obviously (n+~l). Look under Enumeration in the Index for similar problems.
S19666
Solution 1: Let f(x) = v'ax 2 + b. We want to show that f(x) is a contraction, i.e. :30 < C < 1 s.t. If(x)  f(y)1 :; Clx  yl for all x, y E R. In this case, the sequence Xl = C, Xn+l = f(xn) converges to the unique fixed point of f.
If(x)  f(y)1 = Ivax 2 + b  vay2
=
+ bl
valvx
2
+ b/a  Vy2 + b/al
:; val x 
YI·
To see that lv'x 2 + c  Vy2 + cl :; Ix  yl for any c 2: 0, Multiply and divide by the conjugate to get
Iv' x 2 + c + Since
Vy2 + cl'
h/x 2 + c+ Vy2 + cl 2: Ixl + Iyl 2: Ix + YI,
the inequality follows. Since 0 < < 1, f is a contraction. The limit of the sequence {x n } is simply the fixed point of f. Solving x = v'ax 2 + b yields limntoo Xn = Vb/(l  a).
va
Solution 2: we see that
After computing the first several terms of the sequence,
xn+1 = va nc2 + b(l + a + a2 + ... + an  l )
implying that the diagonal is also a permutation of the set {1.. n}. Q3 are all pairwise strangers. Look under Matrix Algebra in the Index for similar problems. Q3. 2.. then P is a stranger of 5 . . Q2. Then we have I j and 1)+2 are disjoint for each j = 1. Exam #21 967 519671 In a symmetric matrix. . Thus. otherwise Ij+1 c 1 j U 1)+2. Q i. 519667 Assume that the set {h} is minimal in the sense that none of the intervals is a proper subset of the union of some others.1) such entries.2. If P is a friend of k other people. At least one of these two sets covers 1/2 of I.2. or 1)+2 C 1jH' either of which violate the minimality assumption. Qj forms a triple of pairwise friends. If Q1. then there must be n copies of each number in the matrix and. so that Ii begins to the left of IiH for all i.k must be at least 3. lim (xn+d = 1 1 Look under Limit Evaluation in the Index for similar problems. . Otherwise.. a 2 Since 0 < a < 1. . at most n . the two must be the same. j). Then we can order the intervals by the value of their leftendpoint. n . the set of intervals with odd subscripts consists of mutually disjoint sets. If each row is a permutation of the set {I. P. Q2. Look under Enumeration in the Index for similar problems. and one of k or 5 . as does the set of intervals with even subscripts. There are n( n . 519668 Choose a person P. .. The remaining copy of each number must be on the diagonal of the matrix. assume that P is a friend of at least 3 other people: Q1.Exam #21967 55 10 O· V c + b() = Ma.. In this case. Without loss of generality.n}.k other people. . .1 copies of each number can paired off. since their union covers all of I. all nondiagonal entries are paired with another.. we may take the limit giving ntcx. Q i and Q j are friends (for some i. then we are done. since n is odd.
519673 Let f (a. b) . that is. ·f ob . b) = O. we have and Subtracting equation (2) from equation (1). Then the equations for the parabolas can be given by and y2 + 2B2x + 2C2 y + D2 = o. assume that the axes of the parabolas lie parallel to the xaxis.ab. Yl) and (X2' Y2) lie on each of these lines. for a . a of < 0 1 b < e a . let (Xl. we obtain 2(Bl . Yl) and (X2' Y2) be the points of contact of the common tangent with the parabolas.56 Solutions 519672 Without loss of generality. since of ob = 10gb .B 2) (Xl. Yi) is given by YYi + Bi(X + Xi) + Ci(y + Yi) + Di = o. b) = ea + b(log b . Further. Y2) + (Dl . the midpoint lies on the line which is the common chord of the parabolas. If f (a. Using implicit differentiation.D 2 ) = 0.a. Since both of the points (Xl. then the result follows. (The common chord is the line segment connecting the two points of intersection of the parabolas. That this chord exists is implicitly assumed here. we have of ob =O·fb= 1 e .::: 1. X2 ) + 2(Cl _ C 2 ) (Yl .1) . Note that if b = e a .) Look under Analytic Geometry in the Index for similar problems.::: 0 and b . then f(a.::: 0. we have the tangent line to parabola through (Xi.
Case 1. with k a positive integer. nor is n equal to 2. thus (. Thus all the odd coefficients must occur for even values of m.) = (. (~) is odd and so the number of odd coefficients in (x + y) n is the same as the number of odd coeficients in (x + y) k. b) has an absolute minimum at b = e ao . b:::. b) 2: 0. is just twice the number of odd coefficients in (x + y)k. x (n .. Since = (n~m) and.m + 1 is even. the result follows. the number of odd coefficients is twice the number of odd coefficients obtained by considering only even values of m in the symbols (. From the facts about of job above.Exam #21967 57 and of O'f b a ob > 1 > e . Further. and only if. then n (n) nm+1(m1 ) = m m is an even integer since n . 00. which contradicts the assumption that n was the least positive integer for which the proposition fails.:.m are not congruent (mod 2). Look under RealValued Functions in the Index for similar problems. and conversely. But as before we find e~~ 1 ) == (~) (mod 2). If m is odd. consider the function f(ao. (.. Case 2. since f (ao. Let n = 2k + 1. fixing ao greater than or equal to zero. Since ao was arbitrary. (That is.. since nand n . Then (. we have f(ao. with k a positive integer. b:::. 519674 Suppose the proposition false and let n be the smallest positive integer such that the number of odd coefficients in the expansion of (x + y)n is not a power of 2. 00. Look under Enumeration in the Index for similar problems. Thus the proposition is true. x m x 2 x 4 x ..m) by removing all odd factors and divisors. By inspection n is not 1. again giving a contradiction.:. Let m = 2q.) is congruent (mod 2) to 2x4x . Therefore. f(ao. for 1 :::.:.:J .~) is odd if. if x and yare congruent (mod 2) then so are hx and ky for any odd integers hand k. Let n = 2k. for 1 :::. e ao ) = 0. b). ·xn 2 x 4 x .) But the latter fraction is evidently (~)..). Thus the number of odd coefficients in (x + y)n.
! 8 B 3 8 c D E . the distance from A to F.. If the square is divided as illustrated. the requirement that Ian I < 2. But using the triangle inequality.. N M L K J o H Q G R A F . AH.58 Solutions 519675 Suppose that z is a root and Izi < 1/3. and BJ are all larger. Look under Complex Numbers or Polynomials in the Index for similar problems. Then implies that or lalz + a2z2 + . 519676 In the illustration below. BH. and the distance from B to G are each V65/8. and Izi < 1/3. AT. we obtain a contradiction. The distances AG. each piece has a diameter of V65/8. + aNzNI = l.
then the set must contain one or both of D and P. or M.00). J. WLOG. so this can't happen. but there are only four. H. L. J. K. if we define Xn inductively by setting . If the set contains none of those points. it may contain P. Look under Geometry in the Index for similar problems. the set has diameter at least V65/8. Q. Again the diameter is at least V65/8. L. Now. but does contain a nonmidpoint of one of the sides. We will now show that this piece has a diameter of at least V65/8. This leaves the possibility of a set consisting solely of the midpoints. H. Finally. R. One of the three pieces must contain at least six of the sixteen boundary points symmetrically arranged in the illustration. K. This leaves the possibility of the set consisting of B. but contains. then xn = L j=n+l 00 xj = Xn + 1a + L j=n+2 00 xj = X~+l + X n +1· So Xn+1 must be the unique solution of this equation. the equation x + x a = y has a unique solution with x > 0. then it must contain one of C. Further. If follows that the sequence {x n } is unique. F. In this case. in which case it has a diameter of at least V65/8. WLOG. monotonic increasing function of x with limxtoo x + x a = 00 and is equal to zero at x = 0. but does contain M. and hence has a diameter of at least V65/8. H. C. the set has diameter at least V65/8. This is because x + x a is a continuous. 519677 First note that for a > 0. P. Thus it assumes every positive real value exactly once for x E (0. or N. Suppose the set contains none of the comer points. G. and one or both of C and Q. If the set also contains G. y > 0. and again the diameter is at least V65/8. or L. D. WLOG point A. M. it must also contain D or F. if the sequence {x n } satisfies the conditions given in the problem. If the set doesn't contain M either. or F. Suppose the set contains a comer point. If the set contains none of those points.Exam #21967 59 Now suppose that the square is divided into three arbitrary pieces. D. say B. comer E then the set must also contain one of N. K. or R. if the set contains none of E. Again the set has diameter at least V65/8. If the set contains any of points F. G. F. Q.
By symmetry. we have that IT(q) .T(q)1 . suppose c > 0. Ip . So as x varies along 5 in a ~neighborhood of m.60 Xn+l equal to the solution of x + x a = x n.T(sn)1 Ip . = j=n+l L xj. let Sn E 5 be such that Ip . Sn ~ m.T(q)IIT(p) .q'l E Q. IIT(p) .qll ::::. we can invoke the triangle inequality to see that IIT(P) T(q)IIT(p) T(q')11 ::::.sI. Look under Infinite Series in the Index for similar problems.q'l for all p.T(q)1 = Ip  ql Look under Geometry in the Index for similar problems.T(sn)1 = + IT(q) snl· .snl + Iq . for each n E N. as n ~ 00. < 2c. Iqq'l < c. for every S E 5.snl < lin.T(q)1 ::::.T(q)IIT(p) .T(q')11 + Iq . IIT(p) . 519678 Suppose that p.xl is a continuous function of x. we see that IT(p) . hence Ip .sl = Iq ...xl assumes rational values.Now. then it is clear that xn = x~+l co Solutions + Xn+l = x~+1 + X~+2 + Xn+1 = . Let 5 be the line segment of length 1 which is perpendicular to pq and has m as its midpoint.q'l· Also.ql.Ip  qll q'll + lip q'Ilp  ::::.q'Ilp . IT(q') T(q)1 ::::.T(q')11 + lip .snl E Q and 1m . Iq . Ip .T(q)Ilp  qll = IIT(p) .q'l < c and Ip . Now. we have IT(p) . Letting c ~ 0. and we see that IT(p) . IT(p) .T(q)1 ::::. IIT(p) . q E R 2 and let m be the midpoint of p and q. .snl + Iq  Letting n ~ 00.T(q')1 ::::. Ip .snl ~ Ip . q E R2.ql· Now. We can choose q' E R2 such that Iq . As a consequence of the previous argument. Such a choice is possible because Q is dense in Rand Ip .
B. QS or and Ri. the rays this possibility is also precluded by the previous discussion. Then the points A. Because no point can be both the initial and second point of such rays. Q. Let S AB be those rays PQ rays PQ or intersecting AB and S BC be those rays PQ intersecting BC. there exists an M such that n M. depending on which side of line P R the point Q lies. Let the other points of these rays be Q.xi. It follows furthermore that if PQ and or if and are both in. QP QR RJ3 QR PR QP QR stJ.Exam #31968 61 Exam #31 96 S 519681 Since the set of Riemann sums is bounded. and S. . P.::: I L:!(ti)(Xi i=1 XiI)I for all ti. then the line through P and Q cannot intersect ABC D in adjacent sides. exactly one of the intersects either AB or BC. A. SAB. The previous discussion showed that or RQ. If rays PQ and say. R. Look under Riemann Sums in the Index for similar problems. then or RQ must be in SBC. 519682 Assume the claim is false for quadrilateral ABC D and five interior points. proving the claim. it follows that some point P is the initial or second point in at least three of these rays. If P and Q are two of the five interior points. and Q form a convex pentagon. where P and Q range over the five interior points. SAB. and D form a convex pentagon. To see this. rays. Thus S AB U S BC contains ten are both in. without loss of generality. R. C. then either the points P. Q. However. suppose.Xil. Q. R. that the line intersects AB and BC. assume S AB contains at least five rays. Thus we have implying that is. Therefore. or must be in S BC. and C or the points P. sR Look under Geometry in the Index for similar problems. Without loss of generality. D. one for each pair of points P. that ! is bounded. say.
1).i nk < ¢(nk) < nkH' . so that by the chain rule we have F'(x) = I'(L(x))L'(x) = tf(tx). Let fn(x) = {nIx.1). This is a special case of Catalan's Conjecture and a general exposition of the progress that has been made on the conjecture can be found in MAA FOCUS. so that F'(x) = tf'(x) also. The only solutions to the second are (2. 0. Now I: 7 fn(x) dx = 1 so that limn+oo J~oo f n= l. 519684 The only integral solutions to the first equation are (1. Since 0 ::::. fn(x) ::::.62 Solutions 519683 Fix t and define L : R n It R n by L(x) = tx. Since dn diverges.0) and (2.en ]. Limit Evaluation or Riemann Sums in the Index for similar problems. Therefore l' (x) is constant and f (x) must be linear. 519686 Consider any two sequences {c n } and {d n } each tending monotonically to zero. Therefore l' (tx) = l' (x) for any t > O. Look under Multivariate Calculus in the Index for similar problems. Look under Infinite Series. vol 21 #5. for each n there exists an integer ¢( n) such that L:j=: dj 2: e where e is any fixed positive number. F(x) = tf(x).3) and (1.(nk)' It is clear that J. By the definition of f. with L'(x) = tI. and L: Cn converging while L: dn diverges. otherwise. Look under Diophantine Equations in the Index for similar problems. 519685 This conjecture is false. with dn > en. where I is the identity matrix. fn on[1. Then L is differentiable on Rn. for all x. ~. Now let F(x) = f(L(x)). 0 uniformly on R. Define the sequence nk recursively by nI = 1 and dnk+l < cq.
}. . it follows that there exists B such that ~Lf(zj."" c</>(nt}. .B) 7f j=1 1 n L cos(Oarg Zj )~O IZj I cos( B . ...Exam #31968 63 Now a solution to the problem can be given by the two series with terms {an} = {Cnl . . j=1 7f 7f Because the maximum of a function is at least its average. each of the series an and bn diverge since they contain infinitely many stretches of terms adding up to more than e.·· . Cn2 . ~ J7r ~ f(z·. Cn2 +1. B) dB = 27f J7I"/2 71"/2 Izl cosBdB = .dn2 ..·· . bn ) converges by comparison with Cn· z= z= z= z= Look under Infinite Series in the Index for similar problems. d n2 + 1 . O}.)..iO < J cos(Oarg Zj )~O L cos(Oarg ZJ )~O Zj. +1.}. .. C</>(n3)' d n4 . . B) = max{lzl cos(B . Cnl +1. Clearly. 7f Therefore. {b n } = {d n" d n. .arg Zj) IZjlcos(argzjB) L cos(Oarg Zj )~O =Re cos(Oarg Zj )~O z·e. d</>(n. C</>(n2)' d n3 . and min(a n .. d</>(n2)' Cn 3 . 519687 f(z. .···. d</>(n3)' Cn4 . Look under Complex Numbers in the Index for similar problems.arg z). Then 1 27f J7I" 71" Izl f(z... .B)dB = ~ ~ J 71" j=1 t hl =~.
then the set of points at the other end of these arcs contain s . that is. t) + r(s. Let n = r(s . Jrj is 0. In particular. k). t) + r(s. By noting that . 2 f "( x ) = . t) blue arcs or r(s. t . Exam #41969 519691 We will show that f(x) = sin x . In the first case these blue arcs. we can apply Rolle's Theorem to verifY that 0. By observation of f' (x) = cos xI + and Jr are critical points of f(x).1) red arcs at v. plus those from v to these points. t . If there are r(s 1. we find that 0. Look under Enumeration in the Index for similar problems. . form a collection of s blue arcs. We will give an existence result for the more general problem of finding r (s. Working inductively on k = s + t. and Jr are the only critical points of f (x). There are either r(s 1. r(s .t. In the second case. ~. A similar argument holds if there are r (s. for what n can we find a set of s blue arcs or a set of t arcs.SIll X +Jr has exactly two zeros on [0.x + Jr 2: x2 ° 2:.) Since f(O) = f(Jr) = and f(~) = 1~. t  1).1 blue arcs or t red arcs among them. by our induction hypothesis. Let v be any point. Jrj. t). we have the result holding for k = 4 by inspection. f" would have a third zero. and let each arc be colored blue or red. as required.1. we show that r(s. Jrj. t) blue arcs. we can see that the minimum value of f(x) on [0. t 1) red arcs at v. (If f' had another zero. ~. t) :::.64 Solutions 519688 This problem asks you to look for the Ramsey number n = r(k.1) be the number of points that we have. Now suppose the result holds for all values smaller than k. ° Look under RealValued Functions in the Index for similar problems. on [0. we already have t points all of whose arcs are red.
But since (p. 519694 Label acceleration. q) = 1 and (p. Pqr = Apqr . to a final value of a = 9 + avo. r) = 1. the amount of time to traverse this distance. The ascent will take the same amount of time as an object travelling on a straight line with initial velocity v = 0 and initial acceleration a = g. going upward. the acceleration begins with a = 9 and continues with a = 9 . r Multiply through by pqr yielding Pqr + Qpr + Rpq = Apqr. So the descent will take the same amount of time as .Qpr . Thus. and 9 the acceleration due solely to gravity. We have 83 = w3 + w3 + 3ww(w + w) = 50 + 3·5·8. 8 = 5. t up . is less than t.Exam #41969 65 519692 Suppose that P p +Q+ q R =A.Rpq = p(Aqr . the amount of time for an object with uniform acceleration 9 to traverse this distance. Let v be the velocity of the ball at any time. Rq). where v = 0 (by Rolle's theorem). a. Look under Number Theory in the Index for similar problems.50 = (8 . in the general direction of travel (downward or upward) and Vo be the upward velocity with which the ball is launched. = 83 .Qr hence p divides Pqr. We wish to determine + w. Since a is continually increasing. so a = g. the ball will always experience an acceleration downward greater than g.av (since the air resistance is now upward. a = 9 + av. so v continually decreases until the highest point of ascent. q divides Q and r divides R. At launch. Now. it must be that p divides P. downward at all points along the trajectory of height h. with a continually increasing through the distance h. Going downward. Look under Number Theory in the Index for similar problems. In the same manner.5)(82 + 58 + 10) has To solve for 8 we note that 0 one real solution. but the velocity is downward). 519693 Let w = {h5 + 5J20 8 = W and w V25  5J20.15 .
519695 Let B denote the angle just above A..B) .(B + A) < 27f/3. with a continually decreasing through the distance h to a final value of a = 9 . . is greater than t.A < B < 7f . the maximum occurs for B outside this range. the time an object under uniform acceleration 9 would require. tdown. To conform to the geometric setup. c = 1. B must satisfY < B < 27f/3 and < 7f . which is equivalent to 7f/3 . It follows that the maximum area of the equilateral triangle is J v. However. we may write x sin a + y cos a as x 2 + y2 sin( a + w). Again. ° [(b  ccos(A + 7f /3)) 2 + (csin(A + 7f /3)) 2] v'3 = """4 (b 2 + c2  2bccos(A + 7f/3)).B)r. and A = 7f .aVj. Using the law of sines to compute the left side of the equilateral triangle. it need not be the maximum value. from which we see that its maximum value is J x 2 + y2.A. tup < t < tdown. ° Look under Geometry in the Index for similar problems.B) +csin(A + 7f/3) cos(27f/3 .B)  c cos(A + 7f /3)) sin(27f /3 . For instance. we find its area to be v'3 [bsin(27f/3B) + CSin(B+A7f/3)]2 4 v'3/2 v'3/2 = ~ [bsin(27f/3  B) .1. if w is such that cosw = x/ Jx 2 + y2 and sinw = y/ Jx 2 + y2. Now.csin ((27f/3 .66 Solutions an object travelling on a straight line with initial acceleration a = g. Look under Differential Equations in the Index for similar problems.(A + 7f/3)) r = ~ [(b sin(27f /3 . The claimed upper bound has been established. Thus. for b = 10. the time required to traverse this distance.
 a1C  ao. 519702 If lei :::. Look under RealValued Functions in the Index for similar problems. Look under Limit Evaluation in the Index for similar problems..a (f(a) + 2 ~ f(a + kh) + f(b)) _ (b 2n k=l ~ 12n ~3 f"(c) for some c E [a.n +1 . Look under Polynomials in the Index for similar problems. If Icl > 1.Exam #51970 67 519696 From the Trapezoidal Rule we have I b f(x) dx a = b . 1.. A complete proof of the Trapezoidal Rule itself can be found many places. n 1  = 1 1 lim n n+ooL k=l 1 (1 1) . Exam #51970 519701 We have n+oo hm ( . the result follows immediately. we have the desired inequality.. then we are done. then we have C n = 1 an1C n1 . 1 + + 2 + .an2c n2  . +) = n+oo ~hm n 2n ~ n +k k=l 1 1. If we set n = 1. b]. Divide by cn  to obtain Take absolute values and use the triangle inequality to see that Since Icl > 1. + 1 n dx o 1+ x = In(2)..n Is. .
Also LCZD = 90°. Suppose the result is true for all sets of points of order e. 2 + (b+ 1)2 = YW. To see that the rhombus is a square.b). a.68 Solutions 519703 Solution 1: Without loss of generality.b. z:(a+~. Hence WXYZ is a parallelogram. or be situated such that arc A divides the remaining arcs into two sets. Y. where e = k. say from PI to P 2 and remove it together with its endpoints. and XY ZW is a square. b + 1). Because of the nonintersection constraint on the arcs. Consider a set of points and arcs of order e = k + 2.0). so a 180° rotation maps the figure to itself. b+~). Next. so we proceed by induction. Thus.l YW. a).a).b. XZ· YW = 0 and XZ . Z. _~). PI and P 2 must either be consecutive points around the circle. we may assume that the coordinates of the vertices of the parallelogram are A: (0. Thus ZY = ZW and LY ZW = 90 0 • Thus W XY Z is a square. Then the coordinates of X. C: (a. In the first . Similarly. Y:(1+a. as illustrated below. so a 90° rotation carries one triangle into the other. Then XZ = We also have XZ = (a. let E denote the intersection of X Z with YW. Pick any arc A. we can color the remaining regions with two colors. D: (a + 1. By our induction hypothesis. the other angles of XY ZW are right angles. B: (1. Then mLW XY = mLW X E + mLY X E = a right angle. ZC = ZD and CY = DW. b). 519704 The problem is trivial for e = 2. Ware X:(~. k even. LZCB = LZDW = 90° + the acute angle of the given parallelogram. Since the diagonals of a parallelogram bisect each other. the length of each side is the hypotenuse of an isosceles right triangle with legs of length a 2 + (b + 1)2 /2 and XY ZW must be a rhombus. w:(a. b + 1) and YW = (b + 1. Hence triangles ZCY and ZDW are congruent and similarly oriented. Ja J Solution 2: Let YW and X Z meet at P.0). Then P is a center of symmetry for the figure. b. Look under Geometry in the Index for similar problems.
So we need 10d . A cA.t. xPy. yPz.6)/10 = 4n (from condition (b)).1 + (_3)1 = 2. 846 satisfies (a) and (b) then it is the smallest such number. hence A = A. since d is the correct number of digits.243 = 234 = 18 ·13 = O. x E A =} :3y E A S. So d . Thus if n = 2(10 6 .1. then it is necessary that 6· 10 d . 7+ (3)3 = 20 = 7. this is equivalent to d1 L 10 k == L( _3)k == 0 (mod 13). 13) = 1. so x EA. Exam #6197 1 519711 For part (a) we have (i) x E A =} xES  =} xRx (with x E A)  =} x E A. k=O Since gcd(9. k=O d1 k=O Now we may simply add alternating powers of 3 until we reach zero: (3)0 = 1. 9 + (3)5 = 9 . 2 + (_3)2 = 7. color one of the two regions created by arc A the other color (from what is was before) and switch colors for all of the other regions on that side of arc A.10 d 62(10d 1) 39 13 d1 Hence it is necessary that 13 divide 10d . color the region created by arc A opposite of the color of the region it is contained in.t. Now. 7 + (3)4 = 74 = 9. xPy ~yPz =} xpz (with z E A). Indeed.1 = 9L 10k == 0 (mod 13). In the second case. (a) and (b) are true by construction.1 = 5 =} d = 6. 519705 Let d be the number of digits in n. It follows that A c A. y E A=} :3z E A S. Multiplying through by 10 yields 6·lO d +(n6) = 40n (note here that n will also satisfY (a)) or equivalently n= 6. Look under Geometry in the Index for similar problems.1)/13 = 153. (ii) By (1). hence A c A.Exam #61971 69 case.1 + (n . .
70
Solutions
(iii) x E Au B =} ::ly E A u B s.t. xPy. If yEA, then x E A and if y E B, then x E B. Therefore AU B c Au B. Also, x E Au B =} (x E A) V (x E B). So one of A or B contains a y s.t. xPy. It follows that x E Au B and hence AU B = Au B.
(b) A = ((u,v) E S : v  y = 3(u  x) for some (x,y) E A} = {(u, v) E S : v  3u = y  3x 1\ x 2 + y2 = I}. First, we determine the values that y  3x can take when x 2 + y2 = 1. Thus we look at the minima and maxima of the two functions VI  x 2  3x and VI  x 2  3x on [1,1]. Since these functions are continuous, they will assume every value between their respective minimum and maximum.
~ (±~  3x) = =f
dx
Vl x 2
x
 3
= 0 =} 9(1  x 2)
= x 2 =} X = ±_3_. v'lO
Checking the values x = ±1, ± in ±Vl  x 2  3x yields a minimum and maximum of  v'lO and v'lO, respectively. Hence y  3x assumes every value on [v'lO, v'lO]. Now we can write A = {(u,v) E S: v  3u E
Jk
= {(u,v)
E S: 3u 
[v'lO, v'lO]} v'lO::; v::; 3u+ v'lO}.
= 3u  v'lO
It follows that A is the strip between the two parallel lines v and v = 3u + v'lO.
Look under Algebraic Structures in the Index for similar problems.
519712
We want to compute 79999 (mod 1000). 79999
== 7 1 (7400 )25 == 7 1 (10 3 k + 1)25 == 7 1 • 1 ==
143 (mod 1000).
So the last three digits are 1, 4, 3. Look under Number Theory in the Index for similar problems.
519713
The derivative of (1  e ,,1 )1 is
ex
1
Exam #61971
71
Note, however that this is valid only for x =I 0; worse yet, (1  e ~l )1 has a jump discontinuity at x = 0.
(1e ~l )1 is not an anti derivative of the integrand on the interval [1, 1], but the function 9 is, where
(b) (1  e ~l )1 { 1 g(x)= 1 + (1  e ~l )1 for x > 0, for x =0, for x < 0.
Hence the integral is g(1)  g( 1) = e':l' A more direct approach would be to express the integral as the sum of two integrals, one on the interval [1,0] and the other on the interval
[0,1].
Look under Differentiation or Integration in the Index for similar problems.
519714
For each t E R, let L t be the line y(x) = t. Clearly the set B = {Lt : t E R} is uncountable, Ltl n Lt2 = 0 for t1 =I= t2, and UtER L t = R 2 . It follows from the latter two statements that for any a E A, there is exactly one line L t for which a E Lt. Let f : A ~ B be the function which takes points in A to the line in B which contains them. Since A is countable and B is uncountable, f cannot be smjective. It follows that there exists an L t E B which does not contain a point of A. Look under Enumeration in the Index for similar problems.
519715
For each positive integer k,
Lk/2J
(2 + Y'2)k
+ (2 
Y'2)k =
f;
C:)
2 k i+\
which is an integer; but < (2  J2)k < 1. Therefore, the fractional part of of (2 + J2)k is 1  (2  V2)k, and its limit is 1.
(b)
°
A similar calculation shows that the fractional part of if k is odd, if k is even.
72
Solutions
The first of these has limit 0 and the second the limit 1. Hence the fractional part of of (1 + y'2)k has cluster points at 0 and 1. Look under Limit Evaluation in the Index for similar problems.
519716
For each group of five robbers, there must be at least one lock to which none has the key, and for any two groups, these locks must be different (else the addition to one group of a new robber from the other would provide a majority no member of which contains a key for such a lock as we have associated with the first group). Therefore there are at least as = 462. many locks as there are groups of five robbers, namely Now consider anyone robber. For each group of five among the remaining 10 robbers, he must have a key to each lock to which they collectively do not. Since each group has at least one such lock, he must have at least as many keys as there are such groups, namely C50) = 252. In order to show that no more locks or keys are required, we now demonstrate that 462 locks and 252 keys are sufficient. In accordance with any onetoone correspondence between a set of 462 different locks and the collection of sixmember subsets of the robber band, we give keys for any one lock to each member of its associated subset and to no other robber. Each minority, being a subset of the complement of such a sixmember subset, cannot open the lock associated with that subset. On the other hand, consider any majority and any lock. Since only five robbers do not have the key to that lock, at least one of the robbers in the given majority does have the key.
C;)
Look under Enumeration in the Index for similar problems.
519717
We define three new items from the given operations as follows:
0=Xn*X,
forallX,and
A C B
iff An B = A.
With these definitions and the axioms given we can show, after a considerable amount of simple but lengthy work, that we do indeed have a Boolean Algebra.
~I.~I) 22' so AI exists and is equal to _~A2 . (b). and D are the same size. C. f has a relative maximum at x = e.B are invertible.B)I(C .Y) = C . The only conditions are that A. If A + B and A . If you start with the standard axioms for a Boolean Algebra.Y = (A .Exam #71972 73 In fact.2A . Then f'(x) = (1 . you can derive (a).2A . X + Y = (A+B)\C+D). B. it is an . and from these equations X and Y can be found. Subtracting the second from the first yields (A . Look under Algebraic Structures in the Index for similar problems. Look under Matrix Algebra in the Index for similar problems.B)(X .D.as x passes through e. the conditions of the problem give an alternate characterization of a Boolean Algebra.D). Therefore. Exam #71972 519721 The following calculation shows that it is sufficient that A + B and A . Supposing the existence of X and Y such that the desired conditions hold. or when x = e. the equations AX+BY=C.Inx)/x 2 and f'(x) = 0 when lnx = 1.B be invertible. BX+AY=D yield (A + B)(X + Y) = C + D. the converse is also true. and (c). Further. Let f(x) = lnx/x. 519722 From the given equation I 519723 = A (_~A2 . then X . Since f'(x) changes sign from + to .
smB/2. rt rt Look under Group Theory in the Index for similar problems. if 9 E K. Thus G cannot be H U K. . If r = cos( B/2). 9. since (r. Thus f(7r) < f(e).= cos 27rcos(B/2) + sm 27rsmB/2 = . 3. 7. a n /(l . 6. ilien r = cos . Let 9 = hk. J..n=1 n=1 00 N 00 an n=N+I and the series converges..an) < 2a n . Since H is proper. 3. So 0<"~<"~+2 " L. Thus p = (p2)7 = (A.4. 8) in cycle notation. K. Q. there exists h E H such that h K. 10. Look under Permutations in the Index for similar problems. 0 < an < 1/2. 8. .1a n L. 519725 Let P denote the permutation made by the machine. 2.B) are different coordinates for the same point. which is impossible. 5. 7.4.. Similarly. 10. 9. and f(x) t x t O.2). so In7r/7r < Ine/e whence 7r e < e7r • Solutions 00 as Look under Matrix Algebra or RealValued Functions in the Index for similar problems.I E K. B) and (r.. there exists k E K such that k H. We are given p2 = (A. 519724 There exists N such that for all n > N. then k = hIg E H.74 absolute maximum since f(x) t 0 as x t 00. 519726 The curves are identical.. J.1a n L.. K. If 9 E H. 37r .. 37rB 3 . then h = gk. sufficiently large. Therefore pI3 is the identity permutation.5. Q. Thus for n > N. 6. Look under Infinite Series in the Index for similar problems. 519727 Suppose G = H U K with Hand K proper subgroups. P is also. which is also impossible... Since p2 is a cyclic permutation.2 .3.
with a variable and () fixed. Thus the area.sin 2 a cot () = ~ sin 2a  sin 2 a cot (). A.AD cot () = cos a . We have = ~~ and BC = AB =OBOA = OC cos a . o Let A be the area of rectangle ABCD.Exam #81973 75 Exam #81973 519731 Using L'Hopital's rule.sin a cot () . 519732 Let a be the angle shown in in the figure below. x+oo lim x(e 1 / x  1) = lim x+oo e1 / x  1 1/x e 1 / X ( 1/x 2 ) lim 1 x 1/ x 2 = x+oo e / = 1. x+oo Look under Limit Evaluation in the Index for similar problems.BC cot () = cos a . Since cot () OC sin a = sin a. of the rectangle is given by A=BC·AB = sin a cos a . . = hm .
+ 1 Y 0 Look under Integration or Limit Evaluation in the Index for similar problems. Further. If three 2's occur replace them by 3 + 3. we get A= IcosO. 2i/2 33 ) where i { i i =4 =2 =0 if n if n if n = 0 (mod 3). =2 (mod 3).1 .1+ y 0 1 11 + 0 and dv = yn dy (1 + y)2 .1dy gives S 19734 Integration by parts with u = 1/(1 + y) 1 Since 1 nyn1 yn dyI o 1+ y . each ni > 4 can be replaced by (ni . yn. the integral on the righthand side of the last equation approaches zero as n approaches infinity. Substituting into our area function. Indeed. S19733 Observe that the maximum occurs when no ni = 1. Thus the maximum occurs when each ni is a 2 or a 3. and only if. with at most two 2's occurring.sin 200 cot O. If some ni = 4. the maximum is: n . replace it by 2 + 2. we find a = 0/2. =1 (mod 3). 2sinO Look under MaxIMin Problems in the Index for similar problems. m > 4. dr Setting this equal to 0. . (m2)·2 > m if. increasing the product.2) + 2 in the partition and the associated product will be increased. Thus. leaving the product unchanged. The desired limit is thus lim n+oo (LI1) = 1/2.76 which we differentiate to obtain Solutions dA = cos 200 .
If a E A : B. Ifx. xb. Look under Polynomials in the Index for similar problems. E A : B). if a E A : B.Exam #81973 77 519735 The quotient A: B is an ideal of R. ab E A=} (ab) E A hence a E A: B.a.d.1)2. ab E A =} r(ab) E A =} (ra)b E A so that ra E A : B (in particular. Now let A be singular. Then = d daf(a) Since f(l) (n + l)a n  (n + 1). Look under Matrix Algebra in the Index for similar problems. It follows that A(Bx) = (AB)x = 0 implying that AB is singular. f(a) is divisible by (a . Let B be singular.y E A: B. then Vb E B. yb E A =} xb + yb E A =} (x + y)b E A. If B is nonsingular. from which we see that AB is singular. and z = a + d. then Vb E B. It follows that A : B is an ideal. we have Bx = O. Then for some x =1= 0. Then (ad)2+d 2 = (a+d)2. 519736 Let f(a) = an + 1 .n(a 1) . Ax = O. Hence Ax = A(By) = (AB)y = 0. y = d. Finally. Then for some x =1= 0. ° hence x + yEA: B. then the product AB is singular. there is a nontrivial solution to By = x. then Vb E B. r E R. 519738 Let x = a . or . 519737 We will prove the contrapositive: If A or B is singular. =} (a)b E A. = f'(1) = 0.
0. 519739 There are (~) = 10 equiprobable ways to choose 3 men from the group. k=O . Look under Probability in the Index for similar problems. z) (3d. Exactly 3 of those choices result in 2 Republicans being chosen (2 Republicans and the choice of lout of 3 Democrats). Exam #91974 519741 (a) For a polynomial p(x). From this. Solutions Thus a = 0 or a = 4d. 5d). It follows that the probability of both Republicans being selected is 3/10. = (d.78 or a 2 = 4ad. the polynomial given in the problem is equal to L xlOOOk(x + 1)k. by the binomial theorem. where d is any integer.(1~00  k)) = Coo:m). and all solutions are (x. Now. First. let [x 7n Jp(x) denote the coefficient of x 7n in p(x). y. 4d. d) or Look under Number Theory in the Index for similar problems. we see that It follows that [X7nJ(XlOOOk(X + l)k) = = 1000 [x 7n (lOOOk»J(x + l)k (m .
(b) From part (a).2m + 1) =(m2 m+1)2+1 = Nl + 1. we see that the sum of all the coefficients is 1000 fo[X m ] = fo 1000 ( 1~1 ) = 2 1001 _l. 519743 If m > 1. Look under Polynomials in the Index for similar problems.1)2 = (m 2 + 1)(m 2  + 1) 2m + 2) +1 = (m 4 . 519742 We have yx = (yx)3 = (yx)(YX)2 = (yx?(yx) = ((yx)2y)X = y(yx)2x = y(yx)(yx)x = xy2yx 2 = xyx 2 = x 3y = xy. Look under Group Theory in the Index for similar problems. . set Then we have (m 2 + l)(ni + 1) = (m 2 + 1) ((m .2m3 + 3m 2 .Exam #91974 79 so Therefore the coefficient of x 50 is C~gl).
Since In(2) R::: .3/4].so (m 2 + 1)(n~ Solutions + 1) = = = + 1) (m 2 + 1)(m2 + 2m + 2) (m 4 + 2m 3 + 3m2 + 2m + 1) + 1 N~ (m 2 + 1)((m + 1)2 =(m2+m+1)2+1 = + 1. then .) Thus. But f(x..x 4 .7. We have f(O. so y(3/4) > 0 means that y(x) > ~ .x on [0. y must have a zero on [0. contradicting our assumption.e3 / 4 < O. let x = 8 and y = ~ where 0 < E < 8/2. then n1 solutions. Therefore dy < .576 1 / 3 = 2. 17 are the smallest If m = 1.. S 19746 This conjecture is clearly false. 0) = 0.y) < O.1.O) > 0 for all x. y(3/4) < 0. Let an = 1/(n + 1).n2 S19744 The point (0. .1. Then f(x.3/4].e x < dx We see that (3 ) .0) is neither a minimum nor a maximum. (Or. S19745 Assume that y is never zero on [0. 24 / 3 = 16 1 / 3 < 17. We note that ~ < 1. Look under RealValued Functions in the Index for similar problems. y(3/4) = 1+ 1 o 3/4 dy dx dx < 1 1 0 3 4 / 1 + eX (3) 4 x dx = 2 e3 / 4 .6 < e.e x y . Hence. Look under Multivariate Calculus in the Index for similar problems.3/4].N1 = 3. 2 . = 2.
with [VI = n. v~. 0. and without loss of generality.{vo. . VI as we move around the circle from Vo to v~... Suppose that Vo and v~ are two vertices joined by a chord through the center of the circle.. the infinite sum is an odd function of a. vi. . L(v~l.Vl. by way of contradiction. . Also let a 2 2' a +x We see. that no two of the vertices can be joined by a diameter through the center of the circle. If a < 0. of orders a and b respectively. . . v~ as we again move around the circle the other way from Vo to v~. v~).Exam #101975 81 (the harmonic series) diverges while Look under Infinite Series in the Index for similar problems. say fJ. We have a + b = n . This chord partitions V . Also note that the angles L(vo. 0. Look under Infinite Series in the Index for similar problems. .V l (VO'Vl'V l ' (2 3 4) .2.VO. Label the vertices of Vl as vi. vi). that f(x)= 0<'" 00 a ~ n=l a2 + n2 < 1 0 00 a a2 + x2 dx=2 7r and the result follows. . Vl has the same angle.Vl. a is odd and b is even. (al .. so consider a > O. Vl and V2 . Let V be the set of vertices. v~} into two disjoints sets. Label the vertices of 112 as v~. so the result follows in this case also. 519752 First we show. odd.. using a lower Riemann Sum. Exam #101975 519751 The result is obvious if a = 0.. .. L(vi. . 0. vi).. l triplets Now each of the at a ') 1 2) Vl.
a = (n . this gives us the result that an arbitrary side of the polygon has length 2r sin a.(3) + a. . each have measure 2(n . V2' 3 4) v2 As before. Therefore. This implies that the triangles 0. and they sum to n.2{3.(3). since a and b are integers. 2 . V2. nl v. v:. Letting the radius of the circle be r. each of these form an angle of measure (3 .82 Solutions (where 0 is the center of the circle) are all equal to 2n .(3) and = (3) +a = n. a .O. . Wi.(3).{3 + ad = (2)(n . (b2 . Now 0 < a < n . Thus If nZ1 is even.( n . a contradiction. Vi.(3) = n. (2 v 2.: nl and 0.(3) and angle L(v~..b = 0.b + 1). v2 . a chord starting a vertex Vo going through the center 0 of the circle and ending at the point w on the circle divides the other vertices into two sets of equal order. we can proceed as before to obtain Inl Inl (2)(n . .L(v~\O.:' are congruent.b + 1 < 2 and therefore.(3) (a . L(v~. vb) Therefore we have = a < 2(n .O.{3 implies that 0 < a .vi). 0.v~). v2 . The angles L(vo..(3 + (2) 2 2 2 2 as illustrated below. v2 . . b(n .. (~) (2)(n (a + l)(n . Now consider the ~ triplets I 2) bI b) (vo. v2 ..v~). Therefore a simple calculation gives us (a + 1) (n . where a =n  nl ..
6+a )=2 (n . Again letting the . This implies that the triangles O.1 (2)(7f. Wi . 0. From this we can see that /'1 nl nl = /'2.. Wi .Exam # 101975 83 nI .l is odd.W ' .V 1 2 )=(/JI.O. W) = /'2. nl nl nl 1 L(V22 .v~ and 0.v~ are congruent.V..l_l and nl_l 0. L(v~.6+a2) as illustrated below. v 2 2 are congruent.W) =/'1. we let L(v~. nl and can proceed as before to obtain 1 2 (n .0.I ~~~~r~~ VI 2 w If n. Wi.1 221 ) 1 1 1 ) (2) (7f. so that triangles 0..
then adding an mth line according to the rules given will yield m more components. then the new line will split m previous components yielding m new components. It follows that T m = T ml + m.84 Solutions radius of the circle be r. . We see that the Tm are just partial sums so that Tm = To + I> = 1 + I> = 1 + m(m+l) .1 lines. this gives us the result that an arbitrary side of the polygon has length 2r sin . where "( = ¢ . Also. w Look under Geometry in the Index for similar problems. 2 m m k=l k=l Look under Enumeration in the Index for similar problems. if there are m .. To see this.1 such lines forming T ml components.c¥. 519753 Let Tm be the number of components yielded by m lines in the plane. If there are m . note that the new line will intersect each of the existing lines at a unique point. Clearly we have To = 1 (the whole plane).(.
We recall that. that 32k . Now consider the case m = k + 1. 2. It's clearthatm = 5 < 8 = 23 andm+1 < m+m = 2m < 2·2=2 = 2m . Since k 2: 4.1 . (3). 5 is divisible by 24. == 2 (mod 4) (b) We note that 31 . we need only check exponents that are powers of 2.Exam #101975 85 519754 (a) Ifm is odd. When m > 4. Look under Number Theory in the Index for similar problems.1 = 80 = 16 . then 3m == (l)m == 2m and is not divisible by 1 (mod 4). Hence.2 == 1 + ·2 k (mod 2 k + 1 ). we have 321 = 9 == 1 + 23 (mod 24).1 (mod 2k). Thus.3 == 1 + 2k. 32"'3 == 1+ 2=1 (mod 2= ). m ::. Squaring. ¢(2m) = 2=1. 32"'3 == 1 + 2=1 (mod 2=) and the claim is proved.(3) = 2m . By Euler's generalization of Fermat's little theorem 3<P(2"') == 1 (mod 2m ). Hence ord2m(3) = 2m2. i. Since 32"'3 == 1 + 2m . .2 == 1 + 22k . ord(3) > 2=3.) Therefore.3 2 . we see that 32k . (This is also easy to see by induction. Therefore the order of 3 modulo 2m . or 4. 2=13= . is a power of 2. 2'" Squaring 32"'3. Next. we see that 32"'2 == 1 + 2· 2m  1 + 22=2 (mod 2=). where ¢(2=) is the Euler totient function. We now prove by induction that for m 2: 4. we have 32"'2 == 1 (mod 2=). 4 and the only values for which 2m l(3= 1) are m = 1.1 = 2 is divisible by 21.1 only if m is a mUltiple of ord 2". Thus.. Assume that the result is true for m = k. When m = 4. by our induction hypothesis. and 34 . As m > 2. we will prove that when m 2: 3.e.1 (mod 2=). (3). so 3m 1 for odd m larger than 1. ord 2".1 = 8 = 4·2 is divisible by 22. this is 3 2k .2 • Since the order of 3 modulo 2m is a power of two. ord 2". m < 2=2.2 + j 2 22k + 2· 2k  1 + 2j2 k + 2j2 2k  1 (mod 2k+l).
. (p. p( x) = x satisfies the conditions.x. p == 5 (mod 6) raised to an odd power. Look under Multivariate Calculus in the Index for similar problems. N) = 1. then n == 3 (mod 4). Similarly... Exam # I 1I 976 519761 By inspection.2. which gives the result. and a(pe) = 1 + p + p2 + . This implies that the primepower decomposition of n contains a prime p. 519762 We are given that n == 11 (mod 12). tdt t = Jo r f(t) dt. Then a(n) = a(pe)a(N). If n == 11 (mod 12). A calculation similar to the above shows that 4Ia(n). and let its degree be n. Look under Number Theory in the Index for similar problems. n+ 1 can be found such that p( ai) = ai.1 or p(3) = 3. e odd. Look under Polynomials in the Index for similar problems.) So we can write n = peN. Thus n+ 1 numbers al. and the primepower decomposition of n contains a prime q. it would follow that n == 1 (mod 6). From p(2) = 2 follows p(22 . of degree n has n + 1 roots and this is impossible. we have Jo r g(t)dt = = = i it x a Jo rl r a f(t) dtdx t ~Ot f(t) dxdt Jo f(t) . . Hence 6Ia(n). q == 3 (mod 4). . (If not. Suppose that there is another solution.1) = (p(2))2 . whence n == 5 (mod 6). raised to an odd power. + pel + pe == 1 1 + 1··· + 1 1 == 0 (mod 6).86 Solutions 519755 By reversing the limits of integration.. p(8) = 8 and p(r) = r implies p(r2 1) = r2 1. i = 1. Thus the polynomial p( x) .
f(l) = f(O + 1) = f(O)f(l).a3.b~. and since the points are not collinear bl . and the system has a unique solution. . and f (x) 7 00 as x exactly three zeros. f has Look under RealValued Functions in the Index for similar problems.h+O Ihl ' we are done. Look under Differentiation or RealValued Functions in the Index for similar problems.Exam #111976 87 519763 The quantity whose absolute value is to be shown to be less than 1 is f' (0).al =I. 519764 f"' > 0 for all x. and f'(1)31n3 .3 . rh + sb 2 + t = bi . 7 00.CI . Since f(O) = f(l) = 0. and f at most three.b2 Thus the determinant is not zero.bl . f' has at most two.a2 C2 .4 < 3. Then ral + sa2 + t = ai . rCI + SC2 + t = ci  c~. and since f(l) =I.4 < 0. x = 2y for some y. Since If' (0) I = Ilim h+O f ( h) I = lim If ( h) I < lim Itan hi = 1 h h+O Ihl . b2 . 519765 Suppose that the equation of the circle is x 2 + y2 + rx + sy + t = O.0. we have f(x) = f(2y) = f(y)2 ~ O. Further. The determinant of the coefficients is (b l al)(c2b2)(CI bl)(b2a2). so f" has at most one zero. S19766 Since for all i E R. f has another zero to the right of 1. Look under Analytic Geometry in the Index for similar problems.
2k) x] since cos a cos b = cos(a + b) + cos(a 2 b) .) = = f (. . 519772 (a) Since. but routine manipulation of the summation. Since L cannot equal 1 since this would imply f (x) ::::. 0 for some x. Look under Limit Evaluation in the Index for similar problems. including shifting an index of summation and adding binomial coefficients. we have cos x = by induction. Exam #121977 519771 f(a) = f(b) =} f(J(a)) = f(J(b)) =} a = b. Now let L = limx+o f(x). we have L = 1. Then 1 = f(O) Solutions = f(x . for n = 1. It follows that if f were not strictly monotone then it could not be injective.x) = f(x)f( x) t L2.) 2~ k=O k=O cos(m . Look under RealValued Functions in the Index for similar problems. it assumes all intermediate values.2k) x + cos (( m .) 2~+l f [(. We therefore assume that Hcos x + cos( x)).2k)x cosm x and consider = 2~ f k=O (.1) .ss f(O) = 1. therefore f is injective on J.2k)x cos x cos (( m + 1) . Since f is continuous. The result now follows after some tedious. we proceed cos(m .
Look under Matrix Algebra in the Index for similar problems. (These are translations followed by reflections. If(t) . 1 1 ((2n)!) 519773 MUltiply the matrix A by the vector 1 = (1.f(O)1 = Also. .1.xl so that any isometry must be of the form f (y) = ± (y .2k (2n)! ° + (2n)7r) n = .I)x = 0 has a nonzero solution and A .10 cos2nxdx=.f(x)1 = Iy .x) for some x in R.k#n 6 ~ (2n) sin(2n.f( t)1 = 2It1 so that one of f(t) = 0 or f( t) = 0 must hold. If(y)1 = If(y) . . 519774 Suppose that f(O) = t.Exam #121977 89 (b) We have 1 r 1 1 2n r . Now for any y in R.22nL (2: ) 10 cos((2n2k)x)dx k=O = !_1 ( 7r 22n k=O.1) to obtain Al = 1. It is also easy to see that anything of that form must be an isometry.I is therefore not invertible.) Look under RealValued Functions in the Index for similar problems. Then If(t)  tl = If(t) . It follows that f (x) = 0 for some x in R. Thus (A .2k)x[1f k 2n .f(O)1 = It I Itl· and If( t)  tl = If( t) .. 22n (n!)2 7r = 22n(n!)2' Look under Integration or Trigonometry in the Index for similar problems.. .
(The hint is verified by picking r 2: max{ai.. Then for each k. + pC=1p=C2. From the hint. MI = 10M2 + d l . 1.ei}. • paN I 2 N + 1 PI P2 ... Exam #131978 519781 To find the decimal expansion of an integer M..=CNp I 2 N 1 since each integer n can be written as a product of our N primes.bi .. . the kth partial sum of the harmonic series can be written n=1 L. .90 Solutions 519775 Assume that there are only N primes... recording the remainders: M = lOMI +do . a contradiction.) Look under Number Theory in the Index for similar problems.. one divides by 10 successively.8.. and therefore converging. or 9 to obtain the expansion . where the di are 0. a finite set. .. we therefore have <IIIp·· i=1 2 N 1 This implies that the harmonic sequence is increasing and bounded above. PN bl b2 bN + . k 1 1 = pa1pa2 •. .
Recall that Ilv x wll gives the area of the parallelogram with sides v and w.Exam #131978 91 To write M in the base 2. and (0. (0. To set this up. and D lie on the axes. + bk . we get n 'L. say at (x. and also S(n) = two expressions. and that B..?=l (n+ 1.j)k.?=l jk. . So 2374(10) = 101111001110(_2)· Look under Number Theory in the Index for similar problems. 0).0. a k + bk factors as a k . C. respectively. 0). one must get at the area of a nonrightangled triangle. we may assume that A is the origin in 3space.3 b2 ak + bk = (a + b)(a k  1 _ + .. Adding these 2S(n) = 2)(n + 1.2 b + a k . one proceeds by an analogous algorithm. 0. z).j)k j=l + II Look under Finite Sums in the Index for similar problems. bearing in mind that the remainders must be 0 or +1. One technique is to use the cross product. y. Then area(ABC) = 2' xy area(ABD) = 2' xz area(ACD) = 2' yz and the area of (BCD) is :2IICD x CBII 1 = :2 V(x y )2 + (xz)2 + (yz)2. 519782 To solve this problem.1 ). Now S(n) = 'L. 1 Look under Analytic Geometry in the Index for similar problems. 519783 We recall that when k is odd.
. n= 1 n=2 n=3 n=4 In view of this.ak) + L (ak . If n = 2m + 1. ::. k=rn+2 k=l .92 Solutions 519784 A first step in discovering the solution of this problem might be to graph the function for small values of n. then 2m+1 f(am+d = L lam+! . f(x) is smallest at the middle a and if n is even. an. the proof will be easier to write down if we reindex the given numbers so that a1 ::. f (x) is smallest in the interval between the middle a's. ..Lak. .am+!) k=l k=rn+2 m 2rn+1 = marn+1 . if n is odd.' . That is. We divide the proof into two cases.. a2. as illustrated. a2 ::. The x for which f (x) is smallest is the median of aI. an.L ak + L ak mam+1 k=l k=rn+2 2rn+1 rn L ak.akl K=l m 2rn+1 = L(arn+1 .
m. we see f(x) 2: f(am+d if x:::.x) + (a m +1  x) 2: O.~ ak . aZ+l for I :::.Lakl k=m+l k=l L if x :::.x k=l k=l+l Z = L ak . x :::.(2m . Look under RealValued Functions in the Index for similar problems. In an analogous way.f(am+d = [ + l)x. where I 2: m + 1.1. .21 k=l+l k=l So that f(x) .21 + l)x 2m+l t 1 m =2 L k=l+l (ak . ao. or if az :::.(2m . x :::. k~l ak . x :::. m . aZ+l. and am:::.Exam #131978 93 Now suppose that az :::. x 2: an. If n = 2m. then 2m m 2m 2m m f(x)=Llxakl=Lxak+ k=l k=l L k=m+l akX= L k=m+l akLak' kl Now if az :::. x :::. we can show 2m m f(x) 2: ak . x 2: an. Then 2m+l Z 2m+l f(x) = L k=l 2m+l IXakl=Lxak= L ak. or if az :::.L ak . then 2m Z f(x) and = L akL ak+(212n)x k=Z+l k=l Analogously. aZ+1 where I :::. am +1. x :::. 0. at+1 and I 2: m + 1.
Since AOC is a straight line and 0 is between A and C. we have 10' BI + 10' DI 2IBDI. B. is contained in the triangle whose vertices are the other three points. say D. IBOI + 10DI = IBDI Thus. this is problem 19784 restated. C. so that IAO'I + 10'01 2 IADI + IDCI· By the triangle inequality. If A. so we then have IAO'I + 10'CI + IBO'I + 10' DI 2 IADI + IBDI + ICDI· Look under Geometry in the Index for similar problems. any point on the middle segment solves the problem. Let 0' be any other point on the plane. B. B. and D lie on a line. Thus the solution should be an experiment which a patient experimenter would expect to finish in a finite amount of time with probability 1/3. (That is. C. there are 2N possible outcomes of flipping a coin N times. and D are not the vertices of a convex quadrilateral. ~ IBO'I + 10'DI· IAOI + lOCI + IBOI + 10DI ~ IAO'I + 10'01 + IBO'I + 10'DI· If A. the triangle formed by this pair and 0' contains the triangle formed by this pair and D. and k/2 N i= 1/3 for any integer k. One such experiment is: . B. If A. For one pair of the points A. and D are the vertices of a convex quadrilateral. 519786 The first step in this solution is to realize that no experiment that always finishes after N tosses of the coin can have probability of success 1/3. suppose AC and BD are the diagonals and that they intersect at O. and C. so let's say AO'C contains ADC. IAOI + 1001 = IACI ~ IAO'I + 10'CI· Similarly.94 Solutions 519785 We distinguish three cases. and 0 = D. C. Let 0' be any point in the plane other than D. then one of the points.
Exam #141979 95 Toss a coin until (a) the coin has been tossed an even number of times and (b) at least one head has been tossed. the monkey) at the time t2.. the monkey) at the time t 1 .. Bo) be the age of the monkey's mother (resp. The experiment is over after two tosses unless both are tails.. B3) be the age of the monkey's mother (resp.. or six tosses.. or four tosses.. . after four tosses unless all four are tails. "'"""' . The experiment will terminate unless the experimenter tosses tails forever. = a fixed difference in ages.. Exam #141979 519791 Let Mo (resp.. an event whose probability is zero. and Lw be the length of the rope on the side of the weight. Lm be the length of the rope on the side of the monkey. L be the length of the rope. let Ml (resp. the monkey) at the time t3. W m be the weight of the monkey in ounces. (4) Ml = ~Bl' . Then we have (1) Mi . (3) Mo = 2Bl. B 2 ) be the age of the monkey's mother (resp. and so forth.Bi = k. Bd be the age of the monkey's mother (resp.. let M2 (resp. Thus the experiment consists of two tosses.. or . The experiment is a success if two heads have been tossed. The probability that the event is a success is given by Pr (success) =L 00 Pr(success given that the experiment requires exactly 2n tosses) n=l x Pr( experiment requires exactly 2n tosses) = 6 3 . the monkey) at the present time.n = .. (2) Mo + Bo 4. and let M3 (resp. Let W w be the weight of the weight in ounces.4 3 n=l 00 1 3 1 Look under Probability in the Index for similar problems.
But by (1) and (2). (6) M3 = 3B3· Since B3 = M3 . Mo + Mo .S(O) = = Jo dt = Jo kbd(t) = k r ds r 1 1 abdt * k* (In(c + b) In(c)) = k* In (1 + n In (d(t)lo 1 = k* In ( 1. d(t) be the depth of the snow. b = ~~ be the rate at which snow falls. So by (5) Bl = Ml . Finally. so k = 1 and Mo = 5/2. xy 2. . if to is the time at which it starts snowing. then O· Therefore .+ n7r Look under Trigonometry in the Index for similar problems.p) = 40 + 40p. 519793 Let V(t) be the volume of snow removed at time t. Therefore = (3/2)k.t~) . The graph thus consists of the two families of lines y = x + 7r + 2n7r and y = x + 2m7r. We have 1 = S(I) . Thus Wm = (5/2) . Then dV = kd(t)ds.. and L = 640 = 15 ft. 519792 If sin x = sin y.k = (9/4)k .k = (5/4)k and by (3) Mo = (5/2)k.96 (5) B2 Solutions = 3M3 . WW = (3/2) ·40 = 60 OZ. then to = c/b. 16 = 40 OZ.. and S(t) be the location of the plow. equation (6) gives M3 B2 = (9/2)k and by (4) Ml = (9/4)k. .sm 2· 7r xy or .k = 4. Suppose t = 0 at noon. p = 3/4 of the rope is on the side of the monkey when the system is at rest since we have 60 + 40(1 . = SIn x . Since dV dt dV ds ds dt' ds dt a kd(t)" Noting that d(t) = bt + c.SIn y x+y 2 = 2 COS x+y.k.= m7r 2 2' where m and n are integers.= . where m and n are integers. a = ~~ be the rate at which snow is removed. so ~~ = kd(t).
Drill the hole along the xaxis. 3 y ++f::I+~ x Look under Volumes in the Index for similar problems. to must be negative and therefore we choose to = 0. to Therefore multiplying corresponding sides of these two equations gives us 3k* In(1  ~) = to 2k* (In 1  Solving for to. Since it started snowing before noon.) 3/2 = 8(2) . as illustrated in the figure below. We convert this to minutes by multiplying by 60. ~) . Look under Differential Equations in the Index for similar problems. 93/2 = 367r.m.y2) dy _ _ 27r (r2 _ y2)3/2Ir (3/2) v?=9 = 47r . we get to = 1~v'5.618 hours. (Letting k* = a/k. Then V = Jr v'r29 27rY' (2V r2 . to get 37 minutes before noon. 519794 Let r be the radius of the original sphere.8(0) = k* In (1  ~) . .Exam #141979 97 Similarly. Therefore it started snowing at 11 :23 a.
16). (3. Exam # 1519 SO 519801 We have hence albl > albl .5). (5.aN+lbN+l N = 2~)anbn . (2. (1. Look under Analytic Geometry in the Index for similar problems. (2.1). 9). (1. (3. (1.9 + 7 + 9. 1 + 9 + l7 /9 J.an+lbn+d n=l n=l Look under Inequalities in the Index for similar problems.4). (3.2). 1 + 9 . (3.1). 13). 2). 19). 8).4). (3. 519796 The set of points consists of (6. 2). (2. (3. 10). 2).3). 2).1). 1 . (3. (2. (0. 7+ This list can be extended to at least 38. (3. 1. 15). 2). . (_1)9 + 7 + v'9.7). 17).3). (1. )9 + 7 + v'9. (0.7 + v'9.5). (4. (3. (1.98 Solutions 519795 The next several positive integers are: 5 = 1.9 6= 7= 8= 9= 10 = v'9. (4. (4. 6).
where the Xi is either 1. Notice iliat this argument assumes that Xo is in the interior of [a. but the same conclusions can be drawn if Xo = a or b.) It seems reasonable to choose W1 = 1 and W1 + W2 + W3 + W4 = 40. Then we can measure the loads I = 1. As I E S implies I E S. then 3 + 3z3 + 3 2 z 4 ) and this divides Z1 and hence Z1 = O.40. the set s = {~X'W' : ~ 'l. Look under Differential Equations in the Index for similar problems. .< 1 _ 1} has exactly 34 = 81 members. Then W4 = 40 . Z3 = Z4 = O. so f (xo) = O. 1. 2.4. it follows that there are exactly 40 positive integers in S and each of these is at most 2::i Wi = 40.. A similar argument shows that f (x) ~ 0 on [a. Therefore 0= f(a) ::::. Now note that if o= L 1 4 ZiWi = Z1 + Z2 W 2 + Z3 W 3 + Z4 W 4. As 5 + 3 + 1 = 9. i=1 1 < _ X· . bj such that f(xo) is the maximum value of f on [a. or on the weight pan to achieve a balance..3.2.Exam #151980 99 519802 The problem is to find the weights W1 < W2 < W3 < W4 such that every positive integer I ::::. So S coincides with the set of integers 1. 519803 Choose Xo in [a.=1 XiWi. 40 can be written in the form I = 2::. That is f (x) ::::. Therefore. not used. O. with each 3(Z2 IZi I : : . 0. bj. To measure a 2 pound load. . Then f'(xo) = 0 and we must also have f"(xo) ::::. f(xo) = !" : : . we can also measure 1=5 if we take W3 = 9. . (A given weight can be placed on the load pan. 0 for all X in [a.2::{ Wi = 27. Similarly. bj and hence f vanishes identically on [a. bj. 0.2. or 1. bj. we need W2 = 3 (I + W1 = W2). bj. Look under Number Theory in the Index for similar problems. So 0 gives Z2 = O.
y)(x.ze n ) is an orthonormal basis with ze i ) = y/llyll.. with respect to this basis.. the matrix I + A has the form (Z(1).:.) A short computation shows that if then the vector Az = (y.ze 2). z)x where (. XnYl X n Y2 XnYn (We can assume Y i= 0.Yn) X2Yn :... .y)(x. .) denotes the usual inner product in Rn. the matrix A = xyt = c y X2Yl XIY2 X2Y2 x.ze 1») +1 0 1 I+A= ( (z(1). .y) (y/llyll.Z(2») (ze1).z(n») o o . Yn ) . 0) : 0 o 0 o 1 so the determinant of I +A is the product of the diagonal entries.100 Solutions S19804 For and yt = (Yl Y2 .y)(x. . Look under Matrix Algebra in the Index for similar problems.y)(x. then. .z(1») +1 (x. as the result is trivially true otherwise.y)(x. With this.y/llyll) + 1 = + 1. Another proof can be given based on the fact that the determinant is unchanged on adding a multiple of one column to another. . one reduces A to a lower triangular matrix and then det(I + A) is easy to compute. that is det(I + A) = = (ze 1). If z(l).
I'1m jb 1 X sinx d b+oo x jb 1 sin x dx X = _ cos x I = XII and Icotbl ::. So we need j exists and is finite. g.Exam #151980 101 519805 ~~~~ If we fix x xo.= ~(Igl. This implies that 1imb+00 lIb g( x) dx exists and is finite by an argument similar to the one used to show that an absolutely convergent series is convergent. Hence l' (xo) = for all xo. 0. b 1 lib ~ 0. divide by y . b co~ x dx _jb x cosb jb cosx dx  x2 so that limb+oo Igl exists and is finite..are increasing functions of b that are each bounded above I: lb Ig(x)1 dx 50 lb :2 dx = 1 ~ I: I: .00]. at x x = 00. (x _ y)2. If g(x) = cosxlx 2. Then g+ 2: 0. and then let y tend to XXo f'(xo) f( Xo ) = d log2 f(x)1 _ x d ::. so that f must be a constant function. ° A similar argument with y ::. Look under RealValued Functions in the Index for similar problems.x.2: 0.. so that g+ and g. = xo.. the integral only check that 101 si~x dx makes sense. 519806 (a) The improper integral has two apparent singularities. Put g+ = ~(Igl + g). g. with y > xo. Now oo sinx d 1 X _ . Conversely.g). 2(xy)2 =} log2 f(x) Iogz f(y) ::. then cos 1 . we get ::. any constant function satisfies the condition. Xo gives f' (xo) If (xo) 2: 0. But as the function = °and f(x) = { sinx ( ' xolO x=o is continuous on [0.xo.
Hence its derivative with respect to a is 0 for any a i=.L 2k = L k + L n + k .O. 1 o 00 sin ax dx x = 1 0 00 sin x dx. 519807 We have = 1 + (~ .O.. 1 o 00 sin ax .k=l k L k=l k=l k=l k=l L n +k' k=l n 2n 2 n 1 n 1 n 1 1 = Look under Finite Sums in the Index for similar problems. if a :::. x Therefore. The derivative fails to exist at a = 0 because of the discontinuity at that point. Then 00.~) + . + (~ 2 2 3 4 4 2n 1 n 2) 2n = L k .both exist and are finite has a finite limit as b 7 (b) Fix a 2 O. Look under Integration in the Index for similar problems. It now follows that 1000 Si~ x dx is convergent. limb+oo I: g. (1000 SinxQx dx) 2 = (1000 Si~ x dx) 2 is constant for a i=..102 by so that Solutions It Igl.dx X = l'1m b+oo lb 0 X sin ax d x = l'1m b+oo lQb 0  sin Y d y Y = 1 00 0 sin x dx. 0. hence limb+oo I: g+. .~) + ~ + (~ .. x Similarly.
. Continuing in this manner. . a1 must be the greatest integer in a1 = 4. and r the rate of consumption of the grass per week. t the number of weeks until the grass is depleted. Go . S the number of steers. ~ = a1 + a! + . r . Go denote the initial amount of grass per acre. 10 + 10 .24)r. 2 7 7 21 8 16 8 a1 a2 or 5 a1 a2 + . and cancel the r.. 6 = S . 7 =+++ . Consequently.. r . For an alternate solution. 519812 Letting A denote the number of acres. (iho = (. Go· 40 + 40 . we find a2 = 2. we have 310 810 3... .Exam #161981 103 Exam #161981 5198 II If then 7· 7 =ao +7"+72 + .. to find S = 88 steers. Look under Systems of Equations in the Index for similar problems. a4 Apparently.= 8 7 = 2. Subtracting equation (2) from (1) gives 40v = 24r. Look under Number Theory in the Index for similar problems. so that ao must be the greatest integer in 281 . 8 = 18 . v . a3 = 4. 3 a1 a2 that is. and subtracting equation (3) from four times equation (1) gives 200v = 384r .. so = 2. 6.. 8. 16 (2) (3) = 12· r . Solve for v. we have the relationship GoA + Avt = Srt.2 + . v .. The information given yields the following equations: (1) Go· 10 + 10· v . v the growth rate of the grass per acre per week..3Sr. 16. Since 7 . ao . substitute. from which we can obtain the result by long division in base 7.
Pr(n = 5) = Pr( n = 6) = 156' Pr( n = 7) = 156' which sum to one. 2 2n . the comer 0.104 Solutions 519813 Draw a plane passing through the center C of the given sphere.1 2__1 (n . Look under Geometry in the Index for similar problems. We have Pr(n = 4) = ~.1) 1 2n 3 ' since 1/2 is the probability of winning the nth game. Thus CO = CQ + QC' + C'O. we have the trace of the given sphere and the trace of the sphere whose radius we seek. v'3 1 r = v'3=3+1 . and the point P of tangency of the given sphere with one side of the comer. If C' is the center of the sphere of radius r that we seek. (n~ 1) . as illustrated below. we know that CO = v'3. and we have 2 teams to choose from. Look under Probability in the Index for similar problems. (n~1)/2n1 is the probability of winning exactly 3 of the previous n . V3 . 519814 The probability of winning the world series in exactly n games is ~. . On this plane. V3 = 1 + r + rV3. Since the radius of the given sphere is 1.1 games. i. then C'O = rv'3. O~~*~~~.1 = r(1 + V3).
dO dO dy = dt dy dt y2 100 + 100 (32t). . Look under MaxIMin Problems in the Index for similar problems. and integrating gives r = K exp(O/V3) for some constant K. Equating the distance s + 2 that the destroyer has gone at the time t of interception with the distance 2ct that the submarine has gone. separating the variables. and then y.jf = 32t. after some simplification we arrive at 3125 + 400t 2 . Consequently. .16t2. If s denotes arclength along the destroyer's path as illustrated below. Since the submarine is moving at a constant rate c. and thus 3(dr)2 = r2(dO)2. Substituting for 1. Taking square roots. 1. 519816 Suppose the path of the destroyer is given by x y = r cos 0. and dy = r cos 0 dO + sin 0 dr. = rsinO. K = 1. Thus s + 2 = 2r. Therefore y2 + 1002 (32)  100 100(2y) dy 32t (y2 + 1002)2 dt = O. By the quadratic formula. then Since dx = r sin 0 dO + cos 0 dr. the distance r that it travels in a straight line is given by r = ct. we find s + 2 = 2ct. Then y = 50 . and so y = 100/3 feet below the observer's eye level.192t4 = O. measured from the horizontal.jf. y2 = 125/24. where r = reO) is some function of O.Exam #161981 105 51981·5 Let y be the position at time t. Since r = 1 when 0 = 0. so ds = 2dr. 0 = tan 1 160' and by the chain rule.
It is known that the coefficients of a polynomial are symmetric functions in the roots. X3 be the roots of x3 + ax2 + bx + c. Then . X2. 111 1 c ==. Xl X2 X3 C . Exam #171982 519821 Let Xl. b'. Let Then we have Let a'. c' be the corresponding coefficients of the cubic whose roots are reciprocals of the original cubic.106 Solutions Look under Analytic Geometry in the Index for similar problems.
the separated subset {2. = _ (~ + ~ + ~) = _ XlX2 + XlX3 + X2 X 3 = ~. Look under EnUllleration in the Index for similar problems.k + 1 places to possibly place the k dashes (this includes before the first slash and after the last slash). Thus there are (n~+l) different separated subsets. Thus the area of the rectangle is (h + h')AB = hAB + h' AB :::. Now draw the smallest rectangle containing all of the points with top and bottom sides parallel to the line AB. Xl X2 X3 XlX2X3 c Therefore the desired cubic is 3 X b a 1 + x 2 + x +. 5. respectively. wlog horizontal. where hand h' are the altitudes of the triangles. c c c Look under Polynomials in the Index for similar problems. vertical sides through A and B. Given a set ofnk slashes. and conditions of this problem. For example. 2 + 2 = 4. 519822 A separated subset of order k from a set of order n can be represented by a sequence of k dashes and n . We consider the quadrilateral formed by ACBD. There will be points C and D on the top and bottom. 519823 Let the two points that are furthest apart be denoted A and B. Look under Geometry in the Index for similar problems. 1.Exam #171982 107 _1_ XlX2 b' and a' = + _1_ + _1_ = XlX3 X2X3 Xl + X2 + X3 XlX2X3 a C . there are n . 519824 f (x) is differentiable at x = 0 if 1/h lim f(h) . of the rectangle. The area of triangle ACB is ~hAB :::. 1 and the area of triangle ABD is ~h' AB :::.k slashes where no two dashes are adjacent. 7} with n = 8 can be represented as /////.f ( 0 ) = lim _e_ h~O h h~O h 2 . Join them by a line.
It follows that the probability that a random cube ends with the digits 11 is 1/100.lOS exists. 1. 6 = 11 . Look under RealValued Functions in the Index for similar problems. Set y = Ilh 2 . 163 = 256· 16 = 6· 16 = 96 = 21. there is only one number satisfying a == 3 (mod 4) and a == 21 (mod 25) (that number is 71). 13 = 1. 3 is the only residue satisfying a 3 == 11 modulo 4. By an easy calculation. then a 3 == 11 (mod 100). Look under Number Theory or Probability in the Index for similar problems. 6. Look under Differentiation in the Index for similar problems. 6 = 16. 519826 If a 3 ends in the digits 11. 519827 Using the disk method for finding volumes of revolution. b + . we obtain the integral 7f f a ( ( a va2 .y2 + b) 2) dy . a 3 == 11 (mod 25) =} a 3 == 1 (mod 5). This is equivalent to the two conditions a 3 == 11 == 3 (mod 4) and a3 == 11 (mod 25). 519825 If a :::.b2 = 7f2 • 47fa a . 11. By another easy calculation. and 21 modulo 25. a = 1 is the only residue modulo 5 which satisfies this equation. By the Chinese remainder theorem. 3 2 3 Look under Volumes in the Index for similar problems. . 21 3 = (4)3 = 64 = 14 = 11.63 = 36 . then f is uniformly continous on its domain. 16. then _1/h 2 Solutions lim _e_ _ h+O h = lim yl/2eY = y+oo o. So a = 21 is the only solution modulo 25. Now. That means we need only check the cubes of 1.
We see then that Sn = Fn+2 where Fi denotes the ith Fibonacci number. x) + ~r(a . Let Ti be the result of the ith flip.1 with no consecutive H's.x) = ~rJ2(a 2 x) + ~r(a  r= ~~~~~ (ax)(ax) x) (a . . T s . S2 = 3. (We have solved a more general version of the problem: if 10 is replaced by an arbitrary n E N.2 with no consecutive H's or start with T followed by a sequence of length n . 519832 There are 210 possible outcomes when one flips a com 10 times.) Look under Fibonnacci Sequences or Probability in the Index for similar problems. then the probability that we stop after the nth flip is Fn_I/2n. Also. Ts = T and we wish to know how many sequences of length 7 have no adjacent H's. T 2 .x) 2 2 ' ax Since we have the relationship ~(a 2 we see that x)(a . It follows that there are 34 possible outcomes which satisfy our condition. Each such sequence must either start with HT followed by a sequence of length n . Therefore Sn = Snl + Sn2.x) + (a  + J2(a  x) 2+J2· . 519833 From the illustration below. we want to maximize f(x) = (a . SI = 2.Exam # 181983 109 Exam #181983 519831 n~~ lim \YrJ> n~~ lim ~(n/2t/2 = n~oo lim v:. The probability that such an outcome occurs is then 34/1024. We want to compute the number of possible outcomes which have Tg = TlO = H and no adjacent H's in T 1 .. Let Sn be the number of sequences of length n without 2 consecutive occurrences of H. T g • Certainly then.• . Hence S7 = Fg = 34..x)x + 7fr2.ri = 00 Look under Limit Evaluation in the Index for similar problems.
the sequence {ti} is monotone increasing. so ti + s for some . 519834 Solution 1.x)x + (2 + V2)2' where 0:::.ti =1= 0 for all i.1) :::. define to = f(t). Finally. Since t is not a finite sum of elements from S. ti+l = ti + f(t . f(x) = x 2(A . since f(t . Letting A = 7f/(2 + V2)2. Suppose that t E R is not a finite sum of elements from S = {1. f is well defined for x > O. 1. we can write + Aa2. since f only attains positive values. The dimensions of the rectangle and circle easily follow.2A). O. 1/4. Every bounded monotone sequence converges. we must have t . x:::.ti) < t . Furthermore.110 Thus we want to maximize Solutions 7f(ax)2 f(x) = (a . 1 gives us f"(x) = 2(A .·· . 1/3. 1/2.1) + x(a . Define f(x) = max{~ E S : ~ :::. x}.A). Now.1) + a(1.ti). Since zero is a limit point of S. we have ti < t for all i so that {til is bounded above. Further this is a maximum since A :::.ti. a a Look under Geometry or MaxIMin Problems in the Index for similar problems.2A)/2(1 .2aA) Differentiating gives us f'(x) = 2x(A .}. which has a critical point at x = a(1 .
 n2  2n + 1 Look under Number Theory in the Index for similar problems..ti < 1.s). and therefore x :s.. Look under Diophantine Equations in the Index for similar problems. diverges. Since the series ~ + + ~ + ..Exam # 181983 III s E R. But now xn > nx n. 1/8. p. satisfies o < p < 1.ti + ~ > .l)n(n + l)(n + 2) + 1 = n 4 + 2n3 =(n2+n1)2. + ynl] > (1)[x n. + xnl] = nx n 1. If s < t. S19837 N is a perfect square.. Case 1: assume that our positive number. y. Itlr which s . 1 1 1 with each bk either zero or one.. II > m lit contradiction. }. ti t S.1 + . a contradiction.. so there exists II t. Then p has a binary representation of the form P = O. then let ~ = f(t . C Solution 2. = (z .. This gives a representation of p as a sum (possibly infinite) of a subset of the numbers {1/2. Without loss of generality we have x :s. Applying case 1 to this q gives the result. ... Case 2: assume that our positive number M satisfies M 2:: 1. y < z < n. Therefore t = s.. . there is a finite sum S = L:~=l 2k~1 such that M = S + q with 0 < q < 1..~.. 519836: Assume that there is a solution with z < n..hb2 b3 ..1 + x n. = b1 "2 + h4" + b38 + . 1/4.1 implies x > n. i Look under Infinite Series in the Index for similar problems. It follows immediately that ti+l . S19835 This is a repeat of problem # 3 on Exam # 7 (1972).y)[znl + yzn2 + . N = (n . From ~his.
therefore P(x) = 0 which implies 2:~=lPk(X) = 1.005. SO we must have g(x) = x. so that x < gm(x) = X. .010.1 with n zeros must be identically zero. .1].1.005. therefore 9 is injective.1 because each Pk (x) is a polynomial of degree n .1. x < g(x) g(x) < g2(X) 1 gml(x) < gm(x).1]. Any injective realvalued continuous function must be strictly monotone on its domain. Since the Xj are distinct.010.112 Solutions 519838 Pk(Xj) First. note that Pk(Xk) = 1 (this is obvious from the definition) and = 0 for j i= k (because of the x . . . Any polynomial of degree at most n . =x =} which is obviously false. k=l then P( x) is a polynomial of degree at most n . Look under RealValued Functions in the Index for similar problems.Xj term in the numerator). y E [0. then for x E [0. 519839 g(X) = g(y) =} gm(x) = gm(y) =} X = y.1].1]. X> g(x) hence x > gm (x) 1 g(x) > g2(X) =} . Look under Number Theory in the Index for similar problems.1]' then for x. Look under Polynomials in the Index for similar problems. Also. it follows that P(x) has n distinct zeros. Now if 9 is decreasing on [0. Exam #191984 519841 By the binomial theorem 1.001 is equal to (1001)5 75(11 5) 135. =} . . Suppose that 9 is increasing on [0. But notice that for each Xj. x < y =} g(x) > g(y) =} g2(X) < g2(y) so that g2 is increasing on [0. Let n P(X) = LPk(X) . =} gml(x) > gm(x). Using the above argument we see that g2(x) = x must hold.
. are unique integers a and b Now since 65 = (11)7 + (1)12 and 65 = 7a + 12b with a and b nonnegative implies a :s. 69 = 3(7) + 4(12). then so does D / n.4. p . 5 slanting right. Lemma 2. Look under Enumeration in the Index for similar problems.) In this question the emphasis is on the nonnegativity because.2.16. 519843 Door D is changed by the nth person if and only if n divides D. If n is a positive integer such that for each k = 0. which we offer without proof. thus D has an even number of divisors unless D is a perfect square.since 66 = 6(7) + 2(12). then there :s... . are open.) = 126 paths. since p and q are relatively prime. Look under Number Theory in the Index for similar problems. 4 slanting left and any such path joins A to B. enable one to do the stated problem with a little trial and error.1. Thus there are (. Corollary. Lemma 1. the rest are closed. 68 = 8(7) + 1(12). Then m = a' p + b' q if and only if there is an integer k so that a' = a + kq and b' = b. every integer can be expressed in the form ap + bq for some integers a and b. 9. Thus. Let m = ap + bq where a and b are integers. q .1 there are nonnegative ak and bk with n + k = akP + bkq. a ° If m is an integer. b are nonnegative integers} generated by p = 7 and q = 12 and asks "What is the largest integer not in S?" (In the general solution here. Doors that have an even number of divisors are closed.Exam #191984 113 519842 Every path from A to B goes along 9 edges. doors 1. 519844 This problem concerns the numerical semigroup S = {ap + bql a. 67 = 1(7) + 5(12). the uniqueness in the corollary shows that it is not possible to give exact change of 65 quanta.1 so that m = ap + bq. those with an odd number are open. . Moreover.9. all variables will be integers and p and q will denote relatively prime positive integers. with :s. Two fairly easy observations.kp. . .. If n divides D. then every integer greater than n is in the semigroup S.
Theorem.1) (q . Theorem. ~~~~ x Because of the independence and uniformity assumptions. 519845 In order for three segments to form the sides of a triangle.) The following theorem. it is necessary and sufficient that the sum of the lengths of any two segments must be greater than the length of the third.1 cannot be written in the form ap + bq for nonnegative integers a and b. the region corresponding to possible triangles has been shaded. the probability that the first point is chosen from the interval [a.c). 71 = 5(7) + 3(12). then there are (p 1) (q 1) /2 positive integers not in S.1) .a) (d . also by Sylvester in 1884. then (p . In our situation.) Look under Number Theory in the Index for similar problems. Lenuna 1 shows 65 is the largest such price. This is actually a special case of an old (1884) theorem by Sylvester. If p and q are relatively prime positive integers and S is the numerical semigroup generated by p and q. The easiest way to find the required probability is to represent the two points chosen in the interval as an ordered pair from the unit square and decide which pairs lead to triangles. More generally. (Proof omitted. and 72 = 0(7) + 6(12). (Proof omitted. but every larger integer can be. the probability that the points x and y are chosen from [0. bJ and the second from the interval [c. 1].114 Solutions 70 = 10(7) + 0(12). this means that none of the segments formed may have length greater than or equal to 1/2. shows that there are 33 prices for which Subsylvanians cannot give exact change. so that (x. y) is in the subset A . If p and q are relatively prime positive integers. dJ is just (b . In the figure below.
then n Pr(z) = LPr(zlxi) Pr(xi). 519846 (a) From the functional equation and the continuity of 1 at 1. when 0 :::.Exam # 191984 115 of the unit square. i=l In our situation.2. •. in order for a triangle to be formed. Thus the probability that a triangle can be formed is the area of the shaded region in the figure. the functional equation shows that Cn+l = Cn +2. the function g(x) = (2/x) cos(2n/x) + 3 has the property that g(l/n) = 2n + 3 for n = 1. l/n]. but certainly. in order for a triangle to be formed.. 1/2. which has probability 1/2 . x :::.x. Let c be the minimum value of 1 on the interval [l/(n + 1). Xn form a partition of the space of outcomes and z is any event.x = x. . . which has probability (x + 1/2) .1/2) = 1 . 1. Recall that if Xl.(x . the second point must be in the interval (x 1/2. Thus. Since 1(x) is continuous on 0 < x < 00. x=1/2 (This might be interesting to verify with a computer simulation. 1/4.1/2 + x). the probability of having a triangle is 11/2 x=O xdx+ 11 (1x)dx=1/4. letting X be the random variable for the choice of the first point. we have x + 00 lim f(x) = xlim + 1 (x/[x + 1])  2 = 1(1)  2 = 3. . it has a finite minimum value on every bounded closed interval.. We may obtain the same result more analytically with the principles of conditional probability.3. the other point must be in the interval (1/2. Since l/(n + 1) equals (l/n)/[(l/n) + 1]. Now. . hence that Cn+1 = C1 + 2n.) Look under Probability in the Index for similar problems. limx+o+g(x) =J +00. X2. . ex) (b) We note first of all that it is not sufficient to observe that 1(1/ n) = 2n + 3 for n = 1. When 1/2 < x :::...1/2). then this becomes Pr(a triangle) = l~o Pr(a triangle X = x) dx.3.. is equal to the area of A.2.
. Look under Limit Evaluation in the Index for similar problems.BBB.. the function f is an arbitrary continuous function on [1. f(6) 21 + 22 = 43. and so on until we arrive at 241 and f(40) = 3' our answer. HH. 1/(1 Solutions + n). BHB.BBB. 519847 It is helpful to consider 25 tons as one unit of weight. ) 3 and f(2k) = 22k+l 3 +1 Look under Enumeration in the Index for similar problems. Let f (n) denote the number of trains of total weight n units: we want to find f (40). BH.1/2).HBB). Thus f(5) = 11 + 10 = 21.00) such that limx+±oo g(x) = 0. (0). (It is easy to prove by induction. f(4) = 11 (BFF. FF.2) where f(l) = 1 and f(2) = 3 implies: 22k  1 f(2k1)= for k = 1. HF.1) + 2f(n . f satisfies f(n) = f( n1)+2f(n2) where f(l) = 1 and f(2) = 3. [1/2.3.116 Thus.1). . . More importantly. F.2. we notice that an n unit train is either an n ..2 unit train with a flat car or hopper car stuck on the end. but perhaps hard to discover. H). BFB. (c) The given conditions relate the values of f on the intervals [1.FBB. if 0 < x::. [1/3. It is easy to find f(n) when n is small: f(l) = 1 (the only train is B). BBH. FH.FB. etc. f(3) = 5 (BF. then f(x) = 3 + (2/x) + g(cot(7r/x)) satisfies the given conditions and any solution can be written in this way for some g.. f(2) = 3 (BB. we have m~n+l f(x) 2: inf Cn = cn+l = Cl + 2n which means limx+o+ f(x) = +00. we want to count the number of trains of total weight 40 units. (0).HB).1 unit train with a boxcar stuck on the end or an n . That is. . A succinct way of say this is: if 9 is any continuous function on (00. f(7) = 43 + 42 = 85. but besides f(l) = 5 and limx+oo f(x) = 3. to each other and specify the values at the end points. that f(n) = f(n ..
16.vi) = u 2 +v 2 . first put x = i. (iii) 2222 5555 is divisible by 7. we have 3333 4444 +4444 3333 == 14444 + == 0 (_1)3333 (mod 7) == 11 (mod 7). the ..a)(i . Look under Number Theory in the Index for similar problems.)( i . Noting that (u+vi)( u. 2.c + e . multiplication of the corresponding sides of the two equalities above yields Look under Polynomials in the Index for similar problems.Exam #201985 117 Exam #201985 519851 We have X4 + bx 3 + CX 2 + dx + e = (x .(J)( i .a)( i . where i = to get A. 519852 Since 3333 == 1 (mod 7) and 4444 == 1 (mod 7). (ii) 11114444 + 44441111 is divisible by 5. 519853 Clearly U is the multiplicative group of units modulo 21.(J)(x . and then x = i.6).6). 1 .a)(x .d)i and = (i . (iv) 8888833333 + 3333388888 is divisible by 11.)(i ...11}.(b .6).)(x . each subgroup to be found is cyclic. In the above equality. Since 6 is a squarefree integer and U is abelian. 8.d)i = (i . 1 .c + e + (b . Some more problems of this type: (i) 22221111 + 11112222 + 5555 2222 is divisible by 3.(J)(i . 4. Hence the subgroup along the solid path could be {I.
1002. or 2 vertices lie on one side and the third on another side. 13.liS Solutions one along the dashed path could be {I. we have Hence. 519855 From an = (n . 100 .20.. there are 4.1003 triangles of the first kind.1003 + 6. The total number of triangles of the type is 4 . 519854 There are 2 types of triangles that can be formed: either each vertex lies on a different side of the square.l)an2]' Iterating this relation n .4. an n! and anl (n I)! L.1)(an l + an2). 19}.3 times. 17}.10.940.. Thus an = n! ( (It) L. We find the number of triangles of the first type as follows: Select 3 of the 4 sides of the square (this can be done in (i) or 4 ways) and then select 1 of the 100 points of subdivision on each of these 3 sides (this can be done in 1003 ways).16.J r=O r! Since al = 0.000. The number of triangles of the second type is found as follows: Select 1 side of the square (4 ways of doing this) and 2 of the division points on this side ( C~O) ways) and then select 1 of the 3 remaining sides (3 ways) and 1 of the 100 points of subdivision (100 ways) on this side.[anl .. Look under Group Theory in the Index for similar problems. Look under Enumeration in the Index for similar problems. we get an . the sum on the left telescopes to %3. 4.:yn r=O . r! (r . and the one along the dotted path could be {I.J r=2 r! L.nanl = .100 2 = 9. 99 .16. 3 . 5.I)! L.99.(n .J r=2 ~ (a r _ arl ) = ~ (It = ~ (It. Thus. e~O) = 6 . Hence the required number is 4.
l. b2 x 2  > [(b 2 1)/b2 Ja 2 .(Of course if neither 2m + 1 nor 2m + 2 is in S. By the induction hypothesis. . Hence a must be in S.x) > (b 2 . Thus the result is true for S. then the result is true for S. If m + 1 is in S. But Sl does not contain such integers. then 2m + 2 is a multiple of S. assume m + 1 is not in S. then since 2m + 2 is a multiple of band b is a multiple of a.. Then Sl has m + 1 integers between 1 and 2m. the result is already true for S ). Let Sl be the set obtained from S by throwing out 2m + 1 and 2m + 2 and putting in m + 1. we have y = 2ax.e. 2ab2 x + (b 2  l)a 2 < 0. Let Sl be the subset of S not containing 2m + 1 or 2m + 2. and thus the result is true for S. some element b of Sl is a multiple of some other element a of Sl. then by the induction assumption. Because the only positive integers m + 1 can divide other than itself are k(m + 1).1 . Substituting in xy we get b2 x(2a . (bl)a (b+l)a b <x< a . (Observe that an is the number of derangements of a set of n objects . Then Sl contains m + 1 integers between 1 and 2m. i.Exam #201985 and n+oo 119 lim an = lim ~ (It n! n+oo ~ r! r=O = e.l)a 2 . Suppose S does not contain both 2m + 1 and 2m + 2.e. If a and b are both in S.e. Suppose S contains both 2m + 1 and 2m + 2. Now a cannot be m + 1.. Assume the result true for n = m. (bx(bl)a)(bx(b+l)a) <0. If b is m + 1. Let S be a set of m + 2 integers between 1 and 2m + 2. the result is true for Sl and therefore for S.. The result is clearly true for n = 1. S19857 From x+y = 2a. where k is an integer ~ 2. S19856 We will prove this result by induction. i. By the induction hypothesis. Therefore. 2m + 2 is a mUltiple of a.) Look under Limit Evaluation or Sequences in the Index for similar problems. Look under Number Theory in the Index for similar problems.
110 Thus. Look under Complex Numbers or Matrix Algebra in the Index for similar problems. the number of favorable cases is Solutions (b + 1)a b (b . The matrix of ¢. o 0 1 +i To find the eigenvalues we set det(A .1)a b 2a b· The number of all possible cases is when 0 < x < 2a .{3I) = 0 = [(1 + i) . and hence is 2a.{3] [{32 . The required probability is therefore 2a/b 2a 1 b· Look under Probability in the Index for similar problems. 519863 ii = (e 7ri / 2)i = e7r/2. b2 .V2i. 519862 Let B = {b 1 . b3 }. 519864 For every four distinct points of the N points there is one and only one pair of segments that intersect and have the given four points as endpoints. . Exam #211986 519861 See Solution #1 on Exam #14.4i] which implies that {3 = 1 + i. with respect to B is given by A= (~~ ~) . {3 = V2 + V2i or {3 = V2 . Look under Complex Numbers in the Index for similar problems.
+ 6x .2 .0) now becomes X2+0 lim f(x~.0). In particular the limit must exist along the path defined by Xl = x~. . then the limit exists along any path to (0. Then the limit of f(xl.2V2x + 2.0). X2) tends to (0.36x = 18x 4 + 24x2 +8 or Look under Polynomials in the Index for similar problems.3 = V2(3x2 + 2).2V2ij3. 24 Look under Enumeration in the Index for similar problems. 519866 (a) If the limit of f(xl. 2 Thus f(xl.12x + 6V2 + 3 6x = 2V2 + 3V2x2 Thus x 3 x6 + 3. X2) as (Xl.2V2x + 2) + 3 6V2 + 3V2x2 .2)(n .2V2(x . (ij3)2 = ij9 = x 2 . x 3 = 2V2 + 3(2( ij3)) + 3V2( ij3)2 + 3.0). x 3 = 2V2 + 6(x .l)(n . Similarly But V'3 = x V2.3) . 519865 Let x = V2 + (V'3).Exam #211986 121 Thus the number of intersections is given by: ( n) 4 = n(n . X2) is not continuous at (0. Squaring both sides gives + 36x2 + 9 + 12x4  6x. X2) = X2+0 lim 2X~4 x = 2 ! 1= o. X2) exists as (Xl. X2) tends to (0. Then x 2 = 2 + 2V2( ij3) + (ij3)2.V2) = x 2 .V2) = 2V2 + 6x  + 3V2(x2 . so  ij9 = x 2 Now 2 .
32 .11. we should replace 11! by a number divisible by 2.. + n 2 ) + (1 + 2 + . We should . Look under Multivariate Calculus in the Index for similar problems.27731 is a solution.X2) is not differentiable at (0. An appropriate idea is present though... !(Xl. 5 . the smallest solution is the sequence beginning with 114.. . 1 519873 If the ball obeys the law exactly.27723... so the sequence 27722. it must bounce an infinite number of times. the kth having k balls on each edge. Look under Enumeration in the Index for similar problems. 11!+3.. but a calculation shows that these numbers are much too large. + (n 2 +n)) = ~((12 + 2 2 22 + . Exam #221987 519871 The wellknown proofthat there are arbitrarily long sequences of composite integers suggests that the solution should be 11!+2. Thus. .. but it is not clear wheilier it bounces for an infinite time. There are smaller solutions.0). .. .11!+11. + n)) 6 =~(n(n+l)(2n+l) + n(n+l)) 2 = "6n(n + 1)(n + 2). For example.122 Solutions (b) Since !(Xl. .. Look under Number Theory in the Index for similar problems.. each layer has _k(k+l)_I(k2 k) 1 + 2 + ..X2) is not continuous at (0.. 519872 Each layer is a triangle of cannonballs. the least common mUltiple of these integers is 23 . 11 = 27720. 7 .. .3.0). + k 2 "2 + cannonballs. The total number of cannonballs in a stack is therefore 1 2((1 2 + 1) + (22 + 2) + .
. we want Pr(Sf n We may expand this as S2 n . that is. Since it is dropped from a height of 16 feet. it will require vT6/4 = 1 second to fall to the ground. this model predicts the ball will bounce slightly more than 5. (In fact.. We wish to find the probability that none of these events occurs..) Look under Infinite Series in the Index for similar problems. )2 )3 . Continuing. decide if it converges.. Sfo)· Pr(Sf n S2 n··· n Sfo) = 1 . we see that the total time is the sum of the series 1 + 2(VS/4) + 2(V4/4) + 2(v'2/4) +"'. and if it converges compute the sum. the constant 1/2 in this problem is called the coefficient of restitution of the ball. the series converges and its sum is 1 1 + v'2 __c::: 11/v'2 v'2+1 v'21' Thus. It then bounces to a height of 8 feet.:'012:'01O Pr(Sj.)  L 1:'Oj. 519874: Let Sj denote the event that the jth senior draws hi1> or her own name from the hat.8 seconds..Exam #221987 123 find the infinite series that represents the total time.[ L 1:'OiI:'01O Pr(Sj.Pr(S1 U S2 U . requiring VS/4 second to rise and VS / 4 second to fall to the ground again... U S1O) = 1. n Sj2) + PreS)1 n S· n S· ) . or 1+ "4 ~J8/2j=1+v'2~ v'2 00 00 ( (2) 1)j Since this is a geometric series with ratio less than 1. this model is in fair agreement with experiment.
.+ . 6. where C' is the more distant intersection of the perpendicular bisector of AB and the circle. and C are on the sphere and and the length of AD is least.368. have coordinates (?. C' is an improvement. Indeed. the unit interval. Thus.. where the logic begins. I! 2! 3! 1O! Look under Derangements or Probability in the Index for similar problems.1~!] 111 1 = . then the 4tuple A.. Clearly. are the vertices of any equilateral triangle with vertices on the unit circle.e. we choose the north pole (0. i. if not. being symmetrically placed. D' is an improvement. if A.124 Solutions But Pr( Sj. are the points + 1 and l.. . i. and at opposite ends of the segment (the Idimensional case). and C are points of the disk that do not satisfY this condition. the other two points. the unit disk. 1) as the first point and note that the center of gravity of the three points must be the origin. B. .. 1/2).. The three most distant points in the unit ball in dimension 2. B.e.. +  519875 A geometric construction: We give a construction that is an induction on the dimension. B.C20) 18~! + C~) 17~! . the above becomes . but the analytic proof below is more concise. This construction and plausibility argument can be made into a proof. To find the coordinates of such points. . B. 1 CI0) :~! . but I will begin here with dimension 1. the two most distant points in the unit ball of dimension 1. B.! + :! . where D' is the more distant intersection of the sphere and the line perpendicular to the disk determined by A. Since the number of ways to choose k indices from lOis e~). the coordinates must be (±J3/2. the idea comes from the inductive step from dimension 2 to dimension 3.. 1/2). (As before. = and Pr( Sjl n S12) = 18 and so forth.. assuming A. The four most distant points in the unit ball in dimension 3 are four equidistant points on the sphere. then the triple A..+ . Pr(Sf n S~ n··· Sfo) = ~ .6. C. then assuming A and B are on the circle and the length of AC is least. For me. Since they are on the circle.G~) 11~!] [ = 1 [:! .
1/3) and (±y'273.LXiXj::::. we find S ::::. 1/4). the scale factor must be so the three points are (0.7.. . then i#j ::::. y'279. the others have coordinates (7. V5/48. 1/4).0. we see Now. that. 1/4). for all i. (0. and the scale factor must be VI5/16. .0.j But so . since the third coordinate is 1/3.) Again assuming the north pole (0.1) to be one of the points.. Moreover.'" . either all of the terms are 2.the solution to the problem is J879. they are a scaled version of the solution of the 2dimensional problem. 5.7.5. 1/4) and the mutual distance is V512.0.Exam #221987 125 and C and through the center. To be on the sphere. V5/48. The solution to the stated problem is five equidistant points on the unit sphere in dimension 4. V5/24.7. J879 (0. Since the sum for S contains 20 terms. 40 + 10 = 50. if. That is. the minimum . 1/3). V576.1). be the five vectors above. these three points solve the problem for the disk they determine. VI5/16. j.X5 are any vectors in the unit ball of R 4 . 40 . Thus. (0.V5 An analytic proof: Letting Vl. Assuming the north pole is one of the points. i#j Combining this with the earlier inequality.0. The first three coordinates must be a scaled version of the solution to the 3dimensional problem. (±/578. we find that the other three points must have coordinates (7.2 LXiXj.5 or there is at least one of the terms that is less than 2. that is. if Xl. 1/3).
1 d(VJ) I = 00. the distance along the hypotenuse is J2 times the distance along the northsouth leg. the total change in () is then 11° Exam #231988 S19881 _. we look at the same infmitesimal triangle and relate the latitude and longitude angles. her path is along the hypotenuse of a right triangle with northsouth and eastwest legs and. Since this is true at every instant of her journey. V5fi and the above solution is Look under Geometry in the Index for similar problems. she would fly 2l7r(4000)/4 = 20007r miles. that is.VJ. Thus. Thus. Suppose x(O) = 2 .126 Solutions distance between points is no more than optimal. Let x(t) denote distance between x's car and the finish line. 1r/2 SlllVJ which means Jane crosses the Greenwich Meridian infinitely often. she would fly a quarter of the way around the earth before reaching the north pole. so she flies further. we want the total change in () when VJ changes from 7r /2 to 0. longitude corresponds to the angle (J and (north) latitude corresponds to 7r /2 . since the surface of a sphere is locally like the Euclidean plane. In the usual spherical coordinates with the origin at the center of the earth. her path is northwest. at each instant. that is. Instead. the total distance she flies before reaching the north pole is 2000J27r miles. we want The northsouth leg of the triangle has length 4000dVJ whereas the eastwest leg of the triangle has length 4000 sin VJd(J since the parallel corresponding to VJ is a circle of radius 4000 sin VJ. To see how many times she crosses the Greenwich Meridian. Look under Integration in the Index for similar problems. S19876 If Jane were flying straight north. To be precise. and y(t) denote the distance between y's car and the finish line.
Driver x finishes first. 0> dxl dt t=O 1 b implies b > O. .Exam #231988 127 and y(O) = 5. Look under Group Theory in the Index for similar problems. Then o = x (tx) and a = In 2 together imply tx = 1~b2' Let ty be the time required for driver y to reach the finish line. so that ty = >3b = t x. Let tx be the time required for driver x to reach the finish line. Then x * xI * Y = e * y = y = y * e = y * xI by the given cancellation rule.=} ty = 5b 5 since t. In2 5b. d2x _ (dX)2 dx _ ~ () _ In(at+b) + c.ad=} dt . For t > 0. hence.at+ b =} x t t a But x(O) = 2 =} c = 2 + Let it Inb a =} x(t) = 2 + In(b) Hat. Look under Differential Equations in the Index for similar problems. 3 In 2 > . = dxl dt tl We have 2 1 dxl dt o· 1=2+ In( ~ a b ) and atl 1 +b 11 = 2(T)' implying that a = In 2. dt 2 . 519882 Let x and y denote any two members of the group. a > 0 be such that x(tt) = 1 and. Then dyl dt all t dxl dt t=O 1 b and y(O) = 5 together imply y(t) = 5 Finally. * X =} X * Y = Y * x.
. If the circle has radius a. that 5 5n = 5n + 10kn where k n is a positive integer for each n. 519884 The unit's digit of 512345 = 5... To verify this it suffices to show. + (IOn + 9)2 = 5 5n + 10jn + (12 + 3 2 + . Changing to polar coordinates gives the polar equation r = f(O) = a[yll + cos 2 0 .1]2 = 5 5n + (1On + 1)2 + (IOn + 3)2 + ... So the area of 5 is Jo {'i f(O)2 dO = a 2 2 2 [1. + 9 2) = 5n + 10kn + 10jn + 165 = 5(n + 1) + 10kn +1) 1]2 where kn+1 = k n + jn + 16.128 Solutions 519883 Note that So if sand t are as prescribed. then a Cartesian equation for the curve bounding 5 on the right is (x + a)2 + y2 = 2a2.. 519885 The desired area is 8 times that of the shaded region 5 in shown below. Assuming the result true for n gives 5 5(n+1) = 5 5n+5 = 5 5n + [2(5n + 1)  + . using induction on n. + [2(5n + 5) .v'3 2 + ~l 3 .cosO]. Look under Number Theory or Sequences in the Index for similar problems. 55 = 12 + 32 + 52 + 72 + 9 2 = 165 = 5 + 10(16).
Look under Analytic Geometry or Geometry in the Index for similar problems. that line segment must also be an altitude of 6BPE. p B~'~~~~L To prove this.Exam #231988 129 after much work. . Extend straight lines AD and BC downward until they intersect in a point E. where Band Dare opposite ends of the diameter lying within L.v'3 + iJ . the desired area is 2a 2 [1 . Thus PE is perpendicular to line L. Hence. Since these segments are concurrent with the line segment through BD. notice that line segments EA and PC are altitudes of 6BP E. 519886 First draw the two straight lines BAP and CDP. P E is then a straight line perpendicular to L. Look under Geometry or Integration in the Index for similar problems.
••• } must be a constant sequence. Look under Logic in the Index for similar problems. Conversely.4' Look under Probability in the Index for similar problems. we deduce that 2a = d.x = 8. and HlO = 89.130 Solutions 519887 There are 210 = 1024 possible lOflip sequences. B scored 8+ I + 1.l and Hn = H n.. y = 4. Thus C came in second on the Geometry test. X2. 2. Let Hn denote the number of nflips sequences not having two or more tails in a row but ending with heads and let Tn denote the number of nflips sequences not having two or more tails in a row but ending with a tail. . It follows that A scored 4 + 8 + 8.l + Tn. Hg = 55. Tn = H n. . or 3. 2(a + h + e + b) = d + g + j + c + f + i. . From C's total we deduce that 3z ~ 9.2. if we fix m and let n tend to infinity. Look under Sequences in the Index for similar problems. 519892 Any constant sequence satisfies the hypothesis of the problem. 519893: If we label the figure as shown below and compare areas. The only solution satisfYing all the constraints of the problem is z = 1. .. we see that limntoo Xn = X m • By uniqueness of limits.l + H n. Clearly H2 = 2 and T2 = HI = 1. So H3 = H2 + HI = 2 + 1 = 3. so that z = 1.. Exam #241989 519891 It is easy to show that there are three tests and that x + y + z = 13. and C scored 1 + 4 + 4.l = H n. we deduce that Xl = X2 = .. We need HlO + TlO = 89 + 55 = 144 and the desired probability is /o~~ = . 2(a+h) =d+g+j. For n 2:: 3. so that {Xl. H4 = 3 + 2 = 5.
and this is equivalent to the assertion of the problem. = h + j. 2h = 9 + j. 2(d + a) From these nine equations we deduce that 6j = a + b + c + d + e + f + 9 + h + i.Exam #241989 These equations may be reduced to 131 2a= d. = b + e + h. . Look under Geometry in the Index for similar problems. 2g = i + j. = a 2i 2(f + cO and + d + g. 2c= f. In a similar fashion we have 2b = e. 2(e+b)=c+f+i.
then the probability of obtaining the results noted is ~'G)(~)2(~)1  1 4 Hence it is more likely that the biased coin was the one flipped only once.. + XnYn' Look under Integration or Matrix Algebra in the Index for similar problems.b 2 2 • Look under Volumes in the Index for similar problems. if the biased coin is the one flipped only once. if the fair coin is the one flipped only once. . Look under Probability in the Index for similar problems. + XnYn' The determinant. If we slice the intersection by planes parallel to the yzplane. is 1 + XIYl + .\ = 1 (multiplicity n . 519896 The characteristic equation is Hence the eigenvalues are . the crosssectional areas are all squares (by symmetry). 519895 Let the y..132 Solutions 519894 Assuming independence. Hence the volume is [aa 4b (1 _ ::) dx = 16.\ = 1 + XlYl + . then the probability of obtaining the results noted is  2 9 On the other hand..1) and .and zaxes be the axes of the cylinders.. being the product of the eigenvalues.
so 35 or 75 are the only solutions. N.. However AN or NA might occur twice . IN. Look under Logic in the Index for similar problems. A key ingredient is the fact that the number 72 has three "admissible" factorizations: 4 x 18. A.. there are no 1 digit solutions. D.ao + 1 are ao = 5. where ai E {O. 1. It is easy to check that the only solutions to (10 .Exam #251990 133 519897 The dimensions are 4 x 18.. NA. Clearly. IA is repeated. Since ak 2: 1. the lefthand side of (*) is at least 10 k . Therefore k :::.9 + 1 = 73. AN.al)al = a6 ..g. 4 x 12 and 6 x 8. AI.. NI.9. N.. NO. I. + lOal + ao. Note: The students were told at the beginning of the contest to assume that the sides of Smith's ranch are parallel to the side of Todd county.. We assume + . . e.l ak_l and ak 2: 1... then neither IN. + (10  adal = a6  ao + 1. where the universe U is the set of arrangements of I. 48 = 3 x 16.9} which implies (10 k . Look under Number Theory in the Index for similar problems. Exam #251990 19901 Write n = 10 k ak + 10 k. ' A6 are those arrangements in which each possible digraph lA. NO to the question "is the short side strictly less than half the long side?" Several other possible areas admit triple factorizations. 6 x 12. say. al = 3 or 7.ak)ak + . but here the answers are always two YESes and one NO. If. A and the subsets AI' . and 8 x 9 with answers YES. 19902 This is an inclusionexclusion problem. . is repeated. and AI could occur at most once. (*) The righthand side of (*) is at most 9 2 . nor NA can occur at all. A complete solution requires lengthy discussion to eliminate many potential solutions.
Since the exterior angle to a regular heptagon has measure LGAB = 5. In other words.134 if the tuples IA or NIA were repeated. (IA).2 = f(x) .(IAN). 2.2.. This one refers to the figure below. r s sin 371" . If IA and AN are repeated... 7 . 3 = 468. since f(x + 2) C2 = f(x). 3. l' is differentiable and f"(x) = f'(x It follows that f(x) = Cle 2 = Cl and C2e2 = + 1) = f(x + 2) . Since LGAF subtends an arc of J. Note that lUI = 7!/[(2!)3 . and LKAL = so LAKL = LALK = The law of sines applied to r and to sand l gives 2.. 519903 Since f' (x) = f (x + 1) . and. D. Thus the total number of possibilities (by symmetry) is assuming Ai n Aj is possible. D.7 cos J. which can be done in 3!/(2! ·1!) = 3 ways.6 . N.sin 571" ' 7 7 so r= sin ]I sin 371" 7 7. I!] = 630.2. N in some order which can be done in 5!/ [(2!)2 . If IA is repeated. 519904 A variety of approaches work. 30 + 6 . its measure is J.sin 271" 7 7 and s 1 sin]I . Thus the desired answer is 630 . sin 271" sin 571" 7 7 This may be reduced to (among others): sin ]I sin 371" 7 7 sin2 271" 7 7r 1 2cos. 2 + cle x + C2ex.. we find that 6 of the pairs Ai n Aj are possible and none of the 20 tuples Ai n Aj n Ak may occur simultaneously.f on the circle. then we wish to write (IA). Look under Enumeration in the Index for similar problems. I!] = 30 ways. Solutions Continuing in this way..f = LGAK = LLAB.f' _ _7_ cos 271" cos]I . so Cl = C2 = o. then we need to arrange (IAN).
519905 Construct the Riemann sum 2 n tt 1 yn 2 i=1 j=1 + ni + j = 8 ~:. Look under Geometry in the Index for similar problems. which intersect at (±1. Look under Limit Evaluation or Riemann Sums in the Index for similar problems.719). and dropping a perpendicular from K to AG. so the vertices of the hexagon are (0. n n 2 1 1 n 2 J1 1 + i/n + j/n 2 ' which in the limit.:.0). A routine calculation . gives us Jo Jo [1 [1 v'1 dXdY +x +y 1 = [\2J1+x+y)l y =odx Jo = 10 1 (2v'2 + x 3 _ 2v'1 + x) dx = i(3V3 . ±~).4V2 + 1) 3 = i(3 3 / 2 2.::::0 .23/ 2 + 13/2) (. etc.Exam #251990 135 Other proofs use the law of cosines.~x2 and y = 1 + ~x2. and the points show that they are y = 1 . The parabolas must have the form y = ax 2 + {3. ±1). (±4.:. 519906 Insert a coordinate system centered at the center of the hexagon.
1/2]. On [1/2.9V3 6 Look under Analytic Geometry in the Index for similar problems. clearly symmetric with respect to both the xaxis and the yaxis. V3/2). + 95 ) = 96 . so between 1 and 1. i 8V2 . Y = ±V3(x+ 1). 8.1/2]' Y = V3/2. 2. V3/2) and (±1.. On [1.0). so there are 96 1 = 531.000. 7.440 numbers without 1's and 468. 4. 3. Exam #261991 519911 The curve.440 numbers without 1's and 468. S19912 The number of kdigit numbers without any 1's is 8 X 9 k . Y = V3(lx). 9 is 96 .1 = 531. Look under Analytic Geometry in the Index for similar problems. area ARD + area BRG = K/2 = r2 + Wl + rl + r4 + W4 + r5. The enclosed area (that of six equilateral triangles of unit edge length) is 6 x 1/2 x 1 x 1 x V3/2 = 3V3/2. 6.000 inclusive there are 8(1 + 9 + 92 + . The curve is a regular hexagon of edge length 1 with vertices (±1/2.560 numbers with 1 'so Alternate Solution The number of ways of filling six spaces with digits 0. (±1/2. Look under Enumeration in the Index for similar problems.9 2V3 8V6 .000. Let K be the area of the parallelogram ABGD. On [1/2. 5.l .000. = K/2 = r3 . But 000000 is not between 1 and 1.136 Solutions shows that the area between the parabolas is V6 and that of the hexagon is 6· so the total area of the shaded region is "'1.. lies on or between the lines x = ±l. 1]. 519913 Label the areas of the eleven regions inside the parallelogram as illustrated below.560 with l's. then (1) area ABR (2) + W2 + W3 + W5 + b.
b. we get red area less the blue area is K/2 regardless of the choices of points F.2ca n= 4 Indeed.4bc suggests taking a 2 + b2 + c2 . Look under Enumeration or Geometry in the Index for similar problems. The identity a 2 + b2 + c2 . so T3 = (4) (5) Wl + b + W4' From (1) and (2). . Q.2ab .Exam #261991 137 r~~c A~~~~ (3) area QAB = area QAF.2ab . so the sum a + b + c is even since parity is not affected by a sign change in a summand.2ca = a 2 + b2 + c2 . and R.2ab + 2bc .2bc .2bc .2ca .b (from (3)) in the right side. 519914 The sum a + b + c is even. T3 +W2+W3 +W5 +b = T2 +Wl +Tl +T4 +W4 +T5. Substituting T3 = Wl + b + W4 from (3) in the left side of (4) and Wl + W4 = T3 . Using K as the sum of the reds + sum of the "whites" + blue. ( a+b+c)2 =n+ bc 2 ' ' and ( ab+c)2 =n+ ba 2 ( a+bc)2 = n+ abo 2 Look under Number Theory in the Index for similar problems. z= Wi + 2b = z= Ti .
138 Solutions 519915 Remark: Ifn = a 2+b 2 and m = c2+rf2. From the example 32 +4 2 = 52 = 25.) t n 1 + 1 + . and so on. it is not possible to have four consecutive positive integers which are sums of two squares. namely 82 + 15 2 = 172 = 289. n2 (c) In any set of four consecutive positive integers one must be congruent to 3 modulo 4. (a) Take a = c and b = d. = _1_. n 2 .) 1 + l/n 1 + 2/n 1 + n/n n k=l 1 1 + k/n = Jo [1 1 1 +x dx = In(l + x)l~ = In2. n In2 Look under Limit Evaluation in the Index for similar problems. 12 +5 2 = 26. one can generate infinitely many more by taking (3k)2 + (4k)2 = (5k)2 = 25k 2 and 12 + (5k)2 = 1 + 25k 2.!. take 72 = 62 + 62. then nm = (acbd)2+(ad+bc)2. the product of two sums of squares is also a sum of squares. 519916 We have n+oo Hn 1) (. It follows that if n is a sum of two squares.. n 2 + 1 are consecutive sums of two squares..74 = 52 + 72. Any Pythagorean triple will serve as a starting point.+ . (b) If n . Therefore.+ . lim Hn n+oo . This integer cannot be a sum of squares since a 2 + b2 == 0. 1 or 2 mod 4 for any i~tegers a and b. Thus. + .. 12 + 172 = 290. 32 + 32 = 18 leads to infinitely many more.. n. then (a 2 .1.) ( (. + :_1. n + 1 are consecutive sums of two squares.73 = 32 + 82. For starters.b2)2 + (2ab)2 = (a 2 + b2)2.!. Therefore.) 1 1 1 .. then 1 = (n l)(n + 1). The first example 12 + 4 2 = 17.. 1l mn = 1· lmn ( n+oo n n+ 1 n+ 2 n+ n = lim n+oo  lim n+oo (. then n 2 is a sum of two squares.. Look under Number Theory in the Index for similar problems.
I)}. the answer is t. This happens with probability Therefore. i=l n 1 n Look under Probability in the Index for similar problems. To avoid three points in a row. (0.cOS(t))2 + (sin(t))2 dt = = 127r J2(1. (1. (0.: :. 1). two rows would contain just two points each. column. (1. or diagonal (broken diagonals included) of a 3 x 3 square array completely occupied.0).2.1). 1). 127r J(l. Look under Analytic Geometry or Enumeration in the Index for similar problems. .3.1).cos t for t ::. If only four or fewer squares were occupied by the nine points... then one or more would be occupied by at least three points which would have a lattice point as centroid. each can be identified with one of the nine points in the square array S: {( 1. (1. (1. 1). Look under Integration in the Index for similar problems.Exam #271992 139 519917 By reducing the coordinates of each of the nine given points modulo 3. It is now easy to show that it is impossible to place the fifth without completing a row. (possibly broken).: . Exam #271992 519921 The curve is a cycloid which can be parameterized as x = t . The existence of three lattice points with a lattice point as centroid is equivalent to having a row. In n + 0. (0.sin t and y = 1 . 519922 Observe that the ith tallest person is viewable if and only ifhe is in front of all the people 1. Therefore.. (1.cos(t)) dt 127r V4sin2(t/2)dt = 4cos(t/2)1~7r = 8. L . or diagonal.0). 27r. column.5772157.0). iI. at least five of the squares in the 3 x 3 square array must be occupied. .. . The arc length is given by °: ..
For n = 3. {e. contrary to the stated condition. and not 2.1 letters to everyone else. For n = 2. whereas Z4 only has 2. Let each letter be represented by an arc in a directed graph.) of these. For n :s. every subgroup contains the identity element. For n = 4. This gives a bound of 2n . e. First. we can find at most L t=l (7 ~ :) = 2 2k  1 = 2n . There are (7~.1 selected from G . For any even n = k 2k. We shall now show that no group of order n ::::: 5 has so many subgroups.1 more letters are required.140 Solutions 519923 Select a head gossip Va. for we need a sub graph of arcs unilaterally connected with directed paths from every vertex into whichever vertex becomes fully informed. That is just barely enough. this yields at most 8 (7 ~ n< k 22k 2n 2 1 = subgroups. We claim that this total of 2n .2 which we have already achieved above. a. proper subgroups. First have all n 1 other people send a letter with their scandal to Va. Z2 works because it has 1 subgroup. Zl has no proper subgroup. For any odd n = 2k + 1. so at least n . That is.2 subgoups. But the first forces . 4. Look under Enumeration in the Index for similar problems.{ e }. 519924 For n = 1. this is true of Z2 and K 4 . then have Va distribute the assembled information via n . and the largest subgroup order i is at most l ~ J.1 letters have been sent. namely Z2 and K 4 . a} and {e. The Klein 4group works. but only if every set of order l ~ J is a subgroup. But for n ::::: 6 this requires every subset of order 2 and 3 be a subgroup. Z3 fails because it has 1. Thus we have found 2 groups that satisfy this condition.2 letters is the minimum. we need 4 subgroups. so the number of subgroups of order i cannot exceed the number of subsets of order i . Then each subsequent letter can inform one more person. b} must both be subgroups. No one can know all the scandal until at least n .
' A (b) = (1 1) (4 0) (2 1) 3 2 0 9 3 Clearly all four choices of signs in C = (~2 Therefore. a contradiction. 519926 (a) The characteristic polynomial of A is X2 13X +36 = (X 4)(X . . Look under Integration in the Index for similar problems.Exam #271992 141 the order of a to be 2 while the second requires the order of a to be 3. Therefore I = Vi.9). We find that the eigenvectors for A are C~/) and C2). 3 This gives p=(1 2 1) 3 and therefore and pl = (2 1) 1 1 . = _r2 He 10 (Xl = H. all four choices of B = PCpl give :3) yield C 2 = (~ ~). Look under Group Theory in the Index for similar problems.
2x = L2 = !.2 x 2 x+o+ Since e. 519927 Solutiou 1 x/VI . there is no maximum. namely.2x2 A =1= o. L = ±(I/V2). = ± 2ni ( 0 2i) .e 2x2 Since y to get x 2 is a continuous function.2x2 ~ 1. 519928 We want integer solutions to a2 + (a + 1)2 =  b2 . But this also equals L/2 provided L find L = so that. We Look under Limit Evaluation in the Index for similar problems. as above. n we may have B Thus. These are Solutions B = 5 1) ( 6 0 ' ( 5 6 1) (13 0 ' 30 12' 5) ( 13 30 15 ) 12 . we can find at least 2(n + 1) different B's. for any i with 0:::. this reduces to . we may square Now this has the form ~ so we may apply L'H6pital's Rule to get lim x+o+ 2x 2 4xe. This can be rewritten as 1 = 2b2 (2a+ I? .142 as desired. (c) If A has a repeated eigenvalue. x+o+ lim VI  e. 2 Therefore L = ± (1/ V2). Solutiou 2 Since the form is lim g. for example if A = 2n I. i :::. 0 Look under Matrix Algebra in the Index for similar problems. Therefore L = +(1/ V2). But L 2: 0 since the fraction is positive on the domain.2x2 2xe. 2x provided this limit exists. we apply L'H6pital's Rule to get VI e.
we can show that Cn satisfies the same recurrence.CnI' The roots of x 2 .c2 • The first solution is bl = Cl = 1. c) is the smallest solution not on our indexed list. we have a Pell equation 1 = 2b 2 . cn). Namely Cn+! = 6cn . Suppose that (b. C2) = (5. c). In general. Upon applying the reverse recurrence we find that (3b . .1 + 3Cn l) = 3bn + 8bn .vs)n. we can generate another solution because Thus. Consequently. the list includes every possible solution..( J22 1) (3 .7). Can we find an explicit solution of the simultaneous recurrence? Applying the recurrences repeatedly we find bn+1 = 3bn + 2cn = 3bn + 2( 4bn . C4) = (169.Exam #271992 143 Upon replacing c = 2a + 1. C3) = (29. + . cn) = (3b n+ 1 . (b 4 . namely r = 3 ± VS.2c. We find that bn = (24J2) (3+vs)n+ (2+4J2) (3vs)n.1 + 18cn· Now nine times the first equation minus the last yields Similarly. 4b + 3c) is another smaller solution that also cannot be in our list. 4bn+ 1 + 3cn+d.239).2Cn +l. (b 3 .1 + 6( 4bn + 3cn ) = 21bn + 8bn.41). C n = ( J22 ) (3 + vs)n 1 . But this contradicts our selection of (b.. we find the solutions (b 2 .. given a solution (b n .6x + 1 are associated with this recurrence. Can there be any other solutions? We can also write the recurrence in reverse as (bn .
12ae (2b)2 . . 519932 For simplicity of notation.144 Solutions and recall an = en. > O.l = ( \1"24. Look under Polynomials in the Index for similar problems. + 2bx + e)(6a) < (6ax + 2b)2. P" (x) = 6ax + 2b. Look under Diophantine Equations in the Index for similar problems. and pili (x) = 6a. Exam#2S1993 519931 Let P(x) = ax 3 + bx 2 + ex + d. The following inequalities are equivalent: pili (X ) P'(x) 2P'(x)P"'(x) 2(3ax2 < ! . the last inequality holds and the result follows. (P" (x ) ) 2 2 P'(x) .4(3a)e < 4b2 . Since P' (x) has distinct zeros.1 ) (3 + V8)n _ ( \1"24+ 1) (3 _ V8)n. < (p"(x)) 2. Then P'(x) = 3ax 2 + 2bx + e. f(fg!)f = fgf Look under Group Theory in the Index for similar problems. f = gff = (fg!)ff = f(gff)f = fff = = ff(gff) =g. we write f and 9 for f(a) and g(a) and omit parentheses. 36a 2x 2 + 24abx + 12ae < 36a 2x 2 + 24abx + 4b2 .
) Thus. 0(2) = 22 . A2 = (XT . 2 must follow it. . for any orderly permutation of {I.1 .. .. Look under Permutations in the Index for similar problems. (k + I)}. P.. r = XT .+l = . Therefore we have 1 = trace of A = trace of (r . k} must begin with 1 or k. .. Since both 12 and 21 are orderly. it has at least one nonzero row.. we see that 0 (k + 1) = 2k. Look under Matrix Algebra in the Index for similar problems. 'n<a <2 so ~. 2. Since.) This process thus creates 2k distinct orderly permutations of {I. completing our induction. (k + I)}. (XT .. one can show by induction that n an n 1 2 . we can reverse the process and obtain the orderly permutation of {I. C equals the trace of C . . Wherever the 1 is. of {I. . .k} which created it. r. 519934 Let O(n) be the number of orderly permutations of {I. . and so on and inductively... and each of its rows is a multiple of r. j must follow the 1..1 and note that any orderly permutation of {I. XT) is a 1 x 1 matrix. or 2) increase each number in P by 1 and place 1 at the beginning. 2. (k + I)} as follows: 1) place (k + 1) at the begining of P. with k = 4 and P = 1423. (Let j be the first number. 2. . . . . n} we create two orderly permutations of {I.Exam #281993 145 519933 Since A has rank 1... (The last equality holds since (r ..YH_n_ 2 . B. Now assume that O(k) = 2k . (For example. 2. (r· XT) . 2. n + an. 2. XT) = r . n}. r) = XT .. r) . Now.. for any p x q and q x p matrices Band C. we create 51423 and 12534. . .. r = A. r.1 = 2. Therefore there is a 1 x n row vector X for which A = XT . the trace of B . We will prove by induction that O(n) = 2n . 3 must follow the 2. . 2. 519935 Solution 1 Since a. XT. a.) For any orderly permutation.
since (A n /2 n ) is an integer. 1 2 Look under Limit Evaluation in the Index for similar problems. .a. using an 700. the an's are increasing and the calculations above show that so an 7 00. it follows that Look under Number Theory in the Index for similar problems. + an .146 Solutions Solution 2 Since a. in turn. + an + an 1 Next since al = 1. Next. since 0 < (J3 . Note that is divisible by 2n+l. va.an a.+l = a. 1)2n < 1. Finally.  one has an+1 an = va. + an. 519936 First note that there are integers An and En for which and further that Thus. shows  an n 7. + an . the same calculations show that This. An is divisible by 2n.
and 2C In case (6) we have A +B = 401. 4) A < B = C = D < E. among A. =C= +D = 791. A 2B = A + D = 546. W < X < Y < Z among A. E then there would be at most three distinct sums of pairs of them: X + X < X + Y < Y + Y. B. and B +E = 836. B = 273. C. D. 1) A = B = C < E. D = E = 418. D. which yields = 128. B. Therefore there are exactly three distinct integers among A. A = = C = 273. B. from which we obtain A = 128. E and we have the following six cases: < D < E. E then there would be at least five distinct sums of pairs of them: W + X < W + Y < X + Y < X + Z < Y + Z. there are three collections of integers satisfYing the given conditions: A = 128. 2B = 546. is odd. E = 563. (Why?) This yields A = 128. D = E B = 273. and 2D = 836. D. B +C = = 836. Also. 6) A < B < C = D = E. B 128. B = C = D = 273. E = 563. 2) A = B = <C D The first three cases are impossible since the smallest sum 401. in case (5) we have = 273. 691. Look under Logic or Systems of Equations in the Index for similar problems. 418. B 273. C. 3) A = B < C < D = E.Exam #281993 147 519937 If there are only two distinct integers. A + C = 546. B = C = D Similarly. A and = 128. we must have A + B = 401. if there are at least four distinct integers. B A + B = 401. In case (4). . 5) A < B = C < D = E. X < Y. A + E = 691. C = D = E = = 418. C = D = E = 418. C. In summary.
+ (2)n/2 n n =n+~+ Now take nth roots. and z. and 6 are replaced by w. Exam #291994 519941 Using L'H6pital's Rule. ... we see that the area of abed is i~.+ . Finally. ( . since Ba = £ = 6 and Aa = £ = 3' 2 ba 1 ad 2 LAaB contains 6~3 = 9 units. +:. 519942 We have (1 2 + J2/n)n = 1 + n~ + n(n 21) .: 2)n/2 >n. x. then the area of abed is wy(w + x)(y + z) xz + (w + x)(y + z)· Look under Geometry in the Index for similar problems. We will determine the area of ABCD in units.148 Solutions 519938 Let the area of abed be one 'unit'. we see that LAdD contains 5x4 = 10 units . Noting that 6AdD = 6CbB and 6AaB = 6CeD. y. and ~~ = ~ = 4. since area of ABC D = 48. · 11m Jo fX et2dt 2 x+oo eX / x = x+oo x·2xe x eX l' 1m ~2~2 e x2 x2 = x+oo 2X2  lim x2 1 1 2"' Look under Limit Evaluation in the Index for similar problems.. .. Since ~~ = ~ = 5. we see that ABCD contains 2 x 10 + 2 x 9 + 1 = 39 units. Look under Inequalities in the Index for similar problems. Similarly. Note: In general it can be shown that if 1.. 5. 2.
y2 y x Let f(p) = Inp2 . 1.p + pl. 2A = C. implies 2e = 5 + e. B. Look under Analytic Geometry in the Index for similar problems.1 = 1/8 and it follows that. c. 1. There are five cases: = =1 abed = 2 abed = 3 abed abed = 4 abed implies e = 4 + e. 1. a contradiction. (A/2.2. and =5 implies 5e = 5. 519945 Let the integers in nondecreasing order be a. .2. Look under Geometry in the Index for similar problems. 3}. 2}. Look under Number Theory in the Index for similar problems. 1.Exam #291994 149 519943 The inequality is x2 X Y In<. substituting their coordinates in x 2 + y2 + Ax + By + C = 0 determines rational values for A. are both rational. impossible. 519944 If C is the largest and A the smallest angle of the triangle. and C. but since d gives e = 2 and the second is impossible. implies 3e = 6 + e. Then. it follows that f(p) < 0 when p > 1. since A and C are between 0 and 7f.B /2). 519946 Suppose that the circle contains three distinct rational points. 1. Then l' (p) = _p2(p .2. so e = 5. Since f(l) = 0. e = 2 is impossible.1)2 < 0. Then abede a + b + e + d + e ::. the first = 8 + e. {1. 3. The three solutions are thus {1. . 5e so abed::. for p > 1. and {1. But then the coordinates of the center. implies 4e = 6 + e or 4e = 7 + e. 5}. Look under Inequalities in the Index for similar problems. b. d. 5. so e = 3. then the law of cosines gives cos C = 1/8 and cos A = 3/4. This implies the inequality. and e. Then cos 2A = 2 cos 2 A .
this yields lui + Ivl = IZ1 + z21 + IZ1 .v1 2 ) = 21z112 + 21zr from above. Let 8 be the given sum. and v = Z1 .(zr  z~)1 = IZ212. Hence z~l.1. Then w is a root of x 3 . Since both bases are nonnegative. we see that (lui + IV\)2 = (IZ1 + z21 + IZ1 . with w i=. 8 _ 1  ~ ( 3k+ 1) ' 3n 6 k=O k=O 8 = I: 2 3 (3k : 2)' Use w3 = 1 and the Binomial Theorem to get + 1)3n = 8 + 8 1 + 8 2 .1)(x 2 + X + 1) = 0. luvl z~l.z~. 21z112 + 21z212 = IZ1 + z21 2+ IZ1 . and lul 2+ Ivl2 = 2(1u + vl 2 + lu .1 = (x . 2zl./zr .z~1 = IZ1 + z21·lz1 . S19952 Let w be a complex cube root of unity.z~. 1 = Izr .z21 which establishes the required result.1.z21)2.150 Solutions Exam #301995 S19951 Let u = Z1 + /zr . But. By the Parallelogram Identity. Look under Complex Numbers in the Index for similar problems. Since w i=. (1 + w)3n = 8 + w8 1 + w2 8 2 . by the Parallelogram Identity. Noting that Izr . (lui + IV\)2 = lul 2+ 21uvl + Ivl 2= 21z112 + 21z212 + 21zr  Again. (1 + w2)3n = 8 + w28 1 + w82.z21. Then u+v = u .z212. we must have w2 + w + 1 = 0.z~. (1 .v = 2/zr .
2) = !' (1) = !' (... Thus N = 4 . + p . 1· 5(p3)/2 2 . This shows that 20p divides N.2P+1  (y'5 .1 . Look under Number Theory in the Index for similar problems.. Look under Finite Sums in the Index for similar problems.. + p . 5(pl)/2 + P . p is an integer. and add the three results above to get Hence 23n + 2. Hence 1 < . Then. But ° ° N + f .''. 5(p .2)p. Hence = 0.. since 1 < f .2)P = 2(p .f' < 1. which must be 22P. so that (1+w)3n = (_w 2)3n = (_1) 3n w6n = (_1)n. 519953 ° Let (y'5 + 2)P . 5(p3)/2 + . 2 . 3 . 2P. where N is an integer and < f < 1. . 519954 Let Then !. and(1 + w 2)3n = (_w)3n = (_1) 3n w3n = (_1)n. 5(p5)/2 + .2 . since < y'5 . (1)n S = :3.2 < 1. + pp  But p is an odd prime implies p divides pr for every r such that 1 :::. we have < f' < 1. We can easily verify that °= f (1) = f ( .Exam #301995 151 Use 1+w+w 2 = 0.f' < 0.1.. since w3 = l. 22 . 1 .(x) = 12(x 2 + X + 1)2(2x + 1)  27(2x(x + 1) + 2x 2(x + 1)). 3· 23 . r :::. since p is odd.~) = f' (.2P+ 1 = N + f... 5 + 2P) . and 1 < f .f' is an integer. Let f' = (y'5 .!' = (y'5 + 2)P . p . We need to show 20p divides N.2).~) = f (.2P+ 1 f ..f' < 1.
. 2n+ 1)n(n+l)/2 N< ( .. and their arithmetic mean equals 1 + (2 + 2) + (3 + 3 + 3) + . (aV3) =b+~=c 2 2V3 1 aV3 2 ="2 a ' 2. The area of the triangle aV3 . As the leading coefficient of f(x) is 4. their geometric mean is N2/n(n+l). Then there are n(n + 1) 2 1+2+3+"'+n= factors in N.1)2(2x + 1)2(x + 2)2. N 2/ n (n+l) < 2n + 1/3.. 519955 Let N = 11 .~.3 Look under Inequalities or Integration in the Index for similar problems. Then G is b+! 3 units above L. nn. + n) 2 n(n+l) 2(1 + 22 + 32 + ... 519956 Let G be the centroid of the triangle. Hence. This establishes the result. and 2 are double zeroes of f (x) and that f (x) has no other zeros..33 . we now see that f(x) = (x .2 2 .square umts.. . Look under Polynomials in the Index for similar problems.= 4.152 Solutions Since f (x) is a sextic. . we conclude that 1. + (n + n + . that is.. + n 2 ) n(n + 1) 2n(n + 1)(2n + 1) 2n+ 1 3 6n(n + 1) But the geometric mean of k unequal positive nUlllbers is always less than their arithmetic mean.
S19957 Solution 1 Let G be the centroid of the triangle. If L is the middle point of BC. then OL. t. Observe that the origin. 3~ "3~ t i=l i=l Clearly. 0. a sin Oi).sm 2 . Since. it proves that 0. C be respectively (a cos 0i. we see that the coordinates of Hare (t.lBC also. G. and H be the orthocenter. 0. Let H be the orthocenter of the triangle. Look under Volumes in the Index for similar problems. (02 3 +0 )) 2 cos 2 ' sm . the required volume V= (27rc) a 2 J3 4 = 7ra 2 J3 (b+~) 2 2J3 7ra 2 J3 7ra 3 =+ 'f 4 or = 7r: (2J3b + a) cubic units. G. H are collinear and OG : OH = 1 : 3. Hare collinear and that OG : OH = 1 : 3. we see that the circumcenter of the triangle is the origin. is the circumcenter of triangle ABC.e. 3."" acoso' "" asino' .Exam #301995 153 By Pappus's Theorem.) Let A. 2.lBC. Then G has coordinates 1 3 1 3 ) ( .. 0. i. In fact. 2 cos 2 2 . i = 1. a cos Oi. Hence AHIIOL. a sin Oi) . But AH. +0 ) 0 )) (02+03) . B. Solution 2 (Solution 2 does not use the property of the Euler line quoted and used in Solution 1. by looking at the coordinates of the vertices. by geometry. AH is parallel to the vector \ II I cos \ II I cos (02+03) (02 3 (02 3 (02 3 0 ).
These two lines meet at H(x. the parametric equations of AH are { Solutions = aCOSDl + tcOS(<>2t<») . t E R. t Thus. = Since y is obtained from x by replacing cos with sin. x = a cos ctl + 2a cos ( D3 . Similarly. where = aCOSDl + aCOSD2 . D2 ) = a cos ctl + a cos D2 + a cos D3. the parametric equations of BH are X { = aCOSD2 + UCOS(<>3t<») Y = a sin D2 +usin(<>3t<>3) . D2 ) cos ( D3 . we see that y a sin Dl + a sin D2 + a sin Thus the coordinates of H are D3. = 2asm (ct 1 2 D2) sm (Dl + D2) . Look under Geometry in the Index for similar problems. = 2acos D3 . D2 ) cos ( ct3 . 2 and tsm .154 Hence. (D2 + D3) 2 = =  3 usm (ct + Dl ) 2 . a sin Dl + asinD2 2a sm (ct 1 2 . y). . t Y = a sin Dl + t sin( <>2t<>3 ) X E R.D2 2 . D2) cos (Dl + D2) ' 2 Solving for t we find.
(2n + 1)(n· 2n 2n .k + 1)(2n .3{ 4(4n 2 + 6n + 2) . obtained by putting x = 1 in the binomial expansion of (1 + x)n and in the results obtained by differentiating with respect to x once and twice. + "2 n (n  1 = 2n. + = = (~)(n+1)(n+2)} = t (~}2n ~ ta (~) ~ ta (~)(2n+ ~ ta (~)k(k ~ k=O .k + 2) {(2n + 1) .2n: 1 ta (~)(2k 1) + 1) = ~(4n2 +4n+ 1 +2n+ 1) ta (~)  (2n+ 1) ta (~)k 1)2 n.3(9n 2 + 19n + 8)..2 + 1 ~ ta (~)k(k 1) 1) = "2 (4n 2 + 6n + 2)2n .1)} 1)2 . .x)3 (~(n) x k) {1~ + 1)(k + 2)x k} . ~ k "2 ~(k The coefficient of x2n is ~{(~)(2n+1)(2n+2)+ (7) (2n)(2n + 1) + .Exam #301995 155 519958 We have (1+x)n = (1.n} = Note: We have used the following results..4n(2n + 1) + n 2 .k}{ (2n + 1)  (k .
(area(EBCD)) 2 = 11 V2 16 . Exam #311996 519961 The problem is equivalent to finding digits (integers from 0 to 9) a. so that = L..9. But for this problem.c)/10 is a mUltiple of 5. .::: 32. 519962 Let ABCDE denote the pentagon P with LA = 90°. 224 Squaring and adding both equations. (b ± c)/10) must be 0 or 1 and a = 0.6 = 37 numbers of this form and thus 36 others besides 43 2 = 1849.c) is divisible by 100. a + (b . Thus one of a + (b + c)/10).5. Note that area(EBCD) = area(BCD) + area(DEB) = 1/2sinC + V2/2sinE. k=O ~ k 1 (1 _ x )2 oo ~ kl . 1 1 (1.x )3 = 2 L: k(k1)x k=O . by the law of cosines. kx k=O k2 1 1. Then the area oftriangle EAB is 1/2 and quadrilateral T = EBCD has sides y'2. lOa + b . 1. we have also assumed that Solutions Ixl < 1. one of the factors must be divisible by 25. =2 2cosC =3 2V2cosE so 1 y'2 1 .5. 1. and in fact by 50 since the two factors differ by the even number 2c.cos C . Since band c are digits. then 5 divides their sum 20a + 2b and therefore divides b.x and = L.5.9 or b = 0. If 5 divides both factors.::: v1000 and n 2 .c2 is divisible by 100. x. l. BD2 Also.2 cos(C + E). we see that (lOa + b + c)(10a + b . b. We may assume that P is convex. Look under Number Theory in the Index for similar problems.156 Of course.4. c such that n = lOa + b .. There are 30 + 13 . Otherwise.. 1.. Factoring. so either a = 4.cos E = .
the remaining 25 cats profit. f(2 3 . Hence the nth cat will get f (n) = (a .Exam #311996 157 This is clearly maximized when C + E = n. 2. If n = 2a rr~=l p~i. 3b4 + 3b2 =  3. b6  4b4 + 4b 2 + b4 b6 b6   2b 2 + b2 = 3. 2 Look under Geometry in the Index for similar problems. n is odd. 3. 6. b ::::. b ::::. the cat breaks even. (c) The total amount the rats left with is 100 2) _1)kll1~0 J = k=l 68. If a = 1. (b) To maximize f(n). b ::::. 5. Equivalently. and there are 25 of these. 3b4 + 3b2 (b 2 1 = 2.32 ) = 6. The cat loses money if a = 0. . 1)3 = 2.1) rr~=2 (ai + 1). 519964 Let the two x coordinates of the square's vertices be a and b with a> b. b ::::. a and if2 a b ::::. note that if a < 2. then the number of even divisors is a rr~=2 (ai + 1) and the number of odd divisors is rr~=2 (ai + 1). Subtracting. and f(22 ·3·5) = 4 gives the 96th cat to gain the most with $8. (a) Number the cats from 1 to 100. b2 = 1 +v2. he'll lose a dollar. f(n) ::::. Finally. 1. he will get a dollar and for each odd divisor. (ab)(a 2 +ab+b2) = 3(ab) so a 2 + ab + b2 = 3. For each even divisor of n.4. Substituting a = b3 . for a = 6.2b. Then a 3 3a = b3 3b = ab so a = b3 2b and b = 4aa 3 . Look under Number Theory in the Index for similar problems.. a 3 = 4ab and b3 = a+2b. so 50 cats lose money. i. f(2 5 ·3) = 8. 100. and consider the nth cat. 3.e. So calculating f(2 6 ) = 5. (which is true if and only if EBCD is a trapezoid) and the area of EBCD is Jll 519963 16 + yf2 . then b ::::. 25. 12.
nPn = (1/2)4n(2n)2 = (2n)!(2n)! n n n (2nn!)4 = n.='''.l) 1 ·3·5· . P.1 2n . 519966 Such an event is mathematically possible if out of the 2n children. Thus r encloses a region of area ~ (VB)2 = 4. 2n 1 .. with 7fr2 = 4.2 2n 2n 22446 ..:::. 2 . so r= and the length of 'Y is vf41ir 'Y 1 1 = (27fr) = (27f) ~ = Vi· 4 4 7f Look under Geometry in the Index for similar problems. 2n 2 .3~ = 6 + {I108 (1 = {l216 + {I108 so A + B = 108. 3 .l) . 4 ....ISS The area of the square is Solutions (b 3 _ 3b)2 = b2(b 2 _ = 3)2 + ?'2)(?'2 . 3 .. 3 ... (2n . 2 .. can be written as n are males. say of radius r. 2 .. 3 . 6 ... 6 .1 n 2n . 4 .. Therefore r is a circle. By the isoperimetric theorem.. Thus. 4 . (2n . 6 .4?'2) = 6 . (2n . exactly = (~) 2n The expression n . (2n . Pn 2 e:).. 2n 13355 2n . 2n . 5 . Look under Analytic Geometry in the Index for similar problems...2)2 = (1 + ?'2)( ~ + 4 . 2n (2·4·6···2n)2 (2·4·6··· 2n)2 = n . if a closed curve r encloses a fixed area...::'::' 1·3·5· .l) =n=''~'~ 2 . gives a closed curve r bisecting the area of a square of side length VB.:::: 1 . the length of r is minimized when r is a circle. 4 . 2n 1 . 519965 Reflecting ABC about the yaxis and then the result about the xaxis. 5· .l) 2 .
LYi = 1 i=l 4 and LY. 9.=l Xi = 25. 4.. . then we must have three odd squares which add to 27.. we have Y~ + 3 == 43 mod 8 whence Y4 must be a multiple of 4. Y2. X4} is statistically equivalent to {I. and the answer is 1/ yI7f. 6} then 4 LXi = 4(1 + 9 + 9 + 6) i=l 1 4 1 so that ~. Y3) must be odd and one (say Y4) must be even. Since each iYii < 7. = 43.. X3... An alternate solution can be found using Stirling's Formula: Since n! ~ (~)ny'2nn. 3.. S 19967 If {Xl. i=l 4 Since squares are congruent to either 0 or 1 mod 4. . Pn ~ fir· Look under Limit Evaluation or Probability in the Index for similar problems. I} or {4. If (J" is the standard deviation. This is possible in two different ways: 27 = 25 + 1 + 1 = 9 + 9 + 9. Checking to see if any of the signed sums of the elements from {5. i=l 4 Letting Yi = Xi  6. then and we must have LX: = 12 + 92 + 92 + 62 = 199. 2n ( 2n) ~ (~)2nV27f2n = 2 n [( ~ )ny'2nn] 2 y7m whence Vii . If iY4i = 4.. = 22446 n by Wallis' formula. 1. As odd squares are congruent to 1 mod 8.Exam #311996 159 Thus. three of the Yi (say Yl. 9. 3. 2 2 hm np n+oo n 13355 2 = . 3} equals 1. iY4i = 0 or 4. X2.
f(c) f(c) . 1223. and {10. If we have two different digits.0715 Look under Enumeration or Probability in the Index for similar problems. 7. 2. the digits {1. 5} equals 1. Checking to see if any of the signed sums of the elements from {O. 2} generate three combinations 1112.3 + 3 .. c + a) to see that f" (a) = 8. 3. and 1222). 9. 3. 5}.. Then Xi = Yi + 6 = {1. If IY41 = 0.. 6}. and 1233).4 + 1 .5 + 1 + 1 = 1. 1. Look under Diophantine Equations in the Index for similar problems. 3. 3} generate three combinations. then there are 3 C20) = 135 combinations (e. we find 0 + 3 + 3 .160 Solutions we find 5 .a 8. Since a was arbitrary. then there are C40) = 210 combinations. {10. 1123. 2. C30) = 360 combinations (e. 519968 Let (a. we have Y = 4X2 + Ax + B. Look under Differentiation in the Index for similar problems. 7. 9. 7}.f(a) = _. If the four digits are all different. If we have three different digits. we must have three odd squares which add to 43. then there are 3 .g.3 = 1.:b'=c'::_=c'=a=_ b. The total number of combinations is 210 + 360 + 135 + 10 = 715 and the probability is p = 715/10. Finally.1 = 1.J(b)) be arbitrary points on the graph with b > a b • We are given that and let c = at or equivalently.000 = 0. 9. The Xi = Yi + 6 = {11. 3}. f(a)) and (b. Take the limit as b + a (and a fortiori b + c. f(b) . 1122.. Exam #321997 519971 We will consider combinations of the last four digits only.5 = 1. the digits {1. and 4 . 4 . This is possible only one way: 43 = 9 + 9 + 25.g... if all the digits coincide then there are ten combinations. .
< .3\ = 6 and \Z2 .2\ ~ \\Z2\_ 2\ = \\Z\2 .3\ ~ \\z2\ . So \z4 .Exam #321997 161 S19972 We calculate sums directly.3)(z2 .3\\Z2 . S 19974 If x Y = yX for x.2\ = \9 . Further.2\ = 7.3)(z2  2).3\ = \9 . then equality holds.= . It is interesting to note that these numbers appear as the first n numbers in the 3rd diagonal of Pascal's triangle. But we know that \Zl  Z2\ ~ \\Zl\\Z2\\ for all Zl and Z2 in C.. Look under Finite Sums in the Index for similar problems..3\\z2 . Thus...2\ ~ (6)(7) = 42.2)\ = \Z2 . .:::.42 1 2 So letting A = 1/2 we get an upper bound.3\ = \\Z\2 . y > 0.5z 21 + 6\ 21 . Finally. A is the smallest upper bound since if Z = ±3. So \z2 . \z2 . which implies lnx Iny x y . then taking the natural logarithm of each side we obtain y In x = x In y. 2 \z4 .5z 2 + 6\ = \(z2 . By using two wellknown formulas the above is equal to ~ (n(n+ 1)(2n+ 1) 2 6 + n(n+ 1)).2\..5z 2 + 6 = (z2 . S 19973 Note first that z4 . 2 SimplifYing we get our answer n(n + 1)(n + 2)/6. Look under Complex Numbers in the Index for similar problems...
162 Solutions Obviously all the above is true if x y. By Axiom 6 and Axiom 2. lnt + lnx = tlnx. call them AI. namely. y > 1 and m > O. We also note that this is possible only if x. and . C l . The proof that this is the fewest number of points and lines will be deferred to the solution of part (b) below.3J3). Cl. thus giving us 16 points in our solution. Ci. and A4 be the other three points on line Cl. So in particular. In t + lnx = t(mx). there is a line Cl that contains the point Al but does not contain any of the points B l . A 3. By Axiom 1. The six axioms will guide us in constructing what is essentially a 4 x 4 array.lnx) and (y. The fact that we also have 16 lines will become apparent later. and Di for 1 :S i :S 3. Now by Axiom 3 and Axiom 2. We adopt the convention that a line in our system can be described by the four points that it contains. call it rl. and D l . there are exactly four points on this line. there are three new lines r2 = A 2B 2C 2D 2. and D 1 . Hence In y = my and In x = mx. C2 = BlB2B3B4. Let A 2. BI.4). By Axiom 5.) Some pairs include: t = 2 =} (2. t = 3 =} (J3. We now have the 16 points in our system. So suppose x =1= y. we write rl = AlBlClD l . t = 3/2 =} (9/4. Let y = tx where t > 0 and t =1= 1. C3 = C l C 2C 3C 4. Each row (and column) will contain four points. Then if we graph the function v = In u. Let us begin the construction. 519975 We claim that 16 points and 16 lines satisfy the six axioms.lny) goes through the origin and has equation v=mu. Simplifying we get tl!(tl) = x. we now add three new lines. there is at least one line. So all pairs of the form (tl!(tl). Then lntx = mtx. tt!(tl)) satisfy the original equation.27/8). we see that the line through the points (x. and r4 = A 4B 4C 4D 4. B i . r3 = A3B3C3D3. These lines together add a total of nine new points. (If we restrict t > 1 we get all solutions with x < y. Then y = tx = tt!(tl). By Axiom 6.
say PI = P. (ii) C I D 2 A 3 B 4 . The possible lines through Blare: We choose lines (1) and (3) to satisfy Axiom 3 for B I . n > 1. the set consisting of the line and all those parallel to it is a family of parallels. and C4. we must add more lines (this is the tricky part). and (e). We now want to satisfy Axiom 3 for the point B I . By Axiom 3. Y contains exactly n points. By Axiom 1. and r4 and four column lines CI. and no nconjiguration with fewer points or fewer lines exists. (c) (f) A I C2 B 3 D4 . This system of 16 points and 16 lines satisfies all of the axioms. A I D 2 B 3 C4 · Let us choose line (a). As above. the problem statement associates a set of six axioms to n. However. n. We start by satisfying Axiom 3 for the point AI' The possible lines through Al are: (a) (d) A I B 2 C3 D 4 . and let C be an nconfiguration. Let Xl denote x. For n = 1. C2. If an nconjiguration exists.2 ::. Note that this simple construction satisfies all of the axioms except Axiom 3. If C is an nconfiguration and x.'" . Say that a collection of points and lines satisfying these axioms is an nconfiguration. It is now easy to picture the 4 x 4 array: we have four row lines rb r2. Let n be a positive integer with n > 1. y are lines in C. there is a configuration with two points and two lines that satisfies the axioms. Proof Let n be an integer. C3. Then by Axiom 2. there exists another line y that contains p. For each positive integer n.P2. For each line. there is by Axiom 6 a unique line Xi containing Pi and parallel to x. and no configuration with fewer points or fewer lines will do. we can choose line (d) and that gives us four lines that contain point AI. Axiom 2 limits the possible lines that we can choose. The rest of the solution to part (b) follows from the following: Theorem. . (c). we must exclude lines (b). i ::. r3.Pn' For each i. Let x be any line of C and choose a point P on x. Note that Axiom 3 is also satisfied for the points C 3 and D 3 . then there is one with n 2 points and n 2 lines. we say that x and y are parallel if they share no points in C.Exam #321997 163 C4 = DID2D3D4. In a similar fashion we choose the following four lines to complete our system: (i) C I A 2 B 3 D4 . To satisfy Axiom 3. A I C2D 3 B4 .
Thus. Since we discarded an entire family of parallels. and 5. 1). Then x = (Xl. the n lines Xl. The second set of points {x E R2 : Ilxlloo = 1} is the square in the plane with vertices (1. This proves the Theorem. there Camlot be two lines through P and parallel to z. let Q be any point of C that is not on Xl. . If (Yl. We wish to show that these are all the points of C. X2) and y = . . Look under Synthetic Geometry in the Index for similar problems. The resulting nconfiguration then has n 2 points and n 2 lines. it is easy to verify that "equal to or parallel to" defmes an equivalence relation on the set of lines of C. For part (b). then there is one with exactly n 2 points and exactly n 2 lines. and (0. X2 + Y2). It is clear that we may continue in this fashion and discard families of parallel lines until we have a minimum of n families of parallel lines. So since z is parallel to Xl. 4. Therefore C has exactly n 2 points.1 families of parallel lines.. Let k be the number of families of parallels for C. we first consider the 1norm. Thus n 2 is the fewest number of points and lines for an nconfiguration.1). (0. and (1.1). Y2).. 1). To this end. B satisfies Axioms 1. then x + y = (Xl + Yl. Then there exists (by Axiom 6) a unique line z through Q and parallel to Xl.0). so Xi is parallel to Xj when i # j. Since n > 1 and C is an nconfiguration.164 Solutions Using Axiom 6. so Axiom 3 is satisfied. 1). (1. and thus B is an nconfiguration. say at Pj' But then z must be Xj. so Q is on Xj. 519976 Part (a) is a standard exercise in norms. (1. 2. We have shown that any nconfiguration must have at least (exactly) 2 points and at least n 2 lines. It is then clear that every line has exactly n lines in its family of parallels. Each point of B is on at least n lines of B. Again by Axiom 6. Since each point is on exactly one line from each family. (1. necessarily z meets y. Choose any family of parallel lines and discard it to obtain a collection B of n 2 points and k . which is a contradiction. We also showed that if an nconfiguration n exists. Axiom 3 implies k ~ n. B satisfies Axiom 6. Suppose that k > n. xn together contain n 2 points. The first set of points {x E R2 : Ilxlh = 1} is the square in the plane with vertices (1. This shows that C must have at least n 2 lines.0). 1).
(c) R3 = {(Xl. IX2 + Y21}. IX21} + max{IYII. (2). max{IYII. Ilx + Ylloo holds if and only if = Ilxlloo + IIYlloo max{IXI + YII. it . = {(Xl. Look under Analytic Geometry in the Index for similar problems. The solution is easier to view geometrically. Then + Yil :::. max{l xII. IXil + IYil :::. The proof easily follows from the properties (l). this means that x and y must be in the same quadrant. (b) R2 = {(Xl. The diagonals Y = Y = x partition the plane into the following four parts: and IXII}. Geometrically. Thus. IX21}. X2) E R2 : X2 :::. IXII}. (d) R4 = {(Xl.Exam #321997 165 holds if and only if Equality holds in this second equation if and only if Thus. X max{lxII. Xl and YI must have the same sign and similarly for X2 and Y2. IX2 + Y21} = max{lxII. X2) E R2 : Xl :::: IX21}. IY21}· Let IXi + Yil IXi = max{lxI + YII. IY21} = and equality holds if and only if each of the following holds: (1) (2) (3) IXil IYil = Xi and Yi have the same sign. IY21}.X21} + max{IYII. For the oonorm. 519977 The volume of water can be found by multiplying the crosssectional area of the remaining water with the length (4 feet) of the trough. X2) E R2 : Xl :::: IX21}. X2) E R2 : X2 :::: (a) RI Then Ilx + ylloo = Ilxll oo + IIYlloo holds if and only if X and Y lie in the same part. and (3) above.
respectively.a 2). Exam #331998 519981 Here is a Cartesian proof: put P at the origin with AB along the xaxis and C. then. The isosceles triangle has congruent angles of measure a. we calculate distances: PE = xIVl+b 2. respectively. D = (0. Now mUltiplying by the length of the trough. Look under Geometry in the Index for similar problems. where Xl and X2 are the positive and negative (resp. We subtract these areas from the total area of the crosssection. vr2 .) roots of the equation (x .a? + y2 = r2. The coordinates of E and C. which is 7r /2.a)2 + b2x 2 = r2. Note that the slopes of these two lines are opposite since D must lie on the yaxis and LCP D = LDP E. bxd and (X2' bX2).4a . to get 7r /2 . we get the volume of the remaining water 27r . and let the lines P E and PC have equations y = bx and y = bx. One can show that the angle subtended by the arc in the sector is 2a. E in the upper halfplane. Since Xl and X2 are solutions of the same quadratic equation. we get the area of the triangle Atriangle = ~bh = ~(2cosa)(sina) = cos a sin a. PC = x2Vl+b2. To find the crosssectional area of the remaining water (region W) we subtract the areas of the sector on the left and the isosceles triangle in the middle of the figure above. D. we know that their product is the ratio of the . Let C have equation (x .166 Solutions suffices to find the crosssectional area of the remaining water when the trough is tilted through an angle a with the horizontal.4 cos a sin a feet 3 . So the area of the sector is Asector 1 = 2(2a)r2 = a.cos a sin a for the crosssectional area of the remaining water. are (Xl. Furthermore. Using trigonometry. as illustrated. Now.
519983 One direction is obvious: if n = p2 .Exam #331998 167 constant term of that equation to the leading term.q > 1 observe that p .iLl iLl + ikdk2 + iL2 ==F1.(fkl + ik2)!kl = /. Then. Then.q) and both of these factors are greater than 1.r2)(1 + b2 ) = r2 _ a 2 = PD 2 (1 + b2 ) A transformational proof: reflect through AB to obtain 0". Look under Number Theory in the Index for similar problems. then n = (p + q) (p .iLl . so that both their sum and difference are even. . 70m = 7k . 1) is on one of the hyperbolas. Look under Number Theory in the Index for similar problems. Thus. we verify that R .2u)/7. First. PO = XlX2(1 + b2 ) = (a 2 . p and q are nonnegative integers.ikdk2 . In the other direction.q = t. iLl . Then.t)/2. 519982 Let k = n/7 and m = (nwu .q2. hl) is on one of the hyperbolas. that is. 519984 For part (a). if n = 8t with 8 2: t > 1. both 8 and t must be odd. then observe that P E" is a continuation of PC and PO" is a continuation of PE.ikdk2 . with n = p2 . Therefore. D" and E".q2. To see that p . inductively. use induction. Thus. then we set p = (8 + t)/2 and q = (8 . observe that (0. Look under Geometry in the Index for similar problems. we see that all points of':S are on one of these hyperbolas (and they alternate between the two).R2 = ±1.hhl  ILl = (fkl + h_2)2 .1 = ILl + 2ikdk2 + IL2 . assume that (fk2. PE.21u so that k is an integer if and only if m is an integer . Apply "mean proportional" to the chords D D" and 0 E". Since n is odd.
b) which is an integer.1). ao): it is clear that this is still a fIrstquadrant integer point and that bo . . But. 2a (since 5 ± 4/ a2 :s.ao :s. Denote by A 2 (y) the area of the portion of R(r. bo :s. From above. Furthermore. we will repeatedly need the following solution of y in terms of x for fIrst quadrant points (the choice of sign corresponds to the choice of hyperbola): y= x + V5x 2 ± 4 =x 1 + J5 ± 4/ x 2 .2) so that (b o .\y)) lying between h and h. bo) be such a point with ao as small as possible. bo).ao)ao . then we have only a few possibilities to consider (a = 0 and a = 1) with both sign choices). we observe a few preliminary facts: if (a. bo) must have been in F as well. leading to three integer points: (0. but is not in F. Furthermore.(1. fIrstquadrant point on one of the two hyperbolas. (1. We now show that this point is on one of the hyperbolas: a6 . implying by our assumption that it must be in F. all of which are in F. b) is a fIrstquadrant integer point on one of these hyperbolas and a 2:: 2. 519985 Denote by Al(X) the area of the portion of R(x) lying between fo and h. then b 2:: a (since 5 ± 4/ a2 2:: 4). namely. using this. Also.Then. there is exactly one point at which bo = 2ao. if a 2:: 2.(b o . Next.ao.b5 + 2boao  a5 = a6 + boao  b6 = ±1 since (ao.ao)2 = a6 .2).168 Solutions For part (b).ao. ao. 2 2 Now. Look under Analytic Geometry or Fibonnacci Sequences in the Index for similar problems. bo). we know that ao 2:: 2 and that ao :s. ao) has a strictly smaller fIrst coordinate than (ao. then b :s. assume that there is a point (a. then (ao. contradicting our choice of (ao. 1) and (1. 9). if a < 2 and is a nonnegative integer. Consider the point (b o .(b o . 2ao. Let (ao. bo) is on one of the hyperbolas.boao + a5 .
1 and then inverting..8 + 1)a 13 a. So. Now.a 13+1 • Integrating and solving a bit further.1 dt.Exam #331998 169 Furthermore.8+1)a13 13+1· Solving for f:. z.8 + 1 .a)f3· Look under Integration in the Index for similar problems..1(y)) or h in area may be rephrased 13+1 l t1/f3 _ 1/13 1 _ o a y f2 dt  (a . . 519986 (a) An inclusionexclusion argument gives D n =nln(n1)1+ Rewriting this a bit gives (.(~)(n3)1+ . This latter form is useful for part (b): Since e=L·.. we find that x f3 a f3 +1. its limit has absolute value strictly less than the absolute value of the first term which is 1/ (n + 1). we find that l o f 2 (t)dt=y13 1 13+1 (. for n > 0 we have part (b).8f3 fz(x) = (a. ·\<1/2. i=O 00 (_1)i it is clearly sufficient for part (b). i=n+1 Since this latter series is an alternating series with terms decreasing in absolute value. .f:.1(t)dt = io r t1/f3 a 1/f3 . A2(Y) = io r r. the condition that it bisect fo and as A 2(y) = A1U.1(t) y r. to show that \n!/eDn \=\ L 00 (1)in' z.8 + 1 .1)y13 (..)(n2)1..
we are reduced to examining determinant 1 Jordan forms. which has determinant 1 (do this by dividing A by the square root of its determinant. for classes other than the class of the identity... say 0 = BAB. which is possible since we're dealing with complex matrices). yields the following expression for the radius: . Consider the nondiagonal form first. We calculate that a2 e(a + d) + be b(a + d)) d 2 +be' . together with the ratio test method for computing radius of convergence. say A'. .=o (. Look under Derangements in the Index for similar problems. then adding or subtracting I. Generally. There are two such. Dn+In! .~)i hm = hm . This occurs when =Fv'det A is an eigenvalue of A which occurs only when A has a repeated eigenvalue.. (b) This provides us with one or two distinct square roots. n!(n + I)! L:~!OI (_." The way to see this is to take a representative of the equivalence class which has determinant 1 and put it in Jordan canonical form. then(A±I)2 = (tr A±2)A so that one simple method of coming up with two square roots of [A] is to find a representative of the class. one very simple method is to observe that the example generalizes in the following way: if det A = 1. diagonal and nondiagonal. In this case. Is this all? The answer. using the expression above. but occasionally one of the two will be singular. is "yes. Any square roots [D] of [0] correspond to square roots [B. this will produce two distinct valid square roots. An alternate form would be v1AT = [A ± v'det AI].1 DB] of [A]. .! . Here 0 = (~ ~) and we are looking for D = (~ ~) such that ( a e b) 2 d ( D2 = wO for some w. So. one of the two square roots will be singular (and its square will be the zero matrix).1)' ..170 Solutions For part (c).I is in Jordan form.=l. n+oo Dn(n + I)! n+oo (n + l)!n! L:~=o (i~)' = lim 1+ n+oo ( _l)n+I (n + 1) "n L. 519987 For part (a).
we must have a = d. bare Exam #341999 519991 The key observation is that the procedure never alters the parity (even/odd count) of the orange balls. . So we must have b = e = 0. To show that there are infinitely many nonequivalent square roots. A~ = (~ 1) with A =1= AI (otherwise C is equivalent to J). we apply the calculations for part (b) and obtain: b(a + d) = 0. Look under Logic in the Index for similar problems. For part (c). a2 + be = d 2 + be. [(~ l/Jx)] = e(a + d) and [(~ 1~v1)] which are equivalent to the two square roots computed by the simpler method above. we have b( a + d) = e( a + d) = 0. consider the matrices arbitrary complex numbers. The former implies that either b = e = 0 or a + d = 0. So. so that. Each performance results in a loss of one ball from the urn. The first equation (together with one of the latter) implies that e = 0 and a + d =1= O. giving infinitely many equivalence classes of square roots of [1]. a 2 + be = W A and d2 + be = W A1.Exam #341999 171 For this to be equal to (~ ~) we need e(a + d) = 0 and b(a + d) = a2 O. the latter two cannot hold. since this would force WA = WA 1. But. but if a = d. Its color must be orange since the original number of such balls is odd. namely [( ~ In the diagonal case. These are all satisfied whenever a + d = 0. in addition to the obvious fact that [1] is its own square root. a2 = WA and d2 = WA 1. (! ~l) where a. This gives us two classes of square roots. just one is left. C ~) ]. Look under Matrix Algebra in the Index for similar problems. since clearly a 2 = d 2 . Thus we are reduced to 2ab = a 2 which gives us one W =1= + be = d2 + be = equivalence class. eventually.
some via analytic geometry. 2x2 + 2V2x Vx 2 + y2 + x 2 + y2. Look under Geometry in the Index for similar problems. 519994 There is a grain of truth to the claim mentioned in the question. so that the same is true of (i). increasing or not. where A. giving 5x 2 + 2xy + y :s. Iff is a convex jUnction. The simplest and most elegant proof. The standard area formula (~base x height) for triangles then shows that the constant of the problem is the altitude of T. in attempting to eradicate the radicals. B.172 Solutions 519992 Various correct proofs can be given. or at the left. then the Riemann sums (over uniform partitions) do tend monotonically to the integral. This is true whether the sums are evaluated at the righthand endpoints (as they are here). (ii) Cancellation leads to the simpler inequality (iii) Squaring again gives (iv) or (v) Inequalities (i) to (v) are all equivalent. some using trigonometry. (v) is valid. 519993 The inequality may be rewritten as (i) We square both sides. however. with equality only when y = x. is to consider area. and C are the vertices of T. A counterexample to . or whether they are inscribed or circumscribed. We have area of T = area of AB P + area of BC P + area of CAP. Look under Inequalities in the Index for similar problems.
x :::. 519995 We take the dartboard to be the square centered at the origin with comers at (±1. f(1/3) + f(2/3) + f(l) _ 3 Look under Riemann Sums in the Index for similar problems. f(x)= 1. if1/2:::. 1/2. The symmetry of the problem enables us to compute this ratio by restricting our attention to the triangle indicated below. if 2/3 :::. It is clear that if 0:::.x:::. The points closer to the center than the edge are those enclosed by four parabolic arcs as illustrated. { 6x3. f(1/2) and + f(l) = 1/2 2 ' 2/3 . ±1).Exam #341999 173 the claim is provided by the piecewise linear function 0.0) .2/3. x :::. The desired probability is the ratio of the shaded area to the total area of the square. 1. (1.0) p _ _I (1. 1) (0.
courtesy of the cosine.0). q). Its terms lie between those of the (divergent) harmonic series. and only one. k = 4.. so the probability is Look under Probability in the Index for similar problems. 41 5. while Q is the intersection of the arc with the line y = x. We show that the partial sums of our series are not Cauchy: more precisely. but this occurs too slowly for convergence because of the logarithm. :Z::::.:O=l (1)k / k.:O=ll/k. and those of the (convergent) alternating harmonic series.174 Solutions The parabolic arc is the locus of points equidistant from (0. k=e(2n i)" k . obtained by expanding the binomial coefficients as prescribed in the question. ). P is the point (1/2. There is one. so that Q=(}21.x. :Z::::..}21).q2)/2. solution: n = 14.2. The shaded area of the triangle is 4}25 6 The total area of the triangle is 1/2.. .= 2 519997 The series diverges. say stand in the ratio 1 : 2 : 3 if and only if nk nkl 3 and k+l k+2 2 These equations. we prove (2n+i)" e L cos(1og k) > (n=I. we must have q = (1 . There is some cancellation in our series.y2)/2. Letting Q = (q. may be rephrased as n = 3k + 2 and 2n = 5k + 8.. 519996 Three consecutive entries of the nth row. .0) and the line l. J so that the equation of the arc is x = (1. We have x 2 + y2 = 1 .
all of which are greater than 1 Look under Infinite Series in the Index for similar problems. + A m. 519998 The standard formula for the sum of a finite geometric series has a matrix analogue.y2 > M whenever 1 so that cos (log k) .A)x = y. Since 1 .Am = 0 by hypothesis. by (**).1)(1  A)x = 0 no matter what the vector x.. there exist nonzero vectors x and y with (1 . ..0.::: whenever V2 1 The series (*) contains at least terms. (1 +A+ ... But then...Exam #341999 175 To do this. + Am1)y = 0 so that 1 + A + .. + Aml is singular and det(1 + A + . + Aml) = Look under Matrix Algebra in the Index for similar problems.A i=. we observe that cos x . 1 . o. so that (1 + A + ..
In particular. 8.176 Solutions Exam #352000 520001 If Cindy didn't do it. we have x Thus we see that + ffx < 2x  2. Sn+1 :::::: Sn and therefore Sn are increasing. b . Therefore Cindy is the only possibility. 520003 It converges.1. We show the Sn < 2 for all n and thus we have convergence. and 11 are 400. the other three are lying. (b) 1885 = 1111 12 . Look under Logic in the Index for similar problems. Since N < 2000. we have N :::::: 1885 whence b4 > 1885 so that b :::::: 7. The values of Mb for b = 7. 520002 The solution to part (a) is 2000 = 5555 7 . Let the number be N = aaaab = aMb where Then since 1 ::. Look under Number Theory in the Index for similar problems. 1464 and no multiples of these lie between 1885 and 2000. 9. 12 and since 111112 = 1885. Let Since n + vn + 1 > n. 820. 10. 585. b::. This means that Ducky is telling the truth. a ::. . Becky must be lying. 1111. Mb < 2000. For x:::::: 6.
or 15.2 1+ 6 < ~'+)2+V3+ .75793 . Look under Group Theory in the Index for similar problems. A == 1 (mod 3). Thus A = B = 1 and n3 = 2. < 1+~2+)3+V4+V5+v'iO < < ~ 1+)2+ V3+ V4+V9 + +. Look under Sequences in the Index for similar problems.. and can therefore be 1. For these k. by Sylow's theorem.. Clearly nl = 1. VI V2 < VI + v'5 < The actual value is 1.)3 + 3 2.1 . Since nl + n3 + n5 + n15 = 15.. B == 1 (mod 5)..2 . n15 = 15 .3+ y2n < . +Vn2+V2n=4 <~ 1 + )2+ V3+'" + Vn. . Suppose G has A subgroup(s) of order 3 and B subgroup(s) of order 5.. 520004 The order of an element divides the order of G.4 = 8 > O..2+ Vn1 + y2n.. let nk be the number of elements of order k. 5. and both A and B divide 15. Since 3 and 5 are prime.Exam #352000 177 Sn < ~ ~2+)3+'" + Vn. n5 = 4. We'll show that n15 > 0 (an element x of order 15 exists) and so G =< x > is cyclic. 3..
where K denotes the area of the triangle.1).. B . 7f 4 f(x) = (1 . and 11 Fk TI < 1. we have ~(1 .1) = where x = cos BE (1/2. G sm sm ~ sin2BsinBsin3B 7f 7f 7f ~ cos Bsin3 B(4cos2 B1) ~x(l X2)3/2(4x 2 .T)k _l_ T k  V5 V5 so ~ (l)kFk = __2. We use the law of sines.: sm = = 2.x) on (1/2.cosh Vs) + l. by the double angle formula cosh 2x = 2 sinh2 x Look under Fibonnacci Sequences or Infinite Series in the Index for similar problems. .1). sinG = c/2R and the area formula K = ~absinG to write p = (a/R) (b/R) sinG 27f = .178 Solutions 520005 Since ~Tk V5 =  ~(1  V5 T)k is an integer for k 2: 1. Thus we seek the maximum of p = .x 2)3/2(4x 3 . A .) = ~(1. 520006 The desired ratio is p = K / 7f R2.sinh (V5) k=O ~ k! J5e 2 and (~ ~~ ) (~ (li!k Fk ) = ~4 sinh2 (V.f(x).
which can be combined under the Principle of superposition to yield the general solution f(x) = LCieTiX. the characteristic polynomial factors as (D2 + 1)(D .24x4 =0 or x =12 ' 2 4+v'IO the other root being less than 1/2.x 2)3/2(12x 2 1) + (4x 3  x)(3/2)(1. Thus the answer is ig( 1l)3/2(4g 2 7r This can be written as  1).X 2)1/2 [(1. i (1.(~+ ~v'iO))3/2 (4 (~+ ~v'iO) 1) V3 + ~v'iO . y = xv'2 is a counterexample since y = 2 is rational. If x is rational.1) + (4x 3 = (1 .X 2)1/2(_2x) x)( ~3x)] = (1. h 12 7r 3 12 3 12 = 7v'IO 1087r 2 /22 V + 4v'iO (7v'IO 547r 2)( V20 + J2) 1087r 34J2 +5V5 Look under Geometry or MaxIMin Problems in the Index for similar problems.24x4) so 16x 2 .1)(D2 + D + 1)f = 0. Look under Number Theory in the Index for similar problems. . Otherwise. 520007 Consider x = J2v'2.1.Exam #352000 179 f'(x) = (1.x 2)(12x2 . it is a counterexample. 520008 If D f = 1'.1 . which yields five linearly independent solutions.X 2)1/2(16x 2 .
ISO Solutions where ri is a root of the characteristic polynomial (±i. B = A and f(x) = AeX/2[eixv'3/2 _ e. . w. Look under Differential Equations in the Index for similar problems. w). Re(ri) < 0 and f(x) = Ae iw + Be iw . x) for some constant C. with w = (1 + v3i)/2. 1. Since limf(x) = 0.ix v'3/2] = Ce x/2 (sin V. Since f(O) = 0.
19713. 19687. 19906. 19717 Analytic Geometry 19672. 19824.Index by Problem Type Algebraic Structures 19711. 19984 Arclength 19921 Binomial Coefficients 19996 Complex Numbers 19675. 19911. 19881. 19763. 19917. 19976. 19803. 19946. 19796. 19862. 19765. 19951. 19782. 19886. 19863. 19816. 19964. 19968 Differential Equations 19694. 19973 Derangements 19874. 19986 Differentiation 19664. 19793.20008 181 .
19704. 19695. 19984. 19962. 19902. 19885. 19967 Enumeration 19665. 19746. 19785. 19714. 19993 Infinite Series 19677. 19981. 19944. 19928. 19753. 19943. 19913.19823. 19716. 20005 Finite Sums 19783. 19668. 19842. 19875. 19853. 19932. 19864. 19957. 19873. 19686. 19684. 19834. 19893.20006 Group Theory 19727. 19977. 19685. 19952. 19872. 19688. 19882. 19822. 19992. 20005 . 19703. 19682. 19678. 19751. 19813. 19938. 19886. 19742. 19676. 19724. 19924. 19854. 19752. 19917. 19836.20004 Inequalities 19801.182 Index by Problem Type Diophantine Equations 19661. 19833. 19955. 19674. 19971 Field and Ring Theory 19735 Fibonnacci Sequences 19832. 19997. 19913. 19972 Fundamental Theorem of Calculus 19883 Geometry 19662. 19847. 19923. 19965. 19904. 19942. 19912. 19807.
19936. 19786. 19755. 19991. 19926. 19844. 20006 Miscellaneous 19667. 19916.20001 Matrix Algebra 19671.19693. 19933. 19811. 19937. 19897. 19766. 19715. 19754. 19983. 19744.19734.19705. 19961. 19855. 19982. 19885. 19896.20007 Permutations 19725. 19837. 19896. 19884. 19843.19685. 19802. 19995 . 19743. 19921.19987.Index by Problem Type 183 Integration 19713. 19925.19998 Max/Min Problems 19732. 19945. 19927.19862. 19852. 19915. 19876. 19866 Number Theory 19663. 19963. 19814. 19966 logic 19891. 19934 Probability 19739. 19737. 19955. 19775. 19901. 19826. 19985 limit Evaluation 19666. 19841.19738. 19832.19701. 19857. 19846. 19941.19772. 19894. 19905. 19762. 19722. 19781. 19914. 19804. 19845. 19953. 19721. 19966. 19815. 19773. 19935. 19731.19712. 19826. 19831.19871. 19874. 19833. 19806. 19887. 19734. 19922.20002. 19856. 19791.19692. 19795 Multivariate Calculus 19683. 19971.
19696. 19838. 19741. 19884. 20003 Solving Equations 19974 Systems of Equations 19765. 19821. 19851. 19761. 19812. 19763. 19723. 19792 . 19771. 19774. 19865. 19691. 19736. 19702. 19954 RealValued Functions 19664. 19764. 19994 Sequences 19855. 19685. 19892. 19931.184 Index by Problem Type Polar Coordinates 19726 Polynomials 19675. 19905. 19839 Riemann Sums 19681. 19805. 19745.19673. 19937 Statistics 19967 Synthetic Geometry 19975 Taylor Series 19958 Trigonometry 19772. 19784. 19825.
Index by Problem Type 185 Volumes 19794. 19895. 19827. 19956 .