Arml: 1995-2003

AMERICAN REGIONS
MATH LEAGUE
&
ARML POWER CONTESTS
1995- 2003
by Donald Barry and Thomas Kilkelly

Cover by Heather Barry
Copyright 2004 by ARML. All rights reserved.
Published by ARML. Website: htqJ:!!www.arml.com
Printed in the United States ofAmerica.

Contents
Preface: American Regions Math League·--------···--------------········-·--------····-···-•···----------------------------- 1v
Preface: ARML Power Contest __ ··-----------········---------·----·--·---------------·-------------------------------------------- vi
Acknowledgements: Problem writers and reviewers _____________________________ ·······---------------------------------- vm
Prize winners lX
AMERICAN REGIONS MA TH LEAGUE: 1995 - 2003
1995 ARML 1
Team questions and solutions __________________________ ···--------------------------------------------------------- 3

Power question and solutions _____________________ ······-----···-·--·------------------------------······---------- 8
Individual questions and solutions __________ ·-·-··-----------··---------------··-····---------····---··---------- 15
Relay races and solutions. ___________________________________________________________________________________________ _ 19
Tiebreakers and solutions 23
1996ARML 25
Team questions and solutions _____________________________________________________________________________________ _ 27

Power question and solutions __________________________________________________________________________ ··---------- 32
Individual questions and solutions __________ ··----------------------------------------------------·····---------- 41
Relay races and solutions ___________________________________________________________________________________________ _ 45
Super relay and solutions. ___________________________________________________________________________________________ _ 48
Tiebreakers and solutions _______________________________________ ···----------------··-------------------------------- 53
1997 ARML 55
Team questions and solutions ___________________________________________________ ._________________________________ _ 57

Power question and solutions. _____________________________________________________________________________________ _ 62
Individual questions and solutions ______________________________________________________________________________ _ 71
Relay races and solutions __________________________ ···------------------------------··-··-··---··•·········-···------ 77
Super relay and solutions·---··-·-···-----------··---··---------------··------------------------------------·--------- 80
1998ARML 87
Team questions and solutions _____________________________________ ···---------------···-·-------------------------- 89

Power question and solutions ________________________________ ··----··---------------···--·------------------------- 94
Individual questions and solutions ____________________ ---------------------------------__________________________ 101
Relay races and solutions.____________________________________________________________________________________________ 105
Super relay and solutions.____________________________________________________________________________________________ 108
AMERICAN REGIONS MA TH LEAGUE: 1995 - 2003
1999 ARML 115
Team questions and solutions______________________________________________________________________________________ 117

Power question and solutions____ . ________________________ . ________________ . _____________ .. ______________ .__________ 122
Individual questions and solutions ______ --·-··-·--·-·····-----·--·-·--·-·---------·--···--------------------·---- 129
Relay races and solutions_____________________________________________________________________________________________ 133
Super relay and solutions_____________________________________________________________________________________________ 13 6
2000 ARML 143
Team questions and solutions_____________ _______________ ________ ____________ ____ _________________ __ __ ____________ 145
Power question and solutions______________ ---------------------------------------------------------······--··-··-- 150
Individual questions and solutions_______________________________________________________________________________ 161
Relay races and solutions_____________________________________________________________________________________________ 165
Super relay and solutions_____________________ ------------------------------------------------------------------------ 168
2001 ARML 175

Power question and solutions_______________________________________________________________________________________ 186
Relay races and solutions __________________________ ------------------------------------------------------------------ 197
Super relay and so 1utions__ _______ __ _______________ ______________ ___ ___ ______________ ________________ ____ __________ ___ 2 00
2002 ARML 207

Relay races and solutions_____________________________________________________________________________________________ 22 7
Super relay and solutions_____________________________________________________________________________________________ 23 0
2003 ARML 237

Relay races and solutions_____________________________________________________________________________________________ 25 7
Super relay and solutions_____________________________________________________________________________________________ 260
11
ARML POWER CONTEST: 1995 - 2004 267
1994-95 Color Transformations ___________________________________________________________________________________ _ 269

Solutions _______________________________________________________________________________________________ _ 271
1995-96 Induction _____________________________________________________________________________________________________ _ 274

Solutions _______________________________________________________________________________________________ _ 276
Rook Polynomials__________________________________________________________________________________________ 280

Solutions 283
1996-97 Rotating Decimals__________________________________________________________________________________________ 286

Solutions 288
Regular Closed Linkages_________________________________________________________________________________ 290

Solutions 292
1997-98 Factorial Polynomials ____________________________________________________________________________________ _ 298

Solutions _______________________________________________________________________________________________ _ 300
Integer Geometry___________________________________________________________________________________________ 304

Solutions 306
1998-99 Unit Fractions________________________________________________________________________________________ . ______ _ 309

Solutions ________________________________________________________________________________________ .. _____ _ 312
Chromatic Polynomials___________________________________________________________________________________ 315

Solutions 321
1999-00 Twenty-five Point Affine Geometry _________________________________________________________________ _ 324

Solutions _______________________________________________________________________________________________ _ 328
Square Sum Partitions_____________________________________________________________________________________ 331

Solutions 333
2000-01 Slides, G!ides, and Rolides _____________________________________________________________________________ _ 336

Solutions _________________________________________________________________________________ . _____________ _ 339
Pythagorean Trip Jes_________________________________________________________________________________ .. _____ 342

Solutions 345
2001-02 Cevians________________________________________________________________________________________________________ _ 350

Solutions _______________________________________________________________________________________________ _ 354
Insane Tic-Tac-Toe _______________________________________________________________________________________ _ 359

Solutions _______________________________________________________________________________________________ _ 363
2002-03 Three Addition Problems-------------------------------------------------------------------------------- 369

Solutions _______________________________________________________________________________________________ _ 373
Number Theoretic Functions--------------------------------------------------------------------------- 386

So lutions _______________________________________________________________________________ ... ________ . ____ _ 390
2003-04 Errors in Mathematical Reasoning____________________________________________________________________ 394

Solutions 397
Mathematical Strings _______________________________________________________________________________ .______ 399

Solutions 401
lll
The American Regions Math League Competition
ARML is like no other mathematics contest. After months of planning and preparations, tryouts and practice
sessions. busloads of students stream onto three co Hege campuses, turning sedate institutions into beehives of
excitement and anticipation. New friendships are made; old ones are renewed. Superb mathematics students from
across the country are drawn together by their love of mathematics, eager to measure their abilities against other
talented students as well as against a collection of truly challenging non-routine problems. ARML is distinguished
by the fact that it creates communities of mathematics students and teachers, that it values and honors talented and
worthy students often passed over by a society focused on media heroes. ARML's format provides for a variety of
problem-solving situations and the fine problems that distinguish ARML provoke and promote mathematical insight
and inventiveness, opening up new avenues of research as well as intriguing aspects of familiar material. I'll never
forget the thrill I experienced the first time I saw an auditorium at Penn State packed with 900 students, all
absolutely quiet, all completely intent on solving a problem. I'll never forget the explosion of joy and excitement
that followed the announcement of the answer. It is an experience that repeats itself year after year, a confirmation
of our hopes and dreams as teachers of mathematics.
ARML was founded in 1976 as the Atlantic Region Mathematics League. It was an outgrowth of NYSML,
the New York State Mathematics League, founded in 1973, and designed to serve as a competition for the best of
the teams in the math leagues in New York. Both were the joint vision of Alfred Kalfus and Steve Adrian. Alfred
served as the first president of ARML and Steve. along with Joe Quartararo (from 1976 to 1982). did the
organizational work, site preparations, and publici~r'- Marty Badoian as vice president and Eric Walstein also played
crucial roles in ARML's early days. The impetus for ARML came when Marty Badoian brought a Massachusetts
team to NYSML in 1975 and they did so well that an expanded competition seemed desirable. ARML was
conceived as an interstate competition covering the eastern seaboard, formed through the joint action of the New
York State Mathematics League, the New England Association of Mathematics Leagues, and leagues from New
Jersey, Pennsylvania, Maryland, and Virginia. But ARML is a great contest and it soon attracted math students and
math teachers from all over the country. By 1984 ARML was renamed the American Regions Mathematics League.
Currently, it takes place simultaneously at three sites. Penn State University in State College, Pennsylvania, the
University of Iowa at Iowa City. Iowa, and San Jos,a! State University in San Jose, California. It brings together
some 1500 of the finest young mathematicians in the United States and Canada. In the past teams from Russia have
competed and currently teams from Taiwan and the Philippines are taking part. Taiwanese educators even created a
similar competition for schools in Taiwan called TR.ML.
ARML is a competition between regions. A region may be as large as a state-there are teams from Texas,
Minnesota, and Georgia, it may be half a state-eastern and western Massachusetts field teams, it may be a county
such as Suffolk county in New York, it may be a city such as Chicago or New York, or a region may even be a
school. For example, Thomas Jefferson High School in Fairfax, Virginia has sent some remarkably successful
teams. Each team consists of 15 students and a region may send more than one team.
lV
The contest consists of 6 parts. First is the Team Round. A team's 15 students are in one classroom, they
receive 10 problems, they have 20 minutes to solve them as a group. and they may use calculators. Typically.
problems are divided up so that every problem is being worked on by at least one student. answers are posted, and
hurried consultations take place if there is disagreement. Next comes the Power Question. Teams are given 60
minutes to solve a series of in-depth questions on one topic and/or prove a number of theorems on that topic. The
teams' papers are graded by a hardworking group of teachers tucked away in a comer of the auditorium. Following
the Power Question, the teams come together in a large auditorium for the Individual Round. Here they solve 8
questions, given in pairs with 10 minutes for each pair, right answer only. Neither calculators nor collaboration are
allowed. Initial problems are easier, but an easier problem is generally paired with a more difficult one. We try to
write problems so that 80-90% of the students can solve the first one and less than 5% can solve the last one. The
next round is the Relay Round. The teams of 15 are divided up into 5 groups of 3. Each I st person in a group gets
the same problem, each 2nd person gets the same problem, and each 3rd person gets the same problem. The first
person's problem has all the information necessary to solve it, but the second person's problem requires the first
person's answer, and the third person needs the second person's answer. None of the three knows the other's
questions. A well-written relay problem enables the second and third team members to do considerable work while
waiting. They may even be able to discover a candidate pool oflikely answers to their problem.
Those four rounds are the only rounds that count for the Team and Individual competition. Scoring is as
follows: each correct answer on the Team Round earns 4 points. Sometimes a team gets all 40 points, but 36 has
typically been the top score and the average for all the teams has been between 20 and 24. The Power Question is
worth 40 points; often one or more teams earn a perfect score. Each individual problem is worth 1 point per
contestant, meaning that the team can earn as many as 8 · 15 = 120 points on this round. Generally, 90 is a great
score. On the Relay Round, only the third person's answer is scored. If the group of 3 gets the problem correct
within 3 minutes they earn 4 points, if within 6 minutes they earn 2 points. Thus, each relay is worth a maximum of
5 · 4 = 20 points; a score of 14 is quite good. The top team scores for the entire contest fluctuate between 160 and
190 points although the 1995 and 1997 contests were particularly difficult, causing top scores to drop into the 120's.
The Tiebreaker Round follows the relays. ARML only gives 3 top individual prizes and usually a playoff
must take place between the top scorers. It is a dramatic moment. At each site the contenders come to the front of
the auditorium, they receive the problem at the same time that it is flashed on an overhead screen. Each student is
timed and the winners are determined by who gets the correct answer most quickly. Sometimes several tiebreakers
must be given to determine the 3 winners. The Super Relay. a competition that first took place in 1996, is the final
round and it is just for fun. It is a relay race for the whole team involving 15 questions. The team that gets the
correct answer to the last question is the winner, earning a round of applause and bragging rights 'ti! next year. To
keep the Super Relay from getting bogged down, we usually slip one or two problems into the relay that can be
solved without requiring the previous person's answer. In the 1996- 1999 Super Relays, students passed answers
from position 1 to 15. Starting in 2000. students passed from both ends into position 8. That student's question
required two answers in order to be solved.
V
ARML is quite an undertaking. It brings together great students, it is one of the few mathematics contests
that involves exciting travel. meeting lots of other students, and renewing friendships established during summer
programs. It relies on the work of a large number of dedicated adults who do all the organizing and spend hours
finding or writing problems to use in practice sessions. It is a contest that promotes and demands creativity and
imagination. The students who take part in ARML are very experienced problem solvers, quick and insightful.
Those of us who write problems for ARML respect the abilities of our participants and, consequently, we spend
hours developing problems that spring from high school mathematics, yet are out of the ordinary, problems that can't
be attacked in rote fashion. We don't just take a theorem and write a problem that employs it. We ask questions, we
imagine situations we'd never thought of before, we try out this idea and that possibility, we run into dead-ends, and
we occasionally stumble across a really neat idea and that's the one that makes it into an ARML competition. There
are many different topics on each year's ARML, but more important, there are a wide variety of ways of thinking.
This publication makes available the problem sets and solutions for the 1995 to 2003 ARML contests. Three
earlier publications contain ARML problems from its beginning in 1976 to 1994 and NYSML problems from its
beginning in 1973 to 1992. The first was NYSML-ARML Contests 1973-1982, published in 1983 by Mu Alpha
Theta. The second was NYSML-ARML Contests 1983-1988 by Gilbert Kessler and Lawrence Zimmerman.
published by the National Council of Teachers of Mathematics in 1989. The third wasARML-NYSML Contests
1989-1994 by Lawrence Zimmerman and Gilbert Kessler published by MathPro Press in 1995.
ARML has an executive board that organizes each contest. Alfred Kalfus was president from 1976 until 1989
when Mark Saul of Bronxville High School, NY took over. Mark retired in 2001 and Tim Sanders, director of the
Great Plains Mathematics League, took over. Starting in June of 2004, J. Bryan Sullivan will be the president.
Current board members include Marty Badoian, Steve Adrian, Linda Berman, J. Bryan Sullivan, John Benson, Josh
Zucker, Paul Dreyer, Kristie Sallee, and Amy Gibbs. Marty, Steve. and Bryan have been involved with ARML
since its inception.
Don Barry has been the head author since 1995 and during that time he has received a number of great
problems, solutions, and a lot of spirited help from Paul Dreyer, Ed Early, Zuming Feng. Zac Franco, Chris Jeuell,
Paul Karafiol, Rick Parris, Ashley Reiter, Leo Schneider, Ravi Vakil, Eric Wepsic, Elizabeth Wilmer, and Andrei
Zelevinsky. The following teachers at Phillips Academy in Andover, Massachusetts have helped enormously in
editing, reviewing, and occasionally suggesting problems: Frank Hannah, Chris Odden, David Penner, Bob Perrin,
and Bill Scott. Our work has benefited from and, we hope, built upon the fine work done by ARML's previous
problem writers including Steve Conrad, Dr. Norman Schaumberger, Dr. Erwin Just, Irwin Kaufman, Howard
Shapiro, Harry Ruderman, Gilbert Kessler, and Larry Zimmerman.
I would like to dedicate this book to my wife, Roxy, whose cheerful patience, support and encouragement
have been so invaluable and to Bob Perrin whose insights and brilliant explanations have helped make each and
every contest so much better.
Don Barry
Phillips Academy, Andover, MA 01810
dbarrv(alandovcr.cdu
Vl
The ARML Power Contest
In 1995, under the leadership of its president, Mark Saul, ARML introduced the ARML Power Contest.
Modeled on the Power Question which involves, as Mark noted, "cooperative effort in exploring a problem situation
through the solution of chains of related problems", ARML began to offer a mail-in competition. In November and
in February, each participating team receives a set of problems based on a major theme. Some of the problems
require just a numerical answer, some require justification, while others require that a proof be written. The
mathematics of the problems has been geared so that students in an honors class, a math club, or on a math team can
have a unique problem solving and mathematics writing experience. There is no limit to the size of the tean1, but the
time for solving the problem set is limited to 45 minutes. Calculators are permitted, but no other outside aids are
allowed. Initially, mathematicians from Thomas Watson Laboratories of IBM, the National Security Agency, and
Los Alamos Laboratories graded the papers. Currently, the papers are now graded by the author and some of his
colleagues. There are 40 points per contest; the team score is the sum of the scores of the two contests. A team may
represent a region, not just a school, but the contest must still be taken by all members of the team at the same time
and place. The contest has grown steadily. As of the 2003-2004 academic year there were 43 teams from the
United States, Canada, and Bulgaria participating.
The Power Question and the Power Contest are both designed to simulate actual mathematical research
activity. This is not easy. As Mark noted,
"Power Contest problems are difficult to write. They must provide meaningful problem situations
both for the novice and veteran mathletes. They must attract schools with strong traditions in
mathematics competitions, yet offer experiences for students new to such events. Thus, they must
build mathematically significant results out of mathematically trivial materials."
Roger Sadlowsky, a ARML coach from Minnesota, coordinated the first few years of the competition. The initial
authors were ARML coach Thomas Kilkelly and ARML alumni John Tillinghast and David Miller. Two weeks in
the Soviet Union, working with Soviet mathematicians, teachers, and students, provided the necessary ideas for the
writers to produce the first few years of problems. This effort was underwritten by Best Practices in Education, a
non-profit foundation dedicated to finding exemplary education practices around the world and adapting them for
American educators and students. In 1998 Sadlowsky retired, Tillinghast and Miller went on to graduate studies and
Kilkelly volunteered to take over as coordinator and author.
I would like to thank Mark Saul for his encouragement and trust, Peter Moxhay and Gail Richardson from
Best Practices in Education for their vision in providing funds for the trip to Russia, John Tillinghast and David
Miller for their insight and enthusiasm in getting the project started, Aloysha Belov, Grisha Komdakov, and Dmitry
Fomin for their Russian hospitality and mathematical creativity, Evgenia (Jenny) Sendova (Bulgaria) and Bogdan
Enescu (Romania) for their continuing enthusiasm and support, and Dr. David Cervone and Dr. Steve Heillig for
their proof-reading and exchange of mathematical ideas. Special thanks to Carla, my wife, whose patience, support
and encouragement were and are essential for my participation in this project. I want to say that I really enjoy the
research, problem solving, mathematical writing challenges the competition provides.
Thomas Kilkelly, Wayzata High School

4955 Peony Lane North, Plymouth, MN 55446.
kilkelly@ties2.net
Vll
Problem Writers and Reviewers for ARML 1995 - 2003
1995: Don Barry, George Markowsky, David Penner, Bill Scott, Eric Wepsic, and Andrei Zelevinsky.
Reviewer: David Penner.
1996: Don Barry, Zac Franco, Bill Scott, Eric Wepsic, and Andrei Zelevinsky.
Reviewers: Frank Hannah. David Penner, Bob Perrin, Leo Schneider, and Bill Scott.
1997: Don Barry, Zuming Feng, Zac Franco, Ravi Vakil, Eric Wepsic, and Elizabeth Wilmer.
Reviewers: David Penner, Bob Perrin, and Andrei Zelevinsky.
1998: Don Barry, Zuming Feng, Zac Franco, Ravi Vakil, Eric Wepsic, Elizabeth Wilmer, and Andrei Zelevinsk-y.
1999: Don Barry, Paul Dreyer. Zuming Feng, Zac Franco, Chris Jeuell, Ashley Reiter, Leo Schneider,
Ravi Vakil, Eric Wepsic, and Andrei Zelevinsky.
Reviewers: David Penner and Bob Perrin
2000: Don Barry, Paul Dreyer, Ed Early, Noam Elkies, Zuming Feng, Zac Franco, Chris Jeuell, Ashley Reiter,
Leo Schneider, Ravi Vakil, and Eric Wepsic.
2001: Don Barry, Paul Dreyer, Ed Early, Zuming Feng, Zac Franco, Chris Jeuell, Leo Schneider, Ravi Vakil, and
Eric Wepsic.
Reviewers: David Penner and Bob Perrin.
2002: Don Barry, Paul Dreyer, Ed Early, Zuming Feng, Zac Franco, Chris Jeuell, Paul Karafiol,
Leo Schneider, and Eric Wepsic.
Reviewers: David Penner and Bob Perrin.
2003: Don Barry, Paul Dreyer, Ed Early, Zuming Feng, Zac Franco, Chris Jeuell, Paul Karafiol, Chris Odden,
Rick Parris, Leo Schneider, and Eric Wepsic.
Reviewers: Bob Perrin and Sam V andervelde.
Problem Writers for the ARML Power Contest

Thomas Kilkelly has written all the contests save for Color Transformations and Rotating Decimals which were
written by David Miller.
Vlll
ARML Prize Winners,· 1995-2003
1995 Team Competition
Division A Division B
1. New York City A___________________________________ 126 1. Chicago B ___________________________________ 82

2. Texas Gold____________________________________________ 123 2. All Pennsylvania_________________________ 75
3. Thomas Jefferson HS A 119 3. Missouri______________________________________ 71
3. Rhode Island 71
1995 Individual Competition
1. Daniel Stronger_____________________________________ New York City A ________________________________________________________ 7

2. Chris Chang _________________________________________ Northern California A_____________________________________________________ 6
3. Michael Korn ---------------------------------------- Minnesota Gold__________________________________________________________ 6
1. San Francisco Bay Area A_____________________ 179 1. Upstate New York A ___________________ 129
2. New York City A___________________________________ 167 2. New York City B ________________________ 127
3. Thomas Jefferson HS A_________________________ 158 3. Massachusetts E 125
3. Montgomery A_______________________________________ 158
1. Nathan Curtis---------------------------------------- Thomas Jefferson A______________________________________________________ 8

2. Lorentz Shyu ____________________________________ San Francisco Bay Area B ------------------------------------------------- 8
3. Matt Gealy ________________________________________________Howard County--------------------------------------------------------- 7
1. Minnesota Gold_____________________________________ 125 1. Suffolk A____________________________________ 90

2. New York City A___________________________________ 116 2. Howard_______________________________________ 89
3. Texas Gold 114 3. Georgia B____________________________________ 88
1. Davesh Maulik ---------------------------------------------- Nassau A-------------------------------------------------------------- 7

2. Li Chung Chen __________________________________San Francisco Bay Area A------------------------------------------------- 6
3. Mile Colsher Wisconsin Red 6
IX
ARML Prize Winners: 1995 - 2003
1. Massachusetts A____________________________________ I 71 1. San Francisco Bay Area B___________ 138

2. San Francisco Bay Area A·---------------------· 167 2. Phillips Exeter Academy_____________ 122
3. Montgomery County A__________________________ 157 3. Taiwan Yale________________________________ 121
4. New York City Y ________________________ 116
1. Gabriel Carroll __________________________________ San Francisco Bay Area A------------------------------------------------- 8

2. Melanie Wood ------------------------------------------ __ Indiana Gold ____________ --------------------------------------- ________ 8
3. Brian Ginsberg ------------------------------------------ ____ Chicago A ___________________________ --------------------------------- _ 8
1. San Francisco Bay Area A·---------------------· 187 1. Taiwan c ____________________________________ 158

2. Massachusetts A____________________________________ 181 2. Mercer County____________________________ 145
3. Thomas Jefferson HS 178 3. Washington Gold________________________ 140
4. Massachusetts E 137
1. Gabriel Carroll __________________________________ San Francisco Bay Area A------------------------------------------------- 8

2. Reid Barton ______________________________________________ Massachusetts A _______ _____ __________ ________ ___ ______ ________ _________ 8
3. Lawrence Detler ______________________________________ New York City A ------------------------------------------------------- 8
1. Chicago A______________________________________________ 172 1. Connecticut A_____________________________ 127

1. San Francisco Bay_________________________________ 172 I. Peninsula South Bay____________________ 127
3. New York City A___________________________________ 1. 71 3. Iowa A________________________________________ 125
3. Northern California 125
I. Tiankai Liu ___________________________________________ San Francisco Bay A ----------------------------------------------------- 8

2. Gabriel Carroll ______________________________________ San Francisco Bay A ----------------------------------------------------- 8
3. Sasha Schwartz ________________________________________ All Pennsylvania -------------------------------------------------------- 8
X
ARML Prize Winners,· 1995 - 2003
1. San Francisco Bay A_··-··············---····---·· 191 1. Michigan Reals···············----···-···· 160

2. Massachusetts A.·--•-····-·············--··-------· 185 2. Ontario B 1 ··························-··•·-·· 134
3. Thomas Jefferson HS A 175 3. Chicago C·---·····················--····--·· 120
I. Gabriel Carroll -······--•················---···-·San Francisco Bay Area A--------··--···--··-···············--··-----····· 8

2. Gregory Price --·····-··--··--···------···--········Thomas Jefferson HS A·---··---·······-··---------···-··-·-·----···------ 8
I. Thomas Jefferson HS A _____ ···--------··---·-·- 190 I. New York City S·--····-••·············-- 122
2. Massachusetts A ____ ·······------------·····-······· 172 2. Southern California ___________ ··--···•·· 114
3. Chicago A__ ···········-····················-·····••··-· 161 3. Thomas Jefferson HS B_______ ···---·· l 07
1. Ruozhou (Joe) Jia --·-····-·---------------------···-------Chicago A-------------·---···----··-----------------·----····--·------ 8

2. Anatoly Preygel ------------···-··-···-·----------Montgomery County A ······-------------·······-·····----------···-·-··· 8
3. Jeffrey Amos ·-····---····------······---·---------·---·Kansas Regional ····-······--------·----··--·-··----·-·---·--------····· 8
I. Thomas Jefferson HS A-------·-----····----·--- 155 1. Connecticut A ______________ ··-·····---·--- 112

2. San Francisco Bay A_····-------------·-----····-· 153 2. Lehigh Valley _______________ ·-·--··---·--- 98
3. Chicago A·----------·····------------------······--·--- 14 7 3. Missouri A_-----------------·······--------- 95
3. Nebraska 95
1. Anders Kaseorg ···-----------------··---···---··-------_North Carolina-------------··----------·--------------····-------------- 8

2. Jonathan Lowd --··------··--------------··--··--San Francisco Bay Area A·-------·----------------------··---------------- 7
3. Bob Hough ··----------------·············--------------··Michigan Reals -····------------·------·---····-·------------------····· 7
XI
ARML Power Contest Prize Winners
1994-95 1. Vestavia Hills H.S., Vestavia Hills, AL

2. St. Paul Academy, St. Paul, MN
3. Louisiana ARML, LA
1995-96 I. Stuyvesant H.S., New York City, NY

2. Mounds View H.S., Arden Hills, MN
3. North Carolina School of Science and Math, Durham, NC
1996-97 1. Stuyvesant H.S., New York City, NY

2. Vestavia Hills H.S., Vestavia Hills, AL
3. St. Paul Central H.S., St. Paul, MN

2. St. Paul Central H.S., St. Paul, MN
3. Evanston H.S., Evanston, IL
1998-99 1. Lynbrook H.S., San Jose, CA

2. North Carolina School of Science and Math, Durham, NC
3. Evanston H.S., Evanston, IL

2. Gunn Math Circle, Gunn H.S., Palo Alto, CA
2. Wayzata H.S., Wayzata, MN
2000-01 1. Phillips Exeter Academy, Exeter, NH

2. Sofia Math Circle (Bulgaria)
3. Stuyvesant H.S., New York City, NY
2001-02 1. Thomas Jefferson H.S., Fairfax, VA

3. Phillips Exeter Academy, Exeter, NH

3. Phillips Exeter Academy, Exeter, NH
2003-04 1. St. Louis Math Circle, St. Louis, MO

2. Academy for the Advancement of Science & Technology, Hackensack, NJ
3. Honorable Vincent Massey SS, Ontario, Canada
Xll
1995
'T'eam 'R,ound ............................................................................................ 3

Power Qyestion......................................................................................... 8
'lndividua( 'R,ound----------------------------------------------- --- ------- -------------- ------------ 15
'R,e(ay 'R,ounc(............................................................................................ 19
'T'iebreakers 23
-1-
THE 20TH ANNUAL MEET
ARML celebrated its twentieth year of competition by expanding to a new, western site.
The University of Nevada at Las Vegas (UNLV) joined the existing ARML sites at the
University of Iowa and Penn State University on June 2nd and 3rd for the 1995 meet. New
teams from Idaho, Nevada, Utah, and western Pennsylvania joined the ranks, and the
former state team from California was replaced by six teams, three from northern
California and three from southern California. Including five alternate teams, a total of
eighty-five teams and over 1275 students participated. With the addition of the west
coast site, some logistical changes were required. Specifically, the teams at UNLV did
both the Power and Team rounds on Friday night, while the Iowa and Penn State teams
ran the meet as always.
Susan Heicklen of Central Pennsylvania received the Samuel Greitzer Distinguished

Coach Award. She has been a key figure in establishing and maintaining Penn State as a
successful ARML site.
Eric Walstein, the ARML regional representative from the Mid-Atlantic states and coach
of the Montgomery County, MD teams received the Alfred Kalfus Founder's Award. He
has been the coach of Montgomery County for all 20 years of ARML's existence and was
a primary force behind ARML's ongoing exchange with Russia.
The following received the Zachary Sobol award for outstanding contributions to their
ARML teams:
Matt Ahart North Hollywood, CA-organizer of a new team

Franz Ed Boas San Diego, CA-organizer of a new team
Ben Stevens San Diego, CA-organizer of a new team
Lauren Williams South Bay, CA-organizer of a new team
Mathew Crawford Alabama-two year captain

Scott Kempen Wisconsin-captain and 4 year participant
Robbie McNerney Wes tern Massachusetts-captain and six year participant
2
ARML Team Questions-1995
T-1. Let .d fl<;;_ represent a three-digit base IO number whose digits are A, B, and C with A 2> I. Compute the
minimum value of .d fl.<;;_ - ( A 2 + B 2 + c 2 J.
T-2. In MBC, side BC is the average of the other two sides. If cos LC AB , compute the numerical value
AC
of cos LC.
T-3. Compute the number of distinct ways in which 77 one-dollar bills can be distributed to 7 people so that no
person receives less than $ I 0.
T-4. Starting at the same time at comer M, Alison and Ben run in opposite directions around square track MNPQ,
each travelling at a constant rate. The first time they pass each other at comer Pis the tenth time that they
meet. Compute the smallest possible ratio of the faster person's speed to the slower person's.
T-5. Determineallintegervaluesof0 with0° :S 0 :S 90° for which (cos0 +isine/ 5 is a real number.
T-6. Assuming the expression converges, determine the largest integer n with n :S 4,000,000 for which
Jn + Jn + ~ - i s rational.
1
T-7. A trapezoid has a height of 10, its legs arc integers, and the sum of the sines of the acute base angles is 2.
Compute the largest sum of the lengths of the two legs.
T-8. Points A, B, C, and D lie on the given circle.

If AB= 8, AP= 2, and PC= 4, determine the p
ratio of the area of quadrilateral PAEC to the

area ofMAE.
T-9. Points C and D lie on opposite sides of line AB. Let Mand N be the centroids of MBC and MBD
respectively. If AB = 25, BC = 24, AC = 7, AD = 20, and BD = 15, compute MN.
T-10. If log 10 14 = x, log 10 15 = y and log 10 16 = z, then determine the number of elements in
S = {log 10 1, log 10 2, log 10 3, ... , log 10 100} which can be written in the form ax+ by+ cz + d for
rational numbers a, b, c, and d.
3
ANSWERS ARML TEAM ROUND- 1995
1. 27
5
2. -
7
3. 1716
11
4.
9
5. 0°, 12°, 24°, 36°, 48°, 60°, 72°, 84°
6. 3,998,000
7. 441
5
8.
16
39
9.
5
10. 46
4
Solutions to the ARML Team Questions - 1995
T-1. !00A+lOB+C-A 2 -B 2 -C 2 =(IOOA-A 2 J+(10B-B 2 J+(c-c 2 J. Weminimizethesumby

minimizing each term. Since lOB - B2 = (10- B)B ~ 0 for BE {o, 1, ... , 9} with equality iff B = 0, choose
1
B = 0. Since the graph of C( I - C) is a concave down parabola with vertex at C = 2, the graph is decreasing
for C > ½and the minimum occurs at C = 9. Since I 00A - A 2 = A( I 00 - A) is a concave down parabola
with vertex at A = 50, we minimize the expression by maximizing the difference between A and 50. i.e.,
set A = I. Thus, the minimum value of the difference is I 09 - ( I 2 + o2 + 9 2 J = I 09 - 82 = [n].
.
T-2. Usmg the Law ofCosmes, c
. 2
= a2 2
+ b - 2ah cos C = a2 2
+ b - 2abl b
(c) ,., ,.,
a- + h- - 2ac. Since
2
2c(b + c)
b+c
a= - - then c
2 '
2
( b; c) + h2
2
Simplifying, we obtain 7 c 2 + 2bc - 5b 2 =0 ➔
(7c- ShXc + b) = O,so i = ~-
T-3. 1st Method: Assume that every person has received $10 and it remains to distribute $7 among 7 people.
Imagine the seven dollars laid out on a table. The first person puts a marker after collecting her money, the
others do the same except for the last person. A distribution, for example, could be represented by the
diagram and table below:
person 2 3 4 5 6 7
receives $2 $0 $1 $0 $0 $3 $]
Hence, the number of distributions is the same as the number of placements of 7 bills and 6 bars in a line
of 13. Choose 6 locations for the bars: 1 f'6 = I17161.
2nd Method: Let x. denote the number received by the ith person for i = 1, .. , 7. The number of
I
distributions is the number of integer solutions of the equation x 1 + x 2 + ... + x 1 = 77 for xi ~ 10 . Letting
xi - 9 = Y;, we obtain the equation y 1 + y 2 + ... + y 7 = 77 - 7 · 9 = 14 with each Y; ~ I.
5
Setting z 1 = y 1 , z 2 = y 1 + y 2 , ... , z 6 = y1 + ... + y 6 , we then obtain I < z 1 < z 2 < ... < z 6 :5: 13.
The number of distributions is simply the number of six-tuples {z 1, z 2 , ... , z 6 } where each six-tuple is a
combination of 6 integers from I to 13 inclusive. Hence, 13c6 = 1716.
T-4. I st Method: Without loss of generality, let the distance around the square be 400. At the first meeting the
faster person will have traveled 200 + d, the slower 200 - d; at the I 0th encounter the distances will be
2000 + I 0d and 2000 - 10d. The distances arc equivalent to 5 circuits plus or minus 1/2 a lap or 3/2 a lap,
etc. The speeds are most nearly equal if the faster person has traveled 5.5 laps and the slower person 4.5 laps.
. o f thcir
T h e sma 11 est possible rat10 . speeds 1s
. -5.5 = ~11
- .
45 9
2nd Method: Let MN+ NP = I, let x be the distance traveled by the slower person between consecutive
meetings, let 2 - x be the distance traveled by the faster person. Then O < x < I and x must satisfy
9
!Ox = 2k + I for some integer k. The largest possible x satisfying this equation is x =- and the ratio
10
. 2-x 11/10 II
IS - - = --
X 9/10 9
T-5. (cos 0 + isin 0) 75 = cos(750) + isin(750). This is real iff sin(750) = 0, meaning that 750 = 180k for
intcgra I k. Thus, 0 = -12k and this. implies

. that k.1s d'1v1s1b
.. 1e by 5. Let k Sm, then 0 12m and so we have
5
the following values for 0: I 0°, 12°, 24°, 36°, 48°, 60°, 72°, and 84°1.
T-6. Let x = Jn + Jn + ~ , then x 2 = n + x, so n = x 2 - x for x rational. But since n is an integer and
n = x(x - I), then xis an integer as well and if n is to be as large as possible but less than or equal to
4,000,000, then x = 2000, making n = 2000 · 1999 = 13,998,0001-
T-7.
. . . 10 10 I +
Smee sm0 + sma = - +-,then - = 10 - - g1vmg
(n mj ..
m n 2 n·m
m 10 n
mn - 20n - 20m = 0 ➔ mn - 20n - 20m + 400 = 400.
'L.
This factors as (m - 20Xn - 20) = 400. The sum (m + n) is e a
maximized when one of the factors equals 400 and the other equals 1. Set n - 20 = 400, giving n = 420 and
m -20 = I, giving m = 21. Thus, m + n = 14411.
6
T-8. In this problem we will use the expression a(MA £) to represent the area of !iBAE.
PA· PB= PC· PD ➔ 20 = 4 -PD ➔ PD = 5 and DC= I. Since MAE - !iDCE, then
a(MAE) AB 2 a(MBC) PC 2
- - - = - - = 64. Since MBC - MDA, then----= - - = 4. Let a(PAEC) = K
a(MJCE) DC2 a(!iPDA) PA2
and a(liBAE) = 64N, making d._!iDCE) = N. Since a(!iPBC) = a(PAEC) + a(!iBAE) and
a(!iPBC) K + 64N 60N
a(!iPDA) = d._.PAEC) + d._!iDCE), then - - - = - - - = 4, making K = - - .
a(!iPDA) K + N 3
a(PAEC) (60N)/3
So, = ~-
a(MDE) 64N
T-9. Since 20 2 + 15 2 = 25 2 = 24 2 + 7 2 , both MBC and MBD C 7

are right triangles, so C and D lie on the circle with diameter
AB and center P. By Ptolemy's Theorem,

(AC)(BD) + (BC)(AD) = (AB)(CD), making
7 · 15 + 20 · 24 117
CD = ----- = - . The centroids Mand Narc the
25 5
intersection points of the medians and hence, Mand N trisect
CP and DP respectively. MN II CD and since
MP NP l CD = '35919.
PC = PD = 3' MN 3 W
.
rcwntten as n = 14a b
• 15 · 16 · 10
c d
= 2
a+ 4c+d h
·3 · 5
h+ cl a .
· 7 . Son has no pnmc factors other than 2,
3, 5, and 7. Conversely, ifn has the form 2P · 3q ·Sr· T~ for some non-negative integersp, q, r, ands. then
we can always find rational a, b, c, d such that p = a + 4c + d. q = b, r = b + d, ands = a; i.e., we solve
. . p-s+q-r b f
these equations and obtam a= s, b = q, d =r - q, and c = .....___....__. So the num er o
4
logs that can be written as a sum of the given logs is simply the number of elements in { 1, 2,3, ... , 100} that
havenoprimcfactorsgreaterthan7. Compute 100- lOOj + llOO]

[lU 13 + ... + llOO]l
97 whcrc[x]standsfor
the greatest integer function or write down all positive integers from I to I 00, count l and all those with
factors of only 2, 3, 5, 7 to obtain~-
7
ARML Power Question - 1995
Consider a collection of piles ofbananas consisting of one pile of9 bananas, one pile of6 and one pile of 2. Such a
collection, C 1, could be expressed as (9, 6, 2). Obtain a new collection, C 2 , by harvesting C 1 where harvesting is
defined to mean remove one banana from each pile to form a new pile. Thus, c2 = (8, 5, 3, I) and, if c2 is
harvested, we obtain c3 = (7, 4, 4, 2).
I. a) Let c1 (8, 5, 2). Determine C 2 , C 3 , c4 , and C I Off
b) Given the collection M 1 (7, 6, 5), determine M 1995 . Explain how you arrived at that result.
2. Prove that the collection ( k, k - I, k - 2, ... , 3, 2, 1) remains fixed after harvesting.
3. Prove that any collection of piles whose total number of bananas is 6 can be reduced to the collection
(3, 2, I) by successive harvesting. Determine the maximum number of harvests required.
4. Denote by c( nk, nk- l' ... , n2, n1) a collection of k piles of size nk through n1 ordered such that
nk 2'. nk _ 1 2: ... 2: n1. Suppose nk + nk- I + ... + n 1 = 7 . Determine the number of collections C such
that the 1995th harvest yields the collection (4, 2, I). Justify your answer.
5. Let C = (n), that is, let collection C be one pile of nbananas. If C is harvested successively, the collections
will eventually fall into a periodic sequence. That is, for all k sufficiently large, Ck = Ck+ P. If the harvest
is fixed after some point, then p = I. Determine all values of n such that p = 1995. Explain your reasoning;
you need not prove your result.
8
ARML Power Ouestfon - 1995
Define the graph of a collection as follows: write the collection of piles in

4 ~
non-increasing order from left to right, then draw a column of squares for each
3
pile. The graph of(4, 2, 2, 1) is shown at the right:
2
- - ~
I
1 2 3 4
Let the level of a square denote the sum of the coordinates of its center. Let the level sum of a collection be the sum
of all the levels of the squares of the collection.
6. Does there exist a collection of piles different from (n, n - 1, ... , 3, 2, I) such that after harvesting once,
the level sum remains the same?
7. Find the level sum of C = (n, n - l, n - 2, .... , 3, 2, I).
8. Show that the level sum of any collection does not increase following a harvest.
. . . . r(n + I)
9. Prove that 1f a collection of piles con tams 2 bananas for n > 2, then successive harvesting will
eventually lead to the collection (n, n - l, ... , 2, I).
. . . . r(n + 1)
10. Prove that a collection contammg one pile of ......;..____ bananas will be reduced to the collection
2
r(n - 1)
(n, n - I, .... , 3, 2, 1) in exactly - - - harvests.
2
9
Solutions to the ARML Power Question - 1995
harvest is fixed or repeats with a period of I.
lb. M 1(7,6,5) ➔ M 2 (6,5,4,3) ➔ M/5,4,4,3,2) ➔ Mi5,4,3,3,2,I) ➔ Ms<6,5,3,2,2,l) ➔
M 6 (6,5,3,2,l,l) ➔ M 7 (6,5,4,2,l) ➔ M 8(5,5,4,3,l) ➔ M 9 (5,4,4,3,2) ➔ M 10(5,4,3,3,2,l). Since
M9 = M 3, harvesting (7, 6, 5) is periodic with period of 6 beginning with M 3 , making M 3+ 6k = M3 .

Since 1995 = 3 + 6(332), M 1995 = M 9 = (5, 4, 4, 3, 2).
2. To harvest (k, k - 1, ... , 3, 2, 1}, we reduce each pile by 1, obtaining piles of size k - I, k- 2, .. , 2, I, 0, and
one new pile whose size equals the number of piles, namely k. This yields the original collection
(k, k- 1, ... , 3, 2, I). Thus, the process terminates whenever (k, k - I, ... , 3, 2, I) is reached.
3. The following diagram represents all collections of piles with 6 bananas and the harvest of each. The
maximum number of harvests required to reach (3, 2, 1) is 6.
(2, 2, 1, 1)
J,
(4, 1, 1)
J,
(3, 3) (1, 1, 1, 1, 1, 1)
t ~ (6)
(2, 2, 2) (2, 1, 1, 1, 1)
>
(3,/1,1)
(3, 2, 1)
~(5,1)~
(4, 2) ~----------
10
4. There are l 5 collections with 7 bananas. If harvested successively all yield (4, 2, 1). Since (4, 2, I) ➔
(3,3, I) ➔ (3,2,2) ➔ (3,2, I, I) ➔ (4,2, l),harvestinghasaperiodof4once(4,2, !)has been reached.
If a collection yields (4, 2, I) on the 3 + 4k harvest, then it will yield (4, 2, I) on the 1995th harvest. There
are 4 collections with that property: (7), (3, 3, I), ( 4, 3), (2, I, I, I, I, I). The following diagram shows all
collections of piles with 7 bananas and the harvest of each:
(1, 1, 1, 1, 1, 1, 1)
✓
(7)
(2, 1, 1, 1, 1, 1) ~ 1 (2, 2, 1, 1, 1) (2, 2, 2, 1)
(3, I, I, 1, 1) ~
(6, 1)
t l
(5, 1, 1)
l
(4, 1, 1, 1)
(5, 2) ~✓
l /(3,3,1)~
(4, 3)
t
(4, 2, 1) (3, 2, 2)
~ (3, 2, 1, 1) ~
5. As shown in #3 and #4, the period (6) yields is I, the period (7) yields is 4. We form a table
showing n and p:
n 2 3 4 5 6 7 8 9 10 11 12 13
p 2 3 3 4 4 4 5 5 5
k(k+l) . . k(k+I)
Whenn=l,3,6,10, .. 2 ,thepenod1sl.Forn= 2 +mfor0<m:Sk,theperiod
1994 · 1995
appears to be k + l . If the period is 199 5, then k = 1994 and n will range from - - - - + I to
2
1994 · 1995
- - - - + 1994, that is, from 1,989,016 to l,991,009 inclusive.
2
11
6. Consider harvesting a collection of I pile of two bananas: (2) ➔ ( l, I). The level sum for (2) is
(.!.2+2.!.J + (.!.2+2~J = 3 and the level sum for (I, I) is (.!.2+2.!.J + (i

22
+ .!.J = 3.
7. Clearly all the squares in the graph of Care of a level less than or equal to n, and for each k = I, 2, .. , n
2 2 2
the graph contains exactly k squares of level k. Thus the level sum of C = I + 2 + ... + n =
rt,n + I)( 2n + 1)
6
8. Consider a harvest without reordering, i.e., let the pile corresponding to the number of piles be listed first.
Thus(6,5,4) ➔ (3.5,4,3).
a
~
b g a ~ a
C h b g b g b ! g
- -------3' - - - -~
d m C h I 0 C h I
e J n d i m k d i m m
f k o e j n f e j n j n;
f k 1
I
o
The center of each square is shifted one to the right and one down. Thus the sum of the coordinates of the
centers of the squares above the x-axis remains constant. Suppose there were r piles before harvesting.
Before harvesting, the sum of the x-coordinates of the centers of the first row of squares equals
1 3 2r - 1 r2 1
-2 + -2 + ... + -2- = -2· The sum of the corresponding y-coordinates equals r • -2 . If the first row
becomes the first column, then the sum of the x-values equals r • ¾and the sum of they-values equals
l 3 2r - 1 r2
2 + 2 + ... + - 2- = 2 . Thus, the level sum is unchanged. lf the first column contains at least as many
squares as any other column, the harvest is properly ordered and the level sum is the sum as before
harvesting. If the first column contains less than the second or third, etc., as in the diagram above, then order
the collection by shifting squares to the left. In this, case squares [QI and [Ji] are shifted to column l and square
[Kl is shifted to column 2. A shift to the left reduces each x-coordinate of the center by l, thereby decreasing
the level sum. Thus, the level sum is constant or decreases.
12
• . . . . 11(11 + 1)
9. Let S = (n, n - I .... , 2, I). Suppose that there exist collectJOns conta111111g - - - bananas that do not
2
lead to S by successive harvesting. Let C be such a collection with the additional property that it has the
minimal level sum of all such collections. Since C does not lead to S, no collection derived from C by
harvesting does either. Additionally, the level sum of C cannot decrease by harvesting since C is supposed to
have minimal level sum. Consider the proof of#8: the only case in which the level sum of C remains constant
under harvesting is that in which the first harvested column contains as many squares as any other column. In
that case, the individual labeled squares move either one square SE or the maximum number of squares NW as
when, for example, the center of d moves from ( i

¾. J to ( %,½Jto ( ¾, "iJ. Under this movement, the
squares cycle with period equal to their respective level; they move along diagonals and since the level along a
diagonal is constant, the level of each square remains constant.
~
~
a f cl
h d e a h f
C e II C h cl C e a I
With these observations in mind, we derive a contradiction. Since C t (n, 11 l, ... , 2, 1), C is not equal to
r(n + 1)
the set of squares with level less than or equal ton. But the number of elements in C equals - - ~ so C
2
must have an element y not included in Sand S must have an element x not included in C.
Intuitively, y is an extra square and xis a hole. Shown is C = (3, 3, 3, 1) and S = (4, 3, 2, I).
X
------·---·
.V ''
If the level of y is greater than n + I, there is some square y' below yin C with level exactly n + I. But y' is
not an element of S because the level of y' is greater than 11. Similarly, if the level of xis less than 11, there is a
square x' of level n above x also not in C but in S because x' has level n. By the logic of #8, under successive
harvesting y' rotates with period 11 + I, and x' rotates with period 11. After n + I harvests, x' moves one square
SE, y' stays where it is. Therefore, after at most 11(11 + I) harvests, y' is directly above x'. This means a square
of C is directly above a hole of C. This is a contradiction.
13
. k(k+I)
We prove by induction that after j = 2 + m harvests where m :S k, our collection consists of:
,i.._n + 1)
one pile of 2 - j bananas
m piles of k + I, k, ... , k - m + 2 bananas
k- m piles of k m, ..... 1 bananas.
. . . _,i.._n+l)
The base case,j = 0, is obvious, since ifj 0, then k m 0, y1eldmg one pile of - - - ' - bananas.
2
We divide the inductive step into three cases: i) m 0, ii) 0 < m < k, iii) m k.
i) If m = 0, we begin with I pile of r(n 2+ I) - j bananas and k piles of k, k - I, ... , I bananas.
Harvesting yields 1 pile of ,(\+ l) - j - I, k- l piles of k - I, k- 2, ... , 1, and I pile of k + I bananas.
Rewritten as I pile of r( n + I) - U + I) , k - 1 piles of k - I, ... , I, and I pile of k + I. we see that

2
m = I, k is unchanged andj is increased by 1.
ii) If 0 < m < k, harvesting yields I pile of r(n 2+ I) - j - I bananas, m piles of k, ... , k- m + l bananas,
k- m - I piles of k- m - I, ... , I bananas and 1 pile consisting of I + m + k- m = k + I bananas. This

gives m + I piles of k + I, k, ... , k- m + I bananas. This is exactly as expected,! and m both being increased
by 1.
iii) lf m = k, harvesting yields I pile of ,(\+ I) - j - l bananas and m + I piles of k +I, ... , l
bananas. Again,j and m are increased by 1.
,i.._ n - I) . . . r( n + l) n( n - I)
After - - - harvests, k = n - I and m = 0. This results m I pile of - - - - - - - =n bananas,
2 2 2
zero piles of k + I, .. , k- m + 2 bananas, and n - I piles of n - I, .. , I. This yields the desired collection:
(n, n - l, ... , I).
14
ARML Individual Questions - 1995
1-l. Compute the largest prime factor of:
1-2. Triangle ABC is similar to 11MNP. If BC 60, AB = 12, MP = 8, and BC > AC > AB,
compute the sum of all possible integer values for the ratio of the area of MBC to the area of 11MNP.
I- 3 . . a II positive
D etermme . . pnmes
. p sue h th at p 1994 + p 1995 1s
. a per1ect
,.. square.
1--4. Let . A = .AAAA .... Compute the number of distinct ordered triples ( A,B, c)
with /4 B, C E {o, l, 2, ... , 9} such that k is an integer if
k = _.A_B_C_+_._A_C_B_+----=.B=-A_C--=+,,_._B-=C=-A_+_._C_'A_B_+_.C_B_A
.A+.B+.C
1-5. Compute the number of distinct planes passing through at least three vertices of a given cube.
1-6. Determine all positive integers k ::S 2000 for which x 4 + k can be factored into two distinct trinomial factors
with integer coefficients.
1-7. Let f(x) = ~ + c. Determine all real values of c such that the graphs of f(x) and its inverse f- 1(x)
intersect in two distinct points.
1-8. For x and y in radians, compute the number of solutions in ordered pairs (x, y) to the following system:
sin(x + y) cos( x + v)
15
ANSWERS ARML INDIVIDUAL ROUND- 1995
1. 73
2. 930
3. 3
4. 999
5. 20
6. 4,64,324, 1024
7. -2.25 < c ~ -2 or its equivalent ( -2.25, - 2]
8. 3990
16
Solutions to the ARML Individual Questions - 1995
312 - I
1-1. The given expression equals 3 11 + 3 1 O + 39 + ... + 31 + I The numerator factors as
3- I
(3 6 + 1)(3 6 - 1) = (3 2 + 1)(3 4 - 32 + 1)(3 3 -I )(3 3 + 1) = 10 · 73 · 26 · 28. The largest prime factor
is[TI].
l-2. Clearly 12 <AC< 60. Bythetrianglcincquality,60-12 <AC< 60+ 12 ➔
48 < AC< 60. Thus, -

( 48 )
8
2
<
area of M BC
area of /'J.MNP
60)
< .-
8
l 2
➔ 36 <
area of MBC
area of /'J.MNP
< 56.25. Thus, the
(37 + 56)20
ratio of areas lies in the set {3 7,38, ... , 56}. The sum of all such ratios equals 2 = 19301.
1-3. .
Inthe genera I case, smce p 2n + p 2n + I = p 2n( I + p ) = k 2 1or
c k an mtegcr,
• clearly 1 + p must be a perfect
square. Setting I+ p = t 2 ➔ p = t 2 - 1 = (t + l)(t - 1). This is prime if and only if t - I = 1, making

t = 2. Thus p = [].
. 100 A + I OB + C
1--4. Smee . ABC = 999 and A, B, and Care each in the hundreds, tens, and ones place twice in the six
222A + 222B + 222C 2A +2B+ 2C A

decimals, the numerator equals - - - - - - - - = Since . A = 9 , the denominator
999 9
equa1s A + B9 + C . Thus, the quotient

· . . Iy equa Is 2 fior aII tnp
surpnsmg . Ies o f A, B, an d C except (0 , 0 , 0).
The number of triples equals ( 10)( 10)( 10) - 1 = 19991.
1-5. The planes are of three types: 1) Face planes such as DCGH, I for each face for a total of 6.
2) Opposite edge planes such as BCHE, 1 for each pair of opposite edges for a total of 12/2 = 6.
3) Face diagonal planes such as ACH fom1ing the base of pyramid D-ACH. Each comer can serve as a
vertex of such a pyramid, so there are 8 such planes. Total: 6 + 6 + 8 = [2m.
B C B C B C
D
A A
A~ G G G
E H E H E H
17
1-6. Note that x 4 + 4 = x 4 + 4x 2 + 4 - 4x 2 = (.x-2 + 2 r-

1
(2x) 2 . The latter expression factors as
( .x2 - 2': + 2 )( x 2 + 2x + 2) . Hence, set x 4 + k = ( .r2 + n t- 2nx 2 = ( x 2 + n f- (xFnt
( x2 - £n + 11)( x 2 + - ~ + n ). Since .Ji;, must be an integer and 11 2 = k, then J2./k must be
4r 44444 .
integral. Let k = 4m . Then 2-t m must be integral. Thus, m = I , 2 , 3 , 4 , 5 , ... makmg
k = 4,64,324, 1024,2500, ... Choosel4,64,324,1024'.
1-7. /-I (x) = (x - c) 2 - 2. The graphs off and f-l are monotonically increasing and are symmetric over
y = x. Thus, they intersect on y =x and the number of intersections equals the number of solutions to
f(x) = x. Set (x - cl - 2 = x ➔ x 2 - (2c + l)x + c 2 - 2 = 0. There arc two solutions if the
discriminant is positive. Thus, (2 c + 1) 2 - 4( c 2 - 2) > 0 ➔ c > -2 25. As c increases the number of
solutions goes from Oto I to 2 to 1. The transition from 2 to 1 occurs when c = - 2. Thus, 1-225 < c < -21.
1-8. sin(x + y) -cos(x- y) =0 ➔ (sinx ·cosy+ cos x · siny)-(cos x ·cosy+ sinx · siny) =0 ➔
(sinx)(cos y- siny) - (cos x)(cos y - siny) =0 ➔ (sinx - cos x)(cos y - sin y) =0 ➔ sinx = cos x or
1t 1t . l 9951t
sm y = cos y. Therefore, x = 4 + n1t or y = 4 + 111t for nan integer. The radius of the circlc is - 4-
. 11·mes at x
For x c:: 0, t11ere are vertlca = -1t , -51t , ... ,-
l 9931t
- ior a tota 1 o f 499 vert1ca
c: . 11·mes mtersectmg
. . th e
4 4 4
circle at two points, making 998 points of intersection. Similarly for y c:: 0 there arc 499 horizontal lines
. 3n: 7n: 19957t
intersecting the circle at 998 points. For x < 0 there are vertical lmes at x = - - , 4 ' ... ,
4 4
There arc 1995 - 3 + 1 = 499 vertical lines, 498 of which intersect the circle in 2 points, the other is
4
tangent. Thus, there arc 997 points of intersection. Similarly, for y < 0 there are 997 points of intersection.
The total number of points of intersection equals 2(998) + 2(997) = 139901.
18
ARML Relay #1 - 1995
RI-I. Compute the number of distinct 11 letter palindromic permutations of the letters of MISSISSI PP!.
K
Rl-2. Let K = TNYWR. If the regular octagon ABCDEFGH has sides equal to - in length, compute the value
30
2
of AC - AD.
R 1-3. Let M = TNY WR. Let S be a unit square subdivided into 2 211
n
congruent squares. The centers are all connected as shown in the ' '
.--
----,'' --- .----
'' ''
'
-
' '
- --~---~----4---- '
diagram at the right for s2 . Compute the least value of n such that the , ' '
' ' '
' '
----►--- ➔ ----+--' -
sum of the lengths of the connecting line segments exceeds 1o4 M . ' '
ARML Relay # 2 - 1995
R2-l. LetABCDEFbe a regular hexagon. lfthe ratio of the area of region ABC£ to the area of the hexagon in
simplest terms is '7;, compute a+ b .
R2-2. Let K = TNYWR. Compute the number of sets of K consecutive integers in { I, 2, 3, ... , 50} such that the
product of the elements of each set of K integers is divisible by both 12 and 21.
R2-3. Let M = TNYWR. Given points D(O, M), 0(0. 0), and E(M, 0), equilateral triangle ABC is inscribed in
MJOE such that C is the midpoint of DE and AB is parallel to DE. The length of AB can be expressed
in simplest form as a( ./i - /;). Determine ( a, b, c).
19
ANSWERS ARML RELAY RACES- 1995
Relay #1:
Rl-1. 30
Rl-2. 1
Rl-3. 14
Relay #2:
R2-l. 5
R2-2. 32
R2-3. (16, 6, 2)
20
Solutions to ARML Re/av #1 - 1995
RI - I. M must lie in the 6th position. To its left arc permutations of SSIPP, to its right are mirror images of letters
on its left. Thus, Mand letters to its right are fixed . The number of permutations is 5! = 00.
2 !·2 !·l !
Rl - 2 . Let :o ~ x . Then AC 2 ~[ x + ;J + [;;]'

A
'
'
~
'
X
'
B x/✓2
- - - -1
~~•
:x1✓2
I
C
'
.Ji'[x+ Ji,) =~.[2' + 1). SinceK '
Also, AD= = 30,thenx = I so
X
AC 2 - AD D
I
Rl-3. There are 2 n horizontal line segments of length I - - . The sum of the vertical line segments equals a
2n
1
horizontal line segment which equals I - - . The sum of the vertical and horizontal lengths equals
2n
( 2n + 1)[ 1 -
2
1n J= I "" 2 n . Since M = I then we want to find n such that 2 n > IO 4 . We
2 n - :;;
IO I3 14 ,,---;;,i
have 2 = 1024, so 2 = 8192 and 2 = 16384, son = LHJ.
Solutions to ARML Relay #2- 1995
area of ABCE 4 2 A B
R2- l. From the diagram, - - - - - - - - - - The sum
area of ABCDEF 6 3
equals 2 + 3 = 13].
F
E D
R2- 2. Since K = 5, we seek the number of sets of five consecutive integers in S = { 1, 2, ... , 50} such that the
product of the elements of each 5-tuple is divisible by 2 2 · 3 · 7. Any five consecutive numbers is divisible
by 4 and 3 so each of the desired 5-tuples must contain a multiple of 7. For each multiple of 7 from 7 to 42
there are five 5-tuples in S containing the multiple of 7 in Sand for 49 there are two such 5-tuples in S. Thus,
there are 6 · 5 + 2 = [TI] sets of five integers the product of whose elements is divisible by both 12 and 21 .
21
Solutions to ARML Relay #2 - 1995
R2-3. Let OA = OB= x giving AB = AC = x.Ji.. Draw OC 1-DE. Since D
. AC AD
mLACD = 60°, then by the Law ofSmes, -- =- - ➔
sin 45° sin 60° M
AD sin 45°
AC=----
AM
Ji . Since AD= M - x, then
sin600 E
M(./3 - 1)
AC= x.Ji.. Solving for x, we obtain x = 2 making AC = x.Ji.
M(-16 - Ji)
2 . SinceM = 32,AC = l~iti -Ji.). Since AC= AB,theansweris (16,6,2)1-
1
22
ARML Tiebreakers- 1995
I. The center and radius of circle Mare M(O, 0) and 10 respectively. The center and radius of circle P arc P(8, 0)
and 2 respectively. A circle with center Q is internally tangent to circle Mand externally tangent to circle P.
Compute the maximumy-value of the locus of points Q.
2. A square is inscribed in a circle of radius I. Circles P and Qare the largest circles that can be inscribed in the
indicated segments of the circle. The line joining the centers of circles P and Q intersects the square in points
A and B. Compute the length of AB.
-----c- ~\------
,• P
23
ARML Tiebreaker Solutions - 1995
I. Since the radius of circle Q = IO - MQ = PQ 2, then
MQ + PQ = 12. This implies that the locus of Q is an ellipse with
focal points Mand P. Thus, the equation of the locus of Q is
(x - 4)2 v2
- - - + -·- = I . If x = 4, then y = w20 = 2y 5 .
C: I r:1
36 20
2. Half the side of the square is ~ . Thus,

Ji. 2--J2
]--
CP = DQ = 2 = - 4- = AC = BD. This
.
gives AE =EC-AC= -
./2. - 2-/2 3/2-2
--- = ---.
2 4 4
Since AB
r: · AE, then AB
= y 1. =
~
CL.J-
2- ,.
24
'Team 'R,ou n/_____________________________________________________________________________________________ 27
Power Qyestion _________________________________________________________________________________________ 32
'lndividua[ 'R,ound-------------------------------------------------------------- ------------ --------- 41
'R,e [ay 'R,oun/_____________________________________________________________________________________________ 45
S~er 'R,e[ay_________________________________________________________________________________________________ 48
'Tiebreakers 53
-25-
The 2F' ANNUAL MEET
ARML celebrated its twenty-first year of competition with its largest competition ever.
The second year of the University of Nevada at Las Vegas site saw an increased presence
of strong California teams and the first western winner ever. New teams from
Washington and California Davis/Sacramento joined the ranks. The former Northern
California teams were replaced by three teams from the San Francisco Bay area and the
Metro New York teams were replaced by the Westchester, Suffolk, and Nassau teams.
We were again honored with the presence of a team from Russia. Including eight
alternate teams, a total of eight-nine teams and over 1335 students participated.
Kay Tipton and Dorothy Wendt of the Alabama team received the Samuel Greitzer
Distinguished Coach Award. They established the team in 1988 and have finished in the
top 15 every year except 1992 when the team couldn't make the trip.
Barbara Rockow received the Alfred Kalfus Founder's Award. Barbara has been a long
time ARML supporter and has been the corresponding secretary as well as a force behind
the scenes these past 12 years.
Michael Colsher of Wisconsin received the Zachary Sobol Award given for outstanding
contributions to his ARML team. Michael helped organize and train his team.
26
ARML Team Ouestions-1996
T-1. A cylindrical container IO units high and 4 units in diameter is partially filled with water. The cylinder is
tilted so that the water level reaches 9 units up the side of the cylinder at the highest but only 3 units up at the
lowest. Compute the volume of water in the cylinder.
T-2. Imagine 173 unit squares arranged in a row, 6 of which are shown. If a rectangle consists of a single square,
or a combination of consecutive squares as indicated by the shaded region, compute the total number of
rectangles that can be formed.
. 73 37
T-3. Let Q( x) be the quotient when 3 7 x - 73x + 36 is divided by x - I . Compute the sum of the
coefficients of Q( x).
T--4. A positive integer N is self-descriptive if each digit of N appears as often as the value of the digit. For
example, 212 is a three-digit self-descriptive number. Compute the sum of the digits of all the six-digit
self-descriptive numbers .
T-5. Let M = I I ..... I 2 where Mis a base l O integer consisting of 299 l's followed by a 2 in the unit's place.
Let N be a positive integer. If the product (M · N) has K times as many digits as N,
compute the largest possible value for K.
T-6. Given two concentric circles with center 0, points A and B lie on the inner circle, point Plies
on the outer circle. PA and PB are tangent to the inner circle at A and B respectively.
If tanLAOB - tan LAPE = -8/3, the ratio of the radius of the large circle to the radius of the small circle
can be written as ix where x is an integer. Compute x.
T- 7. An icosahedron is a regular polyhedron whose faces are 20 congruent equilateral triangles. Moving on the
edges of an icosahedron, compute the number of shortest paths from a given vertex A to the vertex opposite A.
T- 8. Distinct lines L 1 and L2 with positive slopes m1 and m 2 respectively, pass through P(3,IO).
The area of the triangle formed by P and the x-intercepts of L 1 and L2 equals 200.
If m1 :S .24, compute the number of integer values m 2 can take on.
A
T-9. In MEC, AB= AC= 115, AD= 38, and CF= 77. In simplest form the
ratio of the area of !1CEF to the area of ADEE can be written as !!!.... .
n
Compute the ordered pair (m, n).
T-10. Let N (x) = [ [ [ [ [ x ] + 6x] + l 5x] + 65x] + l 43x] where the brackets represent the greatest integer
function. If the domain of N is {x: 0 :S x :S 1 } , compute the number of values in the range of N.
27
ANSWERSARML TEAMROVND-1996
I. 24rr
2. 15051
3. 0
4. 1332
5. 301
6. 5
7. 10
8. 6
9. (19, 96)
10. 209
28
Solutions to the ARML Team Questions - I 996
T-1. Since DE = 3, FC = 9 and AE = FB = x, then

3+ x = 9 - x, making x = 3, so that the height AD of
A
the water is 6. Thus, the volume= n( 22 ) · 6 = l241tl.
2nd Method: Imagine that the water in the tilted cylinder is frozen. Form a congruent solid and join the two
to form a cylinder of diameter 4 and height 12. Its volume is 481t, so half of 481t is the answer.
T-2. A rectangle is formed when any 2 of the 174 vertical segments are chosen. Thus, 174c 2 = I150511.
T-3. Let P(x) = 37x 73 - 73x 37 + 36 = 37( x 73 - 1)- 7\ x 37 - 1) which equals
37( x - 1)(x 72 +x 71 + ... + I)- 73(x - 1)(x 36 + x 35 + ... + I). Thus, Q(x) equals
31( x 72 + ... + I) - 73 ( x 3 6 + ... + I). The sum of the coefficients of Q( x) is Q( I) which equals
37(73)- 73(37) = [Q]. Note: The original problem involved 19x 96 - 96x 19 + 77, but we chose to suppress
our calendrical instincts.
6!
T-4. There is l self-descriptive number with six 6's giving a sum of 36. There are 5 !_1! = 6 numbers consisting
6!
of five 5's and one 1 giving a sum of6(26) = 156. There are 4 !- 2 ! 15 numbers with four 4's and two 2's
. . a sumo f 15( 20) = 300, an dth ere are - -

g1vmg 6 ! - = 60 numbers wit
"hht ree 3's, two 2's, an d one I,
3 !-2 !-I!
giving a total of60(14) = 840. Answer: 36 + 156 + 300 + 840 = 113321.
10300+ 8
T-5. SupposeNhasndigits,making N < 1011 . WriteM= - - - - IfMNhasKndigits.then
9
IOKn-l s NM. Thus 9- IOKn-l s (10 300 +8} N < (10 300 +8)·10 11 = to 300+n + 8· IOn_ It
301
follows that Kn - I < 300 + n, making K < - + I. K is clearly maximized when n = I and N = 9.
n
making K = 1301 I.
29
8
T-6. Let mLAOP = a, then tan2a- tan(l80°- 2a)
3'
8 2 tan a 4 A
giving 2 tan 2a ➔
3 I - tan 2 a 3
p
2
2 tan a - 3 tan a - 2 = 0 ➔ (2 tan a + I)( tan a - 2) = 0.
Choose tana = 2. Without loss of generality, let AP = 2,
OA = 1, making OP = ./s. The ratio of OP to OA = ./s.

Ans: BJ.
T- 7. Consider the icosahedron labeled as indicated at the left below. There is I way to get from A to one of the 5
vertices C which are one edge-length away from A. There are two ways to get from each C to each vertex D,
one edge-length away. From each vertex D there is I way to get to B. Thus the number of shortest paths
along the edges from A to B equals 5 · 2 = [IQ_]. Shown to the right is a different way to visualize an
icosahedron that could be used to solve the problem.
C D
10
T-8. In general, y- 10 = m(x - 3) implies an x-intercept of 3 - - . The base of the triangle equals
m
3 - -10 - [ 3 - - 3] = 10 - I - - 1 . Thus, 200 equals -1 · IO · I 0 giving

m2 m1 m 1 m2 2
m2 -ml
4 = . Since m 2 is an integer and m 2 > m 1, we obtain 4
m 1m2
. m2
we obtam = m1 . Since m1 ::; 0.24, solving for m2 yields m2 :S 6. Hence m2 = I, 2, .. , 6.
4m 2 + 1
The answer is ~-
30
Solutions to the ARML Team Questions~ 1996
T-9. Draw DT parallel to BC, let EC= x, BE = y. Since DB = TC= CF, A
AD AB 38 115
then DT = 2x. Since - - we obtain - Solve
DT BC 2x X + )'
96x
for y and obtain y = - . Since BC bisects the altitude from F to DT,
19
the altitude from D to BE equals the altitude from F to EC,

consequently the ratio of the area of /'J.CEF to the area of /'J.DBE equals
EC• X 19. Ans: ~ )
= - 19,96.
BE y 96
BD AF CE 19 . .
2nd Solution: By Menelaus' Theorem: - I which yields CE 96 by substitution.
DA FC BE BE
AC FE BD . .
Also by MT we have CF · ED · BA = 1, y1eldmg FE = ED.
1 .
a(/'J.FCE) 2 CEXEF) sm LCEF
CE 19
Thus, - - -
~/'J.DBE) BE 96
(½JBE)( ED) sinLDEB
T-10. first,notethatforpositiveintegernandrealnumbcrr,[n+r] = n + [r]. Applied to [[[x]+6x]+15x] we
obtain [[x]+[6x]+ 15x] = [x]+[6x]+[15x]. Applied to N(x) we obtain
N(x) = [x] + [6x] + [ 15x] + [65x] + [ 143x]. For positive integers a and b, expressions [ax] and [bx] increase
1 1
in value by I at intervals of:; and b
respectively, taking on integer values from Oto a and Oto b on
0 :-S x '.S I. We would expect N(x) to take on the 231 values from Oto I + 6 + 15 + 65 + 143 = 230. but
whenever x = p withp and q relatively prime and q dividing both a and b, [bx] will cause an additional
q
3
increase of 1, thereby causing Nto skip an integer. For example, 13 divides both 65 and 143. If x = 13 . both
[65x] and [143x] increase by 1, causing N( x) to increase by 2. We say that [143x] causes an additional
increase or skip at x = 1~. So, [6x] causes a skip atx = 1, [1sx] causes a skip at
x = i,¾ and 1, [65x] causes a skip at x = i ,¼, i ,1, and 1, and [143 x] causes a skip at
1 12
x = 13 , ... , 13 , and 1. There arc 22 skips so N takes on 231 22 = !2091 values.
31
Let the number I be placed at each end of a stick, forming list L1 as shown:
List L11 + 1 is formed from list L11 for 11 ::::: I by inserting between every pair of consecutive terms ai and ai + 1
in L11 the sum a.I + a.+

I 1. Lists L2 , L3, and L4 are shown:
L2: 2
L3: 3 2 3
L4: 4 3 5 2 5 3 4
I. a) Prove that every positive integer appears in some L .

11
b) Let H(n) be the number of times 13 appears in list L . Compute the values of H
11
for n = 2, 6, 10, 12, and 100. Justify your answers.
2. a) Compute the first seven members of L 1996 .
b) If 1995, 8, and 1997 form three consecutive terms in list L , compute the value ofn. Justify.
11
3. Determine with proof a formula in terms of n for the number of terms in L .

11
4. Determine with proof a fommla in terms of 11 for the average of the terms in L11 .
5. Number the elements in the lists from left to right. The 2 in L3 is in the 8th position and the
rightmost 3 in L4 is in the 17th position. Compute the value of the element in the 524,320th position. Show
your reasonmg.
32
6. I , and F
11+ 2 = F 11+ 1 + F17 for 11 > l. Show that F
17 +I
is the largest element in L .
11
b) Determine with explanation the least value of 11 for which 1996 could possibly appear in L .
11
You need not prove that 1996 actually appears in that list.
7. Show that two consecutive elements of any list L are relatively prime.
11
8. Show that numbers a and b can be consecutive elements in at most one list L .
n
9. Let N(t) be the number of times the number t appears in a list. Compute the maximum value of N( 1996).
Prove your answer.
I 0. Prove that if 11 is even, then L has twice as many odd terms as even.
11
33
la. L always begins with l, n. This is clearly true for n = l. If it's true for n, then L I begins
n ~
with l, n + l, n. Hence the result follows from induction.
lb. Shown are lists L1 through L6. Since the lists arc symmetric we need only double the number of l 3's that
appear in the first half.
LI
L2 2
L3 3 2 3
L4 4 3 5 2 5 3 4
LS 5 4 7 3 8 5 7 2 7 5 8 3 7 4 5
L6 6 5 9 4 l l 7 10 3 l l 8 13 5 12 7 9 2 9 7 12 5 13 8 l l 3 10 7 l l 4 9 5 6
Clearly, H(2) = 0 and H(6) = 2. In L7 we obtain 4 additional 13's from 9 + 4 and 10 + 3, making
H(7) = 6. In LS we obtain 4 additional 13's from 7 + 6 and 2 + l l making H(8) = lO. Since if two
consecutive elements of a list sum to more than 13, no number generated between them will equal 13, we can
write half of each successive list, omitting all numbers greater than 13 as well as all numbers sandwiched
between them.
L7 ➔ { l, 7, 6, 13, 4, 13, 13, 9, l l, 2} H(7) 6
Ls ➔ {l, 8, 7, 13, 6, 13, 13, 13, l l, 13, 2} H(8) 10
L9 ➔ { l, 9, 13, 13, 13, 13, 13} H(9) 10
LIO ➔ { l, 10, 13, 13, 13, 13, 13} H(l0) 10
LI I ➔ {!, ll, 13, 13, 13, 13, 13} H(l l) 10
Ll2 ➔ {l, 12, 13, 13, 13, 13, 13} H(l2) 10
Ll3 ➔ {l, 13, 13, 13, 13, 13, 13} H(l3) 12
Wepickupnoadditional l3'safter L13 ,so H(l00) = 12.
34
Solutions to the ARML Power Question - I 996
2a. L1993 = { 1, 1993, ... } . Therefore, L 1994 = { 1, 1994, 1993, .. } , making
L 1995 = { I, 1995, 1994, 3987, 1993, .. }, so L1996 = { I, 1996, 1995, 3989, 1994, 5981, 3987, ... }.
Alternately, formulas can be developed for the clements in each position:
List 3rd Element 4th 5th 6th 7th
L4 3 5 2 5 3
LS 4 7 3 8 5
L6 5 9 4 11 7
L7 6 11 5 14 9
L n- 1 2n-3 n-2 3n-7 2n-5

n
Thus, L1996 1995 3989 1994 5981 3987
2b. If 1995, 8, 1997 appear consecutively in L , then 1995 - 8, 8, and 1997 - 8 must appear in L 1.
n -
Continuing backwards, 1995 - 8 · 249, 8, 1997 - 8 · 249 = 3, 8, 5 must appear in Ln_ 249 .
Since 3, 8, 5 appears in L5, then n - 249 = 5 implies n = 254.
3. Let f(n) be the number of terms in L . We show by induction that f(n)

n
= 2n-l + 1. Clearly,
f(I) = 2!-l +I= 2. Assume f(n) = 211 -l +I. In creating L11 + 1, we put a number in each space
between consecutive numbers, and since by assumption there are 2 n- 1 spaces in L , then
n
Alternately, ifwe ignore the right-hand 1 in each list, the number of terms doubles each time since each
element receives I new right-hand neighbor. L 1 has I term (ignoring the right-hand I) so Ln has 2 n- 1
terms. Counting the right-hand 1 gives 2 n- 1 + 1.
35
4. Let s(n) denote the sum of the terms in L . We note that s(I)
11
= 2, s(2) = 4, and s(3) = 10, so it seems
.
reasonable to conjecture that s(n) = 311-I + 1. Clearly s(l) = 31-1 + 1. Assume that s(n) = 311-I +I.
The new terms added to form L I represent sums of two consecutive terms in L11 and each member of L11
11+
except the first and last I appears in exactly two of these sums. Hence, s(n+ 1) = s(n) + ~s(n) - I) =
311-I + I
3(s(n)) - 2 = \ 3 11 - 1 + I ) - 2 = 3( 17 + l)-I + 1. Thus, the average of the terms in L equals 1
11 211- +1
Alternately. the sum of the terms in L I consists of each term in L 11 exactly three times except for the
11+
last 1 which appears only once. Thus, s( n + I) = 3(s ( n)) - 2 · 1 and the induction follows.
Alternately. we note this pattern:
s(l) 2 2
s(2) 3 · 2- 2 4
s(3) 32 · 2 - 3 · 2 - 2 10
s(4) 3 3 • 2 - 32 · 2 - 3 · 2 - 2 28
s(n) 2( 311- I _ 311- 2 _ ... _ 31 _ 3 o)

z[ 3n- I _ 3•-: -I]
and the proof can proceed by induction.
5. The sum of the number of elements in lists L1 through L11 = (I+ I) + (2 + I) + ( 4 + I) + ... + ( 211 -l + I) =
( 211 - I) + n. If 11 = 19 the sum is 524,306. Thus we seek the 14th element in L2 0. We can generate the
first 17 elements in L2 0 from the first two clements in L16 since the increase in numbers of elements in
L1 through L5 from two to seventeen is reflected in the increase in elements between any two consecutive
numbers in subsequent lists. Since L16 begins I, 16 then L 17 begins 1. 17, 16. L 18 begins 1, 18, 17, 33, 16,
L19 begins 1, 19, 18,35, 17,50,33,49, 16,andso L20 begins 1,20, 19,37, 18,53,35,52, 17,67,50,83,33,
82, 49, 65, and 16. The 14th element is 82.
36
6a. By induction. In L1 the numbers F 1 and F2 appear consecutively and F2 is the largest number in L1.
Similarly, F7 and F3 appear in L2 with F, being the largest. Assume that in list L the numbers F and
~ ~ n n
F appear consecutively with F being the largest number in L . In list L the numbers
n+ I n+ I n n+ I
F +F
n n+ 1
= Fn+ 2 and F
n+ I
will appear consecutively. Since each term in L
n+ I
is the sum of a term in
L and L , the largest of which, by assumption, arc F I and F , then F 2 , the sum of the two largest
n n- 1 n+ n n+
terms in L and L , is the largest term in L 1.
11 n- 1 11+
6b. Since F 1 7 = 1597 is the largest term in L 16 , the least value of n such that 1996 first appears in Ln must be
greater than or equal to 17. Note: 1996 first appears in L1 8 as the sum of 741 and 1255 as well as the sum of
765 and 123 I. We obtained the result by writing a calculator program, but we're still looking for a short
analytical proof that n = 18 is the answer.
7. The elements of L1 are relatively prime. For the inductive step, note that one of the two consecutive
elements (the lesser, in fact) must have been present in a previous list. Call the elements a and b with b the
lesser. In general, the greatest common divisor of a and b is the greatest common divisor of a and a - b .
But band a - bare consecutive in the previous list. If band a - bare relatively prime, then so are a and b.
8. If a and bare consecutive terms in L with a > b, then in L we have the consecutive terms a - band b.
n n- I
Similarly, if b > a - b, then in L11 _ 2 we have the consecutive terms a - b, and 2b - a. In this way we can
trace backwards in a unique manner until we arrive at L1. For an example, see problem 2b. Since the
number of steps in such a process is unique, a and b cannot be consecutive elements in more than one list.
Alternately, consider all ordered pairs of consecutive numbers (a., b.) or ( b., a.) with b. ::::; a. such that the
l l l l l l
pair is consecutive in lists L and L . Let a 1 be the smallest of all such a .. Working backwards,
n m 1
band a 1 - b are consecutive in L11 _ 1 and Lm- I , but a 1 - b < a 1 , which is a contradiction.
9. Consider the pair of consecutive elements (a, b) or (b, a) with b < a. From problem 7 we know that
a and bare relatively prime. From Problem 8 we know that the pair appears at most once. We must show that if
the greatest common divisor of a and b is 1, then a and b appear as consecutive elements in at least one list L .
11
37
Proof by induction. For a= l the statement is clearly true. Suppose the gcd of a and b is l. This implies that
the gcd of a and a bis l. Ifwe assume that band a - b appear as consecutive elements in some list and are
relatively prime, then a and b will appear as consecutive elements in the next list and be relatively prime as
well.
This result, together with problems 7 and 8, implies that a pair of numbers band 1996 can be consecutive in
at most one list and that any and all numbers relatively prime to 1996 will be paired with 1996 in some list.
Thus, the number of times 1996 can appear is equal to the number of numbers relatively prime to and less
than 1996. Answer:
1996 1996 1996
1996- -----+-- 996
2 499 998
10. In terms of parity, for n even, L looks like 0E0 0E0 0E0 .... 0E0, while for n odd, L looks like 00E 00E
n n
00E ... 00E 00. We conjecture that for n even, there are twice as many odds as evens, and for
n odd, there are 2 more than twice as many odds as evens, but the proof of the former requires that we show
that such patterns are, indeed, generated. Thus, we have the following proof:
Proof: Since L2 consists of { 1, 2, 1}, the proposition is true for n = 2. Assume that for L2n there are twice
as many odd terms as even; thus, there are 1 ( 2 2n- I + l ) even terms and 2 ( 2 2n- I + l ) odd terms. Since
3 3
each even term in L2n has an odd term on either side of it, the even terms will generate twice their number of
odd terms in L2n + 1,i. e., the number of new odd terms in L2n + 1is 2(½(22n- 1+ 1)). Each pair of
consecutive odd numbers will generate an even number. Thus, the number of consecutive odd pairs in L2n is
i(
22n - 1- 2( 22n- 1+ 1)) and that equals the number of new even terms in L2n + 1. This gives
3I ( 2211 - I 1) even terms and 31 ( 2211+ I + 4)1odd terms 111· . . that the number of new odd
L2n + 1, 1mplymg
tem1s .
111 . (1(·211
32 -
L 217 + 2 1s 2 1j .
I) and the number of new even terms 1s 2 2n - 2 (1(2n
32 - 1J
1) . This gives a
total of I ( 22n+ I +I)1 even terms and 2 ( 2 211+ I + I ) o dd terms m

3 3 · L n+ .
2 2
38
Alternately, let A denote the number of odd/odd consecutive pairs and B denote the number of odd/even
n n
or even/odd consecutive pairs in L . Looking at list L 1, if we have odd/odd, then the next list L has an
n n- n
even inserted, giving two even/odds. lfwe have an odd/even or even/odd in L n- 1, then L11 has an odd
inserted, making an odd/odd and an even/odd. Thus, An = Bn- I and B11 = 2A 11- 1 + B11- 1 , giving
B = B 1 + 2B 2 . Recalling that the Fibonacci sequence F = F 1 + F 2 has the general formula

11 n- n- 11 11- n-
Fn = JsI ( --
i+-.'5
2-
r: :11
- Js -- 2-
r:;n expressible as ci(r n)+
l [ I --.'5
ci n
s ) where rands are roots of
t 2 = t + I, we conjecture that B11 = B11 _ 1 + 2Bn- 2 has the general formula C i(r 11
) + Ci( sn) where
rands are roots of t 2 = t + 2. Thus r = -I ands= 2 making Bn = C 1(-1( + C 2 (2t. Since B 1 = 0 and
2 1
B2 = 2, we obtain -C 1 + 2C2 = 0 and c 1 + 4C2 = 2, giving c 1 = 3 and c2 = 3 . Thus,
Bn = ¾(2(-l)n + 2n) and An = ¾(2(-1 t - l + 2n- I). The number of even terms En is half the number
of even/odds, so E = -l B = -1 ( (-1)
n 2 11 3
n+ 2 n-1) and the number of odd terms O ts· 11
1 3
( 2 n - 1 +1)- I ( (-1) 11 +2 11 - I ) = 2 ( 2 11- I )+I- I (-1).
3
1
n lnparticular,forevenn,
3
0
11
= -2 ( 2 n-
3
I) + -32 = 2 · E n '11 3
j
and for odd n O = -2 ( 2 n- I + -4 = 2 · E + 2 .
3 11
Alternately, since in L there are 2 11- 1 spaces and every consecutive pair spans a space, the number of
11
odd/odd consecutive pairs plus the number of odd/even or even/odd consecutive pairs equals 2 n- I. Thus
A +B =2n-l_SinceA =B 1,thenB 1 +B =2 11 - 1.Weobtain:

11 11 n 11- 11- 11
B
n
2 11 - I - (s )
11-I
2 11-1_( 2n-2_ 8 n-2 )
39
Since B1 = 0, we have a geometric progression with n - 2 terms, a first tcnn a equal to 2 11 - I and a common
3ll(2 - I)n + 2nl) , and we can proceed as before.

. 1
2 . Thus, B =
rat10 r = - -
11 -1
1--
2
40
1-1. The sum of 19 consecutive positive integers equals p 3 , where p is a prime number. Compute the smallest of
the 19 integers.
1-2. The roots of ax 2 + hx + c = 0 are irrational, but their calculator approximations are 0.8430703308 and
- 0.5930703308. If a, b, and c arc integers whose greatest common divisor is I and which satisfy a> 0,
Ib I S 10 and Ic I S 10, compute the ordered triple (a, b, c).

1-3. The years from 1900 to 1999 are written down consecutively and then pluses and minuses are placed
alternately between the digits as shown:
1+9-0+0-1+9-0+l- 1+9-0+2 ..... -1+9-9+9 K
Compute the value of K.
1--4. Given the four-digit base 10 number J fl(..Q. with A -:f. 0, let the palindromic distance L be defined by
L = IA - DI+ I B - CI- Compute the number of positive four-digit numbers where L = I.
1-5. Shown is the complete graph ofy = .f(x), a polynomial function of
degree 10 whose domain is restricted to { x: l ::; x ::; 5} .
Function f is symmetric about x = 3. Compute the number of

solutions to the equation f ( x) = f (f ( x)).
(1, -200) (5, -200)
1-6. The sides of a regular pentagon are extended to form a five-pointed star. If the ratio of the area of the
pentagon to the area of the star equals sin 0, for 0° < 0 < 90° , compute the value of 0.
1-7. Consider the following sequence of 250 numerals in base b: I h, 2 b, 3 b, 4 b' ... , 249 b' 250 h. If h = 25,
compute in base 10, the largest difference between the values of consecutive numerals.
1-8. In trapezoid ABCD. with AB < DC, the sum of the areas of regions I and II is 1996. If the lengths of bases
AB and CD are integers and the distance between them is an integer, compute the minimum area of ABCD.
41
ANSWERS ARML INDIVIDUAL ROUND - 1996
1. 352
2. (4,-1,-2)
3. 802
4. 332
5. 4
6. 18°
7. 391
8. 4491
42
1-1. Letnbethemiddleterm,giving(n-9)+(n 8)+ .. +n+ . . +(n+8)+(n+9) = p 3 . Thus 19n =p3

makingp = 19 and n = 192 . The least term is 361 - 9 = 3521. 1
1-2. Multiplying .84307 by -.59307 gives -.4999995 ... and the assumption that band c arc small implies a
1 1 .. 2x I 2
product of - 2. Adding the two roots gives 4. The quadratic 1s x - 4 - 2 = 0 ➔ 4x - x- 2 = 0.
Thus(a,h,c) = [(4,-1,-2)1-
I-3. There are 100 dates. Adding the l's in the thousand's place gives 1 + 99(-1) = -98. There arc one hundred
9's in the hundred's place for a total of 900. The sum of the ten's place numbers will be
-(10·0 + 10·1 + ... +10·9) = -10(0+1+ .. + 9),butitwillbecancelledbythesumoftheunit's
place numbers: 10(0 + l + . . + 9). Thus the sum is 900 - 98 = 18021.
I-4. The conditions A - I DI = 1 and B- C = 0 give 9 ordered pairs for (A, D) from (1, 0) to (9, 8) and 8
from (I, 2) to (8, 9) for a total of 17. There are ten possibilities for (B, C) from (0, 0) to (9, 9), making a
total of 17 · IO = 170 possible numbers. The conditions A - D = I CI
0 and B - = I give 9
possibilities for (A, D) from ( 1, 1) to (9, 9), 9 possibilities for (B, C) from ( 1, 0) to (9, 8) and 9 possibilities
for (B, C) from (0, I) to (8, 9), for a total of 9 · 18 = 162 possible numbers. The grand total: 1332:.
I-5. Let a and b be the two solutions to f(x) = x.

Clearly there arc two solutions to both /( x) = a and
f(x) = b. Ans: @l. a~+--------1---
I-6. Without loss of generality let AD = 2, making AB = 1. The ratio of C

the area of the pentagon to the area of the star equals
a(MOB) OB OB tan 54°
- - - - - - since
c.(MOC) oc OB+BC tan54° + tan 72°
OB BC . b .
tan54° - 1- and tan 72° - . Convertmg, we o tam
I
sin 54°eos 72° sin54° cos 72°
sin 54 ° cos 72° + cos 54 ° sin 72° sin(54° + 72°)
sin54° cos 72° sin54° sinl8°

= sin 18°. So, 0 = 118°1.
sin 126° sin54°
43
2nd Solution: The star consists of six congruent 36-72-72

triangles with area R and two congruent 36-36-108 triangles with
R +2S
area S. The desired ratio= - - - Let BC = DC = I, then
6R + 2S
/s -1
BD = AD = - - and S : R = AD : DC = - - : 1 giving
/s - I
2 2
. . . R + 2S b .
Su bstltutmg mto - - - , we o tam
6R + 2S
/s- I which equals sin 18°.

- 4-
I-7. There arc three possibilities as indicated by these examples. First, 188b - 187 h =
( b 2 + Sb + 8) - ( b 2 + Sb + 7) = I. Second, 190b - I 89 b = ( h 2 + 9h + 0) - ( b 2 + 8b + 9) = b - 9.
Third, 200 b - 199b = 2b 2 - ( b2 + 9b + 9j = b 2 -9b -9. lfh = 25, the difference in the values of
consecutive numerals can be 1, 25 - 9 = 16, or 25 2 - 9 · 25 - 9 = 139 l I.
1-8. Let m and n be the lengths of the two bases AB

and DC with m < n, and let h be the height of the
trapezoid. The distance from P to AB = km and
the distance from P to DC= kn since
h 1 hm 2
MPB - ACPD. From km+ kn = h, we have k = - - . Thus area (III) equals -2 (km)m
m+n 2(m + n)
l hn 2 I
andarea(IV) = -(kn)n= - - - . TheareaofABCD = -h(m+ n) =
2 ~m+~ 2
hm 2 hn 2 I 99~m + n)
1996 + - - - + - - - . Solving for h, we obtain h = - ~ - - - . The area of ABCD equals
2(m + n) 2(m + n) mn
998(m+n) 2 (m n) m n
= 998 - + 2 + - . The minimum of - + - = 2 when m = n, but here m < n. To
mn n m n m
minimize 998(: + ; j, assume that m and n are relatively prime, thus m + n and m · n are relatively prime
implying that both m and n must be divisors of 1996, namely 1, 2, 4, 499, 998, or 1996. Choosing m = I and
n = 2, !!!:.. + !!.. = ~, making h = 2994 and the area of ABCD = I449 II.
n m 2
44
ARML Relay #I -1996
RI-I. In trapezoid ABCD, AB is parallel to CD and angles Band Dare acute. If sin LB = cos LD,
compute mLA - mLB in degrees.
T
Rl-2. Let T = TNYWR and set K = Square LMNP is
15
inscribed in right triangle ABC as shown. lf PN = K,
compute the product (AP)(NC).

p N
Rl-3. Let T = TNYWR and set K = !_. There are Kpositive integers in an arithmetic progression with
6
common difference I. If the sum of the terms is 75, compute the value of the first term.
R2-1. Let n be a positive integer. Ifthc number of integers in the domain of y = log((! - xXx - n))
equals 2n - 6, compute 11.
R2-2. Let T = TNYWR. Shown is the graph of the first quadrant portion of a relation. The relation is symmetric
over the x-axis and over the origin. Given A(O. T). Bl 32T, 3; J, and C(T, 0), compute the area bounded by
the graph.
B(l.5T. l.57)
---1
A (0. 7)
I
I
C(T. 0)
R2-3. Let T= TNYWR. Compute the largest integer 11 with O < n :S T such that the triangle with
sides n, 211 - 3, and 3n - 9 has integer area.
45
Relay #I:
Rl-1. 90°
Rl-2. 36
Rl-3. 10
Relay #2:
R2-1. 4
R2-2. 96
R2-3. 74
46
RI-I. Since L.B and L.D arc acute and sinL.B = cos L.D, then
mL.B + mL.D = 90°. But mL.A = 180° - mL.D.
so mL.A - mL.B = (180° - mL.D)- (90° - mL.D) = l90°i.
Rl-2. We have T= 90andK = 6. Since MPL- /1..MNC.

AP MN
- = - giving (AP)(NC) = (PL)(MN) =
PL NC
(PN ) 2 = K 2 = lill. Note that AP· NC is invariant.

p N
(2a+ K - l)K
Rl-3. We have T = 36 and K = 6. Letting a be the first term we have - - - - - = 75. Thus,
2
Kand (2a+ K I) are factors of 150. Possible factor pairs (K, a) are (2, 37), (3, 24), (5, 13). (6, 10),
and(I0,3). SinccK = 6,thena = [Q].
R2- l. The integers in the domain of the log function are 2, 3, ... n - I, making a total of n - 2 integers.
Thus,n-2 = 2n-6yieldsn = @l.
R2-2. We have T = 4. The area of the first quadrant region is
2( ¾•T J 3 T) =
2 ½T 2 . Symmetry over the x-axis and origin implies l.5T
symmetry over the y-axis as shown in the graph at right. The total
3 2J
area = { 2T ) = 6T 2 . .
Smee T = 4, total area =
1nL7
L2fu.
R2-3. We have T = 96, so O < n S: 96. The scmiperimeter is 3n - 6. By Heron's formula the triangle's area
equals J(3n - 6)(2n - 6)(n - 3)3 = 3(n - 3) J2(n - 2). Clearly, the area is integral whenever
(n - 2) = 2m 2 . Since n = 2m 2 + 2 ::S 96, then m S: 6 making n = [HJ.
47
ARML Suqer Relay- 1996
Note: Pass answers from position 1 to position 15.
I. If the area of a triangle with base 2N and height N - 2 is equal to N, compute N.
2. Let T= TNYWR. Compute the slope of the line passing through (T, I) and (I, T 2 ).
3. Let T TNYWR. If y = Tx 2 + T 2x + C and the x-coordinate of the vertex equals C,

compute the value of C.
4. Let T= TNYWR. If (2 + Tz) 2 = a+ bi, compute the value of a+ b.
5. Let T = TNY WR. Pass back the digit in the units place of the product 199T • 279T - 1 .
3 2 3
6. LetT=TNYWR. If T -3T +3T- l = K ,computethcvaiueofK.
7. Let T = TNYWR. If Tis the number of sides of a regular polygon, compute in degrees the positive difference
between the sum of the interior angles and the sum of the exterior angles.
8.
'\
Let T = TNYWR. Let K = I..._
60
If K is the height of an equilateral triangle, let the area
of the triangle be A. Compute - 9 -.

A-'3
9. LetT=TNYWR. Ifsinr'cos286°-cosr'sin(-1O6°) =sin0 forO° :s 0 :s 18O°,then
if i = ~ , compute cos 0 + isin8.
48
ARML Super Relav - 1996
10. I.ct T~ TNYWR. Compute[~]
3x-T T+x
I I. Let T = TNYWR. Compute the value of x satisfying: 2x - - -
2- = 7 - - 5- .
12. Let T= TNYWR. lftan 9T = cot 8 for -90° < 8 < 90°, compute 8.
13. Let T= TNYWR. A line passes through the point (- 2, 9) with slope T. Compute the x-intercept
of the line.
14. Let T = TNYWR. If the ordered pair (x, y) is the solution to the system below, compute (x - y).
x +y T+ 7
Tx - 4y T-8
15. Let T = TNYWR. A square floor is covered with square tiles. If the number of tiles in the two diagonals is
2 r 2 + 5. compute the number of tiles on the floor.
49
ANSWERSARMLSUPERRELAY- 1996
1. 3
2. -4
3. 2
4. 8
5. 9
6. 8
7. 720°
8. 16
9. i
10. - 1
11. 11
12. - 90
13. -1
14. 4
15. 361
50
Solutions to the ARML Super Relay - 1996
T2 - l
2. m = - - = { T + 1). Since T = 3, m = E1J.
I- T '
-T2
3. The x-coordinate of the vertex --= -T So C = -(-4) = [2].
2T 2 2
4. (2 + rf = ( 4 - T 2 ) + 4Ti. Hence a+ h = 4 + 4T - T 2 which for T = 2 equals[[.
5. Modulo 10, 19</ -299T-l =9 2T-I =9 2T+I =8iI' ·9= 9. Hencethelastdigitis[2landthe

value of Tis irrelevant.
7. The sum of the exterior angles is 360°, the sum of the interior angles is ( T - 2) 180°, so
(T-2)180° - 360° = l80°(T-4). For T = 8, the difference equals 1720°1.
720 K2 A/j 144 = rm.

8. K=
60
12. A -J3 so -9- 9
9. sin T cos 286° - cos T sin (- I 06 °) = - sin T cos I 06 ° + cos T sin I 06 ° = sin ( l 06 ° - n.
Since T = 16, 0 = 90°. We have cos 90° + isin 90° = [I].
51
I
10. 2 = tJJ. Note: For real numbers other than 0 the answer is I, for pure imaginary numbers the
I
answer is - 1. Unfortunately, for the clever ARML participant, both answers lead to nice results down the
line, and so a guess of I would not have led to the kind of rotten answers down the line that wrong answers
usually result in.
II. 10(2x)-5(3x-D = 70-2(T+x) ➔ 5x+5T = 70-2T-2x ➔ 7x = 70-7T ➔ x = 10-T.

Since T = - 1, x = [].
12. sin9T
- cos-
- = - 0 ➔
. 0 sm9T
sm . = cos 0 cos 9T ➔ cos 0 cos 9 T- sm
. 0 sm.
. 9T =0 ➔ cos( 0 + 97) =0
cos 9T sine
➔ e + 9T = 90° + 180K ➔ e = ix10° - n + 180K ➔ e = ~10 - n. Since T = 11, e = l-9°1.
9
13. Ify-9 T(x + 2), then for y = 0, x =- T - 2 . Since T = - 9, x = I - 2 = t]].
14. Multiplying the top equation by T and subtracting the bottom equation yields
(T+4)y = T 2 +6T+8 =(T+4)(T+2). Thusy = T+2. So,x+T+2 = T+7,givingx = 5.

Then x - y = 3 - T. Since T = -1, x - y = @].
15. Let n be the number of rows in the square. Ifn is even, the number of tiles in the diagonals
is 2n; if n is odd, the number is 2n - l. Since 2r2 + 5 is odd, we have 2n - 1 = 2r2 + 5 ➔

n = T 2 + 3. Since T= 4, n = 19, and the number of tiles is 13611.
52
ARML Tiebreakers - 1996
I. For integers x and y with I < x, y :S I 00, compute the number of ordered pairs (x, y) such
2
that logx y + logy x =3.
2. Compute the number of distinct paths not passing through point (2, 2, 2) which travel from point (0, 0, 0) to
point (4, 4, 4) in 12 steps, changing a coordinate by I at each step.
3. Compute the value of:
53
I. log v + 21og x =3➔ log y + - - 2 =3 ➔ (log v )2 - 3 log v + 2 =0 ➔

x - y x Jo g.:\__v x- x-
(logx y- I )(logx y- 2) = 0. Thus, Jogx y =I or logx y =2 ➔ y =x or y = x2 . The solutions to the
first equation are the 99 ordered pairs from (2, 2) to ( I 00, I 00); the solutions to the second are the 9 ordered
pairs(2,4),(3,9), (4, 16), ... ,(10, 100). Thus,thereare99+9 = II08lorderedpairsofsolutions.
2. 12 ! -( 6 !
4!-4!-4! 2!·2!-2!
J·(2!-2!-2!
6 ! J = 1265501.
. 2 2 2 n(n+l)(2n+l) .
3. Smee I + 2 + ... + n = 6 , then each term of the sum can be wntten as
n(n + 1)(2n + I) 2n + 1
- - for n = I to n = 60. Thus, the sum can be written as
,(n + I) 6 6
54
1997
'Team 'Round............................................................................................. 57
Power Qyestion ......................................................................................... 62
'lndividua( 'Round--- ----- ------------------------------------------------------ --.......... -------------- 71
'Re(ay 'Rounc(............................................................................................ 77
Suyer 'Re(ay................................................................................................. 80
'Tiebreakers 85
-55-
THE 22 nd ANNUAL MEET
ARML celebrated its twenty-second year of competition with new teams from AAST,
i.e., the Academy for the Advancement of Science and Technology, and Southwestern
Pennsylvania. Including eight alternate teams, a total of eighty-nine teams and over 1335
students took part in this year's competition. This year's contest was a very difficult one
and included problem #2 on the Individual Round, a question that surprisingly only 8
students nationwide were able to solve, and the equally infamous problem #8 on the
Individual Round, a question that only 6 students solved correctly, making it perhaps the
most difficult ARML problem ever. It was reported that #8 spawned an hour-long lecture
at MSOP that summer.
Andre Samson of Thomas Jefferson High School in Virginia received the Samuel
Greitzer Distinguished Coach Award. Andre was coach of the Fairfax County ARML
team from 1976 to 1992 and then coach of the TJ ARML team starting in 1993. His
teams have finished in the top 4 in each of the last ten years. T J is the only school ever to
winARML.
Sam Baethge of Texas received the Alfred Kalfus Founder's Award in recognition of his
years of service to ARML in Texas. He has been the leader of the Texas teams since
1983, the sole organizer much of the time, the Southwest regional director of ARML, and
a problem poser for AHSME and AIME.
David Cooper of Suffolk County and Richard Haynes of Maine received the Zachary
Sobol Award in honor of their outstanding contribution to their ARML teams.
56
ARML Team Oz._.,1e=s'=ti=on==s"'=-==='cl==9"=9==7========================
T-1. A pan of length 25 cm, width 20 cm, and height 15 cm is filled with water to a depth of3 cm. Lead cubes of
edge 4 cm are placed flat on the bottom. When the 11th cube is placed in the pan, all cubes are completely
covered by water for the first time. Compute 11.
T-2. Equilateral triangle ABC is inscribed in circle 0. Let D and Ebe midpoints of
- - DE
AC and AB. If -
a+ib
can be written as - - - , for integers a, b, and c in
EF C
simplest fom1, compute the ordered triple (a, b, c).
T-3. Let P be the parabola with vertex at the origin and directrix y = -1 . Compute the number of lattice points
on P whose distance from (0, I) is less than or equal to 197.
T---4. In an ARML relay, the first person in group A passed back the correct answer while the first person in
group B passed back an answer that was 5 more than it should have been. Problem #2 read:
"Let K = TNYWR. If the sum of the roots of Kx 2 + 4x + c =0 is S, compute K+S."
Both second persons in groups A and B correctly found an answer using the value of Keach had received, but
surprisingly, they both found the same value for K + S. Compute all values of K + S.
T-5. Note that IO is 9 more than the sum of the squares of its digits. Compute the sum of all other positive
two-digit base IO integers which are 9 more than the sum of the squares of their digits.
2
T-6. In MBC,Dlieson AC so that mLABD = mL.DBC = 8. lfAB=4,BC= 16,and BD = --,then
cos 8
a
BD = .[i for relatively prime integers a and b. Compute the ordered pair (a, b).
A
T-7. In MBC, mL.C = 90°, AC= 3i3 and BC= 6i3. Line segments
MN and OP are parallel to BC. lfthc areas of MNM, MNPO, and M,..-<------,N
OPCB form an increasing arithmetic progression, then the length of NP

B---------~n C
has a greatest lower bound K. Compute the value of K.
T-8. Let distinct x and y be drawn from { 1, 2, 3, ... , 99 } such that each ordered pair (x, y) is equally likely.
Determine the probability that the sum of the units digits of x and y is less than I 0. If the answer is 1for
relatively prime a and b, compute the sum a+ b .
I
T-9. Let P(x) be a polynomial whose degree is 1996. If P(11) =- for 11 1, 2, 3, ... , 1997, compute the value
11
of P(l998).
T-10. Let ABCD be a regular pyramid whose edges have length 5. Consider the set of pyramids of the form AMNP
where Mlies on AB, N lies on AC, and Plies on AD such that AM, AN, and AP E { 1,2,3, Compute 4}.
the sum of the volumes of all such pyramids.
57
ANSWERS ARML TEAM ROUND- 1997
1. 8
2. (1, 5, 2)
3. 29
4. ±3
5. 300
6. (8, 5)
7. 3✓2-3
8. 7477
1
9. -
999
250./2
10.
3
58
T-1. 3 · 20 · 25 + 64 n > 4 · 20 · 25 ➔ 64n > 500 ➔ n > 7 .8125 ➔ 11 = [[.
T-2. Extend DE until it meets the circle at G. Let AB = 2x making A
AE = EB = DE = x. Let GD = EF = y.
Thus, DE : EF = x : y. By the Power of a Point Theorem,
2 2 7
AE · EB = GE· EF ➔ x = (y + x)y ➔ x - .')' - y~ =0 ➔
]2 x 1+
( -;) --:;-1 =0. Solving for -;,we obtain - -,so(a,b,c) = 1(1,5,2)1.
-J5
2
T-3. The equation of the parabola is y = x: and xmust be even, so let x = 2k. The distance from q 2k, k 2 ) to
k 2 + 1 ~ 197. Thus, k 2 ~ 196 making
k E {-14,-13, .. , 13, 14},givingl22Jlatticepoints.
4 4
T-4. Since S = - - then K + S = K - - . If K is the correct answer, then K + 5 is the incorrect answer and we
K' K
have: K - .i_ = (K + 5) - - 4- ➔ K 2 + SK + 4 = 0. Thus, K = -1, or - 4. If the second person in A

K K +5
receives - 1 or - 4, then the second person in B receives 4 or 1 respectively, and S for A equals 4 or 1 while S
for B equals -1 or - 4 respectively, yielding K+S = ~-
T-5. Let 10a + h represent the number. We seek solutions to 10a + b = a 2 + b 2 + 9. Complete the square and
multiply by 4 to eliminate fractions. Thus, (2a - 10) 2 + (2h - 1) 2 = 65. Since 8 2 + 12 = 7 2 + 4 2 = 65,
we pair 2a - 10 with ± 8 and 2b - 1 with ± 1 to obtain a = 9 or 1 while h = I or O giving 91, 90, 11, and
I 0. Similarly, pair 2a - IO with± 4 and 2b - 1 with± 7 to obtain a = 7 or 3 while b = 4,
giving 74 and 37. Hence, 11 + 34 + 74 + 90 + 91 = 13001.
Alternate solution: From I 0a + b = a2 + b 2 + 9 we obtain (a - I )(9 - a) = b( b - I). To obtain the solutions

check which of the following products can also be written as b(h- 1): 0 · 8, I · 7, 2 · 6, 3 · 5, or 4 · 4.
Clearly, 0 · 8 and 2 · 6 work. Now choose all combinations of a and b.
59
T-6. By the Angle Bisector Theorem, if AD= x, then DC = 4x. By the B
2 4 2
Law of Cosines, 16x =--+ 256 - 2 · 16 · - - · cos 0.
cos2 0 cos 0 4 16
2 4
Thus, I 6x = --
7- + 192 . Likewise,
cos~ 0 A~---~--4x---~C
D
2 4 2 4
x = 16+---2·4·--·cos0 7 . Substituting into the first equation we obtain
cos2 e cos 0 cos~ 0
I6x 2 = x 2 + 192 . giving x = /s ➔ the answer is the ordered pair 1(8, 5) !.
T-7. LetAN=xandNP = y ➔ MN= 2x and OP= 2(x+ y). Then A

X
d....MNM) = x2 , c(MNPO) = (x+ Yr7 7

-x-, and M
2x y
N
0 p
2(x + y)
c( OPCB) = (3./3}2 -(x + y) 2 . The condition that these areas form an
B C
increasing arithmetic progression yields a(MNPO)- a(MNM) = a(OPCB)- a(MNPO). Thus,
0 < (x + y) 2 - 2, 2 = 27 - 2(x + y) 2 + x 2 . Solving, we obtain (x + y) 2 - x 2 =9 ➔
y = J9 + x 2 - x which can be rewritten as y = ci ,

9 + x2 +X
signifying that y is a decreasing function
. 2 2 2 2
of x. Now we seek bounds on x. Smee (x + y) > 2x and (x + y) = x + 9, then
9 + x 2 > 2x 2 ➔ x < 3. Since y is a decreasing function ofx on the interval (0, 3), the greatest lower bound
is atx = 3, i.e., at ./Is - 3 :3/i. - 31.

=
T-8. There are 9 ways for x to have Oas a units digit and then all 98 choices for y work. For units digits n between 1 and
4, there are 10 choices for x and 98- !On choices for y. For example, ifx = 12, we delete 12 and all 20 numbers
ending in 8 or 9. For units digits n from 5 to 9, there are again 10 choices for x but 99- !On choices for y since in
this case we don't delete the number itself. Adding these yields 9 · 98 + 10(88 + 78 + 68 + 58) +
l 0( 49 + 39 + 29 + 19 + 9) = 5252. This double counts (x, y), so there are 2626 distinct pairings which work. There
are a total of 99C2 = 4851 pairings. The probability is 2626 . The sum of 2626 and 4851 is 17477 I.
4851
60
I
T-9. If P(n) = - for 11 1,2, ... , 1997,thcn xP(x)- l hasrootsx = 1,2, ... , 1997. Thus.
11
xP(x) - I = c(x - I)(x - 2) · ... · ( x - 1997) . If x = 0, then - I = c(-1997 !) so c =

1997!
Therefore, 1998 ·P(l998) - I I

--(1997!) ➔ P(1998) = - 2 = -. ~
1997 ! 1998 999
T-10. Consider the analogous case in 2-space: MBN lies within equilateral A
triangle ABC of side 4. If N is moved to N', then the height of
I
MEN' = - - times the height of MBN. If Bis moved to B'. the N
AN
B C
I
height of MB'N' = - - times the height of MBN'.
AM
MB'N' is a unit triangle whose area is (_I_] _(-1-1 times the area of MBN. Alternately. the area of
l AM) AN)
MEN = (AM)(AN) times the area of unit MB'N'.
In the three-dimensional case, let the volume of A MNP be V.. If M, N, A

l
and Pare moved so that AM = AN = AP = l, then AMNP is reduced

to a unit pyramid of edge-length 1, and its volume is reduced by
(A~)· (A~)· ( AIP). Let the volume of the unit pyramid be v0 .
V. B C
Then V0 = AM. AN. AP ➔ ~ = (AMX ANX AP)V0 .
The sum of the volumes of all the pyramids is the sum of all possible products ( AMX A NX AP)V0 for
AM, AN, AP E { 1, 2, 3, 4}. This sum can be conveniently expressed as v0 (1 + 2 + 3 + 4 )3 = I OOOV0 .

It remains to calculate the volume of a regular tetrahedron of side l. Its base has area ~ and its height can
be found to be 12 ,
yJ3
so its volume = ~3 Bh ./2.
=
12 '
making l OOOV =
0
1
25 0-/2, 1-
3
61
ARML Power Question -1997
Definition #1: if ti.ABC is equilateral and ti.DAB, ti.DEC, and !lDAC are all isosceles right triangles with
the right angle at D, then DABC is a comer pyramid with apex D.
Definition #2: a cornered polygon is one of the following:
a) An equilateral triangle with a comer pyramid completely covering it.
b) A regular hexagon divided into 6 equilateral triangles with comer pyramids completely covering each triangle.
c) A square to which is first attached a pyramid whose base is congruent to the square and whose faces
arc equilateral triangles. Each triangular face of the pyramid is then cornered. See def. 2a and problem 2.
d) A regular pentagon to which is first attached a pyramid whose base is congruent to the pentagon
and whose faces are equilateral triangles. Each triangular face of the pyramid is then cornered.
Definition #3: a cornered polyhedron is a polyhedron whose faces have been replaced by the corresponding
cornered polygons.
I. Let DABC be a comer pyramid with apex D. If Mis the midpoint of AB, P is the foot of the altitude from D
to ti.ABC, and AB l, compute:
a) DM b) MP c) DP d) cosLDMP
2. On the left is square ABCD of side l. Pis the center of the square and Mis the midpoint of AB. On the right
is the cornered square ABCD. Points F, G, H, I are the apexes of the comer pyramids. Compute:
a) cosLFMP. b) cosine of the angle formed by planes FEB and GEB.
c) perimeter of FGHI. d) volume of the cornered square ABCD.
H
E
I G
{' D
,.
' ,,,.,.,,,,,,,, ,. ,,
,,,,,' p ',,
' ,,
C
------3,> ------3,>
A M B A M B A M B
3. Let P be a right pentagonal prism whose 5 lateral faces are congruent squares. Prove that it is possible to
comer P; i.e., show that the cornered faces do not overlap. The proof may use calculator approximations.
62
ARML Power Question - I 99 7
4. Shown below is a squat but instructive polyhedron. Prove that it is impossible to comer this polyhedron: that
is. show that the cornered faces do overlap. The proof may use calculator approximations.
5. The pattern at the right can be folded into
a soccer ball-like polyhedron called a
buckyball. Compute the number of its
edges and vertices.
6. Compute the number of faces, edges, and vertices of a cornered buckyball.
7. A regular tetrahedron is a pyramid all of whose faces are equilateral triangles. Prove that a cornered regular
tetrahedron is a cube.
8. A regular octahedron has 8 faces which are equilateral triangles with 4 faces meeting at each of its 6 vertices.
Prove that a cornered octahedron has double the volume of the original octahedron.
9. Cover each face of an octahedron with a regular tetrahedron and then comer the resulting figure. Prove that
the cornered figure is a cube.
I 0. Let P1, P2 , P3 , ... be a sequence of polyhedra where ~ is a regular tetrahedron of side l. Create
polyhedron P 1 from Pfl as follows: To each equilateral triangle face of Pfl , attach a regular tetrahedron
fl+
with vertices at the midpoints of the sides of the faces of P .

11
a) Compute the number of faces of P .

fl
b) Compute the surface area of P .

fl
c) Find the limit, as n goes to infinity, of the volume of P .

n
I
d) Prove that every P fits inside a cube of side C: .
n y2
63
D
I. a) DM is an altitude of isosceles right triangle DAB, so DM =
2
../j CP 2 I B
b) MC =- and - = - , so MP = - - .
2 MP 1 2-J3
C
) MP 2 +DP 2 = DM 2 ' so DP= JJ...-
4
l
12
MP I
d) cosLDMP =- = -. A
DM i3
2. a) Pyramid EABCD is the union ofthc four comer pyramids PABE, PBCE, PCDE, and PDAE. Define 0 to
1
be the base angle of a comer pyramid so that mLFME = mLEMP = 0. By (Id), cos 0 = .[3, but then
E E
}[)
____ ,,.::
A ,,:'_ ••• ---

:,. C
M
/_l p
p
b) Let Mbc the midpoint of EB. Because triangles FNA, PNA, PNC, and GNC arc corresponding cross-
sections of congruent comer pyramids, mLFNA = mLPNA = mLPNC = mLGNC = 0, so
7
9
c) Usingiliediagramin2b,FN =GN={soFG = [½J2 +[½]'-2 [½)[½][-:] 2./3

-by
3
· By symmetry, the perimeter

th e Law ofc osmes. · o f square FGHI is -
s/i.
-.
3
64
Alternately. let J be the projection of point F onto the plane ABCD. The distance JM equals
MF · cos LFMP = -I •--1 = - -1 ; tI1e negative

. sign
. means that J 1s
. outs1"d e square ABCD. Thus,
2 3 6
FH
4
= 2(JM + MP) = 3 and the penmeter of FGHJ 1s - 3-.
. . . 8-h.
d) The volume of the cornered square is 8 times the volume of a comer pyramid with base side I giving
1
8 · - · (bh)
8
= - ·-
/3 · - I ./2_
=- .
3 3 4 ./6 3
Alternately. the volume of the cornered square is twice the volume of pyramid EABCD giving
I
2---1--=-
1 £
3 Ji_ 3
3. The question of whether one can comer a given polyhedron arose in modular origami. It is relatively easy to
fold comer pyramids from 3 small squares of paper, but which polyhedra can then be built from comer
pyramids? For more infomrntion, see Multidimensional Trans.formations: Unit Origami by Tomoko Fuse,
Japan Publications, 1990.
The pentagonal prism at the right has two types of edges, those that
A C
join two squares such as AB and those that join a square and a
pentagon such as AC. First, we consider edges like AB. The solid
angle at AB is I 08° and we hope to add two cornered squares at AB.

B
By (2a), the base angle of a cornered square is arccos (-½) = 90° + arcsin( i). With a calculator we can
approximate arcsin ( i) as 19 .4 71 °. Or we could note that arcs in ( ij < arcsin( ½J= 30°. Either way, we
can conclude that 108° + 2[90° + arcsin(½)] < 360°, so there is no overlap.
65
Alternately. note that we can fit in the belt of cornered squares only if the sum of all their base angles is less
than the sum of the exterior angles of the pentagon which is 5(360° - I 08°) = 1260°. The sum of the I 0
base angles of the cornered squares is !Oarccos (-½J= + 900° IO arcsin( iJ. Because
arcsin( i)"' (-½) <

19.471 ° < 30°, it is true that !Oarccos 1260°.
We still need to show that there is no overlap at edges like AC. The r---------
,
I
I
solid angle at AC is 90°. We hope to add a cornered square and a
--~/
cornered pentagon at AC. The pyramid attached to the pentagon

during the first stage of cornering has a shorter altitude than the
pyramid attached to a square during the first stage of cornering since
tan 2(54°)
the two altitudes have lengths ----"--'- "' .5257 and
4
J5.2 "' .7071 respectively. During the second stage. comer pyramids are attached in both cases. Thus, the
base angle of a cornered pentagon is smaller than the base angle of a cornered square (the angles are
approximately 91.113° and l 09.471 °, respectively). The total of the angles at AC is bounded by
90° + 2(120°) = 330°, so there is no overlap. Thus, it is possible to comer a pentagonal prism.
4. Let P be the given polyhedron. Notice that the band of squares
around the "equator" of P forms a right decagonal prism. Let AB be

A
an edge joining 2 of these squares. The solid angle at AB is 144°. If
it were possible to comer P, then the cornered squares attached to the · - - - .. f - - - 1 _____,B 1 _ / :-
~ /
squares adjacent to AB would be able to fit around AB.
By (2a), the base angle of a cornered square is arccos (-½ J. With a calculator we compute arccos (-½) to
. . 1 I i5 -
be approximately l 09.471 ° or we could note that smcc - > - - - , then
3 4
( 3I)
arccos - (1-Jsl =
> arccos - -4- .
108°. Either way, we can conclude that the total solid angle would be
strictly greater than 144° + 2(108°) = 360° which is impossible. Thus, P cannot be cornered.
66
Alternately. note that all IO cornered squares fit in only if the sum of all their base angles is less than the sum
of the exterior angles of the decagon which is 12 · I 80° = 2160°. The sum of the 20 base angles of the
cornered squares is 20 arccos(-1 j = 20[90° + arcsin( 1 jJ = 1800° + 20arcsin( ½J. Let a = arcsin( 1 J.
2./2.
We know from (3) that a < 30°. Because cos a = - 3-, cos(2a) = \
(8J 7
9 - I = 9 and
49)
cos( 4a) = 2( - - I = -17 . 17
Because - < -
I
= cos(90° - a), we must have a+ 4a =Sa> 90°, so
81 81 81 3
20a > 360°. But then 20 arccos(-i j = 20(90° +a) > 2160°. Thus, the cornered squares must overlap.
5. A buckyball has 12 pentagonal faces and 20 hexagonal faces. Three faces

meet at each vertex and 2 faces meet at each edge, so a buckyball has
5·12+6·20 5·12+6-20
2 = 90 edges and 3 = 60 vertices. W c can
check with Euler's Formula: V - E + F = 60 - 90 + 32 - 2, as it should.

Note that buckyballs are named after R. Buckminster Fuller,
famous for his geodesic domes, and that carbon compounds with this geometry were discovered by the 1996
Nobel Prize winners in chemistry, Robert F. Curl, Harold W. Kroto, and Richard E. Smalley.
6. A cornered pentagon has 15 faces and a cornered hexagon has 18 faces, so

a cornered buckyball has 15 · 12 + 18 · 20 = 540 faces. A cornered
pentagon has 20 additional edges and a cornered hexagon has 24
additional edges, so a cornered buckyball has 90 + 20· l 2 + 24 · 20 810
-
edges. A cornered pentagon has 6 additional vertices and a cornered
hexagon has 7 additional vertices, so a cornered buckyball has
- ~~ 'i0fy:cJ
____ jA;;;:)if-l
60 + 6 · 12 + 7 · 20 = 272 vertices. Ifwe plug these numbers into
Euler's Formula, we obtain V - E + F = 272 - 810 + 540 = 2 as
hoped. The solid pictured, while somewhat more pointed than a cornered buckyball, has the same
arrangement of vertices, edges, and faces.
67
7. Let ABCDEFGH be a cube. Then A CFH is a regular D

M
tetrahedron and BACF, DACH, EAFH. and GCFH arc corner ................
A ..1----
pyramids, so ABCDEFGH is a cornered tetrahedron.
Alternately. start with tetrahedron A CFH and add comer

pyramids BA CF, DA CH, EAFH, and GCFH. Let M be the
midpoint of AC. Then LDMH and LBMF are both base H

~-----"'--_;'
angles of corner pyramids. so mLD MH = mLB MF = e. E F
Thus, mLDMB = 28 + mLFMH where LFMH. the base angle ofa tetrahedron, satisfies
cos LFMll
1
= - = - cos28 and
.
smLFMH
2/2 .
= - - = sm28. We conclude that mLFM H = 180° - 20, so
3 3
mLDMB = 180°, making ABCD a square. Arguing similarly for the other edges of ACFH completes the
proof.
8. LetABCDEFbe a regular octahedron with center G. Then E
GABE. GBCE, GCDE, GDAE, GABF, GBCF. GCDF. and
GDAF arc 8 congruent corner pyramids. When we corner
ABCDEFwe attach 8 more congruent corner pyramids to the C
faces of the octahedron, so the total volume is doubled.
9. The centers of the faces of a cube form a regular octahedron. D

Let A, B, C, and D be vertices of a cube. and let X, Y, and Z be
the centers of the neighboring faces as shown. Let M, N, and
P be the midpoints of AB, AC, and AD respectively. Then

AXYZ is a regular tetrahedron ( attached to face XYZ of the
inscribed octahedron), and MAXY, NAXZ, and PA YZ arc
comer pyramids attached to the outer faces of tetrahedron I ,,
L" _________________ 5,"
I ., "
AXYZ. Together they completely fill the A-octant of the B
cube. Reasoning similarly for each comer of the cube gives the full dissection desired.
68
I 0. a) P1 has 4 faces. As we go from P to P , each face is replaced by 6 faces. Thus, P has 4 · (/ 7- 1 faces.
n n+ 1 n
b) A face of P is an equilateral triangle of side - 1 - which has an area of [ -1

n 2 11- I
-J .[
4 11- I
Jj
4
] = ./3n . Thus,
4
pn has' total su,faee a,e, of { ·•- t l·(~) a Ji,. (% r I
I Jj 2 I
c) The volume of~ , a tetrahedron of side I is - · - · r ,: . Going from P to P 1 ,
3 4 -y6 6-yl.2 n n+
we add 4 · 6 17 - 1 tetrahedra of side length - 1- so vol

2n
(Pn+ 1) = vol (P
11 ) + 4 · 617 - I · [-1
gn
-J .-6./i..
1-
( 3 Jk
17
+ 9./i. ·:; l4
1 1
Summing from 1 to n we have vol (P
11 + 1) = 6./i. 1
Thus, as 11 ➔ 00 , we have
.Ji iJr
1 I
- --+--
~+ - 6./i. 3.Ji.
d) In fact, a cornered P,7 is a cube of side Ji.I . Questions (7) and (9) show this for P1 and P2 , providing
base cases for our induction. We complete the proof by showing that a cornered face of P11 and the cornered
version of the corresponding configuration in P I are the same; thus, the cornered versions of the Pn and
11+
P11 + 1 are the same polyhedron, namely a cube of side Ji.1 .

D D D
~
,' ,,
''
' ',
'' '
'
''
,,, ' ','
: C ',
, '
, '
, '
, '
, '
, '
A ~ B A
M M
Let ABC be a face of P . Let M, N, and P be the midpoints of AB, BC, and CA. Let D be the apex of the
11
tetrahedron with base MNP, so that ABCDMNP is the corresponding region of P . Notice that Dis also
n+ 1
69
the apex of comer pyramid DABC. Let X be the midpoint of AD . Then Xis the apex of the comer pyramids
XAMP and XDMP. Arguing symmetrically for BD and CD, we see that the cornering of the configuration
ABCDMNP is the same as the cornering of ABC.
Alternatively, ifwe let P1 be situated inside a cube of side i21 as C

-=---------------~~
.··•-,·-- N .-·
- .
in (7), then P2 will have as new vertices the rest of the vertices of D
the cube together with the centers of the faces of the cube. Let
ABC be a face of fj . Let D be the vertex of the cube surrounded
by A, B, and C. Let M, N, and P be the midpoints of AB, BC,
and CA. Then DMNP is one of the tetrahedra
added at stage 1. It is situated within the D-octant of the cube in exactly the same was as P1 is situated inside
the original cube. The three pyramids that will be attached to DMNP at stage 2 will lie inside the D-octant,
just as the pyramids attached at stage 1 lie inside the original cube. This is true for each octant of the original
cube and so we can apply the same analysis to each of them at the next state, and so on all the way down.
Since no added tetrahedron leaves its cubelet, no P leaves the original cube.
n
Shown below are fj, P2 , P3 , P4 , and 1s. For n 2: 3, some of the edges of the added tetrahedra coincide.
However, they never overlap along faces or have common volume. If we let P= = LJ ;= 1Pn , then P00
is not
the entire cube. There will be uncountably many points of the cube that are outside all the P 's. Collectively,
n
they have volume 0.
70
C
1-1. Let a, h, c, and 11 be positive integers. If a+ h + c ( 19)(97) and a + n h - n
n
compute the value of a.
1-2. ti.ABC is inscribed in circle 0, the radius of O is 12 and mLABC = 30°. A circle with center Bis drawn
tangent to the line containing AC. Let R be the region which is within ti.ABC, but outside circle B.
Compute the maximum area of R.
I-3. A team wins 3 games, then loses 1, then wins 3 and loses 2, then wins 3 and loses 3, and so on,
each time winning 3 games before losing one more than before. If N is the number of games played,
find the least value of N such that the percentage of wins is below 25%.
1--4. Let .f(x) = 4x - x 2 . Consider the sequence x 1, x 2 , x 3, ... where xi = f( xi- 1) for i > I. Compute the
number of values of x 1 such that x 1, x 2 , andx3 are all distinct, but xi = x3 for all i > 3.
I-5. The sum of 1999 positive numbers in an increasing arithmetic progression is 1. Compute the width of the
smallest interval containing all possible values of the common difference. Do not leave your answer in
factored form.
1-6. Compute the remainder when cot 1997( 1~ J+ tan 199 iJ1~ is divided by 9.
I-7. The diagonals of a convex polygon connect non-adjacent vertices. The diagonals of a regular I 0-sided
polygon are drawn. Compute the number of parallel pairs of diagonals.
1-8. If i Vi, - 1 is written as '(; + (b + '}/;: where a, b, and care rational numbers,
compute the sum a+ h + c.
71
1. 80
108
2.
1t
3. 217
4. 3
1
5.
1997001
6. 4
7. 45
1
8. -
3
72
Solutions to the ARML Individual Questions - I 997
1- 1. .
Smee b =a+ 2n , c = an+ n-,
7
then 2a + 211 +an+ n-,., = 21._a + n) + n(a + n) = (n + 2)(a + n) = 19 · 97.
Thusn+2 = l,19,97,orl843,butifn+2 = lthenn < 0andifn+2 = 97orl843,thena<0.

Thus, n + 2 = 19 ➔ n = 17 ➔ a + 17 = 97 ➔ a = [BQ].
1- 2. Since m.L4BC = 30°, m.L4OC = 60°, making AC= 12. The
I I 2
area of the shaded region - · I 2 · BD - - · re · BD This is
2 12
-6 36
quadratic in BD and takes its maximum at BD
-2rc / 12 7t
2
Maximum area = 6( 36 )- .2:.. · 36
re 12 1[2
= I'OS
re
I· Note: The
maximum occurs when BD lies outside the triangle .
Figures 1, 2, and 3 below show various positions for B. When B is very close to A or C, the shaded area is
relatively small. When Bis on the perpendicular bisector of AC, the shaded area is quite small. So clearly
the shaded area increases until it reaches approximately the position in Figure 2 and then the area decreases
until Figure 3. The maximum area is reached when mLBCA == 32.6°.
B
•0
C C
Figure l Figure 2 Figure 3
73
Solutions to the ARML Individual Ouestio.c'_n'-"'s~-~I9~9~7~================
1-3. After the nth block of3 wins, the lowest percentage occurs at the end of the next losing streak, i.e., after
,Kn+ I) wms 3n 1
I + 2 + ... + n = - - - losses. Thus, - - - - equals - - - - - ➔ 11 = 17. After
n total games n(n + 1) 4
3n+---
2
51 wms 18 = 204 games, tI1e wmnmg

I 7 .-
· an d 51 + - · · percentage 1s
. 25~o.
' 1 Let x betI1e num ber o f Iosscs af tcr
2
51 + 3 1
the next block of 3 wins; we want the least x such that - - - - < - . Solving, x > 9 ➔ x = IO ➔
204 + 3 + X 4
[2 i1] games.
1-4. If x 3 = x 4 , then f(x 3 ) = x3 ➔ 4x3 - x/ = x3 ➔ x 3 = 0 or 3. If f(x 2 ) = x 3 , then
4x 2 - x 2 2 = 0 or 3 ➔ x2 = 0 and 4 or x 2 = I and 3. Reject O and 3 since in those cases x2 = x3 .

If x 2 = 4, then f ( x 1) = x2 ➔ 4x1 - x / = 4 ➔ x1 =2. If x2 = I , then
4x 1 -x 12 = 1 ➔ x 1 = 2±J3. In sum, x 1 =2 generates2,4,0,0, ... and x 1 = 2±./i
generates 2 ±Ji, I, 3, 3, .. Thus, x 1 = 2 or 2 ± Ji and the answer is W.
I-5. If xis the first term and y is the common difference, then x + (x + y) + ... + (x + 1998y) = 1 implies
1999 ( ) 1
- 2 - 2x+1998y =1➔ x+999y= 1999 . Ifgraphed,theintervalfromtheorigintothe
y-intercept contains all common differences. Its width is - - - -

999 · 1999
7t 7t
tan- - tan - /j-1
I-6. Let N = tan 1997( 17t2 1+cot 1997 ( 17t2 ).
)
s·mce t
an
7t
12 = ( l - ~3 2- Ji, then
I+ tani·tan;) I+
for n = I, 2, 3, .. we obtain a repeating sequence ofremainders of period 6: 4, 5, 7, 5, 4, 2, 4, 5, 7, 5, 4, 2, ..
Since 1997 = 5 (mod 6), the answer should be the 5th remainder, namely 4. To prove this use the Binomial
Theorem:
74
Solutions to the ARML Individual Questions~ 1997
(2 - r-3 ) 1997 = 2 1997 _ C _2 1996 . T:3 + 1995 1 1994 ,- , 1993

"' 3 1997 1 "' 3 1997c2 · 2 ·3 - 1997c3 ·- · 3 "1 3 + 199l4 · 2 · 9 - ...
3 I
( 2 + /;)
997 I ')"7
= -..,, + 1997
c
I
. -~, '
199",
.
'"3 + I 997
c·,2 1995 3
2. . + C'2 1994 33
1997 3 .
/;
. + 1997c4 · 2
1993
· 9 + ···
Note that in Nall terms with /3 will cancel and all terms with coefficients 1997c211 for n :::= 2 are divisible
( 1997 1997-1996 1995 }
by 9. Thus N = 2l 2 + 2 ·2 · 3 mod 9) which is congruent to
( 2 1998 + 1997 · 998 · 2 1996 · 3) (mod 9). Smee

. 2 6= I (mod 9), then 2 1998 = ( 2 6) 332
=I (mod 9)
=8 (mod 9), 998 =8 (mod 9), and 2 1996 = ( 2 6)331 •2 4 = 2 4 (mod 9)

Also, 1997 = 7 (mod 9).
Thus N =1 + 8 • 8 · 7 · 3 (mod 9) = 4 (mod 9). The answer is @l.
I-7. There are two types of parallels, those parallel to an edge and the dashed
ones parallel to a segment connecting A. and A . .., . In the

I I+-
diagram there are 3 solid parallels making 3c 2 = 3 pairs of
parallels and 4 dashed parallels making 4 c 2 = 6 pairs. There are
5 distinct cases making the number of pairs of parallels equal to
5(6 + 3) = @31.
75
1-8. Let t = fi_ so that (t l

+ 1) 3 = t 3 + 3/ + 3t + I = 2 + 3t 2 + 3t + = 3( t 2 + t + l). Multiply by t - I:
. wt
3r-:l
3/:
t - t+l
3,: 2 ¥3(,- 2 - t + 1)
ma k mg - I = -'i/3- . Rat1ona
. 1·1ze tI1e d enommator:
. 'i/3
t+l t + I ,2 - t + I t3 +I
t2 - t + I
'J/9 . Replacing t by
3,:
w2 gives
(;z -?/9¥2, + 1
,i - 3. + .!_ [l.
Hence, a + b + c =
9 9 9
=
w·
76
RI-I. The complete graph of Ix I+ l2y I= I is shown on the viewing window ofa graphing calculator. When the
region enclosed by the equation is measured by a ruler, its area is 120 square units. Ifthc Xmin and Xmax
of the viewing window are tripled, the Y min and Ymax arc doubled, and the new graph of the equation is
measured by the same ruler, the area of the region is K square units. Compute K.
T
Rl-2. Let T = TNYWR. If K = 4 , compute the number of real values ofx for which the following equation is
(K+logx)
true: (log x ) · = 1.
Rl-3. Let T= TNYWR. Point N lies on y = x - 1 y=x-1 y = 21 -x

and Plies on y = 21 - x as shown. If the
N
coordinates of Mare (T, 0) and MNPQ is a
rectangle, compute the area of MNPQ.
M Q
ARML Relay #2- 1997
R2- l. Compute the number of integer values of n such that n 3 + n2 + n + 6 is divisible by n + 1.
R2-2. Let T = TNYWR. In parallelogram ABCD, EC is the A B
extension of BC. If the area of MDF is 81 and the

area of !iFCE = T, compute the area of ABCD.
~--------',~--1 C
D F
R2-3. Let T = TNYWR and set K = I...

2
If y = x is tangent to y = x 2 + Kx + N, then compute N.
77
Relay 1:
Rl-1. 20
Rl-2. 3
Rl-3. 32
Relay 2:
R2-1. 4
R2-2. 198
R2-3. 2401
78
Rl-l. From the diagram, 120 = 2ah. Multiplying Xmin and Xmax by k
multiplies the measured length by T. Thus, a ➔ ~, h ➔ ~ and the
Rl-2. WehaveT = 20soK = 5. RegardlessofK,x = JO is a solution. Iflogx = -K,then
(log x )(K+logx) = (log x )0 = I , so x = 10 -K is a solution. If K is odd and log x = -1, i.e., x

I,
= 10
then we have (-1 )2n = I. Thus, there are 2 or 3 solutions depending on whether K is even or odd.
Since K is odd, pass back [].
Rl-3. We have T = 3. Since NM= PQ = T-1, then 21 - x = T-1 ➔ x = 22 - T, giving

Q(22-T, 0). Thus MQ = (22 -T)- T = 22- 2T. Area NPQM = 2(T- 1)(11 - T) = 2(2)(8) = [TI].
n 3 + n2 + n + 6 n 2 (n + l) + (n + l) + 5 2 5
R2-l. = = n + l + - - . Hence, n + I divides 5, so
n+I n+l n+I
n + l = ± l or ± 5 ➔ n = - 6, - 2, 0, and 4, yielding@] solutions.
R2-2. Let DF = a, FC = b, and the heights of MDF and 1'1ECF be ha A B
andhb. Since ~MDF) =½ha2 = 81 and ~MCF) = ½hb 2 = T,
then ( ha 2 )( hb 2 ) = (2 • 81 )(2T) giving hab = l8ll. Since ~-----',----le

D F
a(ABCD) equals (ha)(a + b) = ha 2 + hab = 162 + 18ll, then E
a(ABCD) = 162+ 18i4 = 11981.
R2-3. We have T = 198, so K = 99. Set x 2 + Kx + N = x ➔ x 2 + (K - l)x + N = 0.
For the line to be tangent, this equation must have one solution, so (K - 1) 2 - 4N =0 ➔
2
N (K-1 J
= - 2- = 49
2
= 124011.
79
ARML Super Relay - 199 7
I. , If AB= DC= 24 and BC= AD= 32, compute the

area of MEC.
T 2 2 2
2. Let T = TNYWR and set K =- . If log 2 b - log 2 a = K, then b - a = Ma . Compute M.
100
3. Let T = TNYWR and set K = :i. Compute the largest angle 8 in [0°, 360°) such that cos(K0) = ½-
4. Let T = TNYWR. Let f(g(x)) = Tx + T and f(x) = x + 2T. If g(x) =ax+ b, compute a+ b.
5. Let T = TNYWR. If the side of square ABCD = T + 9 and

the areas of regions I, II, and Ill are in increasing arithmetic AITB
progression, compute the area of region II. I II III
D C
T I+ 7i
6. Let T = TNYWR and set K = -9 . If - - = a+ bi, compute a+ b.
4+ Ki
7. LetT = TNYWR. lfab=T,bc = f,andac = T 2 ,compute la+cl.
8. Let T = TNYWR. If AB= 3T, BC = 4T, AC = ST, and·

AD = 3T, compute the area of MDB.
C
L D
-
A
80
ARML Super Relay - 1997
T
9. Let T = TNYWR and set K = 30 . If 9 · K + AB = BA for two-digit base ten numbers AB and BA,
compute the largest possible value for the two-digit number BA.
10. Let T = TNYWR. If f(x) =~ is defined only on 0 ::; x::; T 2 , compute the number of lattice points on .f
11. Let T = TNYWR and set K =T - 6 . Given isosceles MBC with mLB = 90° and AB = K, circle O is
7
- - a
tangent to AC at D, the midpoint of AC, and passes through B. If the diameter of O is /;;, compute a+ b.
x-T T T-3x
12. Let T = TNYWR. If x - - -
4
= -3 + -6- ' compute x.
13. Let T = TNYWR and set K = 40T- 5. Each figure is a

square with a side half as long as the square to its right. If the
perimeter of the whole figure is K, compute the area of the
smallest square.
14. Let T = TNYWR. A retail store pays a wholesale price of$Tper toy. It prices the toy at $K so that at a 25%
off sale, the store still makes a profit of 20%. Compute K.
15. Let T = TNYWR. A linear function f is such that f(0) =T and f(2x) = 2/(x - 1) for all x. Compute
the area of the triangle formed by the intercepts of f and the origin.
81
ANSWERS ARML SUPER RELAY - 1997
I. 300
2. 63
3. 340°
4. 0
5. 27
6. 2
7. 5
8. 90
9. 96
10. 97
11. 15
12. 3
13. -254
14. 10
15. 10
82
Solutions to the ARML Super Relay- 1997
I. AABC and ~CDA are 3-4-5 right triangles, so AC = 40,
making ..!!.... = ~ ➔ h = 15 ➔ area = .!_ • 40 • 15 = 1300:.

20 4 2
b b K 2 22K. 2 2 2 2 2
2. log 2 - = K ➔ - = 2 ➔ b = a · 2 . Smee K = 3, b = 64a ➔ b - a = 63a ➔ M = [fill.
a a
I .
3. Cos Ke = 2 ➔ Ke = 60° + 360n or 300° + 360n. Smee K = 3, e= 20° + 120n or 100° + 120n.
Thus, the largest occurs when n = 2, and that makes 0 = 1340°1.
4. f(g(x)) = g(x) + 2T = Tx + T ➔ g(x) = Tx - T. Thus, regardless of T, a+ b = [m.
2 (T+9) 2
5. II - d + II + II+ d = (T + 9) ➔ II = -"----- Since T = 0, II = [n].
3
I+7i 4-Ki (4+7K) +(28-K)i 32+ 6K 50 __ '27.

6 - - · - - = ------------------ ➔ a+ b = - - - . Since K = 3 a+ b = - ~
4+Ki 4-Ki 16 +K2 16 +K2 ' 25
2 1 2 2
7. ab· ac · be = (abc) = T •- • T = T ➔ abc = ± T. Since ab= T, then c = ± I. If c = 1, then
T
a= T 2 . Ifc = -!,then a= -T 2 . Since I- T 2 -II= lr 2 +I I, la+cl is fixed. SinceT = 2,
83
. 4 I 9T 2 4 18T 2 3T A
8. sm8 = - ➔ area !-,.ADB = - · 3T ·3T-sin8 = - - --
5 2 2 5 5 D 3T
2T ' ' ,,
Since T = 5, the area MDB = 00. '
C B
4T
9. Before receiving T, try values for K: K = I ➔ 89 + 9 = 98, K = 2 ➔ 79 + 18 = 97, K = 3 ➔
69 + 27 = 96, ... Since K = 3, the largest value for BA is 00.
10. If 5 and Tare relatively prime, then y is integral when xis a multiple of T. On [ 0, T 2 ], x = 0, T, 2T, .. , T · T,
making T + I values for x. If T and 5 are not relatively prime, be patient. Since T = 96, the answer is [21].
AD 2 and AD = ABJi. AB2 A

1 1. Since AE · AB ➔
= -
2- , then AE · AB =
2
AE = AB, so the diameter EF of the circle is also the midline of

2
MBC ➔ EF
Ki2
= --. Since K = 13, EF
13
=ii. The sum = [Bl.
2
T
12. 12x - 3(x-1) = 4T + 2(T-3x) ➔ 9x + 3T = 6T -6x ➔ x =-.Since T = 15,x = [l].
5
13. 4x + 3(2x + 4x + Rx) ~ 4fu ~ K ➔ areF [ : ]' s;n,e T ~ 3, K ~ 115, and thcare, ~ [~]
3K 6T ST 8 24
14. - - - ➔ K - - - - .- = [ill.
4 5 5 5 5
15. lfx = 1, /(2) = 2/(0) = 2 · T. Thus, points (0, T) and (2, 21) lie on the line. The equation of the line is
y = Tx + T and its intercepts are (0, 1) and (-2, 0). Area = .!.. · 2 · T = T = [ill.
2 2
84
ARML Tiebreaker- 1997
1. Let S11 Ji] + ['2"] + [Ji] + ... + [i,;] where [x] is the greatest integer function.
=[
Compute the largest value of k < 1997 such that S1997 - Sk is a perfect square.
85
ARML Tiebreaker Solution - 1997
I. For 1936::; n::; 1997, [.fn"] = 44. So, for 1936::; k < 1997, S 1997 - Sk = 44(1997 k)
4 • 11 • (1997 - k). For the difference to be a perfect square, ( 1997 - k) must equal 11,
giving 1997 - k = 11 ~ k = I 19861.
86
'Team 'Round------------------------ ---- ------- ------ --------- ---- ........ ------------------- --------- ------ -- 89
Power Qyestion __________________________________________________________________ ------------·· ____ ----· 94
'lndividua( 'Round··-········-------------------------------····-------·····----------------------- l O1
'Re(ay 'Rounc( ______ ··-------·---······················-··---······-·-----------······------------········ l 05
Siper 'Re(ay····························--···--------------··········································----··· l 08
'Tie6reakers 113
-87-
THE 23 rd ANNUAL MEET
ARML continued to grow, continuing to establish itself as an intellectually challenging and

thoroughly enjoyable mathematics competition. This year there were 27 teams in Division A,
75 teams in Division Band 9 alternate teams for a total of 111 teams fielding upwards of 1680
students. Competition was very spirited with only 4 points separating the top two teams in
Division A. This year a team from Taiwan took part in the competition for the first time and this
led to an invitation for Mark Saul, president of ARML, to travel to Taiwan to meet with
educators to explore further ways in which Taiwan could participate in ARML.
Ted Alper and Bill Kling received the Samuel Greitzer Distinguished Coach Award.
Ted was a member of the Montgomery County ARML team in 1979 and 1980. In 1994 while
working for the Education Program for Gifted Youth at Stanford, he began recruiting students in
the San Francisco Bay area to form ARML teams. His teams quickly became very successful
and in 1996, the San Francisco Bay Area team won the A Division.
Bill has started math teams throughout his career. In 1982 he organized the first Upstate New
York ARML team and shepherded it through the next decade, eventually creating three different
ARML teams. This year was the last year that he planned on being involved with the team.
Matteo Paris received the Alfred Kalfus Founder's Award. While a junior in high school in
California in 1989, Matteo first heard about ARML and was determined to create a team. Then
the only site was at Penn State, but Matteo and Jeff Wall, another high school senior, organized,
recruited, and trained a team that eventually got to Penn State and finished in the top ten. Since
then Matteo has organized teams, become an Executive Board member as well as a coordinator
for the UNL V site. While a student he also won the Zachary Sobol Award.
Chris Clark of Western Massachusetts and Greg Tseng of Thomas Jefferson HS received the
Zachary Sobol Awards.
88
ARML Team Questions - I 998
T- 1. In acute-angled triangle ABC, mLA = (x + 15) 0 , mLB = (2x - 6) 0 , and the exterior angle at Chas measure
(3x + 9) 0 • Compute the number of possible integral values of x .
T- 2. Twooftheverticesofasquareare A(log 15 5,0) and B(O,log 15 x) forx > I. Theothertwoverticesliein

the first quadrant. Add the coordinates of all four vertices. The result is 8. Compute x.
au±l+I 5
T-3. If a1 = ~ = I and an +2 = a for n 2: I, compute a1 where t = 1998 .
n
T-4. Point Plies in a square ABCD of side 2 so

that PA = PD. lfthe two shaded regions
have equal areas, compute cos LPAD .
T- 5. Line e is tangent to the circle x 2 + .v2 = 1998 at T(a,b) in the first quadrant. If the intercepts off and the
origin form the vertices of a triangle whose area is 1998 , compute the product a · b.
{: : :
2
y
T-6. Compute all ordered pairs (x, y) such that:
2
X
T- 7. Four congruent circles are tangent to each other and tangent to

the edges of a sector as shown. If the straight edges are joined
to form a right circular cone with vertex at P, the radius of the
base would be 2/3 the slant height of the cone. Compute the
ratio of the radius of the sector to the radius of each circle.
T-8. Let S be the set of lattice points in the region defined by O :'S x '.S 3 and O :'S y '.S 3. Triangles are formed
by choosing three non-collinear members of S as vertices. Two triangles are distinct if they share no more
than two vertices. Compute the number of distinct triangles with an area of 3/2.
T- 9. Squares are erected on the sides of right triangle

D
ABC as shown. If BC= 68 and AC= 204,
compute the absolute value of the difference
between the areas of MDE and !1CEF.
T-10. Let f be a function whose domain is S = {1, 2, 3, 4, 5, 6} and whose range is contained in S. Compute the
number of different functions / which have the following property: no range value y comes from more than
three arguments x in the domain. For example, / = {(I, I), (2, I), (3, I), ( 4, 4 ), (5, 4), ( 6, 6) } has the property,
but g = { (I, I), (2, I), (3, I),( 4, I), (5, 3), (6, 6)} does not.
89
ANSWERS ARML TEAM ROUND - 1998
I. 20
2. 45
3. 2
1
4. -
2
5. 999
6. (¾,-HH,~J
7. 3
8. 92
9. 272
10. 44,220
90
T-1. Since an exterior angle of a triangle is equal to the sum of the two remote interior angles, we set
3x + 9 = (x + 15) + (2x - 6), but that is an identity and is of no help. Since the triangle is acute-angled, we
have several conditions to consider. First, mLA < 90° ➔ x + 15° < 90° ➔ x < 75°. Second,
mLB < 90° ➔ 2x - 6 < 90° ➔ x < 48°. Third, since LC is acute, then 3x + 9° is obtuse, giving
90° < 3x + 9° < 180° ➔ 27° < x < 57°. Considering all the conditions, we have 27° < x < 48°, giving
x E { 28°, 29°, ... , 46°, 4 7°} . Thus, the number of integral values of x is 12Ql.
T-2. The sum of the coordinates of the vertices is clearly h C(l,,a+b)
4a+4b= 4(1og15 5+log 15 x) =8. Thus, B(O, b)
log 1s(5x) = 2 ➔ 5x = 15 2 ➔ x = @31. b D(a + h, a)
A(a, 0)
I+! 2+1 3+1 2+I

T-3. Since a1 =I, a.. =I, a3 = - - = 2 a = - - = 3 a = - - = 2 a - - - - 1 and
L I '4 I '5 2 '6 3
I+I
°7 = - 2- = 1, the sequence is l, I, 2, 3, 2, l, 1, 2, 3, 2, ... and is clearly cyclic with a period of 5.
Since 1998 = 3 (mod 5), I 998 5 = 35 (mod 5) = 3 (mod 5) and for t = 1998 5 , a1 is the third element in
the sequence, namely [21
T----4. Clearly, MEA =!).CFD and MMA = MND. Since the B C B C
p
M h-_ _..,...___ -r-l N
shaded regions have equal areas, all four triangles are
congruent, making AB = PA, so MAD is equilateral

A ____ D A _________ D
giving cos LPAD = cos 60° = rn.
- b - a
T-5. Since the slope of OT = - , then the slope of PQ = - - , making the
a b
a2
equation of PQ equal y- b = -ba (x - a). If x = 0, then y- b =b Q
a2 + b 2 1998 -a ( ) . .
➔ Y= b = -b-. Ify = 0, then -b = b x -a g1vmg
x2 + y 2 = 1998
b2 1998
x=-+a
a a a
1 ( 199811998 j
The area of MOQ = 1998 = 2l- 0
- --b-) ➔ the product a · b = 19991.
91
Solutions to the ARML Team Questions - I 998
T-6. Subtracting the bottom equation from the top yields _v2 - x2 = 2 ➔ (x - yX x + y) = -2. Adding the
two equations yields 2x:v + 16 = x 2 + _v2 ➔ x 2 - 2xy + _v2 = 16 ➔ ( x - y) 2 = 16 ➔ x- y = ± 4.

1 . I
If x - y = 4, then 4( x + v)
·
=-2 ➔ x+ v
•
= - -2 . Add mg x- v
·
= 4 and x + •v = - -2'
. . . , f: .
we obtain x = -47 9 f
and - - . I x - v
4 •
.
= - 4 , we obtam x
7
= - -4·, v = -49 m s11111 ar asluon.
Thus, (x,y) = (; ,-; l(-f ,¾ J.

T-7. The perimeter of the cone's base = 27t( ~ J= 4;r . A B
The circumference of the circle containing the
sector is 21tr, so mATB - = 240°. Thus tangents to

each circle form a 60° angle, making PO = 2(OQ).
Since PT = r = OQ + 2 · OQ 3 · OQ, then ;Q = W.
T-8 From each vertical or horizontal strip of three squares,

8 triangles with area 3/2 can be formed, giving
8 x 3 x 2 = 48 triangles.
From each vertical or horizontal strip of 6 squares,

4 triangles can be formed. There are 2 such vertical and
horizontal strips for a total of 4 x 2 x 2 = 16 triangles.
On each diagonal of the 3 x 3 square 6 triangles can be

formed, 3 above and 3 below. For example, AABC,
AAEC, and MDC.
Four triangles can be put in each 2 x 2 square. Since there are 4 such 2 x 2 squares, there are 4 x 4 16
triangles. Total: 48 + 16 + 12 + 16 = 1:22].
92
T-9. From DB 11 AC we obtain LBDE = LEAC and since LDBE =LACE, then
DB AC . D
ti.DBE~ MCE. Thus, - = -. SmceAC = 204,
BE CE
A
DB = BC = 68, and CE = BC - BE = 68 - BE, we have
68 204
➔ BE = I7 and CE = 5 I. Then
BE 68 - BE
DE = J( 4 · 17) 2 + 17 2 = 11.Jli. WEA = /J.CBG by SAS so

DE CE
LEDB =LECF and since LEED =LFEC, then ti.DBE ~ /J.CFE by AA. Thus, BE = FE ➔
11,/u 51 51 4 · 51
-- - - ➔ EF = - - making CF = - - . The difference in areas equals
17 EF .Ju Jli
l[ . .Ju
51 4-51)
2 68 . 17 - ./u 16 · 17 = 12721. It can be shown in all cases that the difference in areas is equal
to the square of the distance from B to DE.
T-10. The sizes of the pre images under f of {I, ... ,6} must partition 6. The excluded partitions are those
containing a 4 or greater, that is 4-1-1, 4-2, 5-1, and 6. There are 6 · 5 · 4 · ( 6 C4 ) = 1800 functions of the
4-1-1 type since there are 6 ways to pick a range element that four domain elements map to, 5 · 4 ways to
pick a range element that one domain element maps to, and 6 C 4 ways to pick four elements in the domain to
map to the single range value. Similarly, there are 6 · 5 · ( 6 C 4 ) = 450 ways to get preimages
of size 4-2, 6 · 5 · ( 6 c5 ) = 180 ways to get 5-1, and 6( 6 C 6 ) = 6 ways to get 6. Thus, there are
1800 + 450 + 180 + 6 = 2436 excluded functions. The total number of mappings from S to Sis
66 = 46656, so the number of functions with the desired property are 46656 - 2436 = 1442201.
93
ARML Power Question - 1998: Meditations on Partitions
For each of the following problems write complete solutions. Justify each answer. Angles are in degrees.
I. Let positive integers A, B, and C be the angles of a triangle such that A ::; B ::; C.
a) Detem1ine all the values that each of A, B, and C can take on.
b) Compute the number of ordered triples (A, B, C) in which B = 70°.
2. In convex pentagon ABCDE, mLA < mLB < mLC < mLD < mLE. Let T = mLC + m LD.
If mLA : mLB : mLC : mLD : mLE = l: 2: x: y: 5, determine the range of values of T.
3. Let a, b, and c be positive integers such that a < 3b, b > 4c, and a+ b + c = 200.
a) Determine the largest value that c can take on.

b) Determine the smallest value that b can take on.
c) Determine the number of ordered triples (a, b, c) in which c = 11.
4. Let a, b, and c be positive integers. If a+ b + c = 85, c > 3a, 2b > c, and 5a > 3b,
prove algebraically that there is a unique solution (a, b, c) to this system.
5. A unit square is divided into 4 rectangles of positive area by two cuts parallel to the sides of the square.
Let a 1 s a2 s a3 s a4 be the areas of the four parts in non-decreasing order. For each
i = l, . . . , 4, determine with proof the range of values for a ..

I
6. A unit cube is divided into 8 parallelepipeds of positive volume by three cuts parallel to the faces of the cube.
Let v1 s v2 s ... s v8 be the volumes of the eight parts in non-decreasing order. Determine with proof the
range of values for v4 and v5 .
94
ARML Power Question - 1998: Meditations on Partitions
7. Let n be a positive integer. Allie and Bob play a game constructing a partition n = a 1 + a 2 + ... + ak with
a 1 ::C: a 2 ::C: ••• :2: a k ::C: 1 . Allie wins if there is an odd number of terms in the partition, i.e., if k is odd; Bob
wins otherwise. Allie begins by choosing a number a1 between I and n - I inclusive. Bob then chooses a
number a2 between a 1 and I inclusive such that a 1 + a2 :,; n. Allie then chooses an a3 between a2 and I
inclusive such that a 1 + a2 + a 3 :,:; n, and so on, with the game ending when the partition is complete.
Determine with proof all n > I for which Bob has a winning strategy.
8. Allie and Bob play a game similar to the one in #7 except that the inequality a. ::C: a. 1 is replaced by
l l+
2a. ::C: a. 1 . Prove that Bob has a winning strategy if and only if n is a Fibonacci number. (You may assume
l 1+
the following: each positive integer n can be uniquely represented as a decreasing sum of non-adjacent
Fibonacci numbers, i.e., 32 = 21 + 8 + 3.)
95
I a. The maximum of C occurs when A = B = 1° ➔ C = 178°. The minimum of C occurs when

A = B = C = 60°. The maximum of B occurs if A = 1° and Bis as close to C as possible.
Since B + C = 179°, B = 89° and C = 90°. The maximum of A occurs when A = B = C ➔
A = 60°. Thus, 1° 5 A 5 60°, 1° 5 B 5 89°, 60° 5 C 5 178°.
lb. IfB = 70°,thenA+C = 110°andC 2'. 70° ➔ A::; 40°. ForA 1° to 40°, B = 70°, and
C = I 09° to 70° respectively. The number of ordered triples is 40.
2. For an upper bound of T consider an extreme case: let x = y = 5, giving lk: 2k: 5k : 5k: 5k for the angles.
If 18k = 540°, then k = 30° ➔ mL.C + mL.D < 150° + 150° = 300°. For a lower bound of T, note that
5mL.A = m LE < 180° so mL.A < 36° , making mL.C + mL.D = 540° - mL.A - mL.B - m LE =
540° - 8mL.A > 540° - 8 · 36° = 252°. This bound is the best possible. Let Ebe very small and let
( mL.A, mL.B,mL.C, mL.D,mLE) = (36° - E, 72° - 2£, 100°,152° + 8£, 180° - :€). These values satisfy the
inequalities and mL.C + mL.D is as close to 252° as we like. Thus, 252° < T < 300°.
3a. If a 2'. 1, then b + c ::, 199 and since b > 4c, then 4c + c < 199 ➔ c < 39.8 ➔ the maximum
of c is 39, achieved when (a, b, c) = ( 1, 160, 39).
. b b b
3b. Since 3b > a, then 3b + b + c > 200 ➔ 4b + c > 200. Smee > 4c, > c 4 ➔ 4b + 4 > 200 ➔
b > 47.06 ➔ theminimumvalueofb = 48,achievedwhen(a,b,c) = (141,48, 11).
3c. If c = 11, then a+ b = 189 and since b 2'. 48, then a ::, 141. All such (a, b, c) satisfy the two inequalities,
so the ordered triples range from ( 141, 48, 11) to ( 1, 188, 11 ), a total of 141.
c c . 9b
4. Given that a + b + c = 85 and c > 3a, then 3 > a giving 3 + b + c > 85. Smee 5a > 3b, then 3a > 5
giving c > 9: ➔ 5: > b. Thus, f + 5: + c > 85 ➔ c > 45. Similarly, 2b > c implies
C C l~ ~
b > - giving a+-+ c < 85. Also, 5a > 3b ➔ - > 2b > c, making a>-. Thus,
2 2 3 10
3c C
- + - + c < 85 ➔ c < 47.22 ➔ c 46 or 47. If c = 46, then a+ b = 39 and 2b > c ➔
10 2
2b > 46 ➔ b 2'. 24. Also, c > 3a ➔ 46 > 3a ➔ a ::, 15. Clearly, (15, 24, 46) works, but if
96
a :S 14makingb 2 25,5aisnotgrcaterthan3b. Ifc = 47,2b > 47and47 > 3agivesb 2 24and

a :S 15, but since a+ b = 38, a =I- 15 so a :S 14, but then 5a is not greater than 3b. Hence, no solutions if
c = 4 7. The only solution is ( 15, 24,46).
I
5. Let 0 < x s y s-. This is equivalent to x :S I - x, y :S I - y, x :S y, and
2 1-y~
therefore I - y :S 1 - x. Then a 1 s a 2 s a 3 s a4 can be written as
"Y s x(I - y) s y(l - x) s (I - x)(l - y). As x, y approach 0, the greatest lower
yt;t;J
X 1-X
bound of a 1, a2 , and a3 is clearly 0, and the least upper bound of a4 = I.
.
Smee a 1 = 4I( a 1 + a1 + a 1 + a1) s 4I(a 1 + °2 + a3 + a 4 ) = 4I , then a1 s 41 with equality when
x 1. Now
= y =2 a3 = 21( a 3 + a 3 ) s 2
1( a 3 + a4 ) < 2I . If y 1, and x approaches 0, then
=2 a3 approaches,
2
but does not equal -1 . Since a2 = x(I - v) s x(I - x) = -I - ( -I - x ) s -1 , then a7 s -1 with equality
2 ' 4 2 4 - 4
when x = y = 2I . Finally, a4 is minimized when x and y are as large as possible, i.e., x = y = 2I , making
l I I I
a4 = 4 . Thus, 0 < a 1, a2 s 4 , 0 < a 3 < 2 , 4 s a4 < I.
6. Method I: Without loss of generality, take x :S I - x, y :S I - y, z :S 1- z, and x :S y :S z which implies

that I - z :S I - y :S 1 - x. Clearly v4 > 0. To find the upper bound, divide the volumes
into 4 pairs:
{xyz,(l - x)(1 - y)(1 - z)} {(1 - x)yz,x(I - _v)(1 - z)}

{x(1 - _v)z,(l - x)y(I- z)} {x_v(I - z),(1- x)(1 - _v)z}
Consider the lesser of each pair. Since v4 is the fourth smallest volume it must be less than or equal to one
such lesser pair member. But the product of the numbers in each pair is x( I - x) y( I - y )z( I - z) . Since
A(I - A) s ..!_ for positive A, the product of each pair is at most _!_, which means that the lesser of each pair
4 64
is at most .!.. . Since, if x = y = z = _!_, then all eight volumes equal .!.. , and the upper bound is attainable.
8 2 8
I I
Thus, 0 < v 4 s8 For v5 ,notethat 4v5 s v5 + v6 + v7 + v8 < !,so 0 < v5 < 4.
This solution cleverly avoids a thorny problem, addressed in method #2, that volumes ( I - x)yz and
97
x( I - y )( I - z) which correspond to v4 and v5 respectively, can switch order, i.e., if x = ¾and y = z = ½,

. I I I 1 I 9
we obtam v4 = -5 and v = - but if x = z = - and v = - we obtain v = - and v -
5 20 ' 2 - IO ' 4 40 5 40
Method #2: The volumes can be written as 1 = v .l}'Z, v2 = ~( 1 - z), v3 = x( 1 - y )z,

v4 = min of ( I - x),vz or x( I - y )( I - z) , v5 = max of ( 1 - x) yz or x( 1 - y )( 1 - z) ,
v6 = (1 - x )J{ I - z) , v7 = ( I - x)( 1 - y) z , and vS = (1 - x )(1 - y)( 1 - z) . Consider the product
than i, with equality at x = y = z = ½. Since in either ( 1- x)yz or x( 1 - y )( 1 - z), one of the variables
can go to 0, then 0 < v4 ::; i. For v5 , if x, y, and z are very small, both (1 - x)yz and x( l - y )( l - z) are
very small, making O the lower bound. For the upper bound, note that
3
1 1 [ 2x+(1-x)+(1-x) ]
x( 1 - y )( 1 - z) ::; x( 1 - x)( l - x) = 2 (2.x( I - x)( I - x)) ::; 2 3 by the Arithmetic-
. equals 4 . () 11
27 . Smee 1 - x yz < 1 · 2 · 2 = 4 , t en by
Geometric Mean Inequality. The last expression lh
v5 = max of ( I - x)yz or x( 1 - y )( l - z) we choose v5 < ¾. This is the best possible bound since if
I I
y = z = 2 and x approaches 0, then v5 approaches, but does not equal 4.
7. Claim: Bob has a winning strategy if and only if n = 2k .

Proof: Consider this as a game of removing matches from a pile of n matches.
Observe what happens in simple cases:
1. If n is odd> 1, Allie has a perfect winning strategy. She takes I match, i.e., a1 = 1, leaving an even
number. Bob is forced to take I match, leaving an odd number, and this continues until Allie takes the last
match.
2. If n = 2, Bob has a winning strategy. Allie must take I match and then Bob takes the other.
3. If n is even, greater than 2 and not a multiple of 4, Allie has a perfect winning strategy. She takes 2
matches, leaving a multiple of 4. Bob can take I or 2. If he takes I, he loses by (I). If he takes 2, then
n - 4 matches are left. If, initially, n = 6, then there are two matches left and Allie can take them both.
98
Solutions to the ARML Power Question - l 99R
Otherwise. 11- 4 matches represent an even number, greater than 2, and not a multiple of 4. so Allie can
repeat the process, eventually reaching 6.
4. If 11 = 4, Bob has a winning strategy. Allie will not take I or 3, she'll take 2, but then
Bob takes 2 and wins.
Now let p(q) representthehighestpowcrof2dividingq. Thus, p(l7) = 2° = 1, p(24) = 8. Consider the

following lemma: if at a certain stage there arc q matches left, and if the next person to play is allowed to
take p(q) matches, then that player can guarantee a win by removing p(q) matches.
Proof by induction: The result holds for q = I since p( 1) = 2° = 1. Assume it holds true for all positive
integers less than q. Suppose it is X's tum. There are q = 2m · s matches left for s an odd number, man
integer ~ 0. X removes p( q) = 2m matches leaving 2 m · s - 2m. If q = p( q) , i.e., ifs = I, then X has
won. Otherwise, although Y is looking at a number of the form 2m(s - 1) which is divisible by a higher
m . m w
power of2 than was 2 . Now Y isn't allowed to remove more than 2 . Suppose Y removes 2 · t matches
fort odd. Then there remain 2m(s - 1) - 2 w · t = 2a · b matches for b odd. Clearly, w = a, and so Xis
allowed to remove p( 2a • b) = 2a matches, and thus, by the inductive hypothesis, X wins.

Applying this lemma to our problem: if n is not a power of 2, Allie can win if she plays perfectly by
removing the highest power of 2 dividing n. If n is a power of 2, and Allie removes a certain number of
matches, leaving q, then Bob is allowed to remove p(q) matches, so Bob can force a win.
8. The following example may help: Let n = 13. Suppose Allie takes 2 leaving 11 = 8 + 3. Bob takes away 3,
leaving 8. If Allie takes away more than 2. she loses immediately since Bob will take the rest. If she takes away
I, leaving 7 = 5 + 2, then Bob takes away 2. If she takes away 2, leaving 6 = 5 + I, then Bob takes away 1.
Allie loses if she takes 2, so she takes 1, leaving 4 = 3 + I. Bob takes away 1 and Allie will now lose if she
removes either 2 or 1. Each time Bob partitions the remainder into non-adjacent Fibonacci numbers and takes
the least of those. If n had been 15, Allie could take away 2, leaving 13 and the tables are reversed, and so she
can now win by following his strategy. If n had been 32 = 21 + 8 + 3, Allie can take away 3 and by taking away
the smallest Fibonacci number in each partition, she will eventually become the second player in a game starting
with a Fibonacci number and will win.
To establish Bob's winning strategy, we must establish 2 things: I) that it is always possible to remove the
smallest Fibonacci number in a partition of a remaining amount, and 2) if a player is able to remove the
smallest Fibonacci number in a partition of a remaining amount, then that player has a winning strategy.
99
Solutions to the ARML Power Question -1998
Lemma I: If s < F., then the smallest Fibonacci number in the partition of F. - s is less than 2s.
I I
(This means that the smallest Fk can be chosen.)
Lemma 2: If a player is faced with q matches and is allowed to remove the smallest Fibonacci number
in the partition of q, then that player has a winning strategy.
These two lemmas solve the problem in this way: if n is not a Fibonacci number, then Allie can remove the
smallest Fibonacci number in the partition of n and force a win by Lemma 2. If n is a Fibonacci number F ,
11
then Allie must remove a numbers which is less than F . By Lemma I, Bob is able to remove the smallest
11
Fibonacci number in the partition of F - s and then he can force a win by Lemma 2.
11
Proof Lemma I: Let F - s

n
= Fa + ... + Fak for non-adjacent decreasing F . We must show that
ai
1
1 1 1
s > -F <==> F - s + -F < F which would follow from F + ... + F + - · F < F 1 .
2 ak n 2 ak n a1 ak 2 ak a1 +
1
But by a short induction and the fact(*) that F11 > 2 · F + 1 we have:
11
F Fq + Fq
q +I - 2 + Fq - 4 + · ·· + Fal - 2( k - I) + Fq - 2k + I
1
~ Fa1 +Fa1 -2 +Fq-4 + ... +~ 1-2(k-l) +2Fa 1-2(k-l) by(*)
1
~F +F + ... +F +--F
q a2 ak 2 ak
Proof Lemma 2: We prove this by induction on q. If q = I the result is clearly true. Assume the result for
all positive integers less than q. Let q = F + ... + F where F are non-adjacent. Then the first player
al ak ai
removes F matches leaving q - F = F + ... + F . If k = I, the first player has won. Otherwise,
ak ak a, ak-1
the next player can remove up to 2 · F matches, a number that is less than Fa _ 1 since
~ k
q- F - s = F + ... + F + T where Tis a sum of non-adjacent Fibonacci numbers smaller than

ak a, ak-2
F . By Lemma I, the first player can remove the smallest Fibonacci number in the partition of Twhich is
ak-1
the smallest Fibonacci number in the partition of F + ... + F +T = q- F - s matches and so by the
q ~-2 ¾
inductive hypothesis that player can force a win.
100
1-1. Numbers are to be placed in the nine squares so that the sum of the elements
of each row and of each column are equal. If I, 9, 9, and 8 are placed as
shown. compute X + Y - Z.
I-2. Coefficients a, h, and c of ax 2 + hx + c = 0 are selected without replacement from

N = {-4.-3,-2,-1, 1,2,3,4}. Computetheprobabilitythatx = I isasolutionofthccquation.
1-3. If the sum of the positive three-digit base ten numbers ABC, CAB, and BCA can be factored into four distinct
primes, compute the largest possible value of the product A • B • C.
1-4. The sides of a non-right isosceles MBC arc sin x, cos x, and tan x. Compute sin x.
I-5. Compute: 199919981997 2 - 2 -199919981994 2 + 199919981991 2 .
I-6. Let x be the remainder when xis divided by n. For positive integral x, compute the sum of all elements in the
n
solution set of: x 5( x5 )5 - x6 - (x 5 )6 + x(x5 ) =0
1-7. Compute the smallest number d such that fewer than half of the positive integers with d-digits have all distinct
digits.
1-8. In concave hexagon ABCDEF, mLA = mL.B = mL.C = 90°, mL.D = 100°, and mL.F = 80°.
Also,CD = FA,AB=7,BC= 10,andEF+DE = 12. Computethcareaofthehexagon.
IOI
1. 15
l
2.
14
3. 648
-I+Js
4.
2
5. 18
6. 1300
7. 5
145
8. = 36.25
4
102
Solutions to the ARML Individual Ouestions - 1998
1-1. The top row and the right hand column give I + Y + 9 = 9 + Z + 8 ➔ y
I 9
Y - Z = 7. Using the left hand column and the bottom row, 9 X z
I+ 9 + A = A+ B + 8 ➔ B = 2. Thus I+ Y + 9 = Y + X + 2 ➔
A B 8
X = 8, makingX + Y- Z = l.12.
1-2. Ifx ~ I isasolutionthena+b+c = 0. There are 8 c 3 = 56 unorderedtriplesofa,b,andc,andofthose,

thefollowingunorderedtriplessatisfya+h+c = 0: (--4, 1,3),(4,-1,-3),(-3, l,2)and(3,-l,-2). Thus,
the probability is .!.

56
= ITT_
l_iiJ
1-3. (lOOA+IOB+C)+(IOOC+lOA+B)+(lOOB+lOC+A) = lll(A+B+C) = 3·37(A+B+C).

The maximum value of A+ B + C = 27, but that is not factorable into the product of two primes.
Let A+ B + C = 26 = 2 · 13, making the sum equal to 2 · 3 · 13 · 37. Let A = 9, B = 9, C = 8,
making A · B · C = 9 ·9 ·8 = 16481.
1--4. If sin x = cos x, then x = 45°, the third side = tan 45° = I, but the sides are ../2.I , ../2.1 , and I, making a
right triangle. If sin x = tan x, then sin x = 0 or cos x =I which makes sin x =0 and no triangle can be
.-
1orme d . Thus, cos x = tan x ➔ cos~" x .
= smx ➔ ·" x
1- sm~ .
= smx ➔ · 2 x + smx-
sm . l =0 ➔
.
smx = 1-i+Jsl
2 .
1-5. Let p = I 99919981994. Then the expression can be written as (p + 3) 2 - 2p 2 + (p - 3)2
p
2
+ 6p +9 - 2 p - +p 2 - 6 p +9 = Ll8.J.
? 1io7
• <; 5 5 5
x 5 = 0, 1, 2. 3. or 4. Hence, from the first equation x = 1-, 2 ,3 or 4 and from the second
equation x = 1. The sum of the elements in the solution set is 1 + 32 + 243 + 1024 = I13001.
103
I-7. The table shows the fraction of d-digit numbers with distinct digits:
d=I 9/9 = I
d = 2 (9/9) · (9/10) = 0.9
d = 3 (9/9) · (9/10) · (8/10) = 0.72
d=4 (9/9) · (9/10) · (8/10) · (7 /10) = 0.504
d = 5 (9/9) · (9/10) · (8/10) · (7/10) · (6/10) = 0.3024 ➔ The answer is [3].
1-8. Method # 1: Since the sum of the interior angles of a hexagon is 720°, then
360° - mLFED = 720° - 3 · 90° - I 00° - 80° = 270°, so mLFED = 90°. Duplicate the hexagon, rotate
each copy through a multiple of90° and obtain the figure shown below, a square of side 12 oriented inside a
square of side 17. Theareaofthehexagonis ~(17 2-122) = 17 = 36.251.

lO 7
F Y A 7
10
)'
10
7
7 10
Method #2: Complete rectangle ABCG as shown. Then
~ABCG) = a(ABCDEF) + a(WEF) + a(MGD) giving G 10-y F y A

~ ,"'"'
1 1 7-y : ,,,,,,,,,, X
70 = a(ABCDEF) + 2-x · (12 - x) + 2 (7 - y)(lO - y) I

I _,_,
,''
I _,_, E
D ,, 7
140 = 2(a(ABCDEF)) + 12x - x 2 + 70-17y + y2 y
~------~'B
70 = 2(a(ABCDEF)) + 12x- x 2 - l 7y + y2. C
Also, since FD is the hypotenuse of both !iFED and !iFGD we have
x 2 + (12- x}2 = (10 - y}2 + (7 - y) 2 which yields 2x 2 - 24x + 144 = 149 - 34y + 2/ which simplifies
tox 2 - 12x = %- 17x + y 2 . Replacing 12x - ./ in the first equation by -( %- l 7x + y2 J gives
70 = 2( a(ABCDEF))-(½- I 7x + y2 J-17 x + y2. The x-terms and the /-terms cancel yielding
5 145
70 + - ~ a( ABCDEF)) ➔ the area of ABCDEF equals - .
2 4
104
ARMLRelay-""=#""'J=-==='=19===9="8===========================
RI-I. In rectangle ABCD, the areas of MDE, t,.CBE, and MEE

form an arithmetic sequence. If the ratio of the area of the
largest region to the smallest can be expressed in simplest
m
tenns as - , compute m + n. D E C
17
R 1-2. Let T = TNYWR. The digits T, T + I, T + 2, and T + 3 are placed at random in the indicated spaces:
I_ 9 _ 9 _ 8 _ . Each digit is used, one per space. Compute the probability that the resulting
number is divisible by 3 or 5.
x I 1 1
Rl-3. Let T = TNYWR. Compute - in simplest terms if x + - =- and y + - = T.
y y T X
R2- l. A box 4 by 6 by 8 is resting on the floor. A box 2 by 3 by 5 is placed on top of the first box forming
a two-box tower. If A is the exposed surface of the tower, compute the least possible value of the area of A.
T-12
R2-2. Let T = TNYWR and let K =- 4- . In the regular n-gon A 1A 2 ... A 11 , the measure o.f LA2 A 1A4 =K .
Compute the value of n.
R2-3. Let T = TNYWR and let f be a linear function with positive slope passing through the origin. Compute
the least value of /(5r) + f- 1(5r).
105
ANSWERS ARML RELAY RACES-1998
Relay #1:
Rl-1. 4
1
Rl-2. -
4
Rl-3. 16
Relay #2:
R2-l. 192
R2-2. 8
R2-3. 80
106
R 1-1. Let the area of AADE = a, the area of f...CBE = a+ d, and the area of AAEB = a+ 2d. Since the area of
AAEB is half the rectangle, the sum of the areas of the smaller two triangles equals the area of AAEB.
Thus, a+ (a+ d) = a+ 2d ➔ a = d ➔ the areas are a, 2a, and 3a. So, 3a/a ➔ 3 + I = @J.
R 1-2. T = 4. Since I + 9 + 9 + 8 = 27, the number is divisible by 3 if

T + (T + I) + ( T + 2) + ( T + 3) = 4T + 6 is divisible by 3. The number is divisible by 5 if the last digit is
0 or 5. Without knowing T, the following probabilities can be calculated:
T T+l T+2 T+3 T T+I T+2 T+3

0 2 3 P=I 2 3 4 p = 0
2 3 4 5 P= 1/4 3 4 5 6 P=I
4 5 6 7 P= 1/4 5 6 7 8 P = 114
6 7 8 9 P=
Since T = 4, the probability equals~-
J Y Y X 1
Rl-3. T = - . Subtracting ~' + 1 = xT from .,\JI + I = - yields - = xT giving - = - 2 = J]].
4 T T y T
R2-l. The minimum exposed area occurs if both boxes arc placed with the side of largest area face down. Note
that the 3 by 5 side of the small box is hidden and also hides an area of 3 by 5 on the top face of the larger
box. The minimum area is:
2(4·6 + 4·8 + 6·8)- 6·8 + 2(2·3 + 2·5 + 3·5) - 2(3·5) = 1192.
R2-2. T = 192 ➔ K = 45°. Since A1A4 II A 2 ~, the measure A 1 ~----A~~·----------

45° 450
. 360
of the extenor angle at A2 equals mLA 1 . Thus, K = --;;
. 360 360
makmg n = - = - = IBJ.
K 45
R2-3. T=8.Given.f:y=1m,f - I :x=my x

➔ y=-.Hencc,f5T
( ) +f - I (5T ) =5Tm+-
5T =
m m
5T( m + !). Since the minimum of m + ! = 2, the minimum of .f( 5T) + f- 1(5T) = IOT = [8Q].
107
Note: Pass answers from position l to position 15.
I. At which value of x does y = (x - IX x - 19) + 7 take on its minimum value?
2. Let T = TNYWR. If a
n
= an-I2 +3T - a 11 _ 2 and all ai are equal, compute a 1998 .
3. Let T = TNYWR. An auditorium has Trows ofTseats and each row is numbered from I to Tby l's.
Compute the number of even-numbered seats.
4. Let T = TNYWR. In rectangle ACDF, AC = 4Tand CD = T. A B C
~
Right triangle FBD is congruent to MEC. Compute the ratio of
the area of the shaded region to the area of ACDF.
F E D
I K
5. Let T = TNYWR and set K =- . If the equation x = -- has a unique solution, compute x.
T K -x
6. Let T = TNYWR. The first term of an infinite geometric series is T and the limit of the sum is I 0.
Compute the common ratio.
7. Let T = TNYWR. Compute the number of factors of ( 10 4 ) · T.
T
8. Let T = TNYWR and set K = 7 . ln square ABCD, DP = BP = 2K,
mLI = mL'.2 = 30° . The area of the square can be written in simplest form
as a+ b./3. Compute a+ b.
108
9. Let T = TNYWR, T -f. 1. lf the ordered pair (x, y) is the solution to the system, compute x + y.
X + 2y 3
Tx+(T+l)y = T+2
10. Let T = TNYWR. Shown are two equilateral

triangles ABC and CDE of side T.
If AD = /ii, compute N.
11. Let T = TNYWR. If sin 0 + cos 0 _!_, compute sin20.

T
12. Let T = TNYWR. Let T be the slope of a line passing through (-14, 16). Compute the x-intercept.
13. Let T = TNYWR. Compute the number of integers from l to HY inclusive that are divisible by 4 or 5.
4T
14. Let T = TNYWR and set AB = - - - I . PC is tangent to
1000
p
the circle, PA = 1, and PC = x. Compute x.
15. Let T = TNYWR. Compute the largest value of x satisfying x 2 - Tx + y 2 - 8 y =8.
109
ANSWERSARMLSUPERRELAY-1-99-8=================~=
1. 10
2. 10
3. 50
1
4. -
4
5. 2
4
6. -
5
7. 28
8. 96
9.
10. 3
8
11.
9
12. 4
13. 4000
14. 10
15. 12
110
I. Minimum occurs at x = _..!:...._ Since y = x 2 -20x + 26, the minimum occurs atx =
2a
mn
a +3T
2. a = ....__ _ - a
n 2 n
➔ 2a
11
= a
11
+ 3T - 2a
n
➔ 3a
n
= 3T ➔ a
n
= T. 0 1998 = [ill
3. If Tis even there arc ( ~) · T even-numbered scats and if Tis odd there arc ( T ~ 1J·T even-numbered seats.
T = IO gives 5 · IO = G:TIJ even-numbered scats.
1
4. Tis irrelevant. Connect B and E. The area of the shaded region lying in ABEF is 4 · a( ABEF). The area of
the shaded region in BEDC is :i • a(BEDC) so the ratio of the shaded region to ACDF is~-
5. Simplifying, we obtain Kx - x 2 = K ➔ x 2 - Kx + K = 0. The discriminant is K 2 - 4K which gives a

1
unique value for x if K = 0 or 4. If K = 0, x = 0. If K = 4, x = 2. Since T = - , K = 4,
4
k . - "57
ma mgx - 1_.'.._J.
6. T
Set 10 =--,then 10-T
r = - - - . Since T = 2, r = ~
- .
1 -r 10 5
4 4
7. T = - and l O 000 · - 8000 = 8 · 1000 2 6 · 5 3 . There are 7 · 4 = [2B] factors.
5 ' 5
8. K = 4. Since DP = BP = 2K, then PE = K, DE = K/3 and A B
l~F
,.,
side DC= KJ3 +K. Then DC 2 = (Ki3 + Kr = 4K 2 + 2K 2../i,.
,., 2
So, a+ b = 6K- = 6- 4 = ~-
D E C
11 1
Solutions to the ARML Super Relay-1998
9. Multiply the top equation by Tand subtract giving 2Tv - (Tv + y) = 2T- 2 which gives y(T- 1) = 2(T - 1)
soy ~ 2. Substitution yields x = -1, so the value of Tis irrelevant and x + y = [I].
3T
10. Drop an altitude from D to point Fon CE. Then AF = - and BF =
r./3 Thus,
2 2
AD
2
= [ 2)
3T 2 [ T-.,3
+ - 2-
. r::]2 2 .
= 3T . Smee T = I and N =
2
AD , then N = [].
11. (sin 8 + cos 8 t'°)
-
T2
➔ sin 2 8 + 2sin 8cos 8 + cos 2 8 = I + sin28 = - 1- ➔ sin28 = - 1-
T2 T2
- I.
Since T = 3, sin28 = ¾-1 = 1-¾I-
12. y - 16 = T( x + 14). Let y = 0. Then x = -

1: - 14. Since T = -¾, x = -16[-f) - 14 = @l.
100 100 JOO

13. If T = 1, then there are 4 divisors. If T = 2 then - +- - - = 40 divisors. If T = 3, then
' 4 5 20
1000 1000 1000 . . . 10000 I 0000 10000
- - + - - - - - = 400 d1v1sors 1f T = 4 then - - + - - - - - = 4000 divisors, and so on.
4 5 20 ' ' 4 5 20
Here T = 4 so the answer is 140001.
14. AB = 15. Since PA· PB = PA(PA + AB) 2

PC . then l 2X) -l(2 + 15
( X )
= x
2
➔
X
2 + 15 = 2x. Thus,
3x
15 =- ➔ X = [ill].
2
15. x 2 -Tx+ r 2 +y 2 r2 ( r )2 + (
y-4 )2 = r 4+ 96 .
2
4 -8y+l6=24+ 4 ➔ lx- 2 Thelargestvalueforx
occurs when the expression involving y is 0, so let y = 4, then x

T+
= -----.
Jr 2 +96
Nice values result
2
if T = 2, 5, 10, 23, .... Since T = IO, then x = [U].
112
ARML Tiebreaker - 1998
I. Let circles O and Q have a common chord PS . If OQ = 324 and MN: NT = 2 : I,
compute OP - PQ.
o-
113
p
I. Let the radius of circle O be Rand the radius of circle Q
be r. Let MN = 2x and NT = x. Then ON =R - x
and QN =r- 2x. Using AONP we obtain
(R - x) 2 + h2 = R 2 and using AQNP we obtain

2
(r - 2x) + h2 = r 2. The two equations simplify to M N T
7 2 2 2
x- - 2R, + h =0 (1) and 4x - 4x + h =0 (2)
Subtracting ( 1) from (2) and dividing by x yields the condition: 3x = 4r - 2R.
Since OQ = 324, then OQ = R + r - 3x = 324 ➔ 3x = R + r - 324.
Thus,R+ r - 324 = 4r - 2R ➔ 3R - 3r = 324 ➔ R - r = 11081.
114
1999
'Team 'Round 117

Power Qyestion ________________________________________________ ----------------------------------------- 12 2
'lndividua( 'Round-------------------··············----····---------··-···------····----·-·-······· 129
'Re (ay 'Rou nc( .. _. ________ ·---··----------------------------------. ·-------······· ····----··-------------- 13 3
Su_per 'Re(aY----·-·----------------·--···········----·-·······----·-----------·-·--------·---------········· 136
'Tiebreakers 141
-115-
THE 2411i ANNUAL MEET
One of the most exciting developments this year was the establishment of a contest modeled on
ARML Taiwan. In cooperation with the Nine Nine Cultural and Educational Foundation,
ARML helped establish the Taiwan Regional Mathematics League called TRML. The first
contest was held in August, 1999. This year there were 24 teams in Division A and 81 teams in
Division B. From these 105 teams some 1575 students participated. For the national title, San
Francisco Bay A narrowly edged out Massachusetts A. The difference was the score on the
individual round. We had three teams from Taiwan competing this year and the C team earned
the top score in Division B.
John Goebel of North Carolina received the Samuel Greitzer Distinguished Coach Award. John
has played an active role in support North Carolina's ARML team and also has been crucial to
the development and success of the North Carolina Mathematics League.
Tom Kilkelly of Minnesota and Richard Kalman of New York City received the Alfred Kalfus
Founder's Award.
Tom has helped immeasurably with the Minnesota ARML team and has helped create and
sustain the ARML Power Contest. He has written most of the contests, he has done much of the
grading, and, of course, he has been responsible for organizing the contest.
Richard published the ARMLog for 10 years and since the mid l 980's he has chaired the Power
Question grading at Penn State. ARMLog provided information about the contest and its results
and also introduced readers to the coaches and organizers of the meet as well as providing
articles on mathematics and mathematics education. Richard did a wonderful job of pulling this
all together.
Kurt Lazaro of Maine received the Zachary Sobol award for his contributions to his team.
116
ARML Team Questions - 1999
C
T-1. If AB= 2,BC = 6,AE = 6,BF = 8,CE = 7,andCF = 7,
compute the ratio of the area of quadrilateral ABDE to the
area of f:i.CDF .
£
F
T- 2. Compute the number of ordered triples of integers (x, y, z), 1729 < x , y , z < 1999 which satisfy:
2
x + ~ + y 2 = y 3 - x 3 and yz + I = y 2 + z .
T-3 . Two identical sheets of paper with 8 equally spaced lines are attached at
comer A . Initially, comers Band C coincide. The top piece is then
rotated until comer C lies on the next-to- last line of the bottom sheet.
If BC = ./is, compute AB.
T-4. Define [n] to be the greatest integer less than or equal to n. Compute the area of the solution set of [x] · [Y] = 2000.
B
T- 5. A digital watch displays the digits from Oto 9 as shown below by displaying some subset
of the seven segments which make up an 8 as shown at the right. lfa randomly chosen
segment fails to light up, compute the expected value of the number of digits that can still
be displayed.
D tJ 8 BH5 5 J BB
T-6 . Let the area of MED = R and the area of
B(O. n)
OCEB = S. For O < m, n < I,
express IR - S Iin terms of m and n. () C(m . 0) D(l. 0)
T-7 . Defineasequenceofintegersasfollows: a 1 = l, a 2 =2,a3 = 4,a 4 = 5, a 5 =7,a6 =9,thenextfourterms

are the next four even integers after 9, the next five terms are the next five odd integers after 16, and so on.
Compute a1999 .
T- 8. There was a chess tournament involving two eighth graders and at least ten ninth graders. Each contestant
played once against each other contestant. Each contestant received two points for a win, one point for a tie,
and zero points for a loss. The two eighth graders amassed a total of 20 points and each ninth grader
earned N points. Compute N.
T- 9. Compute the number of distinct ways one can arrange the numbers 21 , 31, 41 , 51, 61, 71, and 81 from left to
right so that the sum of every four consecutive numbers is divisible by 3.
T-10. Let Nb = I b + 2 b + ... + I 00 b where b is an integer greater than 2. Compute the number of values of b for
which the sum of the squares of the digits of Nb is at most 512.
117
ANSWERS ARML TEAM ROUND - I 999
1. l or l: l
2. 267
3. 7/i_
4. 40
22
5. -
7
6. ½ll-m-nl
7. 3935
8. 20
9. 144
10. 30
118
T- 1. Since AC = BF = 8, EC = CF = 7, and AE = BC = 6, then MEC = MCF by SSS. Subtracting the

area of MDC from each of AEC and BCF gives area of ABDE = area of !-,.CDF. Thus the ratio is
I I or I: 1j.
T- 2. Since (y - x)( x 2 + xy + y 2 j = x 2 + xy + y2, then either y - x = I or x 2 + .xy + y2 = 0, but the latter is
an impossibility for positive x and y. Thus, y = x + I. From the second equation, yz - z = y 2 - I gives
z(y- I)= (y - l)(y + I) ➔ y = I (an impossibility), or z = y + I. Thus, y = x + I and z = x + 2, and
we obtain 1729 < x < x + I < x + 2 < 1999, making 1729 < x < 1997 ➔ 1730 s x s 1996. Thus, there
are 1996-1730+ 1 = 267triplesoftheform (x,y,z). Answer: 12671.
T-3. Method I: Let AB = 7 x ➔ AD = 6x. Then by the Law of Cosines,
BC 2 = 28 = 49x 2 + 49x 2 - 2(7xX7x)cos LBAC. Since cosLBAC = i,

7
we have 28 = 14x 2 ➔ x = ./2. ➔ AB = l1.J2"I.
Method 2: If AB = 7x, then AD = 6x and AC = 7x. By the
Pythagorean Theorem on MDC, we obtain DC = x./Ji. By the

Pythagorean Theorem on MDC, we obtain x = ./2. ➔ AB = 7.[2".
T-4. Clearly [x] and [Y] will be integer factors of 2000. The graph of
[x] = a will be an infinitely long horizontal strip one unit wide,
edgeless above. The graph of [Y] = bwill be an infinitely long

vertical strip one unit wide, edgeless to the right. The graph of the
intersection of each pair of strips will be a unit square of area I with
two edges missing. Thus, each factor pair generates a region of
area 1. Since 2000 = 2 4 · 53, there are 5 · 4 = 20 positive factor

pairs. Combined with 20 negative factor pairs, the area of the
solution set equals @Q].
119
Solutions to the ARML Team Questions- 1999
T-5. For a displayed digit d, let u(d) represent the number of segments that can fail without impairing the display
ofthcdigit. Then u(O) = I, u(l) = 5, u(2) = 2, 11(3) = 2, u(4) = 3, u(5) = 2, 11(6) = l,u(7) = 4,
u(8) = 0. and u(9) = 2. The sum of the number of pairs (segment, digit) such that the segment can be
removed from the digit without impairing the display equals the sum of the above values, namely l + 5 + 2 +
2 + 3 + 2 + I + 4 + 0 + 2 = 22. There are 7 segments so the average number of displayable digits is~-
T-6. Label the regions as shown. Then: A(0,1)

l
1. R + T = area MBD = 2 · l · (l - n) .
l B(O, n)
2. R + U = area M CD = - • 1 • (l - m).
2
l
3. R+S+ T+ U = area MOD = -
2 0 D(l, 0)
Since (R + T) + (R + U) (R + S + T + U) = R - S =
½(1 - n + 1 - m - 1), then IR - SI = ½I 1 - m- n I.
T-7. Method l: By listing the terms out in staircase fashion. it is clear that the 1 = al
nth row en d s m . th e n(n + I) term o f th e

. n 2 an d that t11·1s entry 1s 2 4 =
2 ~
sequence. For example, the 3rd row ends in 9 and this entry is the
5 7 9 = a6
10 12 14 16 = a,o
~ 4) = 6th term in the sequence. We'll call it a6 . If n = 62, then
17 19 21 23 25 = al5
62( 63) . . 63( 64)
-- = 1953 while 1f n = 63, then - - = 2016. Thus the 62nd
2 2
row ends in a 1953 = 62 2 = 3844. As a result. the 63rd row starts with a1954 = 3845 and this gives
a1999 = 3845 + 2(1999 - 1954) = 139351.
Method 2: The sequence could be defined recursively: a1 = 1 and for n 2: l, an+ 1 = I + an if a is a

n
triangular number or 2 + a n otherwise. This simply says that the terms increase by 2 within a row and
increase by 1 between rows. To get from a 1 to a 1999 requires 1998 additions of either 1 or 2. Since a 1999
lies in the 63rd row, it was obtained by 62 additions of 1 and 1998 - 62 = 1936 additions of 2. Thus,
a 1999 = 1 + 62 + 1936 · 2 = 3935.
120
T -8. Let x be the number of 9th graders and let N be the number of points each 9th grader scored. Then there
arc x + 2 competitors and the total number of points is Nx + 20. Since each person plays each other person
(x + 2)(x + 1)
once, there arc 2 games played and at two points per game there are (x + 2)(x + l) total points
possible. Thus, we wish to find solutions to Nx + 20 = (x + 2)(x + 1) for x ~ I 0. Write the equation as
18 = x 2 + 3x - Nx = x( x + 3 - N) . Since x and (x + 3 ·· N) must be factors of 18 and x c:: I 0,

then x = 18, making I = 18 + 3 - N ➔ N = 12Ql.
T-9. We need only consider the problem under (mod 3): we rewrite 21, 31, 41, 51, 61, 71, and 81 as
0, I, 2, 0, 1, 2, 0. Suppose that a1,a 2 , a 3, a4 ,a5, a 6 , a7 is a required arrangement. Then,
=a2 + o 3 + a4 + a5 =a3 + o4 + a5 + a 6 =o4 + o5 + 6 + o7 = O(mod 3). Hence,

a 1 + a2 + a 3 + a4 0
a 1 =o5 , a2 = 6 ,and a3 =a 7 (mod3). Sincethereareonlytwonumbersequalto l(mod3)or2(mod3),

0
then a4 must equal O(mod 3). Thus, there are 3 choices for a4 . There are 3 · 2 choices for a 1, 2 choices from
each of 0( mod 3), I ( mod 3 ), or 2( mod 3 ). There are 2 · 2 choices for °2 , I · 2 choices for a3 , and the rest are
determined. Answer: (6)(4)(2)(3) = 11441.
T-10. Since 100 b = I · b 2 + 0 · b + 0 = b 2 , the sum of I b + 2 h + ... + I 00 b can be expressed in base I 0
b2(b2 + 1) b4 b2
as I+ 2 + ... + b 2 = -~-~ = - + - . Now the trick is to rewrite in terms of base b. If bis even,
2 2 2
we have ( %} 3 + (%} which gives a number whose digits in order are~ [ill~ [Q]. Thus, we want
(% rr +( ~ ~ 512 ➔ i ~ !024 ➔ 4 ~ b ~ 32 for b even, giving 15 values for b. If b is odd, rewrite

the expression as ( b ; 1) b3 + ( b; 1 } 2 giving a number whose digits are I b; 1 11 b; 1 [Q] [Q]. Thus, we
1
wru,t ( b; Ir+ ( b ; ' r S 512 ➔ &2 +Is 1024, m,k;ng ban odd numbcr such th"1 3 s b s 31,
yielding 15 values for b. Answer: 00.
121
ARML Power Question - 1999: Judgment Day
Suppose you arc watching a figure skating contest in which the judges rank the contestants in order. On the left is
your own ranking of the contestants and on the right is the judges' ranking:
Skater: 2 3 4 Skater: 2 3 4
You: 2 3 4 Judge 1: 1 2 4 3
Judge 2: 3 4 I 2
Judge 3: 2 3 4
For each pair of contestants, your relative ranking agrees with that of exactly two of the three judges. For example,
you rate skater 3 better than 4, which agrees with judge 2 and judge 3, but not judge I. To make the final rating, the
sum of the three judges' rankings arc used:
Skater: 2 3 4
Sum: 6 7 8 9
This ranking agrees with yours! This Power Question is devoted to determining why this is the case.
I) Turnabout:
Let S = { l, 2, 3, ... , n}. A permutation of Sis a 1 to I function mapping the elements of S onto the elements of S.
For example, if S = {I, 2, 3, 4}, then there is a permutation p of S such that p(l) = 2, p(2) = 4, p(3) = 1, and
p( 4) = 3. For the purposes of this Power Question we will write permutations by an ordered list of their values.
Thus, for p above, we write p = 2 4 I 3, or simply, 2 4 I 3.
We define pair to mean two-element ordered tuples (j, k) so thatj < k. Note that every instance of the word pair
below refers to this definition and to no other definition. So for the purposes of this Power Question (3, 5) w. a pair
but (5, 3) is not a pair.
Given a permutation p of S, we define the turnabout set T( p) of S to be the set of pairs (j, k) of elements of S so that
p(j) > p(k) . We define the turnabout number N ( p) to be the number of pairs in T( p) .
Suppose p = 3 2 5 4 I. The turnabout set T( p) = {(I, 2),( 1,5),(2,5), (3, 4),(3, 5),( 4,5)}, and the turnabout number
of pis N ( p) = 6. The turnabout set of a permutation represents those pairs of elements that arc out of order and
the turnabout number represents the "out-of-ordcrness" of a permutation.
l. Find the turnabout set T and the turnabout number N of the following permutations:
a) I 2 3 4 5 6 b) 314592687 c) 12345 ... n(2n)(2n-1)(2n 2) ... (n+l)
2. Find an example of:
a) A permutation of I through 5 whose turnabout number is I.

b) A permutation of I through 6 whose turnabout number is 14.
c) A permutation of I through 8 whose turnabout number is 14.
d) A permutation of I through 10 whose turnabout number is 35.
122
ARML Power Question -1999
3. For each of (a) through (c), give an example or prove that no example is possible.
a) Find a permutation of I through 6 such that the turnabout set T consists of all pairs of the form (o, e)
where o represents an odd number and e an even number.
b) Find a permutation of I through 99 so that the turnabout set T consists of every pair of integers of the
form (1 0a + b, l Oa + c), for O -S a, b, c -S 9 with not both a and h equal to O and h < c.
c) Find a permutation of I through 2000 so that the turnabout set T consists of all pairs with the same parity,
i.e., both numbers are odd or both are even.
4. If Tis a turnabout set of a permutation of { 1, ... , n}, let T be the set of pairs that can be fonned from
{ 1, 2, ... , n} which are not in T. Is T a turnabout set of some permutation? Prove your answer.
5. Show that two different permutations of {I, ... , n} cannot have the same turnabout set T.
6. Show that for any integer t such that O :::; t :::; n( n 2- l) , there exists a permutation of { 1, ... , n} whose
turnabout number is t.
7. If Tis the turnabout set of a permutation p, then it can be shown that p( k) equals k plus the number of pairs
(k,j) in r minus the number of pairs (j, k) in T.
a) Verify this fact directly for the permutation 3 I 4 5 9 2 6 8 7 by finding T explicitly and calculating the
right-hand side of the above equality for every k between I and 9.
b) Prove this fact.
II) Fair Play:
Given a set of permutations p 1 through p n define the functional sum of permutations P :::= p1 + ... + p n to be the
functionP(x):::cp 1(x)+ ... +pn(x). Soif p 1 =54312,p 2 =42315,andp 3 =23145,
then P :::= p 1 + p 2 + p 3 = 11 9 7 6 12. Let S = { I, 2, ... , n} be the domain of a functional sum P of n entries. In
the example above, P(l) = 11, P(2) = 9 ,P(3) = 7, P( 4) = 6, and P( 5) = 12. We define the turnabout set of P,
denoted T(P), to be the set of pairs (j, k) of the elements of S so that P(j) > P(k). We define the
turnabout number of P, denoted N(P), to be the number of pairs in T(P).
8. a) Find three permutations of {I, ... , 5} whose functional sum is 9 9 9 9 9.
b) Find three permutations p 1, p 2 , p 3 of { 1, ... , 4} so that no pair appears in more than one turnabout set
of pm , but so that the sum P = p 1 + p 2 + p 3 has a turnabout number N greater than 0. This shows that the
condition that you disagree with exactly one of the judges is necessary.
9. Prove that if p 1 through p 3 are permutations of {I, ... , 4}, so that every pair appears exactly once in a
turnabout set of pm for some rn, then the turnabout set of P = p 1 + p 2 + p 3 is empty. (Hint: use #7.) This
corresponds to the original skater question.
I 0. Generalize problem 9 with proof Maximum points for the best accurate generalization.
123
la. Since p(l) = 1, p(2) = 2, p(3) = 3, p(4) = 4. p(5) = 5,and p(6) = 6,therearcnoordcrcdpairs(j,k)for
which p(j) > p(k), so T =0 and N = 0.
I b. Given: 2 3 4 5 6 7 8 9
3 4 5 9 2 6 8 7
T = {(I, 2), (I, 6), (3, 6), (4, 6), (5, 6), (5, 7), (5, 8), (5, 9), (8, 9)} and N = 9.
Ic. T = {(11+!, 11+2), (11+!, 11+3), (11+!, 11+4), (11+1,211),
(11+2, 11+3), (11+2. 11+4), (11+2, 211),

(11+3, n+4), (11+3, 211),
(211-l, 211)}
There are n - I elements in the first row, n - 2 in the second, and so on down to I in the final row.
(I - n - l)(n - I) n(n - 1)
Thus, N = I +2+. . . +n- I =
2 2
2a. Reverse any two consecutive elements as in 1 2 3 5 4, or 2 l 3 4 5, or I 3 2 4 5. All have N = I.
2b. 6 5 4 3 1 2 yields {(1,2), (1,3), (1,4), (1,5), (1,6), (2,3), (2,4), (2,5), (2,6), (3,4), (3,5), (3,6), (4,5), (4,6)}.
2c. There is more than one correct answer. Here are two: 6 5 4 3 I 2 7 8 or 1 2 8 7 6 5 3 4.
2d. There is more than one correct answer. Here are two: 9 8 7 6 5 4 3 1 2 10 or 10 9 8 7 6 1 2 3 4 5.
3a. 4 l 5263yieldsT = {(l,2),(1,4),(1,6),(3,4),(3,6),(5,6)}.
3b. 9 8 7 6 5 4 3 2 I 19 18 17 ... 111029 28 ... 212039 38 ... 31 30 ... 99 98 97 ... 91 90.
3c. No such permutation exists. Let p = ah c d . .. To have the pair ( 1, 3) we must have a > c. But we can't
have either (1, 2) or (2, 3) so a < band b < c, making a < c, a contradiction.
124
4. I st solution: Consider p = 2 5 1 3 4. T = {(I, 3), (2, 3), (2, 4 ), (2, 5)}, making T = t(l, 2), ( 1, 4),
( 1, 5), (3, 4), (3, 5), (4, 5)}. We discover that q = 4 1 5 3 2 generates T, suggesting that interchanging each
pair of numbers whose sum is 6 generates T. Thus, if p is a permutation defined on : I, 2, ... , n) with
turnabout set T, define permutation q by p(x) + q(x) = n + 1 ➔ q(x) = n + 1 - p(x). We'll show that the
turnabout set R of q is the complement ofp, namely T. lf (j, k) E T, then/ < k and
p(j) > p(k) giving -p(j) < -p(k) ➔ n + 1- p(j) < n + 1 - p(k) ➔ q(j) < q(k) ➔ (j,k)" R.
lf(j,k)" T, thcnj < k and p(.i) < p(k) ➔ -p(j) > -p(k) ➔ 11 + I - A.i) > n + 1- p(k) ➔
q(j) > q(k), putting (j, k) in the turnabout set R of q. Thus, R = T. For completeness, one could verify that
q is I to I and onto. It is I to I since if q(j) = q(k), then n + l - p(j) = n +1- p(k), giving
p(j) = p( k), and since p is I to l ,j = k, making q a I to I mapping. It is onto because for any m in
(1,2 ... ,n).pickavalueof k so that p(k) = n+ 1-m. Then, q(k) = n+ I- p.,k) = m.
2 nd solution: We can reverse the relative order of pairs of numbers by taking their reciprocals. Consider
p = 2 1 4 3. Write the reciprocals .!_ I .!_ .!_ and pair I with the smallest, i.e., .!_, 2 with next smallest and
2 4 3 4
soon,obtainingq = 3412. SinceT(p) = {(l,2),(3,4)}andT(q) = {(1,3),(1,4),(2,3),(2,4)},T(q)is
clearly the complement of T(p). Let p = p(I) p(2) . . . p(n). If Tis the turnabout set of p, then (j, k) E T
if forj < k, p(j) > p(k) . Consider the sequence of the reciprocals of
I I I
p: - - -- . . - - . Pair I with the smallest, 2 with the next and so on, generating a permutation q
p(l) p(2) p(n)
l l
of{l,2, ... n}. lfforj < k, p(j) < p(k),then - - . >-k-,making q(j) >q(k). Thus,(j,k)liesinthe
p(j) p( )
turnabout set of q but not in the turnabout set of p. and we have constructed a permutation of {I, 2, ... , n}
whose turnabout set is the complement of T(p).
5. Suppose T(p) = T(q) forpermutationspandq. Wewillshowthatp = q. Let m = p(k). Then there m-1
numbers f so that p( t) < p(k) which means m - I numbers f so that either (k, €) is in T(p) or ( €, k) is not in
T(p). This defines a relative order for pairs ofp. Since T( p) = T(q), the same statement holds for q and that
means m = q(k). Thus. for each element k of q there is the same relative ordering of pairs asp. Thus, q = p.
125
6. By induction:
]-(]-[)
I) For { l }, p = I is the only pernmtation, making T = 0, N = 0, and clearly, 2 = 0.
2) Assume that for { 1, 2, ... , n - I l there are permutations whose turnabout sets contain pairs
y1e . between 0 an d (n - IXn - 2) me

. Id.mg turnab out num bers Iymg . Iy. ff1ven any of t10se
. 1us1ve 1
2
permutations, say p = a1 '½ . . . an_ 1, place 11 to the far right, producing p = a1 a2 . . . an_ 1 n.
Since n is the largest element, the turnabout number of this permutation is unchanged, making it a number
(n-lXn-2). . . . .
between 0 and - - - - - mclus1vely. The permutat10n with the largest turnabout set 1s
2
p = n- I n- 2 . . . 2 I. Place non the far right of the permutation giving n - I 11 - 2 ... 2 l n. This
doesn't introduce any new pairs into the turnabout set. Now move n back, one place at time, producing
permutations whose turnabout numbers increase by I each time:
(n-lXn-2)
N(n - I 11 -2 ... 2 n)
2
(n-lXn-2)
N(n-1 11 -2 ... 2 n I) +l
2
(n-IXn-2)
N(n -1 n -2 ... n 2 1) +2
2
(n-lXn-2) n(n-1)
N(n n- I n - 2 ... 2 1) 2 +(n-1)= 2
. n(n - I)
Thus, for { 1, 2, ... , n} we have shown that the turnabout number t sattsfies 0 :S: t $
2 .
7a. Given: 2 3 4 5 6 7 8 9
3 4 5 9 2 6 8 7
wehaveT = {(l,2),(l,6),(3,6),(4,6),(5,6),(5,7),(5,8),(5,9),(8,9)}. Thus,p(l) = 3

which equals l + 2 - 0 since there are two pairs starting with I and none ending with 1. Similarly:
p(2) = 2+0-1 = I,p(3) = 3+1-0 = 4,p(4) = 4+1-0 = 5,p(5) = 5+4-0=9,
p(6) = 6+0-6 = 2,p(7) = 7+0-1 = 6,p(8) = 8+1-1 = 8,andp(9) = 9+0-2 = 7.
126
7b. If p(k) = 111, that means there are m - l values I so that p(k) > p( I). Of these, suppose t of these values I
arc less thank and m -- l - tare greater thank. Then there arc k - I - t pairs of ( e, k) in the turnabout set
since there are k l pairs (k, P) and I of them are not in the turnabout set. There are m - l - I pairs (k, f) in
the turnabout set corresponding to those f greater than k so that p( k) > p( (). The desired sum is then
k+(m 1-t) -(k-1-t) = m.
8a. One answer among many: I 2 3 4 5 + 3 4 5 I 2 + 5 3 I 4 2 = 9 9 9 9 9.
8b. Shown is one answer:
P1 2 3 4 T= 0
P2 2 3 4 T= 0
P3 4 2 3 T= {(I, 2 ), (1, 3 ), (I , 4)}
p 6 5 8 11 T= {(I, 2)} and N = I.
9. By problem 7, P(k) p 1(k) + p 2 (k) + p 3(k)
k + # pairs (k,j) in r( p 1) - # pairs (j, k) in r( p 1)

+ k + # pairs (k,j) in r( p 2 ) - # pairs (j, k) in r( p2 )
+ k + # pairs (k,j) in r( p 3 ) - # pairs (j, k) in r( p3 )
3k # pairs (j, k) in r( p 1)
# pairs (j, k) in r( p2 )
# pairs (j, k) in r( p 3 )
+ # pairs (k,j) in r( p 1)
+ # pairs (k,j) in r( p2 )
+ # pairs (k,j) in r( p 3 )
But every pair ( f, m) is in exactly one of the turnabout sets r( p 1) , r( p2 ) , or r( p3) . Set n = 4. Then this sum
yields P( k) = 3k - (k - I) + ( n - k) = n + k + I. Thus, P ~ n +2 n+3 n +4 n +5 = 6 7 8 9 and since

P's terms are increasing, its turnabout set is empty and its turnabout number is 0.
127
2 11 " solution: Since all pairs appear exactly once in a turnabout set of p , all pairs will contribute to the
111
computation of P(k) = p 1(k) + p/k) + p 3 (k) as described in #7. The set of all pairs is
((L 2), (I, 3), (I, 4), (2, 3), (2, 4), (3, 4)]. Thus,
sum of3 /cs + # of times k is first - # of times k is last Total
P(l) = Pi (1) + Pil) + p 3(1) l+l+l + 3 0 6
P( 2) = p 1(2) + p/2) + p 3(2) 2+2+2 + 2 7
P(3) = p 1(3) + p/3) + p/3) 3+3+3 + 2 8
P(4) = p/4) + pi4) + p 3 (4) 4+4+4 + 0 3 9
Thus, P = 6 7 8 9, P's elements are increasing and the turnabout set of P = 0.
10. #I: Suppose p 1 , p 2 , ... , p 111 are m permutations of { 1, 2, .... , n} so that every pair (i,j) appears the same
number of times € in the set of turnabout sets r( p 1), r( p2 ), . . , r( p 111 ). Then P = p 1 + ... + p 111 • Then
the permutation P = p 1 + ... + p 111 is given by P(k) = mk + €(n - 2k +I).
#2: If p 1 through p 11 _ 1 are permutations of {1, 2, ... , n} for n 2: 4 so that every pair appears exactly once
in a turnabout set ofp for some m, then P(k) = (k + l)n - (3k - 1) where P = p 1 + ... + p 11 and the
turnabout set of Pis empty.
Proof: The set of all pairs is {( l, 2), ... , (l, n ), (2, 3 ), ... , (2, n ), ... , (n-1, n )l. Proceeding as in #9 we
have:
P(l) l(n-1) + (n-1) - 0 2n-2
P(2) 2(n - I) + (n - 2) - l 3n - 5
P(k) k(n-1) + (n-k) - (k-1) (k+l)n-(3k- l)
P(n) n(n+ I) - (3n- l) (n- 1)2
Then P = 2n-2 3n-5 . . (k+ I )n - (3k- 1) (n - 1)2. We can show that Pis increasing and
thus has no pairs in its turnabout set by considering the difference of successive terms:
P(k) -P(k - l) = (k + I)n- (3k - l) - ( kn - (3k - 4)) which equals n - 3. This is greater than O for
n 2: 4. Thus, for n 2: 4, T = 0 and N = 0.
128
ARML Individual Questions- 1999
1-1. A pro athlete played for 17 years and earned 72 million dollars. She was paid k million a year where k is an
integer and received an extra one million each year that her team made the playoffs. Compute the number of
years her team made the playoffs.
1-2. In circle 0, PO .l_ OB, and PO equals the length of the

. . PA
diameter of circle 0. Compute AB .
1-3. If a and bare the roots of I Ix 2 - 4x - 2 = 0, then compute the product:
1----4. -
The measure of LPAQ is 60°, AB bisects LPAQ and circles P and Q arc tangent to ~
AB. If the radii of
circles P and Qare I and 2 respectively, compute the distance from P to Q.
logb a 19 b k
1-5. If - then - c . Compute k.
log a 99' C
C
1-6. Let A0 A1 ... A11 be a regular 12-sided polygon whose area is 24. Form convex polygons by first connecting
consecutive vertices, then joining the first and the last vertices. For example, polygons ~ A4 A5 A6 or
A 10A 11 A 0 A 1A 2 ~A4 , as well as the entire 12-gon can be formed. Compute the sum of the areas of all the
distinct polygons that can be formed.
I-7. Arrange the following products in increasing order from left to right:
1000! (400! )( 400!)(200!) (500!)(500!) (600!)(300!)( 100!) (700!)(300!)
1-8. Two congruent rectangles are positioned in a right triangle of legs a and b
as shown. Both rectangles have a vertex on the hypotenuse and a long side kh
on a leg. If a and b can vary and the short side of each rectangle is kb for b
real numbers k, compute the largest possible value fork.
129
ANSWERS ARML INDIVIDUAL ROUND- 1999
1. 4
3
2. -
2
3. !.! = 2.2
5
4. 2,D
80
5.
19
6. 1320
7. 400!400!200! < 600!300! 100! < 500!500! < 700!300! < 1000!
8.
5
130
72 - 11
1-1. 17k + 11 = 72 ➔ k = - - for O::; n::; 17. The only solution is 11 = [4].
17
1-2. Let PO = 2r, OB = r, making PB /s. Since=r -
PC· PD = PA· PB, then r(3r) =PA· (r/s). Then PA= Es r O r C r

D - - - - - - -~--+----~ P
making AB
r:
= ry'J -
3r
Js 2r
= Js . Thus,
PA
AB =
W
liJ- B
4±M4
1-3. The roots are 22 and both are between -1 and 1. Then both 1 + a + a 2 + a 3 + ... and
I + b + h2 + b 3 + ... are convergmg

. geometnc . equa 1·mg -l- an d -l- respective
. senes . Iy. Th elf
. pro duct
1-a 1-b
equals b b where (a+ b) and ab are the sum and product respectively of the roots of
1-(a+ )+a
1 4 2 k.
x- - -11 x - -11 = 0 , ma mg b 4 2 = 1151 = 2.21.
I -(a+ )+ ab
1----
11 11
1-5. Set logb a= 19t and loge a= 99t, giving a= b 19 t and a= c 991 ➔ b 19t = c 991 ➔
9
b 19 = c 99 . Thus, b: 9 = c 80 ➔ !!_ = cSfjl 9 . Thus, k = ~-
c C l.12]
1-6. Shown is convex polygon A 1A 2 ~ A4 A5 . It can be paired with the unshaded
polygon. So every polygon is in a l-to-1 correspondence with the complement

of that polygon's interior in the 12-gon and the sum of the areas of each pair is A1
the area of the 12-gon, namely 24. A pair of polygons is determined by the
border between them. For example, border A 1A5
131
determines the two polygons in the figure at the right. There are 12c 2 = 66 such pairs, but of the 12 pairs of
consecutive vertices A.A. 1 which generate the entire 12-gon, we can use only 1, so the sum of all the areas
I 1+
is(66-11)·24 = 113201.
1000!
1-7. Since 1000c 300 > I, 700 !· 3 00! > I ➔ 1000! > 700! -300!.
1000! 1000!
Since IOOOC500 > IOOOC300' then 500!· 500! > 700!. 300! ➔ 700!-300! > 500!-500!.
500!-500! 500!·(500 · 499 · ... · 301) · 300! (500· ... ·401)(400· ... ·301) .
Since----- - - - - - - - - - - and smce
600 !-300! · JOO! (600-599 · ... · 501) · 500!·300!-100! (600 · ... · 501)(100 · ... -1)
(500 ..... 401X 400 · .... 30n

----------· >
(600 · ... · 501)(100 · ... · 1)
(iJ IOO. (i)l00
( 1:
)
100
> 1, we have
500!-500! > 600!-300!-JOO!.
To compare 400! · 400! · 200! and 600! · 300! · 100!, divide both by 400! · 300! · 100!,
obtaining (400 · ... · 301)(200 · ... ·IOI) and 600 · 599 · ... · 402 · 401 respectively. Both contain 200
terms, but clearly the expression on the right is greater on a term by term basis. Thus,
1400!-400!·200! < 600!-300!-100! < 500!-500! < 700!-300! < 100011.
WQ b M
I-8. Since AWQR ~ AMP R, then - = - , making QR = ka . Thus,
QR a
. MN h .
PQ =a - ka =NV. Smee !1MNS ~ AMPR, - - = -, makmg s
NS a N
kb
kh 2
MN = -. Since PV + VN + NM = h, we have b
a
kb 2 ab - a2
kb+ (a - ak) + - = b, making k = 2 2 . Consider
a ab - a + b V
w
T
ab - a 2 + b 2 b2 kb
- - - - - - = l + - - - . Let t = b - a. Then p R
k ab-a2 a(b-a) Q
a
l (a+ t) 2 a t . a t
- = 1+ - - - = I + - + 2 + - . Smee - + - ;;:: 2 , then
k at t a t a
1
->1+2+2 ➔
1
k;;:: 5, making k ~
W
s·I Ans: l1J
k -
132
R1-1. The sum of the digits of the year 1999 is 28. Let Y be the next following year in which the sum of the digits
is 28. Compute Y - l 999.
T A B
Rl-2. Let T = TNYWR and set k = Trapezoid ABCD is
225
divided into four congruent trapezoids as shown.
If AB = k and DC = 2k, compute the sum of the lengths of
all line segments in the figure.
n/]>--------------~c
')
Rl-3. Let T = TNYWR. In MBC, sin LA = cos LB. If, for x in degrees, mLA = x~ + 5x,
mLB = x2 - I
x - 36, and mLA - mLB I= T - 12 , compute mLA .
R2-l. Let k be the smallest of six consecutive positive integers. If the sum of the six integers is divisible by
three distinct primes, compute the smallest possible value fork.
p
R2-2. Let T = TNYWR. BC is a diameter of circle O and
AP is tangent to circle Oat P. If AP = 2T
A
and AB = T, compute the length of AO.
R2-3. Let T = TNYWR. For x randomly chosen from the interval [o. ~:], compute the probability that
l5x 2 + 3 < 14x.
133
ANSWERS ARML RELAY RACES - 1999
Relay #1:
Rl-1. 900
Rl-2. 30
Rl-3. 36
Relay #2:
R2-l. 15
75
R2-2.
2
4
R2-3.
45
134
Rl-1. For the year 2ABC, A + B + C = 26. This can only be done using one eight and two nines. To
minimize, choose the eight for the hundred's place giving 2899. Then 2899 - 1999 = 19001.
Rl-2. Since AB = k, then AD = BC = EF = k, but if EF = k, A k B

k
then DE = FC =
2 . Thus, the legs of the four congruent
k
trapezoids are all - , and since legs are also bases,
2 D C
k/2 E k F k/2
k
EH= HG= GF =AH= GB=-. Thesumofthe
2
segments is 2k + 3k + { -
k J= -15k . Since k =
900
- = 4, the sum is 00.
2 2 225
Rl-3. If sin A = cos B, then either A = 90° - B or A = 90° + B. Using A = 90° - B, then
( x 2 + 5x) + ( x 2 - x - 36) = 90° ➔ x 2 + 2x - 63 = 0 ➔ x = 7 or - 9 . If x = 7, A = 84 °
andB = 6°. lf.x = -9, then A = 36° andB = 54°. Using A = 90° + B, then
( x 2 + 5x) = 90° + ( x 2 - x - 36) ➔ 6x = 54 ➔ x = 9 ➔ A = 126° and B = 36°.
Since T = 30, then jmLA - m.L.B I= 18° and only A = 36°, B = 54° works. Ans: ~-
R2-I. k + (k + l) + ... + (k + 5) = 6k + 15 = 3(2k + 5). Since 2k + 5 is odd, 2 is not one of the primes. The
next smallest primes are 5 and 7, so set 2k + 5 ~ 5 · 7 ➔ k = IT3i.
R2-2. T = 15. Let OB= r ➔ (2T/ +r 2 = (T+r) 2

3T
implies 3T = 2r ➔ r =2 . Thus, AO= T + r
s; =~-
75 2T 2 1 3
R2-3. T =2 so 25 = 3. Thus, 15x - 14x + 3 < 0 ➔ (5x - 3)(3x - 1) < 0 ➔ - < x < - . Since
3 5
3
5 - 31 = 15'
4 h b b'l' .
t e pro a I Ity IS
4/15
-3- = ~-
i4l
135
AR ML Super Relay - l 999
I. Let N = 3A4A5A where A is a digit chosen at random from {o, 1.2, ... , 9}. Compute the probability
that N is divisible by 9.
2K
2. Let T = TNYWR and setK = 20T. Compute (1 + ;) .
T 2K
3. Let T = TNYWR and set K If the area of the
-16
trapezoid at the right equals A, pass back Ai3.
i 4K
4. Let T = TNYWR. If sin x = T cos x. compute the value of (sec x - r)(sec x + r).
5. Let T = TNYWR and set K = 14T. A convex polygon has K sides. Compute the number of
diagonals of the polygon.
6. Let T = TNYWR. Let p be a prime number and b be a positive integer. If log b = T log p,
then b has X more factors than p does. Compute X.
T 2 K
7. Let T = TNYWR and set K = -4 . If x + 2x + 4 = - - , compute x.
x-2
8. Let T = TNYWR. The vertex of a concave up parabola is V(T, - 16). Compute the sum of
the zeros of the parabola.
136
ARML Super Relay- 1999
9. Let T = TNYWR and set K = IOOT. Let a the sum of the cubes of the digits of a for n :,, l.
n+ 1 n
If a 1 = 22, compute aK.
10. Let T = TNYWR. Given the points A (log 2, log 3) and s( lo~ log T 2 ), log(log T 3 )),
1--,
compute the slope of AB.
11. Let T = TNYWR. The intercepts of the lines x + y =T and x + y =T+ I form a trapezoid.
Compute the area of the trapezoid.
12. Let T = TNYWR and set K = 2T + 4. Compute the value ofx which solves JK - 2x = 2x - I.
13. Let T = TNYWR. The two circles arc concentric. If the ratio of the radius of
the large circle to the radius of the small circle equals T, compute the ratio of
the area of partial ring ABCD to the area of sector BOC.
14. Let T = TNYWR and set K = 4 T. MBC is a right triangle and the radius of its inscribed circle equals K.
If the perimeter of MBC exceeds twice its hypotenuse by N, compute N.
15. Let T = TNYWR. Children are equally spaced about a circle. If a child in position Tis directly opposite a
child in position 58, compute the number of children who are equally spaced about the circle.
137
3
I.
10
2. -64
3. 144
4.
5. 77
6. 76
7. 3
8. 6
9. 217
10. l
3
11. -
2
3
12. -
2
5
13. -
4
14. 10
15. 96
138
I. 3A + 12 = 3(A + 4) must be divisible by 9 ➔ A+ 4 must be a multiple of 3. Thus, A 2,5,or8
and the probability is~-
3. Since AF = 2FE, MFE and 1'1BCD are 30-60-90 triangles, A 2K B
making AE = K Ji. The area of the trapezoid =

2K
I
•h 2K
I
(¾JKJi)(2K + 4K) = 3K 2/3. Since T = -64, K = 4, so

I
K r-' 2K K
F E D C
the area 48/3 . Pass back 48/3 Ji = I !':HI.
4. sinx = Tcos x ➔ tan x = T ➔ I+ tan 2 x = I+ T 2 = sec 2 X. Thus, sec 2 x - T 2 = [I]. Note that the
value of Twas irrelevant in this problem.
5. For each vertex in a polygon of K sides there are K - 3 diagonals. Hence, the total number of diagonals
equals
K(K - 3) . Since K = 14, the result is [TI].
2
6. Since b = pT and p has two factors, I and p, then b has the following T + I factors: 1,p, p 2 , ... , pT.
Thus, X = (T + 1)- 2 = T- 1. Since T = 77, pass back lNJ.
7. (x 2 +2t+4j(x-2)=K ➔ x 3 -8=K ➔ x 3 =K+8.SinceT=76,K= 19 ➔ x 3 =27 ➔

X = []_
8. Equation of the parabola: y = a(x - D -

2
16. Ify = 0, then x = T ± ia.
4
The sum of the two zeros
equals 2T. Since T = 3, pass back~-
9. Since a 1 =22, a2 = 16, '1:3 = 217, a4 =352, a5 = 160,and a6 =217,thesequenceiscyclicwith
a3 n = 217, a3n+l = 352,and a3n+ 2 = 160. Since T = 6,K = 600and a600= 12171.
139
Solutions to the ARML Super Relay - I 999
I 0. Note that the domain of log(log r 3) consists of all T for which log T > 0 ➔ T > I . All these values work
for log( log T 2 ) . = m
Note that as long as T > I , then the value of Tis irrelevant
I l. Area of trapezoid ABCD = c.( MOD) - a( MOC)

A(O,T+l)~
.!.(r+1) 2 -.!..r 2 = 2 T+I_ SinceT= I.
2 2 2 B(O,T)
the area of ABCD = rn. 0

C(T, 0) 1
E D(T+l,0)
12. K - 2x = 4x
2
- 4x + 1 ➔ 4x
2
- 2x + (1 - K) = 0 ➔ x =
2± J4 - 48 · 4(1 - K) =
I± J4K - 3 Try
4
'nice'valuesforK: ifK = l,x = }.ifK = 3,x = !,etc. SinceT = ½,K = 7,makingx = rn.
m c.(sector AOD) m2 m 2 - n2
13. If T = - , then - 2 . The ratio of the partial ring ABCD to the sector BOC is
n c.(sector BOC) n n2
Since T = %,m = 3, n = 2, giving~-
14. Let a and b be the lengths of the legs and c be the length of the
hypotenuse. Since c = a -K + b- K, then 2K = a+ b- c
giving 2K + 2c = a + b + c. Thus, N = 2K. Since T =; ,
K = 5. making N = [ill.
15. Two cases: if T < 58, then there are 58 - T- I children on each side of the circle between Tand 58.
The total is 2(58 - T- I)+ 2 = l 16-2Tfor T :S 29. If T > 58, then there are
2(T-58-1)+2 = 2T-ll6forTc:: 116. SinceT= 10,thenumberofchildrenisll6-2·10 = ~-
140
ARML Tiebreaker~ 1999
I. , = (x + 3)2 + 9 tor
,
Let f (x) 4 x 2'. -3. Compute the shortest possible distance between a point on f
and a point on f- I,
141
1. The shortest distance will connect the points (x, f (x)) and (f (x), x) since each is the reflection of the other
across y = x. Let d = J(x - f( x)) 2 + (f (x) - x) )2 = j f(x) - x I· ./2. To minimized we must minimize
j f(x) - I I
x = x 2 + 5x + ~ I· For the general quadratic function y = ax-2 + hx + c the minimum occurs
-h 5
atx = - . In this case the minimum occurs at x = .=_, and substituting, we obtain a minimum of 5.
2a 2
Thus, d ~ ls../i,I.
142
2000
Power Qyestion _________ ·----------·-----·-------------··-·-·---------------···---····-------------·---· 150

'lndividua( 'Round·····----·---······--·······----·-------···-·········--··--·-·---················ 161
'Re(ay 'Rounc( _____ ····--····--·········--···-·····-·····---·-···--·····----·--··-···----·-··············- 165
Suyer 'Re(aY--··---···········-·--·····-··--··------···--···-----·--·····--···---·-·····---····-···········- 168
'Tiebreakers l 73
-143-
25th ANNUAL MEET
ARML celebrated its 25 th annual competition in grand style with a real nail-biter of a contest.
There were 104 teams with 1560 students taking part at Penn State, the University of Iowa, and
UNL V. For the first time ever there was a tie for first place as Chicago A and San Francisco Bay
A each scored 172 points, just one more than third place New York City A. But there was also a
tie in Division B, the first since 1983. Both Connecticut A and Peninsula South Bay scored 127
points, just two ahead of Iowa A and Northern California. Three teams from Taiwan took part in
this year's contest and did quite well. Taiwan's parallel contest, TRML, involved 80 teams last
year and this year they are expecting some 280. Talk about explosive growth. Mark Saul,
president of ARML, was feted at a banquet and given a Lifetime Achievement Award in honor
of his long and effective service. Chris Clark, formerly of the Western Massachusetts team, was
in charge of organizing special events in honor of ARML's 25 th year and he came up will all sorts
of good ideas, including a lovely folder with the ARML logo on the front, and a collection of the
best of the ARML problems over the years. The following problem was written in honor of
ARML's 25 th anniversary and students enjoyed solving it:
The decimal . AR MLb equals . 25 10 . If b is a positive integer that is as small as
possible, find the sum of A + R + M + L.
Debbie Clyde ofldaho received the Samuel Greitzer Distinguished Coach Award.
Richard Rukin who was instrumental in getting a second ARML site started at the University of
Iowa received the Alfred Kalfus Founder's Award.
Andy Yang of Howard County won the Zachary Sobol Award for outstanding contributions to
his ARML team.
144
T-1. In the diagram, the circle and square have the same center. If the area of
shaded region ABC equals the area of the region bounded by CD and minor
-
arc CD , compute the ratio of the side of the square to the radius of the circle. :c_-J
T-2. Compute the smallest prime number p such that p 3 + 2 p2 + p has exactly 42 positive factors.
T-3. Square A BCD is divided into two regions of equal area by OE. If OD and
.
AD are mtcgers an d -CE = 2000, compute the smal 1est va 1uc of, AD.
BE
T--4. In base 10, in the equation TWO + TWO = FOUR, distinct letters represent different digits and F t 0.
Compute the smallest possible value for FOUR.
T-5. Equilateral triangle ABC has sides of length 6 and its medians. A A', BB', and CC', intersect at D. If three
segments are chosen from among AD, A 'D, BD, B'D, CD and CD to form a non-equilateral triangle of
positive area, compute the area of that triangle.
T-6. Let x be a randomly selected integer from the set ( I 00, IO 1, ... , 999}. What is the probability that
x 2 and (x + I 00) 2 have the same number of digits?
2 3 1999 2000
T-7. Let /( x) = (x - l)(x - 2) (x - 3) •.. (x - 1999) (x - 2000) . Compute the number of values ofx
for which IJ(x) I= 1.

T-8. In rectangle ABCD, G and Hare trisection points of AD, and E and F arc B E F
trisection points of BC. If AB = 360 and BC= 450, compute the

area of PQRS.
A G H D
T-9. For an integer kin base l 0, let z( k) be the number of zeros that appear in the binary representation of k.
n
Let Sn= L z(k). Compute s256 .
k=l
T-10. The unit hypercube in 8 dimensions can be defined as the set of points (x 1, x 2 , ... , x8 ) such that 0 :::: xi :::: I for
i E { 1, 2, ... , 8}. Suppose the hypercube is cut into smaller hyper-rectangular boxes with side lengths of ½or f by
hyperplanes whose equations are xi = ½for i E {l, 2, ... , 8}. Color the hyper-rectangular box with vertex on the
origin red, color those adjacent to red hyper-rectangular boxes blue, and those adjacent to blue hyper-rectangular
boxes red where "adjacent" means bordering along one of the hyperplanes x 1= ½.
If R equals the total volume of
the red hyper-rectangular boxes and B equals the total volume of the blue hyper-rectangular boxes, compute R - B.
145
ANSWERS ARML TEAM ROUND -2000
1. .J;, or ./;,: 1
2. 23
3. 1999
4. 1468
3,/15
5.
4
7
6.
9
7. 4000
8. 4500
9. 777
10.
6561
146
Solutions to ARML Team Questions - 2000
T-1. Since the sum of the areas of the dark-shaded regions equals the sum of the
areas of the light-shaded regions, the area of the square equals the area of the
circle. Thus, s 2 = rrr 2 ➔ s: r = IJ;cl or IJ;c: 1[.
T-2. If p 3 + 2 p2 + p = p(p + 1) 2 has 42 factors when pis prime, then (p + 1) 2 must have 21 factors. Let
a, b, c, ... be the exponents of primes in the prime factorization of p + I. Then (p + 1) 2 =
( P/1 · p/ · p/ ·--r. Thus, (2a + I )(2b + 1)(2c + 1) ... = 21. This is only possible when
p + I has precisely two prime factors making 2a + I = 3 and 2b + I = 7 ➔ a = I and b = 3.
Sop+ 1 has the form (p 1 • p/). Numbers through 22 fail, but 24 = 2 3 · 3 works, making p = [nJ.
T-3. OE passes through P, the center of the square. Let OD = n and AD = m.

A B
PF m 12
Slope of OE -
FO n+ m 12
= _m__ .
2n + m
The equation of OE is
p E
I
G
m/2
mx mn m(n + m) m/2
y = Since GD = - - EC= , and GD= BE, 0
2n + m 2n + m' 2n + m ll D F C
EC m(n + m)
then - = 2000. Thus, n + m = 2000n ➔ m 1999n. lfn = l,thenm = l1999I.
GD mn
T--4. F = I. Clearly, 0 i- 0 since that would make R = 0. Since F = I then O i- 1; try O = 2. This makes T = 6, but
no value works for W. Try O ~ 3. This again makes T = 6 which can't be because R ~ 6 ifO = 3. Try O = 4,
making R = 8 and T = 7. W = 3 works ➔ TWO+ TWO = FOUR = 734 + 734 = I1468 I.
T-5. The medians have length 3i3 and are divided in to a ratio of 2 : 1, so we choosing from lengths of 2./3 and
./i. We can't pick two of ./i and one of 2./3, so our triangle has sides 2./3, 2./3, and ./i. The triangle
is isosceles and the altitude to the base equals (2./3) 2 - (./3 / 2) 2 3-J5
= 2 . The area 1s_ I-34£- I.
T-6. The squares of three-digit numbers have either 5 or 6 digits. If 10,000 S: x 2 < 100,000, then 100 S: x S: 316
and x and x + I 00 both lie in this range if x E { I 00, IO I, ... , 216}. If 100,000 S: x 2 < 1,000,000, then
317 :S x :S 999 andx andx+ 100 both lie in this range ifx E {317,318, ... , 899}. Out of900 three-digit
numbers, there are (216 - I 00 + 1) + (899 - 317 + 1) = 117 + 583 = 700 values ofx satisfying the
condition stated in the problem. Probability = 700

900
= [Il
l2_J
.
147
T-7. A portion of an approximation to the graph is shown

at the right. From l to 2000, between every pair of I •••••••.••••••••••••••••• --···1············· ~....... ..
I 2 3 5 6
consecutive integers, there are two solutions. There
4 7 8
are 2000 - I = 1999 pairs. There is l solution ·--y--·j-·········--···\:··, ········ ........... ..
u
.1 ..•••••••
between O and I and one solution after 2000, giving

\J
a total of I 999 · 2 + 2 = I4000 I solutions.
Note: We should verify an assumption of our solution, namely that f(x) I I > I between every pair of
consecutive integers. Note that if n is an integer drawn from {I, 2, ... , 1999}, then .t-( n + ¾) is a product
of factors that arc greater than I, namely 3/2, 5/2. etc., and factors of l /2 that arise from the terms
( x-n )n and (x-(n+l) )n + I . Weobtainthemostfactorsof I

= 1999. Sincc_/{1999.5) =
2 whenn
(I 998.5) 1(1997 .5l (I 996.5) 3 .. •(2.5) 1997 ( 1.5) 1998 (.5) 1999 (-.5) 2000 , we have a product consisting of
1998 - 1999 3 1
l + 2 + ... + 1998 = - - - - factors greater than or equal to - and 3999 factors of - . Thus,
2 2 2
Although we didn't expect ARML participants to consider the question of whether If(x) Itakes on a value of
I more than twice in a given interval [ n, n+ I ], we'll present an answer for completeness. By Rolle's
Theorem, f( x) has at least one relative extrema in each interval (n, n+ 1). Thus, f( x) has at least 1999
relative extrema. Write f' (x) = f( x) [ -l- + -2- + .. · + -2000]

- - . Add the fractions to obtain a
· x-I x-2 x-2000
polynomial in the numerator of degree 1999. Thus,/' (x) has at most 1999 zeros. Therefore, f(x) has
exactly 1999 relative extrema, one per interval, and consequently, there arc no more than two solutions to
lf(x) I= l in each [ n, n+I ].

T-8. Since BF = AH, BFHA is a rectangle and therefore R is collinear B 150 E 150 F 150 C
with E and G. Draw PR forming congruent trapezoids PREB and

I
PRGA. Since BEGA is a rectangle, PR = 2· BE = 75, making the
scale factor of similar triangles PQR and EQB equal to .!.. . The
2 A G H D
distance from PR to BE is 360 = 180. Set it equal to 3x, making the distance from Q to BE equal to 2x and
2
1
the distance from Q to PR equal to x. If 3x = 180, then x = 60. The area of PQR = 2 •60 • 75 = 2250.
Thus. the area of PQRS = I 4500 I.
148
T-9. Consider this array of256 eight-digit base 2 numbers written with leading zeros:
010 000000002
110 000000012
2 10 000000102
3 10 000000112
25510 llllllll 2
Of the 8 · 256 = 2048 digits in the array, exactly half, I 024, are zeros. Now we exclude all leading zeros.
One number has 8 leading zeros, one has 7, 2 have 6, 4 have 5, 8 have 4, 16 have 3, 32 have 2, 64 have 1, and
128 have none, giving 8 + 7 + 2 · 6 + 4 · 5 + 8 · 4 + 16 · 3 + 32 · 2 + 64 · 1 = 255. Up to 255 10 there arc
1024- 255 = 769 zeros. Add the 8 zeros from 256 10 = 1000000002 , making 1777 I.
Alternate solution: consider, for example, all numbers with a zero in the third position: __ 0 ____ _
Using O's or I's there are 2 5 numbers that can be formed using the places to the right of 0, and since the
number cannot begin with 0, there are 2 2 - 1 numbers that can be formed using the places to the left of
the 0. Thus, there are ( 25 X2 2 - 1) numbers with a 0 in the third position. Therefore, for a zero in the ith
8
position there are ( 2 -i X 2i-l - I) numbers. Wc can determine the total number of zeros by summing up
8 8
the numb er o f num bers wit·h a zero m
· eac h pos1tlon.
·· Thus, ,
£...( 28-i)( 2i-l - 1 ) = £...
, 2 7 - 28-i
i=I i=I
8-2 7 -(2 7 +2 6 +-··+2 1 +2°)= 8-2 7 -(2 8 -1) = 3-2 8 +1 = 769. Nowadd8zeros.
T-10. Let k = the number of edges at each vertex of a given hyper-rectangular box H that have length i. Then H
is red if k is even and blue if k is odd, and the number of given hyper-rectangular boxes with a given value of
149
Shown at the right is an m by 11 array of lattice points. Define an (I, 11)

• • •• .
(m, n)
.
ARML (m, 11) polygon to be a simple closed polygon with the following
properties: • • •• . .
• • • • • •
I. It is constructed by joining all the points in them by 11 array with unit line
segments that are horizontal or vertical. This can be done without lifting
your pencil off the paper.
...
(I, l)
. . . .
(m, l)
2. Each point is joined by a unit line segment to exactly two other points.
Two ARML (m, 11) polygons are distinct if one joins two points that the other
does not. Two distinct ARML (m, n) polygons may be congruent. Define
F(m, n) to be the number of distinct ARML (m, 11) polygons for a given
ordered pair (m, n). For example, the diagram at the right shows all of the
distinct ARML (3, 4) polygons, making F(3, 4) = 2.
I. a) On the 6 by 8 grids provided, sketch four non-congruent ARML (6, 8) polygons.
b) Find the area and perimeter of each polygon.
2. For each of the following values ofm and n, compute F(m, n). (You may do so by sketching all of the
distinct ARML (m, 11) polygons on the grids provided, but do not tum in the grids).
a) m = 3, n = 6 b) m=4,11=4 c) m=4,n=5 d) m =4, n = 6
3. a) Compute F(2, n) for all values of n ~ 2 and justify your answer.
b) Let n be an even integer. Find and prove a formula for F(3, n) in terms of n.
4. Using the data from problem #2 and assuming that F(4, 0) = F( 4, I) = 0, write a recursive formula
for F( 4, n) with linear terms and integer coefficients. No proof is required.
5. a) Form, n ~ 2 with m and n not both odd, develop a formula for the perimeter of all ARML (m, n) polygons
as a function of m and n and prove it.
b) Prove that ifm and n are both odd, m, n > 2, then F(m, n) 0.
150
6. Prove that for every given m, 11 2: 2 where 111 and 11 arc not both odd, the area of any
. b mn
ARML (m, n) polygon is given y the formula 2 - l.
In the following problems let the area of an ARML (m, 11) polygon be represented by K(m, 11).
For these problems m and 11 are not both odd and both are greater than or equal to 2.
7. Let x, y, and z be distinct positive integers whose sum exceeds 100. Find an example of
three ARML (m, 11) polygons of the form (3, x), (4, y), and (5, z) such that K(3, x), K(4, y),
and K(5, z) are an increasing arithmetic progression. Show your analysis.
8. Since ARML (3, 14), (4, 11), and (5, 12) polygons have areas of 20, 21, and 29 respectively, their areas form
a Pythagorean triple. Determine another set of ARML (3, x), (4,y), and (5, z) polygons whose areas are
proportional to a 20-21-29 triple. Show the work that led to the answer.
9. a) Determine with proof all ARML (m, 11) polygons whose area is half that ofan
ARML (m + 7, 11 + 7) polygon.
b) For n, m 2: 2, show that there are no values ofm and 11 such that K(m, m)
10. For a fixed value k, compute the number of ordered pairs (m, n) such that the area of the
ARML (m + k, 11 + k) polygon is twice that of the ARML (m, n) polygon. Use the
prime factorization of k 2 + I = 2a 2 3a3 ts ••• in your answer.
Bonus question: for 4 additional points give a correct proof of the correct formula for problem #4. Do not attempt
this problem until you've completely finished the Power Question.
151
Power Question Answer Sheet: Four ARML (6, 8) Polygons
• •
• • • • • • • • • •
• •
• • • • • • • • • •
• •
• • • • • • • • • •
• •
• • • • • • • • • •
• •
• • • • • • • • • •
• • • • • • • • • • • •
• • • • • • • • • • • •
• • • • • • • • • • • •
• • • • •
• • • • • • •
• • • • • • • • • • • •
• • • • • • • • • • • •
• • • • • • • • • • • •
• • • • • • • • • • • •
• • • • • • • • • • • •
• • • • • • • • • • • •
• • • • • • • • • • • •
152
Power Question: ARML (3, 6) Polygons
...
... ... • ...
...
... • • • •
...
... ..
• • • •
. .
.. .. .. ... ..
...
ARML (4, 4) Polygons
... • • • • • ... . .
• • .... • • • •
... ....
• • • •
. .
•
.... • •
. .
...
. ... .... .. • • • .. • • ....
• • .. • • .. .
. . ....
•
. .. ....
.... •
.. • • • •
...
..
. . . .• .•
• • • •
.. .. ..
• • • •
...
. . .. . ... ... . .
... . . . . • .• .• • •
153
•
• • • .• .• • •
•
• . . • •
• • • .. • • • • • .. • • • • • • •
• • • . . • • • • • • • • • • • • • • •
. . • • • • . . .
• • • • • • • •
• • • • • • • • • • • • . . •
• • • •
• . . •
• . . . • • • • • • • • • • • • • • • • • • •
• .. • • . . .
• • • • • • •
• • • • . .
. .
• • • •
. .
•
• •
• •
• • • •
• •
• •
• • .. •
•
• • .. .. • • . . .
... • • • •
• • • •
. .
• • •
• •
• •
... •
.• .• .• • • • •
. . .. .. • •
... •
• ... . .
• •
.. • •
• . .
• •
. .
• •
• • •
•
• • • . . . .
. . . . ..
•
. .. • •
. .. . •
• • •
... . .
• •
• •
• •
. . . . . . ... •
.... . . .. . . . . . . .
.... ..
• • • •
. . . . .
.... . . . . . . . . . . . .
• • •
. .
• .. •
•
• • • • • • •
• • • • . . . • • • • . . • .. •
... . . . • • • • . . • • . . •
154
. .
. .
... •
..
155
Power Question: ARML (4, 6) Polygons--Continued
. .
. .
..
. .
. .
. . . .
..
. .. . .
. .
...
156
I. a) Shown below are 4 distinct ARML (6, 8) polygons. There arc other possibilities.
r: _➔ ~
• . l L. --.
L • . • -• J
•----- l
L..... .
~-----.
.---. ·:------~ l
I . - -. ~--. !
L:: :--: : • -l •• ,_j
b) The area of each polygon is 23 and the perimeter of each is 48.
2. a) The number of ARML (3, 6) polygons is 4 .

L:
..r --•
l 1:1
T :J
L7
I
.I fr?
L.
.
l l
!
.---
I
•--• LI L:J l-
b) The number of ARML (4, 4)

polygons is 6.
c) The number of ARML (4, 5)

polygons is 14: 3 pairs and 2 quads
depending on symmetry.
i l~1l !! '. l
I~ ! !W
d) The number of ARML (4, 6)
polygons is 37. These are grouped:
[7l :·1
3 singles, 7 pairs, 5 quads, depending [rJ
on symmetry.
u:! : l
LJ
I!!l
[~.~
[:J
[11 lill I !fl
L--~
•
Lrr
r· .
l::J
I
-♦--♦--~i
C.---;
ls l~ .
L~~-'.
Li~ ~~..
I --•
f1J.-,
m -r
'. !~J
r~ L1~ Wj
~ J r:~-
t:
- ~
LI
l -
lru f1~ I I01 [~j 1/J c:__:__:

~
I~~
M
r~1 ~
Ln
c-~ I~ ·7
I
t,j 1 I . 411
trJ ~ I~ ~
:..JU
7 ~r1 L~~
-------, t
c_~ E~ • 1 i
L ..
157
3. a) Rotate the array 90°. Starting from the comers, the path must look
like the diagram at the right. The only choice is to continue to move C. :_]
horizontally, otherwise, some point will have 3 segments joining it to
other points. Thus, F(2, n) = I.
b) Given the left-hand array, point B on the central axis cannot connect A B C A B C
• • •
with£ since A and C would then lie off the polygon. Thus, the
polygon must look like the right-hand diagram. Since n is even,
F• •
E
•D
Fi--i L
G• •
H
•I G•---
H
•I
there are n - 2 points in the central axis other than Band T which
could be connected by a segment, giving - 2-

n-2
segments. A
• •
• •
•
•
•
•
r--•.
- --
segment EH can either be paired with an indent from D, namely

DEHJ, or an indent from F, namely FEHG. Thus, for each of the
• •
• • •
• T• •
•
1::::
T
i
n- 2
- 2- segments there are two choices. The total number of choices and hence, the total number of polygonal
paths is 2(n- 2)12 , making F(3, n) = i(n- 2)12 .
4. We have F(4, 0) = 0, F(4, l) = 0, F(4, 2) = I, F(4, 3) = 2, F(4, 4) = 6, F(4, 5) = 14, and

F(4, 6) 37. Thus, for n 2: 5 we have the following formula which is consistent with these results:
F(4, n) 2F(4, n - 1) + 2F(4, n - 2)- 2F(4, n - 3) + F(4, n -4).
5. a) There are m · n lattice points. Since the segments can be put into a 1-1 correspondence with the lattice
points by associating each segment with its starting point, the perimeter is mn.
b) As the polygonal path is traversed, every segment on which one moves to the right must be matched by a
segment on which one moves to the left since the path finishes at the starting point. Similarly for up and
down movements. Thus, the perimeter is even. But the perimeter equals m · n. and if m · n is even, then at
least one of m and n is even. Thus, if both are odd, there is no polygon.
Note: Since #6 is independent of#5b we could also prove #5b using #6. Since the ARML polygon consists
of joined unit squares, its area is an integer. Thus, for the area, !!!!!. - l , to be an integer, mn must be
2
divisible by 2, making m, n, or both even, but not both odd.
158
6. Pick's Theorem states that the area of a simple c loscd polygon whose vertices are lattice points equals
p
Cf + - - l where cJ is the number of lattice points in the interior of the polygon and p is the number of lattice
2
points on the boundary. In the case of ARML polygons, there are no interior points since every point lies on
mn
the path. Thus q = 0 and p = mn. The area is, therefore, 2 - I.
Induction proof: it is easier to prove the result in the more general case rather than specifically for
ARML (m, n) polygons. Consider any finite simple closed polygon of squares whose boundary consists of
unit line segments that are horizontal or vertical, containing no vertices in its interior. Let its area be K, its
p V
perimeter be p and the number of vertices on the boundary be v. We will prove that K =- - 1 - - I bv
2 2 '
4
induction on K. Let K = 1. The polygon is a square of side 1; clearly p = v = 4 and K =2- 1. Assume the
result is true for K = t and prove it for K = t + I. Consider any such polygon with area t + I. Then this
polygon contains a square with three edges in common with the boundary (see justification below). Remove
this square by erasing three edges and adding the fourth. Then, the area has been decreased by I, the
perimeter by 2 and the number of vertices by 2 and by induction, the result holds.
Justification: pick any square contained in the polygon. If it has 3 edges in common with the boundary, then
we are done. Otherwise, move to one of the adjacent squares. lf the new square has 3 edges on the boundary
then we're done. Otherwise move to one of the adjacent squares, but not the one you just arrived from, and
continue the process. Note that you can never return to a square you have previously visited (or else you will
have traveled around a loop in your polygon which must contain a vertex, contradicting the fact that the
polygon is simple). Then the process must end at some point and the only way it can end is by finding a
square contained by the polygon with 3 edges in common with the boundary.
7. LetthcareaofanARML(m,n)polygonbedefincdasK(m,n). ThenK(4,y)-K(3,x) = Jand
K(5, z)-K(4,y) = d. From #6 we obtain ( i., - 1)-( ~- 1J =d and ( ~- 1J-( ~- 1J =d giving
4y - 3x = 2d and 5z - 4y = 2d. Thus, 4y - 3x = 5z - 4y, giving 3x - Sy+ 5z = 0. We need to choose

distinct x, y, and z such that the areas are in an increasing arithmetic progression. Not all solutions will work.
For example, from #5b neither x nor z can be odd. Additionally, while x = 60, y = 30, and
z = 12 solves the Diophantine equation and has a sum greater than 100, that solution produces areas in a
decreasing arithmetic sequence: 89, 59, and 29. But x = 30,y = 40, and z = 46 works, giving areas of 44,
79, and 114. There are infinitely many solutions.
159
8. Consider K(3. 2r) = 20w, K(4,y) = 21 w, and K(5, 2s) = 29w. Then 3r - I = 201r, 2y ~ I = 21 w, and
5s - I = 29w. Clearly, w must be odd, otherwise y would be fractional. Let w = 2v + I
for v = 0, l, 2, ... Then 3r- I = 40v + 20, 2y- I = 42v + 21, and 5s- I = 58v + 29, giving
40v 58v
r = - 3-+7,y = 2lv+l1,ands = 5 +6. Clearly,vmustbeamultipleofl5,solettingv=l5twe
obtain r = 7 + 200t.y = 11 + 3 I St, ands = 6 + l 74t, giving ARML polygons (3, 14 + 4001),
(4, 11 + 3151), and (5, 12 + 3481). Letting t = I, we obtain the following: (3,414),(4,326),(5,360).
9. a)
(m+7)(n+7)
Setting -'------------'- - 1 = 2 -
J
[11111 - I yields n = -
51+7,n 100
- - = 7 + - - . Thus, m - 7 must divide
2 2 m-7 m-7
I 00 ➔ m- 7 E {I, 2, 4, 5, 10, 20, 25, 50, l 00 l. Solving form and n we obtain the following:
(8. I 07), ( 11. 32), (12, 27), (27, 12), (32, 11 ), (107, 8). We excluded (9, 57), (17, 17), and (57, 9) since in #5b
we showed that at least one of m or n had to be even.
2 2
b) If K(m,m) 1 K(n, n), then m - 1 = 1 [ n
=2 2 22 - 1] ➔ 2(111-? - 1) = n 2 . Thus, n2 is even so
let n = 2k. Then 2(m 2 - I) = 4k 2 ➔ m2 = 2k 2 + 1, making man odd number. But by #5b there are no
ARML polygons in which both dimensions are odd, so there are no square arrays in which an ARML
polygon can be half the area of another.
Setting
( m + k Xn + k) [ mn J
- 1 = 2 - 2 - I gives n =
k 2 + 2 + km 2(k 2 + 1)
k + ---- Let the prime
2 m-k m- k
2 a a a 2 a +I a a
factorization of k + I be 2 23 35 5 ... . Then the prime factorization of 2( k + 1) is 2 2 3 35 5 ... . Let
d = m - k be a divisor of 2(k 2 + 1) and note that any value of d for which either morn or both are even
results in an ARML polygon. There are several cases:
i) k is even. Then we claim that all factors d will give an ARML polygon. First, if dis even, then since
2(k 2 + I)
m = k + d, mis even. If dis odd, then d is even, and since n then equals an even plus an even,
n is even. Therefore, the number N of such polygons is N = ( a 2 + 2)( a 3 + IX a 5 + 1) · · · .
a a a
ii) k is odd. If dis odd, then m = k + dis even. Thus, any factor d of the form d = 3 3 5 5 7 7 ... gives an
2(k 2 + I)
ARML polygon. If dis even, then mis odd. In order for n to be even, d must be odd, implying that
a +I a +I a a
2 2 divides d. Thus, if k is odd, any factor d of the form d =2 2 3 3 5 5 ... gives an ARML polygon.
In counting the number of ARML polygons in this case, there are two choices for factors of 2, either no 2's or
all 2's. Therefore, the number N of such polygons is N = 2( a3 + IX a5 + 1) ....
160
1-1. For l < x < y, let S = {I, x, y, x + y) . Compute the absolute value of the difference between the mean and
the median of S.
1-2. BCD£ is a square, MBC =tiFCD with mLA = 120° A
and AB= AC. If AF= 2000, compute the area of square BCDE. C
1-3. For x, y, z, and w 2: 0, compute the smallest value of x satisfying the following system:
)' X - 2001
z 2y - 2001
w 3z - 2001
1---4. If b = 2000, compute the numerical value of the infinite sum:
0 4o I 41 2 42 3 43
(log b 2) (logb 5 ) + (log b 2) (logb 5 ) + {logb 2) (log b 5 ) + (logh 2) (logb 5 ) + ...
1-5. If, from left to right, the last seven digits ofn! are 8000000, compute the value ofn.
1-6. Let S be the sphere whose equation is x 2 + y2 + z 2 = 25 . Points P( 3, 4, 0) and Q( 3, -4, 0) lie on S.
Compute the number of planes containing P and Q that intersect Sin a circle whose area is an integer.
I-7. The measure of the vertex angle of isosceles triangle ABC is 0 and the sides of the triangle arc sin 0, Jsin 0,
and Jsin 0 . Compute the area of MBC.
E J F
I-8. Square ABCD is surrounded by four congruent rectangles. Let
/c-,-A~-
S = perimeter of ABCD + perimeter of EFGH +AI+ BJ+ CK+ DL. B
Let T = area of EFGH. Let m be the greatest lower bound and n be the least upper
D C
bound of all values of T for which §_ = 12. Compute the ordered pair (m, n). K
T
H L G
161
1
l. -
4
2. 6,000,000
3. 3335
1
4.
3
5. 27
6. 56
8
7.
25
8. (¾,i]
162
1-1.
l+x+y+(x+y)
The mean is - - - - - - -
J
- +
2(x + F) x + V
· , the median is - - · and the difference is -4 '.
[1
4 4 4 2
1-2. Since AB = AC, mLACB = 30°. Since MBC ::c: t.FDC, then A
\
\
mLFCD ~ 30° ➔ mLFCB ~ 60°. Thus, mLACF = 90° and B 30° ~ 30c . C
~ 60°
since AC = FC, MCF is a 45-45-90 right triangle. Let AF = x ➔ \45o
\
\
X F
AC= Use the Law of Cosines on MBC or drop an altitude from A
E ---- D
to use 30-60-90 right triangles to obtain BC =

x.J3
Ji. . Thus, the area of BCD£ = 3x 2 3 · 2000 2
2 - 2
16,000.0001.
1-3. Sincew 2: 0,3z-200I 2: 0 ➔ z 2: 667. Since::= 2y-200l,2y-2001 2: 667 ➔ y 2: 1334.
Sincey = x-2001,x-2001 2: 1334 ➔ x 2: 3335. Thesmallestvalueofxisl3335I.
Alternate solution: multiply the first equation by 6 and the second by 3: 6y 6x 12006
3z 6_1· 6003
w 3z 2001
w + 20010
Adding yields 6y + 3z + w = 6x + 6y + 3z - 200 IO ➔ w = 6x - 200 IO ➔ x = ----
6
. . 20010
Smee w > 0, the mimmum value for xis - - = 3335.
- 6
1--4. I· logb 5 + (logh 2)(4) logb 5 + (logb 2l(l6) log b 5 + (logh 2) 3 (64)logb 5 + .. · =
logh s( 1+ (4 log, 2) + (4 log,, 2) 2 + (4 log, 2) 3 + ... ) 0

log, s[ 1 _ 4 :ogh 2 l 0
logb 5
log 5 [ - - - I - -: = = logb/1 6 5. Since b = 2000, we have log 125 5
h 1ogh b -1ogb 16 b
log -
b 16
1-5. Since n! ends in 6 zeros, n! must have exactly 6 fives in its prime factorization, so we know that 25 :S n '.S 29.
n!
Consider the units digit of - . For n = 25, a quick multiplication by all remaining factors yields a units digit
106
of 4. For 11 = 26, the units digit is also 4, but for n = 27 it is 8. Thus, n = I 27 1.

Note that for n = 28, the units digit is 4 and for n = 29, it is 6.
163
I-6. The largest circle through P and Q is the horizontal circle containing the center
of the sphere. It has a radius of 5. The smallest circle is the vertical circle with
radius 4. Let A be the area of a circle formed by the intersection of a plane and
sphere S. Then l 61t :S A :S 257t giving 50.26 :S A :S 78.54. If A is an
integer. then 51 :S A :S 78 ➔ 28 integral values for A. The plane can tilt
from the horizontal up to the vertical or down to the vertical making a total of
[3:6] planes intersecting Sin a circle of integral area.
')
1-7. BytheLawofCosincs. sin-0 = sin8+ sin8-2sin8cos8. Since sin8 -:t- O,divideby sin8 to obtain
sin8=2-2cos8 ➔ sin 2 8=(2-2cos8) 2 ➔ l-cos 2 0=4cos 2 8-8cos0+4 ➔
5 cos2 8 - 8 cos 8 + 3 = 0. Then cos 8 = I or i5 Reject I since it makes 8 = 0. Thus, sin 8 = ~ and the
5
4 4
area of MBC = ½Jsin 8 · Jsin 8 · sin 8 2 5 5
8 AD AC sin 8 B
Alternate solution: Since sin 2 r:--::, then
~
AB 2AB 2wsmt:J
2 sm 2 r.-:::et:J
· 8 = wsm ➔ 4 sm
2· e.
· 2 8 = sm s·mce sm· e = · 8 cos 8 .
2 sm
2 2
A sin 0 D C
divide by 2 sin~ to obtain 2 sin~ cos~ ➔ tan~ Since 2
2 2 2 2 2
2 tan~
4
sine then sin 8 Continue as above.
2 0' 5
l+tan 2
I-8. Let the side of the outer square be x and the length of each rectangle bey. Then
7 4x + 4v 2
T = x~ and S = 4x + 4y. Hence, 2 · = 12 ➔ y = 3x - x. There are I
X
X
two degenerate cases, one where the interior square disappears, suggested by Fig. 1
i
and the other where the interior square fills the outer square, suggested by Fig. 2. I
X .
In the former, y = -2 and m the latter, .v = x.
X X 2 1 2
Thus. - < y < x ➔ - < 3x - x < x ➔ -<x<-.
t[J
2 2
1 4 [B]l4)
2 3
Thus - < T < - . Ans: - - .

' 4 9 4 '9
Fig. 1 Fig. 2
164
Let [x] denote the greatest integer less than or equal to x. Compute the largest integer n
such that [ !12000 ] > 1.
Rl-2. Let T = TNYWR. Compute the larger solution to (logT x) 2 = lo~ x2 .
Rl-3. Let T = TNYWR. Let n E { I, 2, 3, ... , T 2 }- Compute the number of values of n for which the product
(n2 - 2n + 2Xn 2 + 2n + 2) is divisible by 5.
R2-l. For 0 < x < 2rc, compute the absolute value of the difference between the solutions to the
.
equat10n: sec x = I + cos x + cos 2 x + cos 3 x + ...
2 2
R2-2. Let T = TNYWR and set K = 4 rc If (a - b1{ (a+ b1{ 512, compute the value of a + b .
T
R2-3. Let T = TNYWR. Arcs are measured in degrees.
If
-
mABC
_
mADC
=T
mADC
+ I, compute - - -
mLP
- p B
165
Relay #1:
Rl-1. 10
Rl-2. 100
Rl-3. 8000
Relay #2:
47t
R2-1.
3
R2-2. 8
1
R2-3.
4
166
Solutions to ARAfL Relay # 1 - 2000
nr::::= n . IO II ~
RI - I. Since we want '1(2000 2: 2, then 2000 c:: 2 . Smee 2 < 2000 < 2 , n = llQJ .
Rl-2. (Iogr x) 2 - 2 logr x =0 ➔ (togr x)(Iogr x - 2) = 0 ➔ x = r0 or T 2 . Since T = 10,
the larger value of x ~ 1o2 = }QQ] .
Rl-3. Since T = 100, then n E { 1, 2 .... , 10,000}. Since (n 2 - 2n + 2)(,/ + 2n + 2) = n 4 + 4, then
n4 must end in I or 6 ➔ n must end in I, 2, 3, 4, 6, 7, 8, 9 ➔ n doesn't end in 0 or 5. There are
lO, OOO = 2000 numbers for which n 4 docs not end in I or 6, so there are I 8000 I which do.
5
2 2
Or consider: n (mod 5) n - 2n + 2 (mod 5) n + 2n + 2 (mod 5)
0 2 2
I I 0
2 2 0
3 0 2
4 0
The product of the two numbers in the first row is not divisible by 5; the products in the other rows are.
Since T 2 = l 0,000 and that is a multiple of 5, four-fifths of the numbers in the set will give a result
divisible by 5.
Solutions to ARML Relay # 2 - 2000
l 51t 7C
R2-l. secx = - - - ➔ sccx-1 I ➔ sec x = 2 ➔ cos x = ➔ x = - or-.
1- COS X 2 3 3
The desired difference is ffi.
K 2 K
R2-2. ( ( a - hz)( a 2
+ ln) ) = ( a + h ) , then a- + b
"J 2
= 512 l !K .
Thus, a 2 + h 2 = 512 113 = [[].
R2-3. Let the measure of arc ADC = x. the measure of arc ABC = 360° - x and mLP = 180° -x. Then,
360° - X 360° 360° 180T

T+ l ➔ - - = x. Also 180° -x = 180°- - - - - . Thus,
X T+ 2 T+2 T+2
X
-2 . Since T = 8, the ratio is ~·
- .
180° - X T 4
167
Note: Pass answers from position 1 to 8 and from position 15 to 8.
I. The letters A, R, M. and L represent even digits. A, R, M, and Lare not necessarily distinct and A is not equal
to zero or to L. Compute the positive integral value of b such that JARML b == 2 5 .
2. Let T = TNYWR. but Tis irrelevant here. P( x) == x 3 + 2x 2 - ( 4K + 5)x - 6 has three distinct zeros.
If two of these zeros are the same as those of /(x) == x 2 + 5x + K, compute K.
3. Let T = TNYWR. The centers of three congruent circles lie on the diagonal of
square ABCD. The two end circles are tangent to the square and to the middle
circle. If the radius of each circle equals fl - 1, compute the area of ABCD.
4. Let T = TNYWR. Compute the number of integers which lie in both of the intervals [ T- 9, 2T+ IO]
and [ T-4,2T+ 13].
5. Let T = TNYWR. Suppose T elements (not necessarily distinct) are selected from the set {O, I, 2, 3, 4, 5).
Compute the number of distinct sums of these T elements.
T
6. Let T = TNYWR. Let K = _ 8 Compute the value ofy solving: 6x + 2y K
-7x + 3y 30
7. Let T = TNYWR. Compute the number of distinct T-letter words that can be formed using the letters from
the word ARRANGE.
8. You will receive two numbers. Let Mbe the larger of the two and let m be the smaller. Compute the area of
the set of points (x, y) satisfying: m $ (Ix I- 2000)2 + (I y I- 2000) 2 $ M .
168
15. The letters A. R. M, and L represent even digits. A. R, M. and Lare not necessarily distinct and A is not equal
to zero or to L. Compute the positive integral value of b such that JARMLb = 25.
14. Let T = TNYWR, but Tis irrelevant here. P(x) = x 3 + 2x 2 - ( 4K + 5)x - 6 has three distinct zeros.
If two of these zeros arc the same as those of f(x) = x 2 + 5x + K, compute K.
13. Let T = TNYWR. In MBC, AB = AC = 10 and BC = 2 T. Point Pis chosen at random on BC.
Dis chosen on AB and Eon AC so that PD and PE are parallel to AC and AB respectively.
Compute the perimeter of PDAE.
12. Let T = TNYWR. The number 2000T has x terminal zeros when expanded. Let the rightmost non-zero
digit bey. Compute x + y.
-T
II. Let T = TNYWR and set K = 22 . Points A( I, 2), B(5, K), and C(K, 7) are collinear. Compute
.......
they-intercept of AB.
12T K 2 -I
10. Let T = TNYWR and set K = In parallelogram ABCD, if cos B - 2- - , compute sin A.
13 K +I
9. Let T = TNYWR and set K = 90T. ABCD is an isosceles trapezoid with AB parallel to CD.
If BC = 5(AB), CD = 7(AB), and the perimeter of ABCD = K, compute the height of ABCD.
8. You will receive two numbers. Let Mbe the larger of the two and let m be the smaller. Compute the area of
the set of points (x,y) satisfying: m (I xi- 2000)2 + (I y I- 2000) 2

:<::: :<::: M.
169
1. 8
2. 2
3. 4
4. 19
5. 96
6. 3
7. 84
15. 8
14. 2
13. 20
12. 66
13
11.
4
3
10. -
5
9. 12
8. 2887t
170
3 2 10
I. Ah + Rh + 1'vfh +L = 2 = 1024. lfh = 9 or 10, then A = I, but A is to be even. Try h = 8. Then
5l2A+64R+8M+L = l024andclearlyA = 2,R = M = l = Oworks. lfh = 7,thcn 2662 7 = 1024,
but A = L. 1f h = 6. then 4424 6 = I 024, but A L. Thus, h = [[j, giving J2000 8 = 2 5 , fitting for
ARML's 25 th anniversary.
x3 + 2x 2 - ( 4K + 5)x - 6
2. 7
(x - 3) with a remainder of I Sx + 3K - (SK+ 5)x- 6. The remainder
x- +5x+K
is zero ifK ~ ~.
3. Let r be the radius of the circles. Using 45-45-90 right triangles, AC = 4r + 2r./2.. Since T = 2,
r = ./2. - I, making AC = 2./i( .Ji. + 1)( ./2. - 1) = 2./i. Thus, AB = 2 and the area of ABCD = [±].
4. Common integers lie in [T-4, 2T+ 10] ➔ (2T+ 10)- (T -4) + I T + 15 integers. Since T = 4, then
there are lliJ integers.
5. Using T O's the sum is 0. By replacing each integer n by n + I, we can obtain all sums from T · 0 to T · 5.
Thus, there are ST+ 1 sums. Since T = 19, 5T+ I = I961.
6. Multiply the top equation by 7 and the bottom by 6 and add, canceling the x's and obtaining 32y = 180 + 7 K.
.
Smee K =
96
- = -12 v = -96 = 3 . W
-8 ,. 32
7. If T = I, there are 5 distinct words. If T = 2, there are two words using RR or AA and 5 · 4 = 20 using
ARN GE. Total: 22. If T = 3, then with two R's and another letter we obtain ~ •4 = 12 words. Similarly
2!
for two A's and another letter. Using ARN GE there are 5 · 4 · 3 = 60 words. Since T = 3,
there are 60 + 12 + 12 = [81] words.
171
15. Sec# I above. Ans:[[]. 14. See #2 above. Ans: 0.
13. Tis irrelevant; the result is invariant. Since MEC is isosceles, PE = EC so AE + EP = AE + EC = AC.
Similarly, AD+ DP = AD+ DB = AB. Perimeter of PDAE = AB+ AC = 20 I I.
12. ( 2 · 10 3 r = 2T · 10 3T. Thus, x 3 T and for T = 20, x = 60. Since 2 lO ends in 6, x + y = I66 I.
K -2 5 2
11. mAB = mAC ➔ -- - -- ➔ K - 3K - 18 = 0 ➔ K = -3 or 6. If K = 6, the line is
4 K-1
v = .
x + I an d the intercept .
1s 1. If K = -3, th e 1·me 1s
. v = -- 5x + -13 an d the v-mtercept
. . -
1s
13
. You cou Id
· · 4 4 · 4
pass each back in turn. Since K = -3, the answer isl]]

1131
2 (Kl - 1)2
Since adjacent angles of a parallelogram are supplementary, sin A = sin B. Thus, sin B = I - 2
(K 2 + 1)
12 13 .
Since K = - •- = 3 sm B
13 4 '
9. From the diagram, we clearly have 3-4-5 right triangles, so the height A X B
I
3 5x 5x
: 4x
is 4x. The perimeter is 18x, so 18x = 90 · 5= 54 ➔ x = 3 ➔ I
I
D 3x X : 3x C
height = [ill .
8. By replacing Ix I and I y I with x or -x, y or -y, we obtain a graph consisting of
4 rings formed by 4 sets of concentric circles. As long as Mis small compared

to the distance to the axes, the rings do not overlap. That is the case here. The
radii of the two circles are /;i and .Jt:i, and the area of each ring is
rtM - rtm. Since M = 84 and m = 12, the total area equals 4rt(84 - 12)
1288rcl.
172
ARML Tiebreakers 2000
I. Points A. B, C and Dare the midpoints ofthc edges of the cube as shown. A plane passing through AB and
CD intersects the cube in a region whose area is i6. Compute the surface area of the cube.
------------------
'
2. Let V(2, 3) be the vertex of a parabola whose equation is of the form y = ax 2 + hx + c. If the graph of the
k
parabola has two positive roots, compute the greatest integral value of k such that a = - -
2000
3. Perpendicular chords AB and CD of circle O intersect at P. If the radius of O is l 0 and OP = 4,

2 2
compute AB + CD .
173
I. A plane passing through AB and DC intersects the cube in regular
hexagon of side DC. Let DC = x. Then the area of the hexagon equals
x 2 Jj r,
6 · - 4- = 116 ➔ x =
'fs
i3. Thus, DC =
f"s
i3. The length of the edge
X X
DC ,::
,of the cube equals r:: · 2 = DCy2. Thus, the surface area of the cube
, 112
D C
equals 6(Dc '2}2 = 12~ = ls/ii.
-b 2 V(2, 3)
2. Since - = 2, then b = -4a. Since 3 = a· 2 - ( 4a)2 + c, then
(\
2a
c = 4a + 3. For there to be two positive roots, c must be negative. Thus,
3 k -3
4a+3< 0 ➔ a< - 4 . Hence, 2000 < 4 ➔ k < -1500. The
largest integral value of k is 1-15011.

I __,1
~-·---1
I
3. Let AM = J100 - y2 making AB= 2J100 - y2. Let ND= J100 - x 2 making CD= 2J100 - x 2 .
Then AB2 + CD 2 = 4(100 - y2) + 4(100 - x 2 ) = 800 - 4(x 2 + y2). Since x 2 + y2 = 16, then
AB2 + CD 2 = 800 - 64 = 17361.

C
-------
M p
B
4
N
Q '. X
10 \
'
174
2001
rream 'R.ound ···························································································· l 77

Power Qyestion......................................................................................... 186
'lndividua( 'R.ound ................................................................................. 193
'R,e(ay 'R.oun/............................................................................................. 197
Suyer ·Re(ay................................................................................................ 200
rr1ie6reakers 205
-175-
THE 26th ANNUAL MEET
For the 26th competition, ARML opened up a new site. Since the site in Las Vegas was proving
to be too small, the Executive Board spent the year looking for a new site and finally found the
perfect one at San Jose State University in California. Mark Saul retired as president of ARML
·to take up his new duties as a director of the NCTM and Tim Sanders, director of the Great
Plains Mathematics League and corresponding secretary for ARML, took over as president.
This year proved to be ARML's largest yet. There were 25 teams in Division A, 76 teams in
Division B and 7 guest teams, 6 from Taiwan and, for the first time, a team from the Philippines.
The total of 108 teams meant that well over 1600 students as well as countless teachers and
coaches participated in the contest. In Division A, San Francisco Bay A edged out
Massachusetts A. They were tied after the team round and the Power Question and the relays
equalized each other, but with a narrow victory in the individual round, SFBA gained a 6 point
victory. In Division B, Ontario B 1 won easily after amassing an 18 point advantage in the
individual round. ARML also introduced the ARML song contest in which teams competed to
come up with the best song with mathematical content. It was great fun.
Tim Sanders received the Samuel Greitzer Distinguished Coach Award. Tim is currently the
new president of ARML and has served on the Executive Board as a secretary. He has worked
with the Kansas and Missouri teams.
John Benson of Chicago A received the Alfred Kalfus Founder's Award for his long time service
to ARML. John has been a dedicated organizer of teams and a site coordinator at Iowa.
The following students received the Zachary Sobol Award for outstanding service to their team:
Brentan Alexander Colorado

Andrew Dexter Western Massachusetts
Jeff Huang Chicago
Jennifer Philbrick Maine
176
ARML Team Questions- 2001
T-1. AB1- BC, ED 1- BE, mLABD = 3x, mLJJBC = x + y, and [)
mLCBE = 2v. Compute the value of yin degrees. B C

E
T-2. If each distinct letter in JARML ~ represents a distinct positive digit in base I0, compute the
sum A + R + M + L.
T-3. The Luzors played y games, winning some and losing the rest. They then won 3 in a row and improved their
winning percentage by exactly I 0%. Compute the least value ofy.
T-4. Rhombus ABCD is inscribed in y =x and _ 1· = x 2 with CD lying on y = x and

C
AD parallel to the y-axis. Compute the sum of the coordinates of A. (Diagram /)
not drawn to scale).
T-5. Let T = {1, 2, 3, 4, ... , n} and let S( X) denote the sum of the elements in set X. Suppose T can be divided
into three disjoint subsets A, B, and C whose union is T. If S(A) : S(B): S(C) ~ 3 : 4: 6, then compute the
smallest possible value for the largest element in C.
T-6. Compute the ratio of the volume of a sphere to the volume of the largest regular octahedron
that will fit inside it.
T-7. In a square of side n, counting numbers arc arranged in an inward spiral. For ~2-3 4 5
example, the diagram shows the result when n = 5. If n ~ 27, compute the 16 17 I 18 19 6
15 24 25 20 7
sum of the elements in the diagonal from the top left to the bottom right.
14 23 22 21 8
13 12 11 10 9
T-8. Compute the number of sets of three distinct elements that can be chosen from the set
{2 12223
, , .... , 2199922000}
, sue hht at th e tl1ree e Iements ,.-1orm an mcreasmg
· · geometric
· progression.
·
T-9. Let ABCDEFGH be a rectangular box such that AB = AD = 20 and
mLGAC = 45°. Point Plies on DH such that plane PAC is parallel to
BH . Compute the volume of tetrahedron FPCA.
T-10. A tree is a connected graph with no circuits. A tree contains at least one
node. To the right is a tree with 15 nodes. Define a subtree of a tree to be
a subgraph that is also a tree, i.e., a nonempty subset of the nodes along
with the edges joining them that together form a connected set. Compute
the number of subtrees in the figure.
177
270 -
1. - or24.54
11
2. 18
3. 12
4. 2-J2
5. 9
6. 7t or 7t :1
7. 12767
8. 999,000
9. 2000,/2
10. 750
178
270
T-1. (3x)+(x+y)=90 and (x+y)+2y=90 ➔ 4x+y=90 and x+3y=90. Solvingv ~ - 1-1 or24.54
T-2. Since0 < L < 10, I00OA+ JOOR+ IOM+L < (IOA+J0)2. Drop off I0M+L and divide by JOO to
. 2 ') ')
obtam l0A + R < A + 2A + I. Thus, A- - 8A + 16 > R+ 15 ➔ (A - 4)- > R+ 15.
Since R 2: I, ( A - 4) 2 > 16, making A = 9. Since the square of AL ends in L, then L = I, 5, or 6, making
AL= 91,95,or96. Now, 91 2 = 8281 fails because A andMarenotdistinctand 95 2 = 9025 fails
because R is not positive. But 96 2 = 9216 works, giving the sum 9 + 2 + I + 6 = ~ .
x
T-3. Letxbethenumberofgameswon. Then ( - -
x+3
= --
JI] ➔ v=
33x
Let z = 30-x. Then
y 10 y+3 ' 30-x
33(30- z) 990 2 · 3 2 · 5 · 11
y= --33 - 33 . Since y is a positive integer, z must be a divisor of 990
z z z
such that the quotient is greater than 33. Of these 11 candidates for (z,y): (I, 957), (2,462), (3,297),
(5, 165), (6, 132), (9, 77), ( I 0, 66), ( 11, 57), ( 15, 33), ( 18, 22), and (22, 12), z ~ 22 gives the minimum value of
~
y as L.!L . TIrns, the team improve
. d.its wmmng
. . percentage by exact Iy I 0% as 1t
. went from -8 to -II .
12 15
Note: one coach thought that the problem was ambiguous in that an increase of 10% could mean adding 10% to
the original winning percentage. If the problem is interpreted in that way, we have the following equation:
IOOx IO0(x + 3) 27 y - y 2 . . . .
- - + 10 = ---------- ➔ x = · . lnterestmgly, the smallest solutions to this equat10n are
y y +3 30
x = 6 andy = 12, so regardless of the interpretation, the least value fory is still 12.
T--4. Starting with D(a, a) and C(b,b), we have A(a,a2 ) and ~b,b 2 ). The slope of AB = I, so
= b+a
Since b = l - a, a - a 2 = (l - 2a~. Solving the quadratic a2 - (2./i. + l)a + i2 = 0 gives
(2-h+1) ± Js+4i2+1-4./i. (2 +l)±3

a = -'----..;.._________ = ______..;..__. This gives a =
-h i2 + 2 which we reject since it
2 2
is too large or a = ./i. - 1 "" 0.588 which works, making a+ a 2 = l2-i2 1

•
179
Note: As originally stated, the problem did not specify that AD and BC were vertical. In the midst of grading
the Team round, we suddenly realized that more than one rhombus was possible, as indicated in the diagrams
2
below, and that, therefore, there would be a range of values for a+ a .
Sasha Schwarz found the range of a+ a 2 in the following way: because AB is parallel to y = x , we have
A( a,a 2 ) and B( 1 - a,(] - a) 2 ). The length of AB will range from AE, the perpendicular distance from
A to y = x, to AO, the distance from A to the origin 0. Using the formula for the distance from a point to a
line, AE =
a - a 2 and clearly, OA
/i =
J.a + a
2 4
. Given that AB r::( 1 -
= 1/2 2x) we have
a-a 2 C: 2 4
/i ~ -t2 ( I - 2a) ~ a + a . Using a calculator, we find that 0.3633 ~ a ~ 0.4384, and since
y = a+ a2 is an increasing function on [0.3633, 0.4384], we have the following range of values for a+ a 2 :
0.495 ~ a+ a 2 ~ 0.631. We gave credit to any answer lying in that interval.
n(n + I)
T-5. Let S(A) = 3x , S(B) = 4x,and S(C) = 6x. Now S(A) + S(B) + S(C) = 2 . This must be
divisible by 3 + 4 + 6 = 13 which is prime so either n or n + 1 is a multiple of 13. Thus, n 2: 12. The
smal Iest value C1or n 1s

•
12, mak.mg 3x + 4x + 6x = -12.2- 13 = 78, so x =
6 . Thus, s(C) = 36,
S(B) = 24, and S(A) = 18. Since 1 + 2 + 3 + ... + 8 = 36, it would appear that the smallest possible
value for the largest element in C is 8, but the remaining elements {9, 10, 11, 12} cannot be partitioned into
two sets with sums of24 and 18. Try C = {2, 3, 4, 5, 6, 7, 9}, giving B = {I, 11, 12} and A = {8, IO},
or C = {I, 5, 6, 7, 8, 9}, B = {3, I 0, 11} and A = {2, 4, 12}. These are partitions with the requisite sums.
Thus, the answer is [2J .
180
T-6. Shown is half the octahedron inscribed in a sphere with center 0. Without loss A
of generality, let the radius of the sphere equal I. AO ~ OB = I so BCD£ is a
square with side /2.. The volume of the octahedron is twice the volume of
4
The volume of the sphere equals
3
43 }1)3 -_ 431t _ The ratio

[- . -47t : 4 =
. 1s In or1t: II.
3 3
T-7. Consider the 5 by 5 example in the problem. ln order, as we spiral inward, the numbers arc I, 9, 17, 21, and
25. To get from I to 9 we walk 4 units to the right and 4 units down. To get from 9 to 17 we walk 4 units left
and 4 units up and to the right. To get from 17 to 21 we walk 2 units to the right and 2 units down. To get
from 21 to 25 we walk 2 units left and 2 units up and to the right. ln other words, once we reach the upper
left comer of a square of side n by n, the next number is reached in 2( n - 2) steps and is, therefore, greater
by 2( n - 2) , and the following number is again reached in 2( n - 2) steps and is also greater by 2( n - 2) .
We obtain the following column whose sum gives the desired result:
9 1+2·4
17 1+2·4+2·4
21 1+2·4+2·4+2·2
25 I + 2 ·4 + 2 ·4 + 2 ·2 + 2 ·2
Adding vertically we obtain 5 · I + 4(2 · 4) + 3(2 · 4) + 2(2 · 2) + 2 · 2 = 73.
Applying this analysis to our problem we note that there are 27 numbers to be summed. They can be
arranged in the following column:
+ 2(26)
+ 2(26) + 2(26)
+ 2(26) + 2(26) + 2(24)
+ 2(26) + 2(26) + 2(24) + 2(24)
+ 2(26) + 2(26) + 2(24) + 2(24) + 2(22) etc.
Note that the number of 2(26)'s will be 26 + 25 = (2)(26) - I = 51, the number of 2(24)'s will be
24 + 23 = 2(24)- I = 47, etc. The sum of the 2(26)'s will be 2(26)(2 • 26 - I)) = (2 • 26)2 - 2 • 26, and a
similar pattern will hold for all sums. Thus, the sum equals:
181
27 + (2 · 26 - I )(2 · 26) + (2 · 24 - I )(2 · 24) + (2 · 22 - I )(2 · 22) + ... + (2 · 2 - I )(2 · 2) which is
2 2 ? 2
27 + (2-26) -2·26+(2-24) -2-24+(2-22t-2-22+ .... +(2-2) -2-2
27 + 16 (
13 · 14 · 27 J { 13 · 14
- - 2-
J= ~-~
112767 I.
6
Alternate solution: for n odd, the diagonal elements are given by the first line, their sum by the second:
l,4n-3,8n-15,12n-35, ... , 2(n-l)n-((n-1) 2-1) = n2 , .... , 14n-49,10n-25,6n-9,2n-I.
1+ (2n- 1) + ( 4n -3) +(6n-9) + (8n-15)+ (l0n-25) +(12n-35)+ (14n -49) + .... + 2(n-l)n-( (n -1) 2 - 1)
Note that constants below the middle of the square are perfect squares, but starting in the upper left at 4n - 3 and
going to the middle to 2( n -1 )n -( ( n - 1) 2 - 1 ) = n 2 , the terms are 1 less than a perfect square. Thus, the number
n-1
of terms 1 less than a perfect square is - 2 -. In creating a formula for the sum, we will subtract the perfect
n-1
squares and then add back 1 times - 2 - . Thus, the sum equals:
2 2 n-1 4n 3 - 3n 2 + 2n + 3
1+2n(l+2+3+ ... +n-1)-(1 + ... + (n-1) ) + -
2- 6
Alternate solution: Let T( k) be the sum of the diagonal elements in the spiral whose square's side is 2k -1 .
Here T(l)=l,T(2)=1+5+9=15andT(3)=1+9+17+21+25=73. The problem asks for T(l4). The
square of side 2k + 1 can be obtained from the square of side 2k - 1 by adding 8k to each element and then
wrapping a single layer containing 1, 2, .. , 8k around the square of side 2k - 1, putting 1 and 4k + 1 in the
comers of the diagonal. Thus, from T(2) = 1 + 9 + 5, we obtain
T(3) = 1+(1+16)+(9+16)+(5+16)+(4·2+1) = T(2)+3·16+1+(4·2+1).
In general, T(k + 1) = T(k) + (2k - 1)8k +I+ (4k + 1). Thus,

13 13
T(k+l)-T(k)=16k 2 -4k+2. So, T(14)= LT(k+l)-T(k) I+ L l 6k 2 - 4k + 2
k=O k=l
13 -
1+16 ( - · 14-· 27J
- -4 (13
-·- 14J +2(13) 12767.
6 2
182
T-8. Consider the set s2000 = [ I, 2, 3, ... , 2000 l whose clements are the powers of 2. If the common ratio is 2,
then the desired sets of exponents arc 1-2-3, 2-3-4, ... , 1998-1999-2000. First elements run from I to 1998
so there are I 998 sets of three elements forming an increasing geometric progression with a common
ratio of 2. The largest first element gives the number of sets and equals 2000 - 3 + I = 1998. If the
common ratio is 2 2, then the desired sets of exponents are 1-3-5, 2-4-6, ... , 1996-1998-2000. The largest
first element is 2000 - 5 + I = 1996. If the common ratio is 2 3, the sets arc 1-4-7, ... , l 994-1997-2000
and the largest first element is 2000 - 7 + I = 1994. The largest ratio is 2 999 and its triples are
1-1000-1999 and 2-1001-2000. The largest first clement is 2000 - 1999 + I = 2. Thus, the number of sets
ofthrecis 1998+ 1996+1994+ ... +6+4+2 = 1999,000I.
T-9. Since LiGAC is a 45--45-90 right triangle and AC = 20./2., then GC = the height E
F
of the box = 20./2.. Since both BH and the altitude from P to AC rise at a 45°
angle, PD = half the height or 10/i.. The volume of the box is
20(20)20./2. = sooo/i.. The volume of pyramid PA CD is ( 1l¾· 20 · 20) o./2. B
2ooo'2. 4oooli.
- - - . The volume of FBAC = - - - since its height is twice that of PACD. Using CPHG as the
3 3
base, the he;ght of FCPHG ;, FG ~ 20 ru,d ;ts volume;, [ {- }0(20 - 20/i_ -(½-20 · 10/i_]) ~ 2WJ/z.
Using PHEA as the base and FE = 20 as the height, the volume of FPHEA similarly equals 20oo/2.. Thus,
the volume of FPCA = &ooo./2. 200~ - 4003°/i. - 2( 20ooli.) = l20oo'2.I.
Alternate solution: in the above diagram, let T be the H(O. 20. 20 ✓2 J
midpoint of AC, giving T(IO, IO, 0). Since vectors

--+-----( / ~G(20,20,20✓2)
~ =(-20,20,2o/i.) and TP =(-10,10,z) are parallel, F(20, 0, 20✓2 )
f-
V" j n
__ ---/ PW. 20. z)
o./2.. Using the cross product or the dot product, we

:: = 1 A -
obtain ( ../i, -Ji ,2) for the normal vector to plane PAC.
8(20, O. OJ C(20. 20. 0)
183
The equation of plane PAC is therefore x./i. - JJI + 2:: = 0 . The distance from F to the plane equals
120./i. -o-/2 +2. 20Eil
J = 30. Since PA = PC = IO-J6 and AC = 20./2_, the height
2+2+4
of MAC = 20 and its area is 20Q-J2, giving a volume of 200Q-J2.
T-10. Think of the top node as the root and the other nodes as its descendants, each of which has exactly one parent
and O or 2 children. Let \ be the number of subtrees rooted at a kth generation node. Then h4 = I for the
8 fourth generation nodes. The third generation nodes have two children each with h4 = l subtrees rooted
there. Accounting for the empty subtree, we have b3 = (I+ l ) 2 = 4 possibilities. Similarly, the second
2
generation nodes have b2 = ( 4 + I ) 2 = 25 subtrees and the root has h1 =(25+ I) = 676 subtrees. The total
number of subtrees equals I · 676 + 2 · 25 + 4 · 4 + 8 · I = I750 I.
Alternate solution: Let .f(n) be the number of connected subtrees of a complete binary tree on 2 n- 1 nodes
and let g( n) be the number of connected subtrees of a complete binary tree on 2 n- l nodes which include the
root (top) node. Then/( I)= g( 1) = I since the only connected subtree of a single node is that node itself.
2
We will find recursive relations for .f and g. First, note that g( n + 1) = ( g( n) + 1) because any subtree
containing the root node must have a connected subtree on either side of the root which is either empty (there
is 1 of these) or contains the top node ( there are g(n) of these). Next note that_/( n+ l) = g(n+ I) + ~f(n) since
any subtree included in the count off{n+ I) must either have a tope node or be one of the trees counted by g of
which there are g(n+ I). Or the binary tree does not have a top node and therefore must consist of one of the
right or left subtrees empty (otherwise the tree would not be connected) and the other subtree (left or right
respectively) containing all the vertices of the graph. Since the graph can therefore go in either the left or
right (smaller) subtree, there are 2/(n) such graphs. So, g(2) = 4, g(3) = 25, g(4) = 676 from the first
relationand/(2) = g(2)+2/{l) = 4+2 = 6,/(3) = g(3)+2/(2)=25+2(6)=37,and
/(4) = g(4) + 2/(3) = 676 + 2(37) = 750.
184
Alternate Solution: number the vertices along the left side top to bottom 3, 2, I, and let h(n) be the number of
subtrees whose highest vertex is n. Note that there is another node whose subtree structures below it will be
the same as vertex 2, and that three other subtree structures will be the same as that of vertex I. When
n > 0, the subtree may consist of: I) only vertex n, or 2) of the left (or right) edge leading down from it
together with a subtree structure starting at vertex n - I, or 3) the vertex equivalent to it, or 4) both the left
and right edge and any combination of substructures starting at n - I or its equivalent. Thus,
2
/~n)=l+2h(n-l)+(h(n-l)) whenn > 0. Notcthath(0) = I and the formula gives
2 2
h( I) = I + 2 h( 0) + ( h(0)) = I + 2 + I = 4 , h( 2) = I + h(I) + ( h( I)) = I + 8 + 16 = 25, and
h(3) = l + 2h(2)+ (h(2))2 = 1 + 50 + 625 = 676. The total number of subtrees equals
h(3)+2h(2)+4h(1)+8h(0) = 750.
185
ARML Power Question -2001: Power to the Triamde
Let T = T(a, h, c) be LiABC whose sides BC = a, CA h, and AB = c satisfy both 0 < a S b S c

and a+ h > c.
Define r 2 = T( a2 , b 2 , c2) to be the square of triangle T, provided that such a triangle exists.
Define fl = ~ ia, /h, /;:) to be the square root of triangle T, provided that such a triangle exists.
Use K (7) for the area of triangle T and P(T) for the perimeter of triangle T.
Results from a preceding problem may be used in a succeeding problem but not vice versa.
You may use Fermat's Last Theorem as a reason in any proof.
l. a) Show that the square of an equilateral triangle is equilateral.
b) Prove that the square of a right triangle does not exist.
2. Compute all x for which the area of T = T( 1, 1, x) is the same as the area of T 2 .
3. a) Prove that r2 exists if and only if Tis acute.
b) Prove that Jr exists for all T.

4. a) Prove that all the angles of fl are acute.
I
b) Prove that cos C :S - with equality if and only if Tis equilateral.
2
5. Prove that the largest angle of T is at least as close to 60° as the largest angle of T 2 .
Let X
11
= T(n - 1, n,n + I) for n a real number.
2
6. a) Determine with proof, all n for which (x 11 ) , i.e., the square of X 11 , is a triangle.
b) Note that X 11 = T(I0,11, 12) can be squared, producing triangle T(l00, 121, 144) and that this can be
squared again, producing triangle T(l0000, 14641, 20736). The latter, however, cannot be squared. Now for
an integer p, there exists a real number k, p S k :'.S p + I, such that for all n ::: k, X can be squared at least
n
twice. Find p and justify your answer.
186
ARML Pmver Question - 2001: Power to the Trianrde
7. a) Provide, with reasons, a triangle which can be squared at least 2001 times and at most 2001 times.
b) Let N(n bethenumberoftimesatriangleTcanbesquared. For example, N(x 11 ) = 2. Determine,
with proof, all triangles T for which N ( T) = 00 • Specify side relationships and angle restrictions.
8. a) Prove that if n = .[;, + ib + Jc for integers n, a, b, and c, then a, b, and c must be perfect squares.
b) Using (a) or otherwise, prove that if T = T(a, b, c) is a right triangle with integer sides, then the perimeter
of .Ji- cannot be an integer.
9. Prove that there exists a right triangle T with the same perimeter as ./i-.
10. Define the reciprocal of T(a, b, c) to be T- I = r(.!.,c ..!_)

b a)
. I where .!_c < 1 <
b a
When T- I is itself a triangle, define T to be invertible. Show that if Tis invertible,
then a > (3-/s:

--
2- · c.
187
l. a) IfT= T(a,a,a),then T 2 = T(a 2 ,a 2 ,a 2 ) and T 2 isclearlycquilateral.
b) If T = T( a, b, c) is a right triangle, then a2 + b 2 = c 2 and so T 2 = T( a2, 1/, c 2 ) fails the triangle

inequality.
2. Using Heron's formula, K (I) = 1+ ~ andK(T2 )= 4x2 'i4-x

2xJ( I - 2xJ( 2xJ(x2J = 4x-i4-x"" r-:;...
Setting these equal we obtain J4 - x 2 = xJ4 - x 4 ➔ x 6 - 5x 2 + 4 = 0. Let u = x 2 giving
u3 - 511 + 4 = 0 . By inspection, by using the rational root theorem and synthetic division, or by realizing
that the equilateral triangle T( l, l, l) solves the problem, we know that u ~ l is one solution. This gives
(u-lXu 2 +u-4) =0,makingu = 1 or Ju2 - 1 . Thus,x = I or Jm - 2

1 = 1
2i 2/u
17 -2.
3. a) Given T, then T 2 is a triangle iffthe sides satisfy the triangle inequality, namely that a2 + b 2 > c 2 . By
the Law of Cosines, c 2 = a2 + b2 - 2ab cos C giving c 2 + 2abcos C = a2 + h2 . Thus, a2 + h 2 > c2
iff2abcos C > 0 which is true whenever O < mLC < 90°. Since neither LA nor LB is greater than LC,
then T is acute-angled.
2
b) Given T, then fl is a triangle iff fa+ -ib > /;. Since ( ia + /h) =
a+ 2 ~ + b > a + b > c, taking the square root yields fa + -ib > /;.
2 2 2
4. a) Let 8 be the angle opposite the longest side .[c. Then (/;) = (ia) + (/i) - 2/;i/b cos 0 ➔
a+b-c r::.
cos 0 = , . . Since a+ b > c, then cos 0 > 0, making wT an acute-angled triangle.
2wab
b) Since mLAand mLBarebothsmallerthan mLC,wchave 180° mLA + mL B + mL C '.S 3 mL C.
Thus, mL C 2: 60° with equality when mLA = mL B = mL C.
188
5. Given mL C 2: mL B 2'. mLA, let cos C X stand for the cosine of LC in triangle X. Then we wish to show
that cos C Jr :=:: cos CT" To avoid ugly expressions, let T = T( a2, h 2 , c 2 ) so that Ji = T( a, h,c). By the
Law of Cosines:
a2 + h2 - c2 0 4 + h4 _ c4
2ab 2}b 2
a 3 b+ ab 3 - abc 2 - a 4 - h4 + c 4
2a 2 h2
j
( c 2 + b 2 - ab)( c 2 - h 2 + a 3( b - a) .
- - 2 2 >
- 0 smce c -> b -> a.
2a h
Note: From (4b) we know that the largest angle C 2: 60°. Thus, #5 proves that the largest angle of Ji is
basically closer to 60° than the largest angle of T. lf one also proved that the smallest angle of Ji is at least
as close to 60° as the smallest angle of T, then one would have shown that Ji is closer to being equilateral
than T.
6. a) The square of X n is ( (n- 1/,n 2 ,(n + 1) 2 ) and for a triangle to exist, (n - 1) 2 + n 2 > (n + 1/ ➔
2
n - 4n > 0 ➔ n > 4. But we should check to see if X is a triangle for all n > 4. Here we find
n
that n - I + n > n + I ➔ n > 2, son > 4.
b) The square of (x J2 is ( (n- 1) 4 ,n 4 ,(n + 1) 4 ). We have (n - 1) 4 + n 4 > (n + 1) 4 ➔
4 3 3 2 3 2 .
n -Sn -Sn>0 ➔ n -Sn -8>0.LetAXn)=n -Sn -8.SmceR(8)=-8andR(9)=73
and R is continuous, there is a zero between 8 and 9. For completeness we must verify that R(n) isn't negative
3 2 3 2 2 2
for some larger value of n. For n > 9, AXn) = n - 8n - 8 > n - Sn - n = n (n - 9) and the last
is greater than O for all n > 9. Thus, p = 8.
189
7. a) T = T(3, 4, 5) is a triangle and since it is a right triangle, we know by ( I b) that r 2 doesn't exist. From
(3b) we know that if a triangle exists, then its square root exists. So, starting with T = T(3, 4, 5). take
successive square roots 2001 times. This gives T = , 3

jl 1122001
• 4
112 2001 l/?2001)
,5 ~ , a triangle which can
be squared exactly 200 I times.
b) It seems reasonable that if N ( T) = oo. then c = h, for if c > b 2: a, the continual squaring of c would
eventually make c much larger than the sum of a and b, and the squaring process would eventually end. So
let T= T(a,b,b) andcheckthetrianglcinequality. Clearly, a2k + b 2k > b 2 k and

k k k
b2 + b 2 > a 2 forallkiffb2:a.Thus,N(7)=ooiffTisanisoscelestrianglewith
base angles 2: 60°.
8. a) Suppose .[; + {q = m for integers p, q, and m. Then p = ( m - J-;;)2 = m2 - 2m,Jq + q and q has to
be a perfect square in order for both sides to be integers. Similarly, p must be a perfect square. Call this
result #1. Now, squaring .[; + ib = n - / ; gives 2/;; + 2rlc = n2 + c - a - b. The right side is an
integer and the left side equals ..Ji;ii + J4n 2 c and by# I, both ..Ji;ii and J4n 2c must be integers. Thus,
2n/;; is an integer, making c a perfect square. Returning to .[; + ib = n - /;, the right side is an integer,
so by # 1, both a and b must be perfect squares.
b) Suppose the perimeter of .Ji= r{_.[;, /,,, /;) is an integer. Then by (8a) a= x 2 , h = y2, and c = z 2
·
tior mtegers x. y, an d z. s·mce a, b , an d c sat1s· t·y a 2 + b2 = c 2 , th en ( x 2)2 + ( y 2)2 = ( z 2)2 g1vmg
..
x4 + y 4 = z 4 , a false result. So the perimeter of Ji can't be integral.
9. If the sides of Tare all greater than 1, then each side is greater than its square root so P( r) > P(Jr) . If the
sides of Tare all less than 1, then each side is less than its square root so P( r) < P(/i). To obtain equality
of perimeters, the smallest side should be less than 1 and the largest greater than 1. W c ought to obtain the
simplest solution if the middle side equals l. So, let a = x < 1, b = 1, and c = Q.
190
Solutions to the ARML Power Question - 200 l
Then T= { D) /i = { ix, 1, i1
x, I, and + x 2 ). Let f(x) = P(r)- P(ir) =
(x - /;) + ( D - fiJ). This is clearly a continuous function. Since.f(.7) = -.0208 and
.f{.8) = .05455, the Intermediate Value Theorem assures us that at some value ofx between .7 and .8, the two
perimeters are equal and we have shown the existence of a right triangle whose perimeter equals that of its
square root. That value of xis approximately .72864. Note: b doesn't need to equal I, but it seems
reasonable that it can't grow too large. What is the maximum value that b can take on or what is its least
upper bound?
Alternate solution: Consider right triangle ~ = (1, I, Ji) with Ji: = ( 1, 1, ~ ) . Then P( T1) = 2 + Ji
and P( /ri) = 2 + ~ and clearly P( ~) - r(.Jr:) > 0. Divide the terms of T1by 4, producing right
triangle T4 = (.!.4' .!.4' Ji]

4
with
,, 4
.!. fl.·]-
'T = [.!_2 '2' 2
We have P(T)
4
= 2 +4Ji and P(,,fr)=
4
2 + fi,.
2
Note that P ( )< P (1tr)

~ r4 , since otherwise -2 +-Ji
4 - 2
2 + fi_
-- 2- ➔ 2
,:
+ y2 2 4 + 2-v2
4r: ➔
Ji, - 2 2 ~, clearly a contradiction. Thus, P( ~ )- P(~) < 0. Now consider the function
f ( x) 0 r( Tx) - ~ F,) whcce xis thonumbcc dividing the mig ins I tmns of 7j . !~ - { ; ,.:; , ~ lmd
tr - [ix,
1 ,x
1 I f2_
ix,/; l . Thus,.f{x) -
2 + Ji
x
2
ixfl -
+ (2 + Ji) - x(2 + fi-)/; . .
We have Just seen
that.f{l) > 0 and.1{4) < 0. Since/is continuous for x > 0, then by the Intermediate Value Theorem, there
must be a value ofx such that.f{x) = 0. Here, x "" 1.1460826 and the triangle is approximately
T = T(.872537, .872537, 1.2339).
I 0. For T(a, h, c) we have 0 < a 'S h 'S c and a+ b > c. If T-l is a triangle, then we have
1 1 1 I 1
0 < - ::; ::; - and - +- > - From a+ b > c, we o b tam - + -b
. a #
> 1 ( 1). From
C b a b c a C C
I I I . C c a h
-+- > - , we obtam - + 1 > (#2). Let x =- and y = - , resulting in
b C a b a C C
I X
0 < x :Sy :S 1 (#3). Then(#l)gives x+ y > l and(#2) gives - + l > ➔ y <
y X 1- X
191
Shown at the right arc the graphs of these inequalities:

A(O. l) ~_8_(_11_2,_l_)+/m~~ C(l. 1)
O<x:Sy:S
x+ y > 1
E
X
y <
1- X
The shaded region £BCD represents the solution set. D(O, I)
Only the horizontal and vertical boundaries are included:
To find the coordinates of point E solve the system x + y = 1 and y = - x- , ob tammg

.. x 2 - 3x + 1 =0
1- X
➔ x
3±/s
= -- -.
. 3+./s . ,.
ReJect - - - smcc it exceeds I. Then 1or x > - -
3-/s . . a
= -C ,
2 2 . 2- , Tis mvertible. Smee x
w,have; > 3-/5 _, a > [ 3-/5]c.

C 1+
Note that a > - 2 ifwe let 't stand for the golden ratio, - - - .
Is
"C 2
This Power Question is full of great research possibilities. Here are two additional problems:
I. Let T be a right triangle. With proof, compute the largest value of cos C for Ji.
2. Let T be a scalene triangle in which exactly one side has a length of I. If T and T 2 have exactly two sides of
the same length, then compute the least upper bound for the area of T. Show your work and justify your
answer.
192
4 a 5 . h
1-1. If-- < < - - . compute the number of mtegral values that - can take on.
2001 a+h 2001 a
1-2. A lattice rectangle has vertices at lattice points and sides parallel to the axes. It contains all lattice points in
its interior and on its sides. If the shorter side of lattice rectangle R is doubled, the new rectangle contains
304 more lattice points than R. Compute the minimum possible area of R.
1-3. Compute the largest real value for b such that the solutions to the following equation are integers:
2
( log 210 x 2h ) = log 210 x 4
I-4. The vertices of a regular dodecagon are given by (x; ,Y;) for i = I, 2, ... , 12 in clockwise order.
12
If(x 1,y1) = (l5,9)and (x7'y7 ) = (15,5),compute L(x; - yJ
i == 1
5
]-5. Form> 0, f( x) = ,n,;. The acute angle 0 fom1ed by the graphs of fand f- 1 is such that tan 8
12
Compute the sum of the slopes off and f- I .
1-6. In unit square ABCD = S0 , two pairs of parallel lines are drawn, each A B
making an angle of 8 with a side of S0 . These lines produce an inner
square, EFGH = S 1. The procedure is repeated using SI and angle 8,
.
producing an inner square S,,. The process is continued. If 8 = l 5°,
compute the sum of the areas of squares s0, S 1, S 2 , S3 , ...

D C
1-7. Let [ x] represent the greatest integer less than or equal to x. Compute the number of first quadrant ordered
pair (x, y) solutions to the following system:
X + [ y] 5.3
y + [ X] 5.7
For 0 < x < l, let f(x) = (l + xX l + x 4 )(1+x 16 )(1 + x6 4x 1 + x 256 ) ...

tr½')l·
1-8.
Compute r
193
ANSWERS ARML INDIVIDUAL ROUND-2001
I. 100
2. 288
3. Jw
4. 96
13 -
5. - or2.16
6
6. 2
7. 6
9
8.
64
194
Solutions to the ARML Individual Problems - 2001
2001 a+b 2001 h b

1-1.
4
>
a
> --5
➔ 500.25 > I +
ll
> 400.2 ➔ 499.25 >
a
> 399.2.
b
Thus, 400 S S 499, giving luili] integral values for !!.. .
a a
1-2. If the sides arc n and m with n > m, then the number of points on each side is n + l and m + I. If the side of
length m is doubled, the number of points on each side is n + I and 2m + I. Thus,
(2m+l)(n+l)-(m+l)(11+l) = 304 ➔ mn+m = m(n+I) = 2 4 -19. lfm = landn+l = 304,there

is no rectangle, so consider these 4 pairs of (m, n + I): (2, 152), (4, 76 ), (8, 38), ( 16, 19). The areas arc
respectively 2 - 151 = 302, 4 · 75 = 300, 8 · 3 7 = 296, and 16 · 18 - 288. Answer: [288 [.
There are several values for b that make x an integer. If b = ± I, then x = 2

10
. If b ~
M
± - 3-, then
x = 2 9 . But the largest value of b making the second solution an integer is b = ] .Jlo I which gives x = 2 1.
1--4. Given opposite vertices (15, 9) and (15, 5), the center of the ( 15, 9)
dodecagon is ( 15, 7). By congruent triangles, note that for •·;.··- (15+x, 7+_v)
I
any vertex ( 15 + x, 7 + y) there is a corresponding opposite I _V
vertex ( 15 - x, 7 - y). The sum of all 12 x-coordinates equals I
-y:
I
15 - 12 and the sum ofall 12y-coordinates = 7 · 12. The

I
I
difference is (15 - 7)12 = ~ - (15-X, 7--y) I -_\'
(15, 5)
I
m-- m2 -1 5
I-5. If f(x) = mx, then f- 1(x) = x
m
Using tan8 = gives tan 8 = ( 1 j = ~ = 12
I +m -
m
3
Thus, 6m 2 - 5m - 6 = 0 ➔ (3m + 2)(2m - 3) = 0 ➔ m=- -2 , making the sum equal ~3
·. - .
2' m 3 I~
195
1-6. The first 2 pairs of parallel lines divide the unit square S0 into S 1 and A B
the four congruent triangles DFC, CEB, BHA, and AGD. Consider
WFC. Since DC = 1, DF = cos0 and FC = sin 0, its area is
I
2(cos0 )(sin0), and the sum of the areas ofall 4 triangles is
2sin0cos0 =sin20. Since0 = 15°,thesumoftheareasis
1 I I
sin 30° ~ 2. Thus, the area of s 1 = I - 2 2. Similarly, the
D C
1
area of s2 4 , the area of s3 -1 and the desire d sum 1s
. I + -1 + -1 + = [I].
8 2 4
I-7. Let x = n + 03 and y = m + 0.7 for non-negative integers m and n. Then m+ n =5 and there are 6 pairs
of integers from (0, 5) to (5, 0) for which this is true. The solutions are (0.3, 5.7), (1.3, 4.7), (2.3, 3.7),
(3.3, 2.7), (4.3, 1.7), and (5.3, 0.7). Ans: ~.
4 16 64 2 2 8 32 128
I-8. If f ( x) = ( I + x)( 1+ x )(I+ x )(I + x ) .. ·, then f ( x ) = ( I + x )(I+ x )(I+ x )(1 + x ) · •· and
. 2 2 4 8 16 2 3 4
j(x)f(x )=(l+x)(l+x )(l+x )(l+x )(l+x ) ... = l+x+x +x +x + Thus,
1-x
J( x 2) = f 1
(x)(l -x)
. Letting x = ~, we obtain
8
fl J~ (l l
9
64
·
1· l
8
1-l
8
8
-().Thus,
Sf l
8
9 -I[ S/( 8
64 =f l
318 ) . Answer:
W
LMJ.
More rigorously, let g(x) = f(x )f ( x 2) = ( I + x)( I+ x 2 )(I+ x 4 )(1 + x 8 ) .. •. We claim that g(x) = - 1-
1- x
because g(x)(l-x)=(l-x) ( (l+x)(l+x 2 )(l+x 4 )(l+x)

8...
) = (1-x 2 )(l+x 2 )(l+x 4 )(I+x)
8 ...
4 4 8 n n .
(1-x )(1 + x )(I + x ) .. · = ( 1-x )g(x ). For O < x < I, each expression approaches 1 as n gets
m
· fimite
· 8
·l
· ly large, so g(x)( I - x) = l . Therefore, / -3 }·[-9 = - -
64
J
1 - = -8 , so /( -9 ) = - -
1-(3/8) 5 64
8 -.
5/(3/8)
Therefore, f -1(-5/(3-8/ -8) ) = -649.
196
ARML Re/av #1 - 2001
RI-I. The area of a regular octagon is 2 + 2./J.. Compute the length of a side.
Rl-2. Let T = TNYWR. Compute the number of real roots to x 3 + (T + l)x 2 + (T + l)x + I == 0.
ARML Relay #2 - 200 I
Compute the smallest positive value of x in degrees such that sin x == cos( x 2 ) .
R2-2. Let T = TNYWR. ABCD is a square with

PD = PC = DC. The area of MBC = T. Compute
the sum of the areas of MAB and MDC.
R2-3. Let T = TNYWR. Compute the number of lattice points which lie strictly inside the triangle formed by
x = 0, y = 0, and x + y == T.
197
ANSWERS ARML RELAY RA~C-'E=S==-'-""~"='?0-0~1==============~=
Relay #1:
Rl-1.
Rl-2. 1
Rl-3. 1
Relay #2:
R2-l. 9
R2-2. 18
R2-3. 136
198
Solutions to the ARML Relay #I - 2001
R 1-1. Let the side of the octagon be x. The area of the octagon x/✓2 x/✓2
A~---- X
equals the area of square ABCD minus the 4 corner triangles: ' X ~B
''
N
"?
jc
>-:
2+ 2-J2 . Thus, +
:-,
2x 2 + 2x 2 Ii. = 2 + 2/i. ➔ X = [I] . /]~---- -
Rl-2. By inspection, x ~ - l is a solution, giving (x + 1)(.cx-2 + Tx +I) = 0. The discriminant is T 2 -4.
If T < -2 or T > 2, the quadratic has 2 distinct solutions giving 3 real roots. If T ~ 2. the quadratic
has a double solution of -1 ➔ the cubic's three solutions are -1, giving I real root. If T = -2. the
solution to the quadratic is I, giving 2 real roots, I and -1. If-2 < T < 2, the quadratic has no real
roots, making -1 the only real solution of the cubic. Since T = I, the cubic has just [I] real root.
3
Rl-3. We have log2 _ log3 + log2 _ log5 _ logT __ l + - logT
- . Since T = 1, log T = 0 and the
log 3 log 2 3 log3 2 log2 2 log5 4 1og 3
answer is [I].
Solutions to the A RML Relay #2 - 200 I
R2- l. sin X = sin( 90° - x 2 ) ➔ X = 90° - x2 ➔ x 2 + X - 90° =0 ➔ (x + l 0°X X - 9°) =0 ➔ X = [23.
R2-2. Since PE= PF= .!_AB and AD = BC, the areas of MBC and MAD
2
are both equal to T = 9. Since the sum of their areas equals
X
2Y
A2AB JBC
2[ 1 2(
= 1 AB)(BC) equals half the area of the square equals
X
2T = I 8, the sum of the areas of the other two triangles is half the
D C
area of the square which is~. This solution docs not require that
MDC be equilateral, only isosceles.
R2-3. By inspection the sum is (T-2)+(T-3)+(T-4)+ ... +l = T o
(T-2)(T-l) . 16-17
. Smee T = 18, the answer is - - - = I 136 I.
2 2
. . • • . T
199
I. A square has side of lengths and a regular hexagon has side of length h. The hexagon's area is three times
s4
that of the square. Compute 4 .
h
2. Let T = TNYWR. Compute log8 ( 2 24 f

3. Let T = TNYWR. If (x 2 - Nx + 7 )(2.i + Tx + 9)( 5x 3 + Tx + 13) !Ox 7 + P(x) where Pis a fifth
. . N
degree polynomial m x, compute - .
T
4. Let T = TNYWR. Shown are four semicircular arcs whose centers are
collinear. The radius of the largest semicircular arc is T. Let P be the
p
perimeter of the entire figure. Compute - .
n;
5. Let T = TNYWR. In MBC, AB = AC = 5, and BC = 2T. Compute sinLBAC.
6. Let T = TNYWR and let K = 25T. A line passes through A(0, K) with slope -K. Compute the area of the
triangle formed by the origin and the x- and y-intercepts of the line.
7. Let T = TN YWR and let K = T 2 - 4 . ln the arithmetic sequence {aJ, aK = 200 l. If the common
differenced is an integer, compute the largest value of d such that '½K < 2500.
8. You will receive two numbers. Let a be the smaller and h the larger. Define a sequence of numbers as
follows: s 1 = a,s 2 =b,andforn > 2, sn = asn-l ifnisodd orbsn-l ifniscven. Computcthe
smallest value of k such that sk is divisible by 102001 .
200
15. A three-digit number ABC is called primal if A, B, and C arc primes and are not all equal. Compute the
number of positive three-digit primal numbers.
14. Let T = TNYWR and let K = f. In the arithmetic sequence { ai}, aK = 200 I. If the common difference dis
an integer, find the minimum value of d so that aK + 9 > 10,000.
13. Let T = TNYW R and let K be the sum of the digits of T. There are K people in a room, 4 of whom arc
women. If two people in the room are chosen at random, compute the probability that at least one of them is
a woman.
12. Let T = TNYWR. Let T = ba where a and b are relatively prime. 90

C
Set K = a+ b. Base runner R leaves base B running at K feet/second

and after x seconds the area of MBR is K% of the area of square ABCD.
:1 R
If AB = 90 feet, compute x. A B
90
II. Let T = TNYWR. Compute base b if logh (75T) = 2 + log h 3 + log h 5.
I 0. Let T = TNYWR and let K = T + 2. In the diagram, AD =K 2

and DC = 96. Compute AB.
9. Let T = TNYWR. Let f(x) + f(x +I)+ f(x + 2) + f(x + 3) =T where/is a function defined on the set
of integers. Iff (222) = 60, f(l 776) = 28. and /( 1999) = -83, compute /(200 I).
8. You will receive two numbers. Let a be the smaller and b the larger. Define a sequence of numbers as
follows: s 1 = a, s2 = h, and for n > 2, sn = asn- I if n is odd orbs n- I if n is even. Compute the
smallest value of k such that sk is divisible by I 0 2001 .
201
3
I. -
4
2. 6
1
3. -
2
4.
4./6
5.
25
6. 2./6
7. 24
15. 60
14. 889
3
13.
10
9
12. 1.8 or -
5
11. 3
I 0. 55
9. 50
8. 2002
202
I.
. 3
8Tlog 8 8 = 8T. SmceT = 4
,8T =&].
3. Thefirstpartoftheproductequals 10x 7 +~T-2N)x 6 + .... Sincethecoefficientofthe x 6 tem1is0,
..
T - 2N = 0 g1vmg N T = !Uh
liJ· T e result from #2 is unnecessary.
4. Let the diameters of the three smaller arcs be a, b, and c. Then P = ½(7t • 2T + 1t • a + 1t • h + 1t • c)
7t 7t 1 p Ill
2 (2T + a + h + c) -(2T + 2T) = 2Tn. SinceT = - P = 1tand - = Ll.J.
2 2. n
5. Area = area ➔
1 . 0
-(5)(5)sm
2
J(5+T)(5-T)T·T ➔ sin0= 2 TJ25-T 2 .
25
Since T = l, sin 0 = 14: I-
6. A line with slope -K passing through A(0, K) will also pass through B( I, 0).
The area of the triangle = -21 (l)(K) = K. Since K =

2
4-'6, K2 = \2/61.
7. T = 2.J6 so K = 24 -4 = 20. If aK = 2001, then a2K = 2001 + Kd and if a2K < 2500,
499 . 19 r-;:,-;,i
then Kd < 499 ➔ d < - . Smee K = 20, d < 24 + - ➔ d = ~.
K 20
203
15. The digits could be 2, 3, 5, or 7. There would be 4 · 4 · 4 = 64 primal numbers if A, B. and C could all be equal,
but we reject 222, 333, 555, and 777 obtaining [6oJ primal numbers.
14. Since aK+ 9 = 2001 + 9d > 10,000, then 9d > 7999, making d > 888.8 ➔ d= I889 I. Tis unnecessary.
13. T = 889, so K = 8 + 8 + 9 = 25. The probability that both are men =

K-4
K · (K-1)-4
K _ 1 . The probability that
_ . (K - 4)(K - 5) 8K -20 l½3

at least one 1s a woman 1s 1- - - - - - = - - - . Since K = 25, the answer is - .
K(K-1) K(K-1) 10
12. Tis unnecessary. RB = Kx, so the area of AABR = ½• 90 · Kx = 1: 0 · 90 2 . K's cancel sox =[!]].
2 2 2 2 I'll
II. T = 1.8. logb 75T= logbb +logb15= logb 15b ➔ l5b = 75T ➔ b =ST= 9 ➔ b = Ll.J.
AB and Kare integers, K = 2, 5, 10, or 23, making AB = 20, 55, 140, or 575. Since K = 5, AB ~ ~ .
9. Since /(x)+ /(x + l) + f(x+ 2) + f(x +3) = f(x+ l)+ f(x +2)+ f(x +3)+ f(x + 4), then /(x) = f (x+ 4) ➔
f(x+ 1) = f(x +5), f(x+ 2) = f(x +6), f(x+ 3) = f (x+ 7), etc. Letting /(l 776) = f (x), then
/(222)=/(x+2), /(1999) =/(x+3) and /(2001)=/(x+l). Thus, 28+/(2001)+60-83=T.
Since T = 55, /(2001) = Iso 1.
. 2 22 23 ft
8. Leta= 24= andb = 50. Smee s 3 = ab, s4 = ab , \=ab, s 6 =ab ,we have sk = s 2 t+I =ab or
t t+ I . 3 2 4t I 2t
s k = s 2 t+ 2 = a b fort 2: l. Smee a = 2 · 3 and b = 2 · 5 , s 2 t+ 1 = 2 -3 - 5 and
s 2 t+ 2 = 2 41+I · 3 1 • s2t+ 2 . We h ave p Ienty of2' s so 1.f 10 200I 1s

. to ct·1v1·cte sk' th en 5200I must ct·1v1'de e1t
. her
521 or 521 +2 . If k is odd, then 2t ~ 2001 ➔ t 2: 1001 ➔ k = 2003. If k is even, then 2t + 2 ~ 2001 ➔
t 2: I 000, so k = I 2002 I works.
204
ARML Tiebreakers-2001
tanB 2
I. In MBC, AB = AC = I, BC = k, and - - = 200 I. Compute k .
tan A
2. A circle intersects the graph of y = ax 2 at x = I or 4 and it is tangent to y = ax 2 at a third point.

Compute the x-value of the third point.
I
3. 2, increasing y by 50% decreases x by a factor of k. Compute k.
205
tanB tan 0 tan 8

I. Let mLB = 8, then mLA 180° - 28 and tan A = -tan 20. Thus, - -
tan A - tan 20 2 tan 8
tan2 0
tan 2 0-l J4-k 2 J4-k 2

2 = 2001 ➔ tan2 8 = 4003. Since AD = 2 , then tan8 = k and
1 = 4003 ➔ k 2 = ~
2
tan--, 8 ~ -
4-k
- = -4 - 1 Thus, -
4
7 - --
k2 k2 . r I 001 .
B e__~___e~ c
k/2 D
2
Alternate solution: Use tan A + tan B + tan C = (tan A)(tan B)(tan C) ➔ tan A+ 2tan B = (tan A)( tan B) ➔
2tanB 2 tanB tan 2 B-l

I+ -- = tan B ➔ -- -
2 and continue as above.
tan A tan A
2. Let the circle's equation be (x - h/ + (y - k)1 = r 2 . Expanding and substituting ax 2 for y gives
a2 x 4 + 0x 3 + (I - 2ak) x 2 - 2 hx + h 2 + k 2 - r 2 = 0. This equation's solutions are I, 4, t, and t since the
circle is tangent to the parabola. The coefficient of x 3 equals - ( 1 + 4 + 2t) = 0, so t = 1-; 1·
1
3. ➔ x =- Increasing y by 50%
y3·
206
2002
'Team 'Round-------- -------------------------------------------------------------------------------------- 209

Power Qyestion ......................................................................................... 214
'lndividua( 'Round----------------------------- ----------------------------------------------------- - 221
'Re(ay 'Rounc(............................................................................................ 227
Suyer 'Re(ay................................................................................................ 230
'Tie6reakers 235
-207-
THE 2ih ANNUAL MEET
In 2002, ARML saw 23 teams take part in Division A and 67 teams in Division B. In addition,
Taiwan sent 5 teams and the Philippines sent I. In Division A, Thomas Jefferson and
Massachusetts A matched up on the team and relay rounds, but TJ pulled ahead on the individual
and Power Question for a solid victory. In Division B, New York City S trailed Southern
California by 8 points after the team round but pulled ahead on the individual and relays for an 8
point victory. The new site at San Jose State proved to be an excellent venue for the second year
ma row.
Curt Boddie of New York City received the Samuel Greitzer Distinguished Coach Award for his
years as a coach, problem poser and solver, and for his love of fine literature.
Carrie Kiser Wacker received the Alfred Kalfus Founder's Award for playing a crucial role in
establishing the Iowa site. Until recently Carrie was the conference coordinator at Iowa. In a
much appreciated effort, she was able to get the university to provide air conditioning.
Austin Shapiro of Northern California received the Zachary Sobol Award for his fine work with
his California team.
208
T-1 . Compute the smallest positive base IO integer the product of whose digits exceeds I 000.
T-2. 2000 cards of width 2" arc placed in a row as suggested in the diagram
at the right. The second card overlaps the first card by I /I 000 of an
inch and each successive card overlaps the previous card by one-
thousandth of an inch more. In inches, compute the length of the row.
T-3. If 240 equals abc, the product of positive integers a, band c, compute the number of distinct ordered triples
(a, b, c) such that a is a multiple of 2, b is a multiple of 3 and c is a multiple of 5.
1---------> ............
T-4. AD0C is a secant of a circle with center 0. A lies outside the circle and D and C lie on the circle. ABE is
also a secant of circle O and B and E are distinct points lying on circle 0. If AB = BC, compute, in degrees,
the largest possible integer value of the measure of L.CAE.
T-5. Suppose that palindromes with n digits are formed using only the digits 1 and 2 and that each palindrome
contains at least one of each digit. Compute the least value of n such that the number of palindromes formed
-
exceeds 2002.
T- 6. Let P be a point on side ED of regular hexagon ABCDEF such that EP: PD

........
3: 5. CD meets AB and AP
---,
. areaMMN
at Mand N respectively. Compute
areaABCDEF
T-7. Starting at the origin, a beam of light hits a mirror ( in the form of a line) at A( 4, 8) and is reflected to the
point B(8, 12). Compute the exact slope of the mirror.
T-8. Compute the number of positive integers a for which there exists an integer b, 0 :S b :S 2002, such that both of the
polynomials x 2 + ax + b and x 2 + ax + b + I have integer roots.
T-9. Shown is a square in which the elements in each row, each column, and each a b C
diagonal have the same product N. Let a, b, c, ... , i be distinct integers d e .r

greater than or equal to I. Compute the least possible value for N. g h i
T-10. Let m be an integer such that O < m :S 29 and let n be a positive integer. There exist rectangles which can
be divided into n congruent squares and also into n + m congruent squares. Compute the number of distinct
values of m so that for each of those values of m there exists a unique value of n.
209
1. 2789
2. 2001
3. 10
4. 29 or 29°
5. 21
32
6. -
33
l+Jw
7.
3
8. 44
9. 216
10. 18
210
T-1. The largest possible product for a three-digit number is 9 · 9 · 9 = 729, so the number must have four digits.
The thousandth's digit must be at least 2 since I doesn't change products obtained from three-digit
numbers. Let the number be 2ABC where A· B · C > 500. If A = 6, then B · C > 500/6 > 83, an
500 3
impossibility. If A = 7, then B · C > - = 71 + - ➔ B = 8, C = 9. Answer: 127891.
7 7
T-2. The right hand edge of the pile is at the following distance from the left hand edge:
( I J ( 2 ) ( 1998) ( 1999) 2 . 2000 __

( 1_+_2_+_._
.._+_1_9_99_)
l l l
2 + 2 - I 000 + 2 - I 000 + . . . . . + 2 - 1000 + 2 - I 000 l 1000
2000 . 1999 = 4000 - 1999 = 120011.

4000- 2000
T-3. Since 240 = 2 4 · 3 · 5 we assign a 2 to a, a 3 to band a 5 to c, leaving three 2's to assign to a, b, and c in
some way. There are 3 ways to assign all three 2's to one letter, 1 way to assign a 2 to each letter, and
3 · 2 = 6 ways to assign two 2's to one letter and one 2 to another, giving IO distinct ordered triples
(a. b, c) where a, b, and c > 0. Ans: [ill].
T-4. Let mLA = x. Then mLC =x ➔ m BD- = 2x and by the
Exterior Angle Theorem, mLEBC = 2x ➔ -

m EC = 4x. Thus,
180° = 2x+mBE+4x - ➔ 6x < 180°,makingx < 30° ➔

A
mLA = 129°1.
T-5. We need only consider the first half of the number since the second half will mirror the first half. For an
n-digit palindrome there will be k = [ -n 2+- I] entries in the first half. Ifwe could freely chose l's and 2's
there would be 2 k different palindromes, but we're excluding those two with all l's or all 2's, so there are
k [( n+ 1) / 2) [(n+ I)/ 2] [(11+ I)/ 2]

2 - 2 = 2 - 2 palindromes. Hence, 2 - 2 > 2002 ➔ 2 > 2004.
1] =
Thus, [ -n 2+- 11 ➔ n = 12:I] .
211
T-6. LetAB ~ 8andnotethatAMBC is equilateral and ANAM - ANPD. A 8 B 8 M
ND
Thus - -
'NM
PD
=--
AM
➔
ND
ND+ 16
5
16
➔
80
ND=-. The area of
11
\
\
8\
\
z
/ 8
F C
ANAM = .!.__
2
16 · (16 + SOJsin LM
II
= s( 256 J /j
11 2
= 1024.Jj. The
II
8
D
E
g2/j ./3 ➔ area ANAM
area of ABCDEF = 6 · - - 96
4 area ABCDEF \; N
1024./3 I 11 = ~ 2 .
33
96i3 liiJ
T-7. Let the equation of the mirror DC be y = mx + b. The slopes C B(8, 12)
of the mirror, OA, and AB are m, 2, and I respectively. Since

A (4, 8)
the angle of incidence equals the angle ofreflection, ,.
'
m-I 2-m
tanLCAB = tan LDAO ➔ -- - --- ➔
I+ m I+ 2m
3111 2 - 2111 - 3 =0 ➔ 111 = i 1+ f° . 0
T-8. The discriminant of each quadratic must be a perfect square if each is to be factorable over the set of integers.
Thus, set a2 - 4b = m 2 and a2 - 4(b + 1) = n 2 for integers m and n. Subtracting cancels a2 and 4b
yielding m 2 - n2 = (m + n)(m - n) = 4. Since m + n and m - n are of the same parity, we obtain
m = ±2, n =0 ➔ m2 = 4. Thus, from a2 - 4b = m 2 we obtain a 2 = 4 + 4b ➔ a=± 2 ~ . Let
a > 0. For a to be integral, b + I must be a perfect square, so let b + I = c2 for c a positive integer. Then
both quadratic expressions are factorable since
') 2
x 2 +ax+ b x- + 2c + c - I (x + c- IXx+ c+ 1)
2 2 2
x +ax+b+I X + 2c + C (x + c) 2
Thus, for both expressions to be factorablc it is necessary and sufficient that b + l be a perfect square and that
a= 2.{i;;l. For 0 :::; b :::; 2002, b + I is a perfect square for [J2002] = 44 positive values of b. Thus, a
can take ~ values.
212
T-9. Note that N 4 = (ahc)(dlf)(ghi)N = (ae1)(heh)(ceg)(del) = (m)(hh)(lg)(def)e 3 = (ahc)(def)(ghi)e 3 .
Thus, N = e 3 . We seek the least positive integer e such that ahc = e 3 for distinct integers a, b, c, and e.
Letting one of a, b, or c equal 1, say b = I, then ac = e 3. Now, e can't be prime since if it were then
a= I and c = e 3 or a= e and c = e2, both of which are impossible since a, b, and c must be distinct. Nor
can e = 4 since then a = I and c = 4 or a = c = 2. Thus, the least value for e is 6 making
Note: Such a square exists. Let b = I, making h = 2 2 · 3 2 . Then g = 2 2 2· 3 2T

?
and i = 3 or vice-versa. Setting g = 2, then c = 2 · 3 2 and the rest of the ?

3- 2·3 22
entries are determined in similar fashion, giving the figure at the right.
2 22 32 3
T-10. Without loss of generality, consider a rectangle ABCD that can be divided into n + m squares of side I and
n larger squares of side x. Since the dimensions of ABCD are integral, x must be a rational number. Let
x = p where p and q are relatively prime. Since x > I, then p > q. Since the area of
q
ABCD = (n+m)·I = {qPJ2 , we solve for n to obtain n =

~2
p2 _ q2
~2
= (p- q)(p + q) Since
p and q are relatively prime, neither p - q nor p + q divides q 2 so both must divide m and they must
have the same parity. Ifm has two odd factors, say m = ijk wherej and k are odd, we have
i(j - k)2
p - q = j, p + q =k or p - q = I, p + q = jk giving two different values for n, namely ~ - 4- -
i(jk - 1)2 .
or -"------- respectively. Som must have only I odd factor. If m is to be odd, it must be an odd prime.
4
There are 9 of those less than or equal to 29, namely, 3, 5, 7, 11, 13, 17, 19, 23, and 29. Ifr is an odd prime,
the factors of 2r or 4r with the same parity are (I, r) and (2, 2r) and they both generate the same value for n,
(r 1) 2
namely - . This is an integer since r - I is even. Thus m = 2r = 6, 10, 14, 22, 26
4
and m = 4r = 12, 20, and 28 generate an additional 8 values form. If m = 8r, then (2, 4r) and (4, 2r)
have the same parity and p - q = 2, p + q = 4r gives n = (2r - 1) 2 while p - q = 4, p + q = 2r gives

n = (r - 2) 2 . These arc equal for r = 1, so form = 8 there is a unique value for n. Form = I, 2 or 4 there
are no values for n since both p - q = 1, p + q = 1 and p - q = 2, p + q = 2 yield q = 0. Hence, the total
number of values is 9 + 8 + I = [ID .
213
ARML Power Question - 2002: Power ofAssociation
In all these problems a clique is a non-empty set of students and students may be members of more than one clique.
Also, lowercase letters denote students; uppercase letters denote cliques. For example, C = ahcd is a clique of four
students, a, h, c, and d. We represent a collection of cliques with set notation, e.g. { abcd,ef'gh} is a collection of
two cliques of four students each. We use S to stand for the set of all students, i.e. the student body. There is a finite
number of students.
At Archimedes Academy, the faculty is concerned about students' tendency to form cliques and it hires an
anthropologist to study the cliques. The anthropologist finds that the cliques at AA satisfy the following three
conditions:
A 1. For any two students, there is exactly one clique of which they are both members.
A2. If a student a is not a member of a clique C, then there exists exactly one clique D of which a is a
member and that has no members in common with C.
A3. There are three students that are not all members of the same clique.
a) If S = {a, b, c,d}, determine a collection of cliques that satisfies A l-A3.
b) If S = {a,b, c}, prove that there is no collection of cliques that satisfies Al-A3.
2. If S = {a, b, c,d, e,f, g, h, i} , determine a collection of cliques that satisfies A l-A3.
In problems #3 - #6 assume that C, D, and E are cliques at Archimedes Academy where the collection of cliques
satisfies A l-A3 above.
A clique D will be called exclusive of clique C if either (i) C = Dor (ii) C n D 0.
3. Prove that if C is exclusive of D, and Dis exclusive of E, then C is exclusive of E.
Let IC I denote the number of students in clique C.
4. a) Prove that if C and Dare exclusive, then IC I = ID I-
b) Prove that if C and D are not exclusive, then IC I = ID I-
214
ARML Power Question - 2002: Power o[Association
Let [a] denote the set of all cliques of which ais a member, let [c] denote the set of all cliques exclusive of C and
let I[ a] Iand I[ C] Ibe the number of clements in the specified set. Thus, [a] = { C: a E C} and
[C] = { D : D = C or D n C = 0 }.
5. Prove that I[ c]I = IC I.
6. Prove that if C and Dare distinct cliques, then I [ c] I = I [D] I-
7. Prove that the number of students at Archimedes Academy must be a perfect square.
Faculty at Hausdorff High were similarly concerned and called in the same anthropologist. At HH the
anthropologist found that the cliques satisfy the following four conditions:
HI. For any two students, there is exactly one clique of which both are members.
H2. For any two cliques, there is exactly one student who is member of both.
H3. There exist three students who are not all members of the same clique.
H4. Every clique has at least three students.
8. a) Prove that if the student body has exactly four students, there is no collection of cliques satisfying Hl-H4.
b) Prove that it is impossible for Hl-H4 to be satisfied by a student body of fewer than 6 students.
9. Find one collection of students and cliques satisfying H l-H4 in which each clique has 3 members.
10. ls it possible for a collection of students and cliques to satisfy H l-H3 but not H4? Justify your answer.
11. Let S be a set of students satisfying H l-H4, and let C be a clique in S. Let SC be the set of students in Sand
not in C, i.e. SC = {s E S: s ~ C}. Cliques in SC are formed by removing the members of C from the
cliques in S. Show that SC satisfies the Archimedian conditions Al-A3.
12. If IC I = n for some clique C, find with proof the total number of students at Hausdorff High in terms of n.
215
I. a) Let T ~ [ ab, ac, ad, be, bd, cd}. Then Tis a collection of cliques satisfying A l-A3.
We can show (not necessary for credit) that no clique can have more than two members. First, by A3 no
clique can have four members. Assume that abc is a clique. Then da, db, and de must all be distinct cliques
by Al (since a, b, c, cannot be members of any other single clique). But by A2, if dais a clique, be must also
be a clique, which violates A 1.
Since all cliques have two members, we can represent them as rows and columns but not diagonals in an
array:
~
~
The array yields ab, ac, bd, cd as cliques. So far this is not sufficient because we know that a and dare also
members of a clique. So we arrange a, b, c, d into a second array incorporating ad as a row, also yielding be
as a clique.
b) Let ab be a clique. By A2, c must be a member of a clique that does not include a orb. Therefore, c must
be a singleton clique. However, be is also a clique by axiom Al, so a must be a singleton clique by axiom
A2 applied to be. This is a contradiction because a is now a member of two cliques that have no members in
common with clique c, namely, the clique a, and the clique ab.
2. Note that two-element cliques no longer work: if ab is a clique and cd, ce are both cliques, then A2 is
violated since c is a member of two cliques disjoint from ab. We will look for a collection of cliques each of
which has three elements. Arrange the letters a-i in a 3x3 array alphabetically:
I
a b C
d e f
g h i
216
Using columns and rows but not diagonals, this arrangement yields cliques abc, def; ghi, adg, heh, c/i. Now
a and e also share a clique, so we need to start building a second array with ae as part of a row. To avoid
duplicating the ahc clique, we arrange hand c along the main diagonal, and similarly, d and g outside any
row or column with a. By A 1, b cannot be in the same row or column as e since we already have a clique
with be in the first array. In the array below these restrictions place b in the lower right hand corner.
[[;~
Now the choices for the other three spots are severely restricted: by axiom Al, the upper-right corner cannot
bef (already in def) or h (already in heh) so it must be i. The lower-left corner must befand the middle left
must be h:
a e i
h C d
f g b
Notice that the four cliques containing a exhaust all other members of S, so we're done. Thus, a collection of
cliques satisfying Al-A3 is {ahc, def, ghi, adg, heh, c/i, aei, hcd,fgb, ahf, ecg, idb}.
Note: Should we consider collections of cliques in which some cliques have a different number of members
from others? In 4a and 4b we prove IC I = ID I, so the answer is no.
3. If C = D, then we are done, since D = E ⇒ C = E and D n E = 0 ⇒ C n E = 0, so in either case C is
exclusive of E. If C n D = 0 and D = E, then C n E = 0 and so C is exclusive of E. The final possibility is
thatCnD=0andDnE=0. AssumethatCnE-::f.0. ThenthereexistssomexE (CnE). Nowx~ D
by the hypothesis that C n D = 0. But then by axiom A2, C and E cannot be distinct cliques. So C = E and
so C is exclusive of E.
4. a) Trivial if C = D, so assume C n D = 0. Let the students in clique C be c1, c2 , c3 , ... and the students in
clique D be d 1, d 2 , d 3 ,... . Then c1 and d 1 belong to some clique, call it c1d 1 . By axiom A2, c2 belongs to
another clique that is exclusive of c1d1 . That clique is not exclusive of D by axiom A2
217
Solutions to the ARML Power Question~ 2002
(or else there would be two cliques containing c2 exclusive of D) so it contains some member of D, say ~.
We can continue constructing cliques exclusive of c1d1 until we exhaust the members of C: each such clique
contains a distinct member of D (again by A2). So IC Is; ID I- The same process applied to D shows that
ID I s; IC I- So IC I = ID I- (Technically, we've used one-to-one functions to map C into D and then D into
C, which, in the case of infinite cliques would require the Schroder-Bernstein theorem to infer that
b) Let the students in clique Cbe a, c 1, c2, ... and the students in clique D be a, d 1, d2 , .... We use the same
construction as in 3a, beginning with the clique c1d 1 and choosing cliques containing c2 , c3 , ... each
exclusive of c1d1 . This time the fact that the cliques are not exclusive of Dis guaranteed by problem 3: if
c1d1 is exclusive of some clique containing c2 and that clique were exclusive of D, then c1d1 would be
exclusive of D, which it isn't since it contains d1.
5. Let the students in clique C be c1, c2 , ... , c11 and let d Ii!'. C. Let D be a clique containing c1, d. By problem
4, D has 11 members. Notice that by construction D is not exclusive of C. Now by A2, for each s E D, there
is a clique C that contains sand is exclusive of clique C. We'll show that the cliques C actually form a
s s
complete collection of all the cliques in [ c]. If there were a clique C' in [C], then C' must intersect D. If
not, then C' and Care exclusive and so are C' and D which would make C and D exclusive, but they can't
be exclusive since they intersect. That means that every C' in [ c] intersects with D for one values; it can't
intersect in two or more values by A 1. Moreover, by A2, it is true that each sin D but not in C is a member
of only one clique that does not intersect C. Therefore, the collection of cliques that are exclusive of C
consist of C itself and one class Cs for which every s which is in D but not in C. There are ID I possible
cliques that are exclusive of C, so I[c] I= ID I- Since D and C intersect, ID I = IC I- So
l[c]I = IDI = le I-
6. By problem 5, I[ c]I = IC I and I[D] I = ID I. By problem 4, IC I = ID I. Thus, I[ c] I = l[D] Iby the

transitive property.
218
7. Pick some clique C. By A2, every student at Archimedes Academy belongs to a clique in [c], either C itself
or some clique C. exclusive of C. But by the uniqueness provision of A2, cliques C. arc disjoint. So
I I
Isl= I lcil= I IC I since ICi I= IC I by problem 4. By problem 5, there

those Ci E [cl those Ci E [cl
are exactly IC I summands, so IS I = IC 12 .
8. a) Call the students a, b, c, d. Then by HI and H4, ab is a clique with another member. Assume without loss
of generality that this member is c; by H3 that clique cannot contain d. Now by HI, a and dare members of
a clique ad which contains another member. But that contradicts H 1, since both b and c are already members
of another clique with a.
b) Start with a clique abc by H4. There is at least one other student dby H3. Now ad is a clique with
another member, necessarily neither b nor c, so call it e. Then bd is a clique with another member,
necessarily neither a, c, nor e, so call it.f So at the very least we need 6 students.
9. The axioms are satisfied by the cliques abc, ade, afg, bdf, beg, cdg, cef for S = { a, b, c, d, e ,f, g } .
10. Using S = {a, b, c}, the cliques ab, ac, and be satisfy Hl-H3. Clearly every two students are in exactly one
clique, every two cliques share a student, and there exist three students who are not members of the same
clique.
11. Al. For any two students, there is exactly one clique of which they are both members.
Suppose that a, b E SC and there are distinct cliques C 1and C2 in SC containing both a and b. Then
C1 s C/ where C / is a clique in Sand C2 s c2' where c2' is a clique in S. But then C/ and c2' violate H 1.
219
A2. If a students is not a member of a clique C, then there exists exactly one clique D of which sis a
member and which has no members in common with C.
Existence: Choose a clique Din SC and a students !le' D. We need to find a clique E containing s as a
member with En D = 0. Now for some clique D' in S, DI'.;;; D '. By H2, D' n C contains exactly one
student, call it t. Then by H l there exists a clique E' in S containing sand t and by H2 that clique contains
no other member of D. Therefore, let Ebe the clique in SC that results from removing t from E'.
Uniqueness: Guaranteed by H 1 since in the previous construction E' was unique.
A3. There are three students that are not all members of the same clique.
We show how to find three such students. If a E SC , then for some c E C, ac is a clique distinct from C (by
HI and the fact that a !le' C), so by H4 there is a third student b E ac. But then b !le' C by H2 applied to C and
ac, sob E SC. Now Chas at least three members so pick c'=t- c in C and repeat the process to find a student
d E ac '. But d =I- b because ac and ac' are distinct cliques, so by H2 they have only one student in common
in S, hence only one in common in SC . So a, b, dare distinct students in SC . Since ab and ad are distinct
cliques in S, a, b, d cannot all be members of any other single clique in S (by H2) and so are not members of
a single clique in SC .
12. Since lsc I= IS 1-1 CI we have Is I= I SC I+ IC I- Since Sc satisfies Al-A3, by problem 8,

then ISC I = ID f, where D is any clique in SC . But ID I = n - I, because if we consider the extension D' of
D to S, In' n CI = l by axiom H2. So Is I = (n - 1) 2 + n = n2 - n +I. Incidentally, this shows that the
smallest student body is given by 32 - 3 + I = 7 students, and that the next smallest is 4 2 - 4 + I = 13 students.
220
1-1. For A> B, if dll.d 10 - BAfl 10 is divisible by exactly three distinct primes, compute the largest possible
value of B.
1-2. Starting with the outermost ring, a farmer cuts a circular field in rings of uniform
width. As an example, the diagram shows one and a half trips around the field . If
the width of each ring is 5 feet and if after the farmer has made eight and a half
trips around the field, he has cut half the field , compute the radius of the field in
feet.
1-3. The legs of isosceles trapezoid ABCD are diameters of

tangent circles O and P. If AB = 90 and CD = I 000,
compute the height of the trapezoid.
1-4. Compute the number of ordered pairs of integers (x, y) with I :::: x < y s; 100 such that / + iy is a real
number. Note: i = .J::i.
1- 5. II
Let k be a positive integer. The intersection of the graphs of y < k and y > x contains at least 90 lattice
points. Compute the smallest value of k.
1 1 1 1 1 a
1- 6. Let a be the integer such that 1 + 2 + 3 + 4 + · · · + 22 + 23 = 23 ! . Compute the remainder when a is
divided by 13.
1-7. If f(x) = (x - 3) 2 - 1, compute the set of real numbers such that J(I x I) = If(x) I.
1- 8. AB CDEFG is a regular heptagon of side 100 I. Compute A c 2 - (DE)( FB) .
221
1. 7
2. 145
3. 300
4. 1850
5. 11
6. 7
7. 0 sxs 2 or 4 s x. Alternately, [ 0,2] u [ 4, 00 ).
8. 1002001
222
Solutions to the ARML individual Questions - 2002
1-l. ABA 10 BAB 10 = (IOIA + !OB) -(1018- IOA) = 91(A - B) = 7 · 13(A- B). To maximize B, choose A
as large as possible and minimize A - B. If A = 9, then for B = 7, A - B = 2, a prime. Thus, B = [1].
1-2. Let OA = r, OB = r- 40, and OC = r - 45. Then the area of the large
ring plus half the area of the small ring equals half the area of a circle.
n,) - n(r - 40) 2 + ¾( n(r -40) 2 - n(r - 45) 2 ) = ½m) ➔
? 2 I r 2
2
,.- - (r - 80r + 1600) + (I Or - 425) =2 ➔
r 2 - l 70r + 3625 =0 ➔ (r - 25)(r - 145) = 0. Thus, r = 25 or 145, but 25 is too small, so r = I1451.
1-3. Let AB = x an d DC = y. Then DE = -·V-- -X . The me d"ian

2
f . x+y '
o trapez01d ABCD equals OQ = - 2- , but that also equals ' r
----L--- Q
'
'
' r
X + V .. '
2r which equals AD and BC. Thus, AD =- 2-·, g1vmg
D y
( y -2 x) 2 + h 2 -- ( y +2 x )2 --,- - -'- h2= 4I (4xy) = xy. r

Thus, h = -i xy and since x = 90 and y = I 000,
h = J90000 = 13001. Note that the area of the trapezoid equals ~ ( x; y), the product of the arithmetic
and geometric means of the bases.
I-4. l + iy is real when both x andy are even. There are I0,000 ordered pairs of (x,y) and 2500 of these will
have x and y both even. In 2500 - 50 = 2450 pairs we have x /: y and in ( I /2)(2450) = 1225 pairs we have
x < y. Since i 1 + i 3 = 0, but l + ,5 == 2i, it is clear that /' + iy is also real when x and y arc both odd as
long as their difference is 2 more than a multiple of 4. There are 2500 ordered pairs with both x and y odd
and ( l/2)(2500) = I 250 pairs with x :S y. In halfof those y - x == O(mod 4) and in the other half
y - x == 2(mod 4). Thus, there are ( 1/2)( 1250) = 625 pairs where x and y are odd.
Answer: 1225 + 625 = I I 8501.
223
1-5. lfwe intersect the region with vertical lines drawn through lattice -----------------------------------,--
'
' ,,
........: • • ■ • •,., ... '
points we see that from left to right the number of points will be ....... • ■ • • ■ • ■ •,.,'
I + 2 + ... + (k - 2) + (k - I)+ (k - 2) + ... + 2 + I = '-,,: : : : :,,:,-·'

__
l+(k-2)J 2 ',, ______ ,.,,,'

2( 2 (k - 2) + (k - I) = (k - I) . Or, use horizontal
lines to obtain I + 3 + ... + (2k- 3) = (k - 1) 2 . For (k - 1) 2 > 90, k = IT].
23!(1)+23! .!.J+23!(! +···+23! ...!...J+23! ..!.J

1-6. Since I+ -I + -I + ••• + - I + - I 2 __.....__.___
= ____.....__._ 3 ___......_....._
22 __...._....._,
23 then
2 3 22 23 23!
a = 23 !(l) + 23 { ½J + 23{¾J + •·· + 23 { ; 2 J + 23{ ;3 J. Every term in this expression is divisible by
13 except 23 ! ( - I J . Consequently, the remam

. der when a 1s
. d.1v1.ded b y 13 1st . der w h en -23 ! 1s
. he remam .
13 13
divided by 13. Since the remainders when 14, 15, 16, ... , 23 are divided by 13 are the same, respectively, as
the remainders when l, 2, 3, ... , 10 are divided by 13, the remainders R of the following are the same when
divided by 13 : ( 1 · 2 · 3 · ... · 12 )( 14 · 15 · 16 · ... · 23), (1 · 2 · 3 · ... · 9 · 10) 2c 11 · 12) , and

{(2 • 7) ·(3· 9) •(4 · 10) · (5 · 8)· 6) 2(11 · 12). Note that we have paired up numbers whose product is 1 more
than a multiple of 13. This means that the remainder when a is divided by 13 is given by the remainder when
6 2 · 11 · 12 is divided by 13. Since 10 · 11 · 12 = 1320 = 13 · 100 + 20 gives the same remainder, we need
only look at the remainder when 20 is divided by 13. Answer: [I] .
Alternate solution:
R = ( I · 2 · 3 · ... · 12 )( I 4 · 15 · l 6 · ... · 23) (mod I 3) = l · 2 · 3 · ... · 12 · I · 2 · ... · IO (mod 13) =

(12!)(10!) (mod 13). By Wilson's Theorem, (p- I)! = -1 (modp), giving R = (-1)(10!) (mod 13).
Since 10! · 11 · 12 = 12! =-I (mod 13), then (10!)(-2)(-1) = (-1) (mod 13). Thus, -2R = -I (mod 13)
so 2R = I (mod I 3). making R = 7.
224
Solutions to the ARML Individual Questions ~ 2002
1-7. The leftmost diagram is the graph off( x) = (x - 3)2 - I , the middle is the graph of f(x) I I= I(x - 3) 2 - I I,
a selective reflection over the x-axis, and the rightmost is the graph of f (Ix I) = ((Ix I- 3) 2 - IJ,a reflection
over the y-axis.
i
~ I
vJ
I 8
8
2 -2 2
4 4 -4 4
I 2
The solution set consists of the x-values for which the middle and right hand graphs overlap, namely,
I0<x<2or4<x1.
1-8. In all these solutions note that AB = BC = GA = x, GB = AC = b, and FB CC = a. Thus,
AC 2 - (DE)(FB) = AC 2 - (AB)(GC).
Since ABCG is cyclic, we have by Ptolemy's Theorem, B
A ,· C
AC· GB= AB· GC +AG· BC. The trapezoid is isosceles so AC = GB -------,'--------------_-_..._
2 2 2 _7_ 2 , '. :, ,,,,,,,,,,
__ ,,," ,,-_,, ,,-
giving AC = AB· GC + AB ➔ AC - AB- GC = AJJ = 1001
G'
= I I 00200 II .
Note: this method would establish that for any regular n-gon A 1A 2 A3 ... An, the following relationship
225
Alternate solution using Law of Cosines: Draw A X B

900/7
AH 11 BC making rhombus ABCH. Let GH = y. Since
X h
ABCG is an isosceles trapezoid and each interior angle of
the regular heptagon equals 900/7 in degrees, the rest of

G 360/7 360/7 ~9_0_01_7_ _ _ _~ C
the angles are as marked. y H X
a-----
2 2 ? 900 900 360

Using l',.ACH: AC 2 b =x + x- - 2(x)(x) cos-. Since - +- 180, then
7 7 7
900 360 2 2 ? 360
cos-= -cos- giving h = 2x + 2x- cos - .
7 7 ' 7
2 2 2 2 360 2 360
Using 1',.AHG: AG = x = x +y -2(x)(y)cos 7 . Thus, 0 = y - 2xycos-7- ➔
360 2 360
y = 2xcos- ➔ xy = 2x cos-.
7 7
1 . 2
AC- - ( GC)(AB) = b - (x + y)x = h2 - x
2
- xy l
( 2x 2 + 2x 2 cos 360
7 J- x 2 - l 2x 2 cos -360
7-
J 2
X .
Since x = I 001, the answer is 100200 I.
Alternate solutions using the Law of Sines:

b X A X B
Using /',ABC, = --:--0 and since
sin(l 80 - 28) sm
x(2 sin 0 cos 0) X I "'--- b

sin(l 80 - 20) = sin 28 , b = --'--------- = 2x cos 0 . : 180-38
sine h: I
Using t,.AJC, cos 0 = !➔ z = b cos 8, and using /',AG/

G
28
w
I
I
I
1' z
I
a
w
cos20 = - ➔ w = xcos28. Since a= w+z, a= xcos20+ hcos0 xcos 20 + (2xcos 0) cos 8.
X
Thus, h2 - ax = 4x 2 cos 2 0 - (J"\2 cos 2 8 - I ) + 2x cos 2 0 ) x = x 2 .
-----'- a -_ { 3 sin 0 - 4 sin 3 0)

Or, one could use b = 2xcos0 and from AAGC obtain _a_ = _x_ ---,
sin 30 sin 0 sin 0
3x - 4xsin 2 0 = 3x - 4x(l - cos 2 0) and continue as above.
226
ARML Relay #I - 2002
R 1-1. Suppose that in base I0, ELEVEN stands for a number divisible by 11 . If different letters stand for different
digits, compute the largest value for ELEVEN .
Rl - 2. Let T = the sum of the digits ofTNYWR. In the A M B

diagram, ABCD is a parallelogram, the circle is
inscribed in rhombus AMND, and the areas of the
shaded regions are equal. If NC = T - 41 , compute the
circumference of the circle.
D N C
I+/;
Rl - 3. Let T = TNYWR. Solve for x: - - - = T.
I-ix
R2- l. Ed spends $1.20 every week on noodles. The price of a package of noodles changes each week, cycling
through I 0, 15, and 20 cents in a regular fashion. To the nearest penny, Ed spends N cents on average per
package of noodles. Compute N as an integer without any dollar or cents notation.
R2- 2. Let T = TNYW R. In MBC, AB ..L AC, AD ..L BC, the

area of MBD = 2, and the area of MDC = T.
Compute tan 2 LB .
B D
R2-3 . Let T = TNYWR. Compute the area of the smallest regular hexagon containing the points
7 , 0 ) and B ( log -
A ( log - - lOOT
- , 0) .
lOOT 7
227
ANSWERS ARML RELAYS - 2002
Relay #1:
Rl-1. 989791
Rl-2. 8
49
Rl-3.
81
Relay #2:
R2-1. 14
R2-2. 7
R2-3. 6✓3
228
RI - I. If ELEVEN is divisible by 11, then E - L + E - V + E - N = 3E - (L + V + N) is divisible by I I.

Set E = 9 ➔ 27 - (L + V + N) is divisible by 11. ELEVEN is largest when L + V + N = 16 and when
L - 8,V = 7,andN = I. Thus,ELEVEN=,98979Jj_
Rl-2. T = 9 + 8 + 9 + 7 + 9 + I = 43, making A M B
NC = 43 - 41 = 2. Since 7t(r) 2 = 2r(NC), then '

''
= 2NC , giving = 2r
nr 2nr 4NC . Thus, the circumference '''
'
of the circle is 4 · 2 = [§]. ''7
D N C
2
Rl-3. T= 8. Since I+ ix= T- Tix, then Tix+ ix= T- I ➔ ix= ~: ~ ➔ x = ( ~: ~)
Hence x = fil -
R2-I. Every three weeks Ed buys 12 packages at $.JO/package, 8 packages at $.IS/package, and 6 packages at
$.20/package, making a total of26 packages costing $3.60. Since 360/26 = 13.8, Ed spends an average
of $.14 per package. Pass back CT±] .
R2-2. T = 14. In general, if the areas of fi.ABD and ti.ADC are
m and n respectively, then their bases BD and DC have

lengths mx and nx respectively, since the triangles share an
altitude. Since AD is the geometric mean between BD and
DC, AD = J(rnx)(nx) = x&. Thus, B mx D m
tanB = - -
x& .Jn
-
2
and tan B
11
= - . Since m = 2 and 11 = T = 14, tan B =
2 14
[TI.
mx im m 2
R2-3. T = 1. The smallest regular hexagon will have A and Bat endpoints of
. . IOOT 7
the longest diagonal whose length will be log-- - log - -
7 100T A B
104r2 10472
log-- log - - = 4. Since AB is twice the side x, the side
49 49
of the hexagon is 2 and the hexagon's ,ucea is { 22 :'3) ~ ifs [

I
229
I. Given that 2002 2 · 7 · 11 · 13, compute the number of positive two-digit factors of 2002.
2. Let T = TNYWR and let K = T + 2. The large circle has a radius of K.

The congruent circles centered at P, Q, and R are tangent to exactly two other
circles and circle Q is concentric with the large circle. Points A and B lie on
the large circle so that AB is tangent to circles P, Q, and R. Compute AB.
3. Let T = TNYWR. If la+ bil = T and la+ 2bil = Jr 2 + 96, compute la I-
4. Let T = TNYWR. Four congruent isosceles trapezoids are

arranged as shown. The length of the longer base of each trapezoid
is T more than the shorter base. If the area of quadrilateral ABCD is
the product (20.02)T, compute the height of a trapezoid.
5. Let T = TNYWR. Let K be the integer closest to T. A bag contains K marbles, half of which are green;
the other half are red. Two marbles are drawn randomly and without replacement from the bag.
Compute the probability that their colors differ.
6. The number you will receive should be a fraction. Let ~ represent the simplest form of the fraction and let
T =D - N. For a digit A, if .AT =I- .AT for numbers written in base T, compute A.
. 2004
7. Let T = TNYWR and let K = T + 3. Compute the largest pnme factor of K + 3K 2002 + 2K 2000 .
8. Let m and n denote, respectively, the smaller and larger of the two numbers you will receive. Cubes A and B
intersect each other and their intersection is cube C. The volume of C ism, the volume of A is at least as large
as the volume of B, and the volume of the union of A and Bis n. If the volumes of A and Bare integers,
compute the number of different volumes that cube A can have.
230
15. Compute the largest two-digit multiple of 3 that is one less than a perfect square.
14. Let T = TNYWR and let K = T +I. The diagram shows ten A
congruent squares, each with an area of K. Compute the
exact perimeter of ABCDE.
B
C
13. The number you will receive should be in the form a+ b./i where a and hare integers. If b consecutive
15
integers have an arithmetic mean of 2 , compute the smallest of those integers.
12. Let T = TNYWR. The following system of equations has no solutions except for one particular value of K.
Compute that value.
x +Ty+ 5z = -]
-2x + 14y K + !Oz
11 . Let T = TNYWR. Andy's Ristorante of Miscellaneous Lunches offers a special where you can order an
entree, a dessert, and either a salad or an appetizer, but not both. There are 4 entrees, 3 desserts, 5 appetizers,
2 salads, and each salad will come with one of T different salad dressings. A customer must order an entree,
but need not order any of the other dishes. Compute the number of different lunches that could be served.
10. Let T = TNYWR. Let K = -T . Given 0° < 8 < 90°, compute 8 such that sm
5
. - +K0
2
(8 ) 8
= cos-.
6
(Pass back the value of 0 without the degree symbol).
9. Let T = TNYWR. Since [ log T] I, compute the value of 6 + [ log T 3 ]. [ x] is the greatest integer
function.
8. Let rn and n denote, respectively, the smaller and larger of the two numbers you will receive. Cubes A and B
intersect each other and their intersection is cube C. The volume of C is rn, the volume of A is at least as large
as the volume of B, and the volume of the union of A and Bis n. If the volumes of A and Bare integers,
compute the number of different volumes that cube A can have.
231
ANSWERS AR ML SUPER RELAY - 2002
1. 7
2. 12/i.
3. 16
4. 20.02
IO
5. -
19
6. 4
7. 17
15. 99
14. 100 + 30.J°2
13. -7
12. 2
11. 160
10. g70
9. 11
8. 4
232
I. The two-digit factors of2002 arc 11, 13, 2 · 7 = 14. 2 · 11 = 22, 2 · 13 = 26, 7 · 11 = 77, and 7 · 13 = 91. Ans: [11
2. Let r be the radius of the small circles ➔ the radius of the large circle = 3r. Let M be the point of tangency of
2
circle Q and -AB. Then QM
- .l -AB and from right triangle MQB we have r 2 + (ABJ
2 = (3r) 2 . Since
')
T=7,K=9=3r,thenr=3 ➔ 9+(~Br =81 ➔ AB2 =288 ➔ AB=li2-hl
3. Ja 2 + b 2 = T and Ja 2 + 4h 2 = JT 2 + 96 ➔ a 2 + b 2 = T 2 and a 2 + 4b2 = T 2 + 96. Subtracting
gives 3b 2 = 96 ➔ b2 = 32 ➔ a2 = T 2 - 32. Since T = 12i2, then a 2 = 288 - 32 = 256.
I I = IT6J .
Thus, a
. T E A
4. ABCD is a rhombus whose area is ( l/2)(AC)(BD). Smee AE = 2 , L, ,__1
I I
I
I
•h
I
then BD = T. Leth be the height of the trapezoid. Then '
B' '
~---~---r::'~---- D
(20.02)T = .!._(2h)(7) ➔ h = 120.021. Tis irrelevant.
2
5. After the first marble is selected there are K 1 marbles remaining and ~ of them have a different color from
. . . K/2
the selected marble. So the probability .
1s - - . Smee T = 20.02, then K = 20. Ans: lffO
- .
K - I 19
6. Thus -A + - A + - A + ••• - A ( -1/-T-) = -A- = -I . Poss1"ble

2 ' T T2 T3 - 1 - 1/ T T- I 2
answers (A, T) are (I, 3), (2, 5), (3, 7), (4, 9), ... , (9, 19). Since T = 9, then A = @l.
7_ K2004 + 3K2002 + 2K2000 = K2000(K4 + 3K2 + 2 ) = K2000( K2 + 1)( K2 + 2 ). Since T = 4, K = 7,
and we have 7 2000 (50)(51) = 7 2000 (2· 52 • 3-17). Ans: [I].
233
15. Since 99 = I 02 - 1 and 99 is the largest two-digit multiple of 3, the answer is [22].
2
14. Let x be the side of a square. Then the perimeter is lOx + 3xi2. Since T 99,K = 100,and x = 100 ➔
x = I 0. Answer: I100 + 3o./i. I.
13. Let the integers be n, n + I, n + 2, ... , n + (b - l) ➔ their sum= bn + b(b 2- l). Dividing by b to obtain the
b-1 15 16-b ~,::

mean yields n + - - ➔ n = --. Since T = 100+ 3W2, b = 30 making n = ~ -
2 2 2
. K
12. The second equat10n can become x - 7 y + 5z = - 2 . The equations are equivalent for T = -7 and K = [2l.
The answer is independent of problem 13.
11. If no dessert is a choice, there are 3 + 1 = 4 choices for dessert and 5 + 2T + I choices for salad/appetizer or
neither. The total number of meals is 4( 5 + 2 T + I )4 = 96 + 32T. Since T = 2, the answer is I I 601 .
10. Since cos¾ = sin(90° - ~ J > 0, we have sin(~+ K J = 0

sin(90° - ~J ➔ i +K 0
= 90° - ~ or
i + K 0 = 180°-( 90° - ¾ J. Thus, 8 = ½(90° -K°) or 3(90° - K 0 ). Since T = 160, K = 32, making
8 = 87° or 174°. Since 8 is acute, 8 = 187°1.
9. Since [ log T] = 1, Tis a two digit number and [ log T 3 ] = 3, 4, or 5 for Tin [10, 21 ], [22, 46], and [47, 99]
respectively. Since T = 87, [ log T 3 ] = 5 and 6 + [ log T 3 ] = [I].
8. If A and B partially overlap, then vol (A)+ vol (B) - m = n ➔ vol (A)+ vol (B) = n + m. Since
vol (A) ~ vol (B), then vol (A) ~ n~ m . If A and B coincide, then vol (A) = vol (B) = n but since
n > m, vol (A) > vol (C). Thus, m < n ~ m :S vol (A) :S n. Since m = 11 and n = 17, we have
11 < 14 :S vol (A) :S 17. So, the possible values for vol (A) are 14, 15, 16, and 17. Answer: @l .
234
ARML Tiebreakers - 2002
I. Compute all k such that the equation Jx + k =x- I has two distinct real solutions for x.
2. Shown are three semicircles whose diameters lie on AB. C
If CD_l_ AB, CD 3, and AB I 0, compute the area of the
unshaded region.
B
D
3. There are n triangles of positive area that have one vertex at A(O, 0) and the other two vertices at points with
coordinates in {0, I, 2}. Compute n.
235
I. x+k=x 2 -2x+l ➔ x 2 -3x+(l-k)=O.Thus.9-4(1-k)>O ➔ 4k>-5 ➔ k>- 5 .But

4
some of those solutions arc extraneous once the curve y = J;-;-; slides too far to the left as shown below:
k = -112
Thus, I-¾< k::; -11.
2. Since ABC is a right triangle, AD· DB = CD 2 . Let AD = x, then x(lO - x) = 9 ➔ x = I.
Then the unshaded area equals -

2
1
7t ·
-,
5- - -
2
1 1
7t · ( -
2
)2- 1
- 1t ·
2
(9 J2 = [-91t .
-
2 4
3. There are 3 · 3 = 9 points that can be formed using the coordinates {0, 1, 2}. Eliminate A(O, 0) and from the
remaining 8 points choose 2. This can be done in 8 C 2 = 28 ways. Eliminate the pairs (0, I) and
(0, 2). (I. 1) and (2. 2), ( 1, 0) and (2, 0) since they are collinear with A, leaving 28 - 3 = [23] pairs of points.
236
2003
Power Qyestion......................................................................................... 246

'lndividua( 'Round................................................................................. 253
'Re(ay 'Rounc(............................................................................................ 257
Suyer 'Re(ay................................................................................................ 260
Tiebreakers 265
-237-
THE 28th ANNUAL MEET
This year there were 25 teams in Division A and 72 teams in Division B for a total of 99 teams
involving close to 1600 students. There were teams from India and the Philippines but no teams
from Taiwan this year due to SARS. There was a very spirited competition in Division A for the
national championship. Going into the relays Thomas Jefferson trailed San Francisco Bay A by
6 points, 131 to 137, but with 24 points on the relays, TJ surged ahead, winning 155 to 153. In
Division B Connecticut A won easily with 112 points, but there was spirited competition for
second and third with 3 teams bunched very close together. Lehigh Valley prevailed for second
place.
Sam Baethge of Texas received the received the Samuel Greitzer Distinguished Coach Award.
Sam has long been associated with ARML and in 1997 he received the Alfred Kalfus Founder's
Award in recognition of his stalwart and enthusiastic support.
Tatiana Shubin and Marilyn Blockus received the Alfred Kalfus Founder's Award for their fine
and tireless work in setting up the new western site at San Jose State University.
The following received the Zachary Sobol Award for their outstanding contributions to their
teams:
Dominic Albino Western Massachusetts

Eve Drucker AAST
Robert Ikeda Southern California
David Vincent Phillips Exeter
238
T-1. If 121 h = N 10 and N has 9 factors, determine the least positive integer value for b.
T-2. Starting at I 1:44 A.M., Tony walked for 5 miles. He noticed that his average speed, expressed in minutes
and seconds per mile, was numerically equal to the time at which he stopped if he thought of the hour as
minutes and the minutes as seconds. Compute the time at which he stopped.
T-3. ABCD is a quadrilateral with mLDAB = 90°, mLBCD = 135°, BC = 3, and CD = 2i2. Compute the
maximum possible area of ABCD.
T-4. Let S = { L 2, 3, ... , 24, 25 } . Compute the number of clements in the largest subset of S such that no two
elements in the subset differ by the square of an integer.
T-5. A square is divided into 24 congruent rectangles as shown. On each side of the
dotted line 4 rectangles are chosen at random and colored black. The square is then '' I
'' I
folded over the dotted line. Compute the probability that exactly one pair of black ''
''
' I
rectangles is coincident.
T-6. Let the faces of a unit cube be the six planes x = 0,y = 0,.:: = 0, x = 1,y = I, z = I. Compute the
values oft such that the points A ( I, 1/2, t) and B ( 1/2, I, t) have multiple equal-length shortest paths
connecting them along the faces of the unit cube.
T-7. In trapezoid ABCD, the perimeter is 600, all sides are integers, AB = BC = CD, and AD is the longest
side. If the area of ABCD equals kJk fork an integer, compute k.
value of x + y.
T-9. Assume that as a cubical bar of soap is used, all edges shrink at a constant rate of n units per day. Starting
with a full bar, the soap was used for 6 days and its surface area was cut in half. Starting with a full bar of
soap, compute exactly the time it would take for the volume to become one-eighth of the original volume.
T-10. ARMLovian, the language of the fair nation of ARMLovia, consists only of words using the letters A, R, M,
and L. The words can be broken up into syllables that consist of exactly one vowel, possibly surrounded by a
single consonant on either or both sides. For example, LAMAR, AA, RA, MAMMAL, MAMA, AMAL,
LALA, MARLA, RALLAR, and AAALAAAAAMA are ARMLovian words, but MRLMRLM,
MAMMMAL, MMMMM, L, ARM, ALARM, LLAMA, and MALL are not.
Compute the number of7-lettcr ARMLovian words.
239
1. 5
2. 12:48
41
3. - = 10.25
4
4. 10
5. 224 = .4525
495
6.
Ji -1 3-.Ji
2 2
7. 573
8. 3540
9. 6+3✓2
10. 1435
240
2 2 . 2 2 2
12lh=h +2h+l=(h+l) .Smcc6 =2 ·3 has(2+1)(2+1)=9factorsanditislessthan
2 8 = (2 4 ) 2 which also has 9 factors, then h = ~-
T-2. If he finished at x minutes past 12 then he walked for 16 + x minutes. Setting his average in minutes per
. . . 16+x X
mile equal to the time he stopped expressed as mmutes gives - 5- = 12 + 60 ➔ I Ix = 528 ➔
x = 48 . He stopped at I12: 48 I.
T-3. Break ABCD into triangles ABD and BCD. The area of BCD is fixed and D
equals½· 3 · 2.Ji. sinl35° = 3. By the Law of Cosines,
➔
3
DB 2 = 32 + (2.Ji.)2 - 2 · 3· 2i2 cos 135° = 29 DB= /i9. Since
DB is fixed, the area of ABD is maximized when ABD is an isosceles right A ~ - -- - ~B
triangle, making AD = AB =
m
Ji. ➔ the maximum area of ABD
29
Thus, the maximum area of
4
29 14'1
l_~]
ABCD equals 3 + 4 =
T-4. Note that of any five consecutive numbers k, k + 1, k + 2, k + 3, k + 4 at most two of them can be in S. If
three were in S, either two of them would be adjacent and would differ by l, a square, or they would be
k,k + 2,k + 4. But k + 4 and k differ by a perfect square. Apply this lemma to the subsets {I, ... , 5},
{6, ... , 10}, {11, ... , 15}, {16, ... , 20}, and {21, . .. , 25}. At most two of each set can be in S, so at
most 10 of the elements of {l, 2, ... , 25} can be in S. The set {1, 3, 6, 8, 11, 13 , 16, 18, 21, 23} clearly
satisfies the conditions, so the answer is [IQ].
T- 5. There are [ 1: ) ways to choose 4 rectangles on the left. For exactly one
rectangle to be coincident with one on the other side, we must choose one of
the four possible X's and that can be done in (:) ways. We must choose 3
g '
X
blank rectangles and that can be done in (;) ways. Answer:
241
2
Alternate solution: There arc ( ': J ways to choose 4 rectangles on each side of the dotted line. Letting
lower case letters stand for the rectangles, let A = {a, x, y, z} be the rectangles chosen on one side and
B = {a,u, v,w} be the rectangles chosen on the other. We need x, y,z, u, v, w to be distinct. There
are 12 choices for a, ( 1} Jways to choose x, y, J

and z, and (; ways to choose u, v, and w. Thus,
224
495
T-6. Consider the cube in the diagram to the right
I
and consider the case where t > -
2· Let M, N,
N K B(l/2,1,t)
M
and K be the names of the faces. Unfold the
A (I, 1/2, t)
r+-----ac-----t-~- y
cube in the two ways shown below.
N K
C 1/2 1- I I
M K ,--
A B ,,' B
------ ------ 1/2 ,,
,,
I t
,
I- t ,
,,
1/2 1/2
A~
I
M
Fig. I Fig. 2
In Fig. I the distance between A and B is clearly 1. So, in Fig. 2 the distance must be I as well, but it also
equals -h( ~ - J t since ABC is a 45-45-90 right triangle and AC= ½+ (I - t). Thus,
..[2 (~- t J = 1 ➔ t = 3 - 2Ji. . If t < ½, then 1 - t would give the same two equal length paths.
242
Thus, I - 3-
- -
2
fl: -J2 -
= -
2
- I also works. If t = :;I , the distance from A to Bin Fig. I equals I, while in
~
Fig. 2 it equals J2. Thus, A and B can't lie in the center of a face. Answer: t = I~ ¥ I- or
T-7. P = 600 = 4a + 2c ➔ 300 = 2a + c. The area equals a
.!_(2a + 2c)Ja 2 - c 2
2
= (a+ c)Ja 2 - c 2 = kik ➔ I
a : ✓ a2 - c2
( a - c)( a + c) 3 = k 3. Since a, c, and k are positive integers,
I
I
~-~'~:_ _ _ _ _ ~-I
a - c = m 3 for some positive integer m.
C a C
3
From 300 = 2a+ c and a- c = m 3 , we obtain 3a = 300 + m 3 , giving a= !!!_ + 100. Let m = 3n
3
➔ a= 9n 3 + 100 and c = 100 - 18n3 . For n ~ 2, c < 0, so for n = I, a= 109, c = 82, giving
k = 573.
T-8. Observe that
1 + 2 + 3 + ... + 59
1 l2·3J
-+ - + l3·4J
- - + l4·5J.
- - +···+ l59-60J
-- = 885.
2 2·3 2 ·4 2·5 2 · 60 2
885
Thus, 2·X3·V = 24 885 = ( 233 ) ➔ x = 3(885) and y = 885. Thus, x + y = 135401.
243
T-9. Since in 6 days the surface area 6a 2 becomes 3a2 = { ./i, ·Ji l then in 6 days each side must be i/.Ji.
as long. To reach a volume of one-eighth of the original, each side must be one-half as long. Without loss
of generality, let the bar be a cube of side l. Consider the linear function where edge y is a function of
.
time/. Att=0,y= I an d at/= 6,y= 1/ yL.
,:2 T1esopeo
1 I f th.1s 1·me1s
. ~
l- an d 1tsequat10n1s
· · · i2
6y2
y- 1= r :!! ], .
l Let y- 112 and so Ive fon obJaining I = 6 I + :i.'21 days. This isaboutl 0.24 days.
T-10. The conditions given are equivalent to the condition that a word not start or end with more than one
consonant and not have a consecutive string of three or more consonants. We'll count the number of
possible n-letter words ending in a vowel V , one consonant O , or two consonants T . We include the
n n n
words ending in two consonants because although they are not ARMLovian words, they generate
ARMLovian words by the addition of a vowel at the end.
Note that V1 = 1, 0 1 = 0. Also, v; = 4, i.e., AA, LA, MA, and RA, 0 2 = 3, i.e., AR, AM, and AL, and
T2 = 0 since no word can end in two consonants. In general, V11 = Vn- 1 + 0 n- 1 + Tn- I because adding
a vowel to the end of a word ending in a vowel, l consonant, or 2 consonants will produce an ARMLovian
word. Or one could say that removing a vowel from the end of an ARMLovian word will yield a prefix
ending in 0, 1 or 2 consonants.
In addition, 0
11
= 3Vn- I since adding any one of three consonants to the end of a word ending in a vowel
will produce an ARMLovian word. Finally, T
11
= 3 · 0 11- 1 since any one of three consonants can be added
to the end of a word ending in a consonant. Thus, we can generate the following values:
Continuing in this fashion we obtain:
V
4
= 28 0
4
= 21 T = 36
4
V
5
= 85 0
5
= 84 T
5
= 63
v6 = 232 06 = 255 T6 = 252
v7 = 739 07 = 696 T7 = 765
The number of acceptable seven-letter words is V7 + 0 7 = 739 + 696 I 14351.
244
Alternate solution: Let's determine the number of words by the number of A's in the word:
7 A's 1 word, namely AAAAAAA.
6 A's 21 words since there are 3 choices for the consonant and it can go in one of 7
places. Thus, 3( 7 c 1) = 3 · 7 = 21.
5 A's 171 words. There are 3 · 3 choices for a consonant and the number of choices for their
position is 7 C2 - 2 since they can go anywhere except together in the first two positions or
together in the last two positions. Thus, 9 · ( 7 C 2 - 2) = 9 · 19 = 171 .
4 A's 594 words. There are 33 choices for the three consonants. They can't be together in the first
two positions, together in the last two positions, or be three in a row. lf the consonants were
3 in a row, they could occupy positions I - 3, 2 - 4, ... , 5 - 7 giving 5 possible
placements. lfthey occupied the first two positions but not the 3rd , there would be 4 possible
placements of the consonants. Similarly if they occupied the last two positions. Thus, there
are 5 + 4 + 4 = 13 positions the consonants cannot occupy. This gives
27 · ( 7 C3 - 13) = 27 · 22 = 594 ARMLovian words with 4 vowels.
4
3 A's 648 words. There are 3 ways to choose the consonants. Once chosen we have the following
possible words:
AccAccA, AcAccAc, AccAcAc, cAccAcA,

cAcAccA, cAAccAc, cAcAcAc, and cAccAAc.
This gives 8 · 81 648.
There are no ARMLovian words with 2 or fewer A's since that would put 2 consonants together at the
beginning or end, or would give a run of 3.
Thetotalnumberofwordsis648+594+ 171 +21 + l 1435.
245
ARML Power Question - 2003: It's Those Rahhits Again!
A set of positive integers {x 1, x 2 , ... } is called a Fibonacci set if x 1 < x2 and x 11 = x 11 _ 1 + x 11 _ 2 for all n > 2.
Wc say that {a. h} is a subset of a Fibonacci set Fifa and h are distinct but not necessarily consecutive members
of F. Notice that any two-member set {a, b} of positive integers is a subset of at least one Fibonacci set. Since any
Fibonacci set is determined by specifying its two smallest members, list just the first two members. Thus,
write { I, 2.... } for { 1, 2, 3, 5, 8, 13, .... } .
l. a) Give an example of a set of positive integers {a, b} that is a subset of only one Fibonacci set.
b) For how many Fibonacci sets {x 1, x 2 , ...} is it true that x 3 = 2003?
2. a) Find all Fibonacci sets that have {8. 144) as a subset. Specify each one by listing its two smallest
members.
b) Let a and b be positive integers with 0 < a < b. Prove that {a, b} is a subset of more than one Fibonacci
set if and only if 2a < b or 2a > b.
3. Prove that no Fibonacci set has {60, 117, 174) as a subset.
4. Compute the number of Fibonacci sets where x

11
= 2003 for n ~ 2.
5. Prove that any Fibonacci set has infinitely many subsets of the form {a, a+ d, a+ 2d}.
6. Prove that no Fibonacci set has a subset of the form {a, a + d, a+ 2d, a + 3d}.
246
ARML Pmver Question - 2003: It's Those Rabbits Again!
A set of positive integers {x 1, x 2 , ... } is called a linear sci if there is a positive integer d for which
xn = xn- 1 + d for all 11 > I.
7. a) Prove that the Fibonacci set [7, 11, ... } and the linear set (8, 23, ... ) are disjoint.
b) Is it possible for a Fibonacci set and a linear set to have members in common, but onlyfinilelr many'?
Prove your answer.
8. a) Prove that two Fibonacci sets can have exactly one member in common.
b) Prove that two Fibonacci sets can have exactly two members in common.
9. a) Prove that whenever two Fibonacci sets have more than two members in common, then the set of
common members is itself a Fibonacci set.
b) Prove: Given any Fibonacci sets F 1, F2 , ... , F11 , there is a positive integer that does not belong to
any of them.
I 0. Can the set of positive integers be partitioned into Fibonacci sets? This means finding infinitely many pair-
wise disjoint Fibonacci sets whose union contains every positive integer. Prove your answer.
247
I. a) Without loss of generality assume that a< h. The simplest example is, of course, { I, 2, ... } .
b) Let x 1 be any integer between I and 1001 inclusive and let x2 = 2003- x 1. For all of these sets
{x 1, 2003 - x1, .• .} the third element will be 2003. There are I 00 I such sets, ranging from {I, 2002, ... } to
{ 1001, 1002, ... } .
2. a) There are 9 such Fibonacci sets: { I, 2, ... }, (2, 3, ... }, {3, 5, ... }, {5, 8, ... }, (8, 13, ... :.
(8, 24, ... l. {8, 68, ... }, {8, 136, ... }, and [8. 144, ... i.
Justification: Let x be the term that immediately follows 8 in any Fibonacci set F that contains 8. Then the
members of F following 8 must be x, 8 + x, 8 + 2x, 16 + 3x, 24 + 5x, 40 + 8x, 64 + l 3x, .... Since
x > 8 , 64 + 13x > 168 and we need only consider the first 6 expressions. Setting each in tum equal to 144
we have x = 144, 8 + x = 144, 8 + 2x = 144, 16 + 3x = 144, 24 + 5x = 144, and 40 + 8x = 144.
These yield respectively x = 144, 136, 68, 42 + 2/3, 24, and 13. Thus, 16 + 3x can't serve as a term, giving
at least 5 Fibonacci sets containing both 8 and 144. These are {8, 144, ... }, (8, 136, ... : .
(8, 68, ... }, {8, 24, ... }, and {8, 13, ... i. In the first four sets, 8 is less than half of x, so 8 and x are the
smallest members of F. In the fifth set where x = 13, 8 can have predecessors and F could begin with two
consecutive terms chosen from the standard Fibonacci set { 1, 2, 3, 5, 8, 13, ... } .
b) First, if b -t:- 2a , then either b < 2 a ➔ b - a < a, in which case let F = {b - a, a, b, ... }, or
2a < b ➔ a< b - a, in which case let F = {a, b - a, b, ... }. In other words, b < 2a means that there is
room for a to have a predecessor and 2a < b means that there is room for a term between a and b. In neither
case are a and b the two smallest numbers. Conversely, we will show that if a and bare not the two
smallest numbers, then b -t:- 2a . If F contains a member m between a and b, then 2a < a + m ::; h ➔
2a < h. If F does not have members between a and h, but does have a member x smaller than a, then
b =x+a < 2a ➔ b < 2 a. So, if a and b are not the two smallest numbers, then b -t:- 2a.
3. Referring to the proof in (2b) we cone Jude that because 117 is more than half of 174, no Fibonacci set that
contains 117 and 174 can have members between 117 and 174. Thus, the member that immediately precedes
117 must be 174 - 117 = 57 < 60.
248
4. If x 2 = 2003, then x 1 = I, 2, ... , 2002 all work, giving 2002 sets.
If x 3 = 2003, then we know from ( I b) that there are I 00 I sets.
If x4 = 2003 = 2x2 + x 1, we obtain { 1, I001, ... }, [3, 1000, ... }, ... {667, 668, ... } for a total of
334 sets.
If X5 = 2003 3x 2 +2x 1,weobtain {l,667, ... }, {4,665, ... }, ... , {400,401, ... J fora total of
134 sets.
If x 6 = 2003 = 5x2 + 3x 1 , we obtain {l, 400, ... }, {6,397, ... }, ... , {246,253, ... } for a total of
50 sets.
If x 7 = 2003 = 8x2 +5x 1 , we obtain {7,246, ... }, {15,241, ... }, ... , {151,156, ... } fora total of
19 sets.
If x 8 = 2003 = 13x2 + 8x 1 , we obtain {5, 151, ... }, {18, 143, ... }, ... , {83, 103, ... } fora total of
7 sets.
If x 9 = 2003 = 21x2 + 13x 1, we obtain {20, 83, ... } and {41, 70, ... } for a total of 2 sets.
If x 10 = 2003 = 34x2 + 2lx 1, we obtain only {29, 41, ... }.

If x 11 = 2003 = 55x2 + 34x1 , we obtain only {12, 29, ... }.
Justification: Each equation for xi in terms of x 1 and x2 is a linear Diophantine equation, solvable by
writing x 1 and x 2 in terms of a parameter t. For example, from 2003 = 8x2 + 5x 1 we obtain
xI = 400 + 5 -
3
x2 - 7
3x
= 400 - x2 +
3- 3x
5
2 . Let 3 - 3x2 = 5m form an integer, giving
2m
xI = 400 - x2 + m and x2 = 1- m- 3 Let m = 3t fort an integer and then x 2 =I- 3t - 21 = l - St
and x 1 = 400- (1- 5t) + 3t = 399 + St. Thus, if x 1 = 399 + St and x2 = 1- 5t fort an integer, we find
solutions by choosing all values oft for which O < x 1 < x 2 , namely for O < 399 + 8 t < I - 5t which gives
-49.875 < t < -30.615 ➔ -49 :::; t :::; -31 . Since there are 19 values oft generating proper values
(x 1, x 2 ), there are 19 solutions. Experience with linear Diophantine equations teaches that the x I values
increase by the coefficient of x 2 and the x 2 values decrease by the coefficient of x I and this would provide a
quicker way to generate solutions once one has been found. For example, once we discover that the first
solution to 2003 = 8x2 + 5x 1 is p, 246) then we could write x 1 = 7 + 8k and x2 = 246 - 5k, determine
the appropriate values of k and find the number of solutions.
249
Interestingly, the x 1 in the last ordered pair for x11 is the x_7 value for the first ordered pair for x 11+ I and the
x 1 value for the first term for x I equals x 2 - x 1 using the values for the last term in x11 . That would
11+
imply that x 12 = {17, 12), a contradiction since that makes x 1 > x 2 .
Adding, we find there arc 3551 Fibonacci sets with x

11
= 2003 for 11 > 2.
5. Given any four consecutive members of a Fibonacci set, the third is the average of the I s t and 4 th . In other
a+ (a+ 2x) .
words, given a, x, a+ x, a+ 2x, we have a+ x = 2
. This says that x., x. 2 . x. '" form an
l 1+ . z+.,
arithmetic sequence for any Fibonacci set for any i ::,: I.
6. Suppose there is a four term arithmetic subsequence of a Fibonacci set. Call it { a, a + x, a + 2x, a + 3x} . By
(2b) if twice the smaller is greater than the larger, then there are no intervening terms. Since
2( a+ 2x) > a+ 3x and 2( a+ x) > a+ 2x, then the elements a+ x, a+ 2x, and a + 3x have no intervening
terms. Thus, they are consecutive elements of a Fibonacci set, making ( a + x) + ( a + 2x) = a+ 3x, and
hence a = 0. But this is a contradiction, showing that a four term arithmetic sequence is impossible in a
Fibonacci set.
7. a) Modulo 15, the members of the Fibonacci set are periodic and are congruent to 7, 11, 3, 14, 2, l, 3,
and 4. Modulo 15, the members of the linear set are all congruent to 8. Therefore, there are no members in
common.
b) lt is impossible for a Fibonacci set and a linear set to have a member in common without having
infinitely many members in common. To see why, suppose that dis the positive difference between
consecutive members of the linear set and that all members of the linear set are congruent to k modulo d
where k < d. If the members of the Fibonacci set are calculated modulo d (listing just remainders), the list
will be periodic. This is because there are only d different remainders, hence at most d 2 pairs of consecutive
terms that can occur in such a list. This implies that some pair must appear for a second time. Suppose that
the members are x 1, x2 , x 3, ... , and that xi is congruent to x 11 , and that xi+ 1 is congruent to
x n+ 1 (modulo d) for some indices i < n . If 1 < i, then x.1- 1 is also congruent to x n- I because
x. 1 = x. 1 - x. and x 1 = x 1 - x (modulo d). It follows that the initial pair of x-values must,
1- 1+ z 11- n+ n
in fact, repeat, and so too must every x-value in the list. Thus, if the linear value k appears once in the list,
then it appears infinitely often.
250
8. a) If x 1 = y 1 and x2 < _v 2 < x3 , then the Fibonacci sets {x 1• x2 , . .. } and {.vi' _v2, ...} have only x 1 in
common. In fact, x. < v. < x. 1 < v. 1 holds for all indices I < i. This can be shown by induction.
I 'I I+ '1+
Notice first that x 1 = _v1 and x 2 < _v 2 imply that x 3 < _v3 and that initializes the induction. The inductive
step is also a consequence of the Fibonacci recursion, namely x. < v. < x. 1 < v. 1 implies x. 2 < v. ')
I'/ 1+ 'I+ 1+ •1+~
and v. 1 < x. < y. < x. 1 implies v. 1 < x. 2 . Combine the data to obtain
'I- I I I+ 'I+ 1+
If Fibonacci sets {x 1, x 2 , ...} and {_v 1, _v2, ...} satisfy ·\ < y 1 < xi+ 1 < y 2 < xi+ 2 for some index i, then
the preceding shows that the sets are interleaved-neither set has two members between consecutive
members of the other set.
b) Given any Fibonacci set {x 1, x 2 , ... } • let y 1 =\ and _v2 = xi for any positive index i and any indexj
that is greater than i+ 1. The Fibonacci sets {x 1.x2 , ... } and {_vl'y2 , ...} have only _v1 and _v2 in common
because they interleave x . < y 3 < x. 1 < v4 < x . , after the two common members. The first
J J+ . J+ -
inequality x. < v 3 is obvious, the second follows from v3

j • •
= x.I + x..f < x. 1 + x.
J- J
= x. 1 , the third
.1+
follows from xi-I< y 2 and xi < y 3 , and the fourth follows from _v2 = xi and y 3 <xi+!·
9. a) LetaandbbethesmallesttwointegersthattheFibonaccisetsF = {x 1,x2, ...} andG = {_vl'y2, ...}
have in common with a< b. If a and b arc consecutive members of both F and G, then F and G share all
their members beyond band the desired conclusion follows easily. Suppose, therefore, that a and b arc not
consecutive members of F which implies that a< x. 1 < x.
1- I
=b= y .. There are two cases: x. 1 < v. 1
j /- '.J-
. J- I
Or V. < /- 1.
X.
In the first case the Fibonacci recursion implies that }j < xi+ 1 < y i + 1 < xi+ 2 < Y j + 2 .
In the second case the recursion implies that x. < v. 1 < x. 1 < v. 2 < x. 2 .
I 'J+ 1+ 'J+ I+
In either case, F and G have only a and b in common. It follows that F and G have more than two members
in common only if the two smallest shared members are consecutive members of both sets, in which case the
Fibonacci set generated by the two smallest members equals the set of shared members.
251
b) Herc the following fact is needed: If {x 1, x2 , ...} and {YI' y 2, ...} arc Fibonacci sets and if there are more
than two •v-values between x.I and x.t+ 1, then i = I. In other words, only the initial interval of a Fibonacci
set can include more than two members of any other Fibonacci set. The reason is that the first two of those
y-values would otherwise be greater than xi- I and xi respectively, hence the third y-value would have to be
greater than x. 1 by the Fibonacci recursion.

1+
For any Fibonacci set, the recursion x 11 - xn- 1 = x11-.!.~ shows that each member is equal to the difference
between the next member and the one following. Thus this "gap" is an increasing function of position. No
matter how large n is, the gaps in F1 will eventually be larger than 2n. And as shown previously, each gap
can contain at most two members from each of the other Fibonacci sets, implying that there are integers in
that gap that belong to none of the other F..
I
I 0. The set of positive integers can be partitioned into infinitely many Fibonacci sets. Let F 1 = {I, 2, 3, 5, ...} ,
F2 = {4, 6, I 0, 16, ... } , and F3 = {7, 11, 18,29, ... } . Notice that each ~ for i = 1, 2, or 3 is interleaved with
the other two and therefore the three sets are mutually disjoint. Assume inductively that the Fibonacci sets
F1,F2, ... ,F11 have been chosen by the following procedure: Let m be the smallest positive integer that is not
a member of any of the F.. Thus, every F. has at least one member that is smaller than m. In particular,
l I
m - I is a member x. of a Fibonacci set F . Let

I C
p =I+ x. 1 . The desired F I is defined by its first two
1+ n+
members m and p and it is interleaved with every previously defined Fibonacci set because the interleaved
sets F1, F2 , ... , F each have one member between x. and x. 1.

11 l 1+
252
1-1. 0 is the center of two concentric circles as shown. AB and CD arc chords of
the larger circle and they intersect at Eon the smaller circle. CD intersects
the smaller circle at F. If m AC + mE F = 83 °, compute mDB.
1-2. A triangle with sides a ::; h ::; c is log-right if log( a2 j + lo~ b2 ) = lo( c 2 j. Compute the largest possible
value of a in a right triangle that is also log-right.
1-3. Compute the least prime p such that p - I equals the difference of the squares of two positive multiples of 4.
1--4. Let Nbe a three-digit number that is divisible by 3. One of N's digits is chosen at random and removed.
Compute the probability that the remaining number is divisible by 3. (Note: if a digit is removed from 207
we obtain either 20, 07 = 7, or 27; ifa digit is removed from 300 we obtain either 30, 30, or 00 = 0.)
1-5. There are n triangles of positive area that have one vertex at A(O, 0) and the other two vertices at points
whose coordinates are drawn independently and with replacement from {O, I, 2, 3, 4}. Compute n.
1-6. Let n be an integer. Of all fractions .!_, the fractional part of J123456789 is closest to one such fraction.
n
Compute that value of n.
I-7. Compute the largest factor of I 00I00100 I that is less than I 0,000.
1-8. The graph of f(x) = x 4 + 4x 3 - 16x 2 + 6x - 5 has a common tangent line at x = p and x = q.
Compute the product pq .
253
I. 97°
2. Ji
3. 113
19
4. - = .38
50
5. 256
6. 9
7. 9901
8. -10
254
1-1. Let mAC = a, mEF = b, and mDB = x. Then

180- h
mLEOF = h -----, mLFOB = 180 - h -----, 111LFEO = - - - But
2
a+x
111LFEO=-------, 180-b=a+x-----, x=180-(a+b).
2
Since a+ b = 83°, x = 197°1.
1-2. From log a2 + log h2 = log c 2 we obtain ah= c. Substituting into a2 + b 2 = c 2 yields ( ab) 2 = a2 + b2
-----, a2 h 2 - h2 = a2 . S.mce a <

_ b , rep Iaemg
. b b y a y1e " 11 owmg
. Id s tI1e 10 . mequa
. 11ty:
· a2 a 2 - a2 <
_ a2 .
Thus, a4 :::; 2a 2 -----, a2 : :; 2 -----, a:::; ./i.. Thus, the largest value of a is I.Ji,I. If a= b = ./i., then c = 2
and both the Pythagorean Theorem and log a2 + log b2 = log c 2 are satisfied.
1-3. p - I = (4k) 2 - (4m) 2 = 16(k 2 - m 2 ). If k and 111 are odd, their difference is divisible by 8, making
p = 128t +I. The least prime is 257 obtained fort = 2, k = 5 and m = 3. If one is odd and the other
even, then their difference is odd, making p = I 6(2t + I) + I and the least prime is 113 obtained when
t = 3, k = 4 and m = 3. If both are even, then p = 64 t + I and the least prime is 193 obtained
when t = 3, k = 4 and 111 = 2. Thus, the least prime is I1131.
1-4. There are 300 three-digit numbers times 3 digits = 900 cases. If the numbers are written as 3xx, 6xx, or 9xx
and one of 3, 6, or 9 is removed, then since there are 34 numbers divisible by 3 from 00 to 99, there are
3 · 34 = 102 possible favorable outcomes. If the numbers are of the form xOx, x3x, x6x, or x9x. then
removing 0, 3, 6, or 9 yields a number divisible by 3 in the 30 cases from IO to 99, giving 4 · 30 = 120
favorable outcomes. Similarly for xxO, xx3, xx6, and xx9. Thus, there are I 02 + 2 · 120 = 342 favorable
1 1ty o f -342 = ~- 9 .
. Id.mg a pro b ab·1·
outcomes y1e
900 50
1-5. There arc 5 · 5 = 25 points that can be fon11ed using the coordinates {O, I, 2, 3, 4}. Eliminate A(0, 0) and
from the remaining 24 points, choose 2. This can be done in 24c2 = (24 · 23)/2 = 276 ways. Pairs such as
(0,1) and (0, 2), (I, I) and (2, 2) are collinear with A(0, 0) and must be eliminated. Thus, subtract
3 ·4 c2 = 18 points to eliminate those pairs of points on x = 0, y = x, and y = 0, but we must also
eliminate the two pairs (I, 2), (2, 4) and (2, I), (4, 2) for a total of276- 18 - 2 = l256ltriangles.
Note: if the set was (0, 1, 2} there are 16 triangles. Is there a pattern here?
255
1-6. Let x = J123456789. Then -'10¾ = Jo.0123456789 "" I

10
2
10
3
2 + - 3 + 4 + .... To evaluate this, set
10
I 2 3 4
S =- +- +- +- + ··· , then divide by IO giving
10 2 10 3 10 4 105
S I 2 3 4
-=-+-+-+-+···
IO 10 3 104 105 10 6
I
Thus, .2_ S = _I_ + _I_ + _I_ + ... = ...lillL = _!_ ➔ S I . Th"1s gives
= 81 .
10 10 2 10 3 10 4 I _ _!_ 90
10
x I 100000 I
- "' - ➔ x "" - - - = 11111 + - . Hence, n = [2].
105 9 9 9
Alternate solution: One could observe that 11111 2 = 123454321 and write
I J123456789 + 11111 22222

= ""--""9.
2468
J123456789 - 11111 123456789 - 11111 2
1-7. 1001001001 = 1001 · 106 + 1001 = (1001)(10 6 + 1) = 1001(10 2 + 1)(104 - 10 2 + 1) = 1001 · 101 · 9901
= 7 · 11 · 13 · 101 · 9901. Since no combination of 7, 11, 13, and 101 can generate a factor greater than 9901
but less than I 0,000, the answer is 1990 I I.
1-8. Let the equation of the common tangent be y = mx + b. Consider the function
g(x) = x 4 + 4x 3 - 16x 2 + 6x - 5 - (mx + h). It must have double zeros at x = p and q. Thus,
x 4 + 4x 3 -16./ +(6-m)x- (5+ b) = (x - p) 2 (x -q/ =
x 4 - 2(p + q)x 3 +(/ + 4pq + /)x 2 -2(pq 2 - /q)x+ / / . Setting the coefficients of x 3 equal
gives -2(p + q) =4 ➔ p+q = -2. Setting the coefficients of x 2 equal gives p 2 + 4 pq + q2 = -16 ➔
(p + q) 2 + 2 pq = -16 ➔ 4 + 2 pq = -16 ➔ pq = - IOI. In this case, the common tangent is
y = 46x - 105 and letting p > q, we have p = -I + /ti and q = -I -/ti.
256
R 1-1. Mary insists on sitting next to Bob and Bob insists on sitting next to Jane. Compute the number of different
ways that Bob, Mary, Jane, and 3 other students can sit in a row.
Rl-2. Let T = TNYWR. AD is a diameter of circle O and has

C
T - -
length 3. BA, CD, and BC are tangent to circle Oat A, D,
and E, respectively. If BC = 18, compute the area of ABCD. A --------, D

0
Rl-3. Let T = TNYWR. The circle x 2 + y 2 =T intersects the positive y- and x-axis at A and B respectively.
If the line y = 3 intersects AB at C, compute the larger of AC or BC.
R2-l. A box 5 by 8 by 12 is placed in the comer of a cube of side

14 so that three faces of the box are coincident with three B
faces of the cube. Compute the distance from comer A to

comer B.
R2-2. Let T = TNYWR and let K =T - 7. If log 16 K, logK 16, and x form a geometric progression,
computex.
R2-3. Let T = TNYWR. In AABC, mLB = 90°, and D lies on CB so that AD bisects LCAB and
CD= T ·(DB). Compute cos 2 LCAD.
257
Relay #1:
Rl-1. 48
Rl-2. 144
Rl-3. 9../2.
Relay #2:
R2-l. 11
R2-2. 8
9
R2-3. - = .5625
16
258
Solutions to the ARML Re/av #1 - 2003
RI-I. Treat Mary, Bob, and Jane as a single bloc and imagine that there arc 4 places to sit. Mary, Bob, and Jane
can choose any one of 4 places and Mary and Jane may switch, giving 2-4 C 1 = 8 possible arrangements.
The other students may be seated in 3 ! = 6 ways, giving 6 · 8 = @ID different ways to sit.
Rl-2. Since BE = AB and CE = CD, then AB+ CD = BC= 18. The area
I T
oftrapezoid ABCD is ½(AD)(AB + CD) - • - • 18 = 3T. Since
2 3
T = 48, the answer is 3 · 48 = I1441.
Rl-3. The equation of AB is x + y = Jr and since y = 3 we have A (0. ✓r /

C (x, 3)
c(./r -3,3). Thus, AC= J(.Jr- 3) 2 + (3-./r) 2 =

1Jr-3j'2" and BC=J((Jr-3)-.Jr)2 +(3-o)2 3i2.
Since T = 144, then AC is the longer and AC = [9/2. [.
Solutions to the ARML Relay #2 - 2003
R2- l. The distance from A to B is simply the diagonal of a box whose sides are 14 - 5 = 9, I 4 - 8 = 6,
and 14- 12 = 2. The diagonal has the same length no matter how the smaller box is oriented. That length
is J2 2+6 2 +9 2 = Jui = [I]_
CD AC C
R2-3. By the Triangle Angle Bisector Theorem, since DB = T, then AB = T.
Tx
v I 2 1 2 T+l
Thus cos 28 = ...;;_ = - ➔ 2 cos 8 - 1 = - ➔ cos 8 = --. Ty
' Ty T T 2T
.
Smee T = 8 cos 2 8 = ~·
- .
' 16
y
259
Note: Pass from position 1 to 8 and from position 15 to 8.
I. Let a, b, and c be positive integers such that a+ b + c = 7. Compute the smallest possible value
of a2 + b 2 + c 2 .
2. Let T = TNYWR. A farmer has T animals. Some are horses, the rest are chickens. There are a total of 46
legs. Compute the number of horses.
3. Let T = TNYWR. P and Qare right circular cones with radii of a and b respectively and altitudes
of m and n respectively. The volume of Q is T times that of P and b = ~ Compute : .
4. Let T = TNYWR. Let u(x) denote the units digit of the integer x. Compute the value of
5. Let T = TNYWR. Let K = T + l. Consider the number N = ( 2x X3Y) where x andy are non-negative
integers. Compute the smallest value of N such that N has exactly K positive factors.
6. Let T = TNYWR. In the figure all sides are equal and the angle between each
pair of consecutive sides is 90°. If the numerical value of the figure's area
exceeds its perimeter by T, compute the perimeter of the figure.
7. Let T = TNYWR. If log 2 T = a+ b Iog 2 3 and log 6 T = b + b Iog 6 2 , compute a+ b .
8. Let a and b be the two integers you will receive and let K = Ia - b I- For x in radians, the graphs of y =;
I I
and y = si~ K1t.x) intersect in N points. Compute N.
260
15. A point (a, h) is chosen at random from the region bounded by the coordinate axes and the lines x = 10 and
y = 10. Compute the probability that a+ h S 8 .
14. Let T = TNYWR and let K be the sum of the numerator and denominator of T when Tis written in lowest
terms. Compute the solution to Jx + K = 11 - /;.
13. Let T = TNYWR and let K =T- 13. The radius of a cylinder is Kand the volume of the cylinder is
numerically equal to the sum of its lateral area and the area of the two bases. Compute the cylinder's height.
12. Let T = TNYWR. If !OT - 7 + i is written as a+ bi where a and bare real numbers and i =
T + 2i
.r-i.,
compute a.
11. Let T = TNYW R. Compute the smallest integer solution to .:x-2 + Tx + T =1 .
I 0. Let T = TNYWR and let K = T + 246 . If K is the first term of an arithmetic sequence and 2003 is the tenth
term, compute the common difference.
9. Let T = TNYWR. The large square has an area of T. Each of the small
squares shares a vertex with the large square and has sides of length of 4. s: I
I
Compute the area of square ABCD. c:
8. Let a and h be the two integers you will receive and let K = Ia - b I- For x in radians, the graphs of y = {
and y= lsi~K7tX)I intersectinNpoints. ComputeN.
261
I. 17
2. 6
3. 24
4. 20
5. 576
6. 144
7. 6
8
15.
25
14. 16
13. 6
12. 8
11. -7
10. 196
9. 36
8. 1800
262
Solutions to the A RML Super Relay - 2003
1. Let a::;; h ::;; c. There are only 4 triples satisfying a+ h + c = 7, namely (1, 1, 5), (I, 2, 4 ). ( I, 3. 3 ).
and (2. 2, 3 ). The smallest occurs when ( a, b, c) = (2, 2. 3) and that triple gives a2 + h 2 + c 2 = [I.
2. Since h+c =T and 4h+2c = 46,wcobtain h= 23-T. Since T= 17, then h= [6].
3. From !rch 2 n = 1J.!.)rca 2m and h= 5!.... we obtain (

3 l3 2' l2
!:!..J\ = Ta 2m ➔ n= 4Tm ➔ .!!.... = 4T.
m
n
Since T = 6, then - = [Ml.
m
4. If n is a non-negative integer, then u( 9n j is I if n is even and 9 if n is odd. j

Thus, u( 9n+ 1 + u (9n) =
j
1 + 9 or 9 + 1 = 10. Similarly, u( 7n is I ifn = 4k, 7 if n = 4k +I, 9 if n = 4k + 2. and 3 if n = 4k + 3.
Thus, u(1n)+u(1n+Zj = 1+9,9+1,7+3,or3+7 = 10. Thesumis[2Q]anddocsnotrcquircT.
5. N = 2x · 3Y has (x + l)(y + I)= K = 21 factors. Thus, (x, y) = (0, 20), (2, 6), (6, 2), and (20, 0).
N = 2x · 3Y is least when (x,y) = (6,2) ➔ N = 15761.
6. If x is the side of the figure, then the area equals 5./ and the perimeter equals l 2x giving 5. .2 - I 2x - T = 0.
Since T = 576, then 5x 2 - 12cr - 576 = 0 ➔ (5x+ 48)(x -12) = 0 ➔ x = 12 ➔ perimeter= 11441.
a+log_3h ab b+log.2h h b h b b 2h b
7. T = 2 ~1, = 2 • 3 . Also, T = 6 6 = 6h • 2 = 2 • 3 • 2 . From 2° · 3 = 2 · 3 , we
obtain a= 2b ➔ a+ b = 3b. Since T = 144 = 2 2b • 3b = 16 · 9, we obtain b = 2 making 3b = [§].
263
15. The area under x + y =8 . I

2
.. I 32
1s (8)(8) = 32. The probab11Ity equa s 100 =
187
l1ij·
2
14. Squaring yields x + K = 121 - 22/; + x. Isolate and square obtaining x = l(121-K)
22 l(2828)2 = ITfr.
1 2 2K ..
13. rrK~h = 2rcK + 2rrK.h ➔ h = - - . Assummgmtegervalueswefind (K,h) = (3,6),(4,4),and(6,3)
K-2
making h = 3, 4, or 6. Since K = 3, then h = [6].
(10r 2 -1r+2)+(14-19r); JOT 2 - 7T+ 2

lOT - 7 + i T - 2i
12. -------= Thus, a= - - - - - - Since T = 6, a = [BJ.
T+2i T- 2i T2 + 4 r2 + 4
-r±/r2 -4(T-l) -r±lr-21

11. X=
---"-----'-----'- = = -1 or I - Twhether T 2". 2 or T < 2. Since T = 8,x = t]].
2 2
10. K = -7 + 246 = 239. Let d be the common difference ➔ 2003 =K + 9d = 239 + 9d ➔ d = I1961.
9. Large square's side = ll ➔ area of ABCD = (../i - 2 · 4r = (14 - 8) 2 = 00 since T = 196.
8. The period of y I I
= sinK7tX is ~ = ..!_. If the graphs intersect at
Krr K
(m, n), then n s I ➔ !!!.. s
K
I ➔
m s K. Note that the line intersects the curve twice in each period. There are ....!5.._ = K 2 periods so there
11 K
arc 2K 2 points of intersection. Since the answer from #7 is 6 and the answer from #9 is 36, then
K = 136 - 61 = 30 . Thus, the number of points of intersection is 2 · 30 2 = I18001.
264
ARML Tiebreakers 2003
I. Point D lies on circle Osuch that OD J_ AB. lf AD ~ 40. DC is an D
integer and DC < OD, compute the number of values of DC such that A B
C
the circle's diameter is an integer.
0
2. Let [ x] the greatest integer :S x. •

Compute the least solution to [x] = 2002
X
2003 . Write your answer in the
form 1 where a and b have no common factors.
3. Let N be an integer such that the product 41 · 43 · N can be written as the sum of 6 consecutive positive
integers. Compute the least value for the smallest of the 6 integers.
265
'1 AC 2 A c2 DC 2 + AC 2 AD 2 40 2
I. AC"'=DC-CE ➔ CE=-- DE=DC+--=-------
DC DC DC DC DC
There are 7 · 3 = 21 divisors of 40 2 and we seek all those divisors DC where DC < ¾DE ➔
I 40 2 ~ r: ?
DC < - • - - ➔ DC < 2u,,2 "" 28.28. Thus, DC can be a factor of 40- as large as 28, and so
2 DC
DC = I, 2, 4, 5, 8, I 0, 16, 20, 25. Answer: [2].
n+ h 2002 1
2. Let x =n+ h where n is an integer and O :s; h < I. Then -n- = 2003 ➔ h =- 2003 n. Thus,
2002 2002 2002(1- 2003)

-2002 :S n < -1. Choose n = -2002, making h = 2003 ➔ x -2002 + - -
2003 2003
-4,008,004
2003
3. Let x be the smallest of the six integers. Then x + (x + I) + ... + ( x + 5) 6x + 15 3(2x+5) =
4l ·43·N. SoNisdivisibleby3. IfNwere3,then2x+5 = 41 ·43 1763 ➔ X = 1758 =

- 18791.
2
266
Induction
(CU,.0f~:!tiill
p :DJ u:m i.f!J !D [}LfiJ O ~ □ C 11 ll l]l n~ ~ ~ ® ~
~ lf:td[~;; ., 'iJP~1u, { {~ s, Vllf~rl '~ji!} {irl@ ~ ',r,l ( ,

q
Twenty-five Point Affine Geometry ~ e,I
a:.::i..i
·~...,,,/7',n~ .-.✓/. .-,.'.· .,⇒

(())[/VCW/W' }},;ff-
Li-:J
>-
Ii. ~
~
~ .::=i.;i JI"'~
... (1) ~
Cl) E (2)
....~
0)0 §
! (1) t:<~1
..~
~
.EC,
I,,
.., ::a
C
:» ,.,:r-I C
a Chromatic
•
0,. 0
-· ,.-·
0
fl'
...tD Polynomials
...
:» 0
267
ARML POWER CONTEST: 1995 - 2004 267
1994-95 Color Transformations ___________________________________________________________________________________ _ 269
Solutions _______________________________________________________________________________________________ _ 271
1995-96 Induction ______________ _________________ _____________ _______ ____ __________________ ____________________ _________ 274
Solutions 276
Rook Polynomials__________________________________________________________________________________________ 280

Solutions 283
1996-97 Rotating Decimals__________________________________________________________________________________________ 286

Solutions 288
Regular Closed Linkages.________________________________________________________________________________ 290

Solutions 292
1997-98 Factorial Polynomials_____________________________________________________________________________________ 298

Solutions 300
Integer Geometry___________________________________________________________________________________________ 304

Solutions 306
1998-99 Unit Fractions________________________________________________________________________________________________ 309

Solutions 312
Chromatic Polynomials___________________________________________________________________________________ 315

Solutions 321
1999-00 Twenty-five Point Affine Geometry._________________________________________________________________ 324

Solutions 328
Square Sum Partitions.____________________________________________________________________________________ 331

Solutions 333
2000-01 Slides, Glides, and Rolides ______________________________________________________________________________ 336

Solutions 339
Pythagorean Triples._______________________________________________________________________________________ 34 2
Solutions 345
2001-02 Cevians.________________________________________________________________________________________________________ 350

Solutions 354
Insane Tic-Tac-Toe---------------------------------------------------------------------------------------- 359

Solutions 363
2002-03 Three Addition Problems-------······--------------------------------------------------·---------------- 369

Solutions _______________________________________________________________________________ . _______________ _ 373
Number Theoretic Functions--------------------------------------------------------------------------- 386

Solutions 390
2003-04 Errors in Mathematical Reasoning _________ ·····-------------------------------------·---------------- 394

Solutions _______________________________________________________________________________ . _______________ _ 397
Mathematical Strings _____________________________________________________ ·---------------·---------------- 399

Solutions 401
268
ARML Power Contest- November 1994 - Color Tran~formations
Color Transformations
The Definitions and Notation
Begin with a regular grid composed of squares, each colored black or white. For example:
Figure l
We begin changing this color according to the following rule: In one step we may either reverse the colors of all the
squares in one row or reverse all the colors in one column. So for example, starting with Figure l, in one step
we could get the coloring in Figure 2 by reversing row 2, or we could get the coloring in Figure 3 by reversing
column 3. If we took several steps, the resulting coloring might look quite different fonn Figure 1. See Figure
4, in which we've reversed row 2, then column 2, then column 4.
;W ~
" Figure 2 Figure 3 Figure 4
We use the notation rl to signify the act of flipping or reversing the colors in row i and g_ will refer to flipping or
reversing the colors in column i. We will record a sequence of flips in a flip string. So the flip strings
representing Figures 2, 3, and 4 above would be r2, c3, and r2-c2-c4.
One coloring will be called accessible from another if either coloring can be transformed into the other by a sequence
of flips.
The Problems
1. Compute the results of applying the following flip strings to the coloring of Figure 1, by drawing the resulting
colorings on your answer sheet.
1a. rl-c l-r2-c4 Ib. r2-r l-c4-c l I c. r3-c2-c3 Id. rl-rl-rl-c2-r3-c2-rl
le. r3
~~
2. Construct two different flip strings each of which turns Figure l into the
Figure 5.
Figure 5 Figure 6
3. Starting with Figure 1, construct a flip string which turns the four
squares in Figure 6 marked with an asterisk all white.
269
ARML Power Contest - November 1994 - Color Transformations
flUI
4. Starting with Figure I, construct a flip string which turns the four squares in Figure 7
marked with an asterisk all white.
Figure 7
5. In am by n grid, how many colorings (different colored grids) have the bottom row and the right-hand column
entirely white?
6. Can every coloring be transformed into such a coloring? Why or why not?
7. In a 2 by 2 grid. if the initial coloring contains exactly one black square, how many colorings are accessible
from it? (Proof not required.)
8. In a 2 by 2 grid, if the initial coloring contains exactly two black squares, how many colorings are accessible
from it? (Proof not required.)
9. Show that if we start with any five different colorings of a 2 by 3 grid, we can find two colorings among them,
each of which is accessible from the other.
10. Suppose you start with a grid which is larger than a 2 by 2 and has only one black square. Can you ever get a
coloring that is all black? Why or why not?
Given a coloring, we can try to eliminate as many blacks as possible, seeking a coloring with a minimum number
of black squares. We call this number the minimum of the coloring. For example, the coloring in Figure 8 has
a minimum of 0, since we can eliminate all the black squares by cl -c2-c4.
ll Figure 8
11. Show that all colorings accessible from one another have the same minimum.
12. A given coloring, together with all the those accessible from it, form a coloring class. As a result of problem
12. we can speak of the minimum of a class. What values occur as the minimum of a class for the following
grid sizes:
12a. 2 by 3 12b. 4 by 4 12c. m by n
270
ARML Power Contest - November 1994 - Color Tran~formations
The Solutions
~
la., lb., le. Id., le.
2. Two solutions are: c2-c3-rl-r2 and r l-r2-c2-c3. Many other examples are possible.
3. One possible answer is r2-c3.
4. The only flips that affect these squares are rl, r2, c 1 and c2. In the present state, rl will change two black
squares to two white squares (or zero black squares). lfthe two top right squares are both white, then rl will
tum them both black. If one of these squares is white and the other black, rl simply reverses their colorings.
In all cases. rl changes the number of black squares among the top four by an even number, so that the number
of black squares remains odd. It is not difficult to see that the same is tme ofr2, cl, and c2. Hence any
application of these flips, in any order, will result in an odd number of black squares among the top right four.
Since O is an even number, no flip string can ever result in O black squares and four white squares.
5. In am by n grid, there are 2""' colorings, since each of these mn squares may be colored either black or white.
The number of colorings described in the problem is only im- n<n -ii since there is no choice for the nth row or
the mth column.
6. The following algorithm will tum any coloring into a coloring with the bottom row and right-hand column
entirely white: First examine the bottom square of every column. If that square is black, flip the entire column.
After going through each column, the entire bottom row will be white. Now examine every row except the
bottom one. If the rightmost square in the row is black, flip the entire row. After going through each row
(except the bottom one), the right column of the grid will be all white.
7. and 8. Answer: 8 in each case. Following the line of reasoning described in #4 above, it becomes clear that for
a 2 by 2 grid, the parity (evenness or oddness) of the number of black squares does not change. Therefore a grid
with a single black square can only be transformed into one with oddly many black squares. A quick count
shows that there are eight of these: four with one black square and four with three. Similarly, a grid with two
black squares can only be transformed into one with evenly many black squares. These are the other 8 possible
colorings.
For a full proof, it would remain to show that each of the 8 odd colorings can be obtained from any other, as
can each of the 8 even colorings. This task can be accomplished by actual constmction.
9. We will show there are only four "classes" of colorings of a 2 by 3 grid. Any two colorings within one class are
271
ARML Power Contest - November 1994 - Color Tran.~formations
mutually accessible, and any two colorings in different classes are mutually inaccessible. Therefore, ifwe have
five colorings, at least two must belong to the same class, and so are accessible from each other.
First look at the first two columns of the 2 by 3 grid, and consider these four squares as a 2 by 2 grid in its own
right. We have seen in #7 and 8 that there are only two kinds of these grids: those with evenly many black
squares and those with oddly many. A 2 by 3 grid with evenly many black squares in the first two columns
can be transformed into one with no black squares in the first column, and a 2 by 3 grid with oddly many black
squares in the first two columns can be transfonned into one with a single black square, say, in the upper right
hand position. Furthennore, no 2 by 3 grid of the first class can be transformed into one of the second class.
Having transformed any 2 by 3 coloring as described above, we now look at its last two columns. Again, they
form a 2 by 2 grid in its own right, and so there are two classes it can belong to, according as the number of
black squares is even or odd (if it has evenly many black squares, flips can make it all white, and if it has oddly
many black squares, flips can give it a single black square, say in the upper center of the original grid).
Hence we can classify any coloring in two ways according to its first two columns, and in two ways according to its
last two columns. It is not difficult to see that these classifications result in four classes: odd or even for the
first two columns, and odd or even with the last two. Together with out initial observation, this proves the
assertion.
10. We have seen (questions #6 and #7) that a 2 by 2 grid with one black square cannot be turned entirely white.
We can extend this to any grid by considering the 2 by 2 "subgrid" which includes the single black square.
Considering this as a 2 by 2 grid in its own right, we see that if it has a single black square to begin with, it
must have oddly many black squares after a sequence of flips. The all-black coloring has evenly many black
squares (four of them) in this grid, and so is not accessible from the initial coloring.
11. If coloring Y can be transformed into coloring Z with k black squares, then any other coloring X accessible
from Y can also be transformed into Z, for example, by combining the flip strings effecting the transformations.
Thus the minimum of Y is the same as the minimum of any other coloring assessable from Y.
12. To solve this set of problems, we introduce the notion of a parity lattice. We have
already seen that if a grid has more than two rows or columns, then any 2 by 2 sub grid
always has either evenly many or oddly many black squares, no matter how we transform
the coloring. Thus we can affix a O or a I to every intersection of four squares, according
FIJI
as there are evenly many (indicated by 0) or oddly many (indicated by I) black squares among the four that
share the comer. We call this array, derived from any coloring, its parity lattice. An example, based on the
coloring of figure I, is shown above on the left. What we have just shown is that the parity lattice of a coloring
does not change as its rows or columns are flipped.
272
ARML Power Contest - November 1994 - Color Tran~formations
12a. The possible minimum values are O and I. A 2 by 3 grid has a I by 2 parity lattice, and there are four such:
00, 01, l 0, 11. Thus there are four classes of 2 by 3 colorings. A set of minimal colorings for the classes was
given in the answer to #9.
12b. The possible minima for 4 by 4 colorings are 0, I, 2, 3, and 4. Examples of colorings having these minima
are:
■1--r-
1 ,-71~ ■ I■
, I
__ ,,_I
i
._-=-i
An argument showing that these colorings do in fact have the stated minima, and that no other minima are possible,
follows from the following theorem:
Theorem: For an m by n grid with m = 2p, n = 2q, there is a class with Sas its minimum for each O :S S :Spq.
Proof: Rather than viewing the grid as an m by n table of squares, view it as a p by q table of non-overlapping 2
by 2 subgrids. Since the parity lattice of the table doesn't change when flips are applied, a 2 by 2 subgrid
containing an odd number of black squares will always contain an odd number of black squares, no matter how
-~
the coloring is transformed. In particular, it must always contain at least one black square. So if U and V are
the 2 by 2 subgrids below:
--·-7,-1
"
subgrid U subgrid V
Then a coloring composed of T copies of U and pq- T copies of V will have a minimum of T.
A 6 by 8 coloring with a minimum of 5:
LTI_~··-11
Observe that it is impossible for a class to have a minimum which is greater than pq, since any coloring of a 2p by
2q grid with at least pq + l black squares must have some row or some column which is more than half black
(this is not obvious, but can be shown, for example, by using the pigeon hole principle). By flipping this row
or column you reduce the number of black squares in the coloring. So no coloring with pq + l or more black
squares can be a minimum coloring of its class, and pq is in fact the "maximal minimum."
273
ARML Power Contest- November 1995 - Induction
Induction
The Problems
I. Suppose you are given a square and asked to subdivide it into n non-overlapping squares. Figure 1 shows that
it can be done if n = 4. (It is impossible if n = 5 or n < 4.)
a) Show that a square can be subdivided into
b) Show that a square can be subdivided into

11
11
squares if n = 6.
squares if n = 7.
1-o-1
c) Show that a square can be subdivided into n squares if n = 8.
[_ _[
Figure I
d) Prove that a square can be subdivided into 11 squares if n > 8.
2. Suppose you are given an equilateral triangle and asked to subdivide it into n non-overlapping equilateral
triangles. Figure 2 shows that it can be done if 11 = 4. (It is impossible if 11 = 5 or n < 4.)
a) Show that an equilateral triangle can be subdivided into 11 non-

overlapping equilateral triangles if 11 = 6. /\
b) Show that an equilateral triangle can be subdivided into n non-
overlapping equilateral triangles if n = 7.
/V\ Figure 2
c) Show that an equilateral triangle can be subdivided into 11 non-overlapping equilateral triangles if 11 = 8.
d) Prove that an equilateral triangle can be subdivided into n non-overlapping equilateral triangles if n >8.
3. Suppose you are asked to draw a three-dimensional solid figure with n edges. Figure 3 shows that this can be
done if n = 6. (This is impossible to do if n = 7 or 11 < 6.)
a) Make a sketch of a three-dimensional figure with n edges if 11 = 8.
b) Make a sketch of a three-dimensional figure with 11 edges if n = 9.

Figure 3
274
ARML Power Contest- November 1995 -Induction
c) Make a sketch of a three-dimensional figure with 11 edges if 11 = I 0.
d) Prove that a three-dimensional figure with 11 edges can always be made if 11 > l 0.
4. Suppose you have only 3¢ and 5¢ stamps. Using a 3¢ and a 5¢ stamp, you could come up with postage for 8¢.
Using three 3¢ stamps, you could come up with postage for 9¢. Using two 5¢ stamps, you could come up
with postage for 10¢. It would be impossible to come up with postage for 7¢.
Prove it is possible, using only 3¢ and 5¢ stamps, to come up with all postages greater than or equal to 8¢.
5. Suppose you have only 3¢ and X¢ stamps (where X is not a multiple of 3 ). Find an expression for the largest
amount of postage which you cannot make using the 3 ¢ and X¢ stamps.
6. Suppose you have only Y¢ and X¢ stamps (where Y and X are relatively prime). Find an expression for the
largest amount of postage which you cannot make using the Y ¢ and X¢ stamps.
7. Suppose you have only Y ¢ and X¢ stamps (where Y and X are relatively prime) and w equals the largest amount
of postage that you cannot make using the Y¢ and X¢ stamps, prove that of all (positive) postage amounts, a
and b, where a + b = w, either a or b but not both can be made using Y ¢ and X¢ stamps.
275
ARML Power Contest - November 1995 - Induction
The Solutions
la) b) c)
d) Any subsquare in any of the above three diagrams can be subdivided into 4 squares as shown in Figure I. The
net result would be an increase of three in the total number of subsquares in each figure. Any number n > 8 can
be written in one of the forms 3m + 6, 3m + 7, or 3m + 8, where m is some positive integer. Each of these
forms determines one of the three figures above a, b, and c respectively and m would be the number of
subsquares that must be subdivided into four squares to make the total number of subsquares equal to n.
2a) b) c)
/\
d) Any subtriangle in any of the above three diagrams can be subdivided into 4 triangles as shown in Figure 2.
The net result would be an increase of three in the total number of subtriangles in each figure. Any number
n > 8 can be written in one of the forms 3m + 6, 3m + 7, or 3m + 8, where m is some positive integer. Each
of these forms determines one of the three figures above, a, b, and c respectively, and m would be the number
of subtriangles that must be subdivided into four triangles to make the total number of subtriangles equal ton.
3a) b) c)
d) Each of the above figures contains a vertex of order 3, i.e. a

vertex connected to three edges. As shown in the diagram
below, any vertex of order 3 can be truncated or cut off and the
\
three "dangling" edges connPcted with three edges. The net
result is that the figure now has three more vertices of degree 3
and three more edges.
276
ARML Power Contest - November 1995 - Induction
This truncating process can continue and will always add three more edges onto the figure. Any number 11 > I 0
can be written in one of the forms 3m + 8, 3m + 9, or 3111 + 10, where m is some positive integer. Each of
these forms determines one of the three figures above, a, b, and c respectively, and 111 would be the number of
vertices of order 3 that need to be tnmcated so that the figure has 11 edges.
4. It has already been shown that 3¢ and 5¢ stamps can be used for 8¢, 9¢, and I 0¢ postage. Any number 11 > I 0
can be written in one of the forms 3111 + 8, 3m + 9, or 3m + I 0, where m is some positive integer. Postage in
the fonn 3m + 8 can be made with a five-cent stamp and m + I three-cent stamps. Postage in the form 3m + 9
can be made with 111 + 3 three-cent stamps. Postage in the form 3m + IO can be made with two five-cent
stamps and 111 three-cent stamps.
5. Let f(X) = the largest amount of postage that cannot be made with 3¢ and X ¢ stamps. Experiment with various
values of X to complete this table:
~ 4 5 7 8 10 ...
f(X) 1 5 1 11 13 17 ...
It can be shown that f(X) = 2X - 3 = 2(X - 1) - I.
6. Let f(X, Y) = the largest amount of postage that cannot be made with Y ¢ and X ¢ stamps. Experimentation
again shows f(X, Y) = (X - I )(Y - I) - 1 or t{X, Y) = XY - (X + Y).
X
_J
I 2 3 4 5 6 7 8 9 10
~- - - -
2 I 3 5 7
3 5 7 11 13 17
4 5 11 17 23
5 3 7 11 19 23 27 31
6 19 29
y 7 5 11 17 23 29 41 47 53 f(X, Y)
8 13 27 41 55
9 7 23 31 47 55 71
I
10 17 44 53 71
11 I 9 19 29 39 49 59 69 79 89
12 43 65
13 11 23 35 47 59 71 83 95 107
For given X and Y, call an integer n 2: 0 postahle if n = aX + bY for some nonnegative integers a and b.
277
Then XY - (X + Y) is not postable: Assume XY - (X + Y) = aX + bY. Then XY =(a+ I )X + (h + I )Y. Notice that
if a and h are both nonnegative, this implies that (a+ I )X < XY so that a+ I < Yanda< Y - I. Similarly, h
<X - I. But then, (b + 1) Y= (Y- a - 1) X, implying that YI (Y - a - 1) because Yand X are relatively
prime. But obviously, 0 < Y - a - 1 < Y, so this is impossible.
Proving that any larger number is postable is trickier; it's enough to prove that the next Y integers are postable,
since any larger integer can then be written by adding a multiple of Y. Since X and Y are relatively prime, for
any k in ( 1, 2, ... , Y - I } there exist nonnegative integers a and h such that aX = h Y + k. Clearly a > 0, but
also remember that we can assume a< Y and b < X, because if a> Y then b > X and (a - Y)X = (b -X)Y + I ,
we can continue until a < X and b < Y. Then aX - b Y = k, where b :S X - I,
so aX + (X - 1 - h)Y = XY + k- Y; since a 2: I we then take (a - 1)X + (X - 1 - b)Y = XY -X - Y + k with
both a - 1 and X - 1 - b nonnegative. For k = Y. we get XY - X, which is a multiple of X.
7. First note that (X - 1)( Y - I) is even: X and Y are relatively prime, so at least one of them is odd, meaning that
at least one of X - I and Y- I is even. We now show that the number ofpostable integers n in
{O, 1, ... , XY-X- Y} is exactly equal to (X- l)(Y- 1) / 2.
Let's define three distinct but related sets:

Let N be the set of postable integers n in [ 0, 1, ... , XY - X - Y} .
Let T (for triangle -- see figure) be the set of pairs of nonnegative integers (a, b) such that
aX+bY<(X- l)(Y-1).
Let R (for rectangle) be the set of pairs of nonnegative integers (a, b) such that a < Y - 1 and h < X - 1.
The diagram shows the case X = 4, Y = 5.
'
·~ 4.,r r Sv = 11
1~o",,,~ D14~~D 8 ~-1
a
23
R T
Dg (31 2, 1)
1
a a
5 13 17
278
Notice two fundamental facts. First: Tis a subset of R. because if (a. b) is in T then
aX < aX + h Y < (X - I )(Y - I) < X(Y - I), so a < Y - I; likewise h < X - I. Second. every element of R
defines a unique sum aX + hY, because if aX + hY = a 'X + h'Y then (a - a')X = (h - h')Y so that YI (a - a')
and XI (b- b'); this is impossible ifa and a' (also band b') are in the appropriate ranges. It follows from this
that N has exactly the same number of elements as T, since no element of N can correspond to more than one
element of T.
Now we show that exactly half the elements of Rare in T.
Letr(a, b)=(Y-2-a,X-2-b). This is a 180degreerotationaboutthepoint(Y /2 - l,X/2- ))which,

we will show, maps T to R - T and vice versa.
First, r(a. b) is in R for any (a, b) in R. Since at least one of X - 2 and Y - 2 is odd, no element maps to itself;
but r(r(a, b)) = (a, b), so r breaks up R into doublets of points {(a, b), r(a, b)}. In each doublet, the two sums
together add up to exactly 2(XY - X - Y):
(aX + bY) + (Y - 2 - a)X + (X - 2 - b)Y = (Y - 2)X + (X - 2)Y ~ 2(XY - X - Y).
No element has the sum XY - X - Y, so one element of each doublet has a sum less than XY - X - Y, the other
has a sum greater than XY - X - Y. Therefore each doublet contains exactly two distinct elements of R with no
overlap between the doublets, and the union of the doublets is R; so the number of doublets must be half the
number of element ofR, or (X - 1)(Y - 1) I 2. For every doublet there is one element of T, and for every
element of T there is one element ofN, so the number of elements ofN is (X - l)(Y - I )/2.
However, because XY - X - Y is not postable, it is not the sum of any two postable integers either. Therefore, if
we divide all numbers in {O, 1, ... , XY - X - Y} into pairs (m, n) with m + n = XY - X - Y, each pair has at
most one postable element. Again, XY - X - Y is odd, so each pair has two distinct numbers. But if any pair
did not have at least one postable number, the total number of postables would be less than (X - 1)(Y - I )/2,
since there are (X- l)(Y - 1) / 2 pairs and none of them has two postables! Therefore in every case, exactly
one of n and XY - X - Y - n is postable.
279
ARML Power Contest- February 1996 - Rook Polynomials
Rook Polynomials
The Definitions
A board consists ofan n by m array of squares, some of which are unshaded (available) and others are
shaded (unavailable). Rooks (markers) can be placed on any available square provided that it is non-challenged,
that is, no two rooks are in the same row or column.
In the above board, a single rook can be placed in eight different ways into the eight available squares.
However, there are nineteen different ways to place two rooks onto the above board and only fourteen ways to place
three rooks onto the board. Some of these fourteen ways are illustrated below:
In addition, as shown below, there are only two different ways to place four rooks on the board above so
that no rook is challenging another.
For every board, B, there exists a rook polynomial, Rs(x),
where Rs(x) = l + r I x + r ,;K 2 + r -,,x 3 + ... + r ,;t k + ... , and r k , the coefficient of xk , is the number of
ways of putting k non-challenged rooks on board B. The rook polynomial for the board above is :
R s(x) = l + 8x + 19x 2 + l 4x 3 + 2x 4 • (The first term is one because it is the coefficient of x 0 and there is
one way of putting no rooks on any board.)
The following theorems may be helpful in determining and checking rook polynomials for more difficult boards:
280
Theorem 1. If board B can be partitioned vertically into two boards C and D where the corresponding rows of
C and D do not both contain available squares or if board B can be partitioned horizontally into two boards
C and D where the corresponding columns of C and D do not both contain available squares, then
RB(x) = Rc(x) · RD(x)
Theorem 2. Let B be a board and let s be one particular available square in B. Then let B, be the board
obtained from B by shading in square s and B 2 be the board obtained from B by deleting the row and the
column containing s. Then Rs(x) = Rsi(x) + x ·Rs2 (x)
Theorem 3. Let B be an n by n board with rook polynomial Ra(x) = 1 + r 1x + ri( 2 + r-;< 3 + ... + r11X"
and let B be the complement of B (B is shaded where B is unshaded and vice versa). The number of ways
of placing n non-challenging rooks on Bis n! - (n- 1)! r 1 + (n-2)! r2 - ••• +( - lt(n- n)! rn.
The Problems
1. Determine the rook polynomial for each of the following boards:
~
Cl[J
a b C
2a. The rook polynomial for Board 2a is I + 5x + 8x2 + 5x 3 + x 4 . Demonstrate your

understanding of Theorem 1 by partitioning the board and factoring this polynomial
into two rook polynomials in two different ways.
Board 2a
2b. The rook polynomial for Board 2b is 1 + 7x + 14x2 + 8x3 + x 4 . Demonstrate your
understanding of Theorem 2 by eliminating square s 1, then s2 from the resulting
two, and writing the polynomial in the form:
Ra= Ra1(x) + x ·Rai(x) + x ·Ralx) + x 2 ·Ralx).
Board 2b
281
ARML Power Contest - February 1996 - Rook Polynomials
2c. Demonstrate your understanding of Theorem 3 by determining in how many ways five rooks can be placed on a
5 by 5 board whose five downward diagonal squares are the only squares shaded.
2d. Prove Theorem 2: Ra= Ra 1 (x) + x ·Rai(x).

(Hint: Show that the coefficient of of any xk is the same on both sides of the equation.)
3a. Find the general rook polynomial for an n by n board where all squares are shaded except those in the
downward diagonal.
3b. Write down the general rook polynomial of an by n board of the

type shown, consisting of m available squares (two in each row,
going diagonally down the board). All others squares on the
board are shaded. (Hint: Think Pascal's Triangle.)
..
3c. A board is complete if it contains no shaded squares. Find the rook polynomials for complete I by 1, 2 by 2,
3 by 3, 4 by 4, 5 by 5, and n by n boards.
282
The Solutions
I. a) I + 8x + 17x2 + 8x3
b) l + 8x + 20 x2 + 16x3 + 4x4
c) I + 9x+28x2+35x3+ 15x4 +x5
2a. Board 2a can be partitioned in four ways but because of the board's symmetry there are only two partitions that
produce different polynomial factors:
I + 5x + 8x2 + 5x3 + x 4 = ( I + 4x + 4x2 +x3)( I + x)
'
■I _J
I + 5x + 8x2 + 5x3 + x4 = (I + 3x + x2)(1 + 2x + x2)
2b.
(I + 5x + 7x 2 + 3x3 ) + x(l + 3x + 3x 2 + x3 ) + x(l + 3x + x 2) + x 2(1 + x) =
l + 5x + 7x 2 + 3x 3 + x + 3x 2 + 3x 3 + x 4 + x + 3x 2 + x3 + x2 + x3 =
1 + 7x + I 4x 2 + 8x3 + x 4
283
ARML Power Contest- February 1996 -Rook Polynomials
2c.
Board B Board B
RB(x) = I + 5x + I 0x 2 + I ox3 +5x4 + x5. Therefore, the number of ways of placing five non-challenging rooks
on Board Bis 5! - (4!)(5) + (3!)(10)- (2!)(10) + (I !)(5)-(0!)(1) = 44.
2d. The coefficient of J< in Rs(x) = the number of ways of placing k rooks on B
(# of ways of placing k rooks on B with s not used) + (# of ways of placing k rooks on B with s used)
= (# of ways of placing k rooks on 81) + (# of ways of placing k- I rooks on B2)
= (the coefficient of xk in Rs I (x)) + (the coefficient of J< - 1 in Rs2(x))
= (the coefficient of xk in Rs 1(x)) + (the coefficient of xk in x• Rs2(x))
the coefficient of xk in Rs J(x) + x•Rs2(x).
Therefore, Rs(x) Rs J(x) + x•Rs2(x).
3a. Rs(x) =(I+ x)n Using Theorem I over and over again (n - 1) times, the board can be partitioned into n
boards, each I by I with a rook polynomial I + x. Therefore, the polynomial for the entire board is ( 1 + x)n.
3b. m Rs(x)
I+ Ix
2 1 + 2x
3 I+ 3x + lx 2
4 I+ 4x + 3x 2
5 I + 5x + 6x2 + lx3
284
ARML Power Contest- February 1996 - Rook Po(ynomials
6 I + 6x + I ox2 +4x3
7 I + 7x + I Sx 2 + I Ox3+ x4
8 1 + 8x + 21x2 + 15x3+ sx4
m 1 r
+( )x +(m 2 1)x2 +(m 32 )x3 +... + (m - z+ 1)x k+ .. _
3c. n Rs(x)
l+x
2 I+ 4x + 2x 2
3 I+ 9x + 18x2 + 6x3
4 I+ 16x + 72x 2 + 96x3 + 24x4
5 I+ 25x + 200x2 + 600x3 + 600x4 + I 20x5
Notice that the coefiicients ofx are the square numbers and the coefficients ofx11 are n! (Theorem 3 would prove
this last result.) Therefore, the general term will probably contain squares and factorials. With this in mind, the
table above can be rewritten:
n
l+x
2 l+ 1(2 2)x + 2(1 2¥
3 I+ I (3 2)x + 2(3 2 ¥ + 6( I 2},.3
4 I+ 1(42)x + 2(62¥ + 6(42}J + 24(12)x4
5 I+ 1(52)x + 2(102)x2 + 6(102~ + 24(5 2 )x4 + 120(1 2 ~
Now the squares and factorials are even more evident and also the binomial coefficients appear.
Therefore,RB(x) = 1 +l! (7)\+ 2! C2)2x +3! (~2 r +... +

x3 n! (~Y x"
285
ARML Power Contest - November 1996 - Rotating Decimals
----------------------------------------------------------
Rotating Decimals
Notation
Throughout this problem we are concerned only with rational numbers that can be represented by totally repeating
decimals. For example, t =0.142857. Partially repeating decimals like ¼=0.250 and 7; =0.16 are not part
of this problem. 0d 1d 2d 3 ••• d,,, means the totally repeating decimal with digits d1, d2,dJ, .... dn in the
repetend. The expression, did 2d, ... d n, means the integer with digits d1, d2, d 3, ... ,dn .
The following is a theorem that you may use without proof:
d1d2d3 ... dn
For every totally repeating decimal, 0d 1d2d3 ... d,,= - - -1-0 ....
, _-1- - -
For example, 037 =~ and 0.145 = ~
Definition
For every totally repeating decimal, let's define a rotating function, r, such that
r(Odid 2d, ... d,, _ 1d,) = 0d ,,d1d2 ... d n-1
As examples, r(0.1234) =0.4123 and {tr)= r(0.09) =0.90 =t? .
The Problems
la. Compute repeating decimals for t , ~ ,and ? and then compute r ( t ), r ( ~ ), and r ( ?). Express your
final answers as reduced fractions.
1b. Compute repeating decimals for "t3 and TI and then compute r ( "t3 )and r ( TI ). Express your final
answers as reduced fractions.
2b. Prove: If it is a totally repeating decimal, then r ( it )is an integer multiple of it .
286
2c. Prove: If ,fn is a totally repeating decimal, then r ( ,fn ) = fn for some integer t.
Give a compact formula fort.
3a. Solve the following two multiplication problems by finding the values of the digits a, b, c:, d, and e.
i) a b c de 6 * 4 = 6 a b c de ii) a b c de 4 * 4 = 4 a b c de
3b. Find a solution to the equation, r (x) = 4x. Express your answer as a reduced fraction.
3c. Prove: For each digit L, where L = 1, 2, 3, ... , 9, there exists an integer n such that the multiplication
problem,
dJ d2 d3 ... dn -1 L * L = L dJ d2 d3 ... dn -1, has a solution.
Consider the set of repeated rotations of ? = ~ - i.e. r ( ~ ), r (r ( ~ )), r (r (r ( ~ ))), ...
It is a fact (but difficult to prove!) that the repetend of 2f87 is 243 digits long! You may use this fact in the
following problems.
4a. Prove if 2ffi is in the set of rotations of ~ then u = I mod 9 (i.e. u = 9k + I for some integer k).
4b. Prove the set of rotations of 2f87 is { ~ , ~ , ~ , ... , ~ }.
4c. Prove that each of the 100 pairs of digits 00, 01, 02, ... , 99 appears in the decimal expansion of 2f87 .
Extensions (not part of the contest problem)
I. When is the decimal equivalent of the fraction t (in lowest terms) totally repeating?
2. If gcd(n, 10) = 1 then the length of the repetend of ¼is r, where r is the smallest positive integer such that
lOr=lmodn.
3. T' has repetend of length 3' 2 for any t 2': 2. (A generalization of the fact used in problem #4.)
4. The decimal expansion of -Au

3
contains every possible three-digit sequence and the decimal expansion of +oo,
3
contains every possible l 00-digit sequence.
5. The decimal expansion of the irrational number, f

11= l
~
3
,contains every digit sequence of every length !!!
287
----------------------------------------------------------
The Solutions
la. r(l/7) = r(0.142857) = 0.714285 = 5/7 lb. r(l/13) = r(0.076923) = 0.307692 = 4/13
r(2/7) = r(0.285714) = 0.428571 = 3/7 r(4/37) = r(0.108) = 0.810 = 30/37
r(3/7) = r(0.428571) = 0.142857 = 1/7
dJ d2 d3 ... dn + 10ndn - dn
2a.
10 n - I IO n - 1
10 10 10
IO n - 1 dn dJ d2 .... dn-1
10 IO n - I
r(x)
I I
l.) = iii l+Odn , where dn is the last digit in the repetend of J.m . Since iii+10dn -- 1 10
2b. By 2a, / \in + dn • .1 all we
m'
J. d1d2d3 ... d n
must show is that I + dn * rn is a multiple of I 0. Since m = 0.d I d2d3 ... d n = - - - - - - - -
10 - I
==> 10n-I=m*(d1d2d3 ... d,,) ==> 10n-m*(d1d2d3 ... dn-10)=m*dn+I.
Since the left hand side of this equation is a multiple of I 0, the right hand side must also be a multiple of 10.
fl)_
r\m - 10
I an d
I+ d,, · m. l1f I+ d,, ·m is an integer.
10
2c. By 2b, r(ij,) = ~ = s + 15 ·m · ~- Again s + dn * m is a multiple of IO because
s
TFI = I 0' -1
hand side of this equation is a multiple of 10, the right hand side must also be a multiple of 10. Thus the
r Ia r1or t 1s
. s+d ·m and t is guaranteed to be an integer.
1ormu 10
3a. 153846 * 4 = 615384 and 102564 * 4 = 410256 (See 3b for general method.)
3b. From 3a, both 0.153846 and 0.102564 must be solutions. To help getting these into lowest terms, note that
r(x) = 4x ==> x + d = 4x ==> x = d /39. Knowing this, it is easy to calculate that the first solution is 2/13
and the second is 4/39.
4a. We show ifs= I mod9, then r (s / 2187) is some u / 2187 with the numerator u"' I mod9. This suffices to
288
prove the problem since I = I mod 9 and so all successive rotations of 1/2187 will have numerators congruent
to 1 mod 9. Supposes= I mod9. Then r(~) =s+~~87 d - 2fITT. We must show s+fJ 87 ::cc I mod 9.
Since s = 1 mod 9 and 2187d = 0 mod 9 , s + 2187d = I mod 9. Since IO and 9 are relatively prime, (s +
2187) /JO= I mod 9. Sor (s I 2187) has a numerator that is congruent to l mod 9.
4b. From the fact stated with the problem, there are 243 distinct rotations of 1/2187. Since there are 243 integers of
the form I + 9k in the set {I, 2, 3, ... , 2186) and 4a states that only such integers may appear as numerators of
the rotations of 1/2187, by the pigeon-hole principle, the rotations must be exactly the set
{(9k + I) I 2187 with k= 0, 1, 2 .... , 242}.
4c. By 4b, the elements in the set of rotations of I /2187 are spaced 1/243 apart. Since I /243 < l /100, there must
be at least one element of the set in any interval of length 1/100. So some rotation of 1/2187 must fall in the
interval [ 0.d Jd2, 0.d 1d2 + 1/100). So for each digit pair d1d2, there is a rotation of l /2187 which begins with
0.dJd] ... . So dJd2 must appear somewhere in the expansion of 1/2187.
Supplementary Notes
I. In this problem we dealt with only those fractions that have totally repeating decimal expansions. It is an
interesting question to ask which fractions have this property. If * is in lowest terms and *
= 0 Li 1d2 ... d,,,
th a_ d 1d2 ... d,,

and bf 10" - I , which implies every divisor of b also divides 10" - I . Since 21 10" and
en b - 10"- I
51 10" , we know that 2 does not divide 10" -1 and 5 does not divide 10" - I . Sob cannot be I or a
multiple of 2 or 5 if i is to have a totally repeating decimal. But are these conditions enough to guarantee
that i will be totally repeating? See any university number theory textbook (e.g. The Higher Arithmetic by
Davenport) for the complete discussion and proof. Other interesting facts you might want to try to prove is that
i has a terminating decimal expansion if b is divisible by no primes other than 2 or 5 and i is a partially
repeating decimal if b is divisible by 2 or 5 and some other prime factor not equal to 2 or 5.
II. The difficult fact you were allowed to use in problem 4 holds more generally: r, has a repetend of length 3'- 2
for any t ~ 2. As examples, ~ = .i, 'ff = .037, and ITT = .012345679 . Equivalently, the smallest integer n
for which 3' I t 0" - 1 is 3, _2 • Restating once more, the smallest integer n for which IO"= 1 mod 3' is
n = 31 - 2 • This generalization allows us to prove many statements similar to problem 4c. Since ~ > :1,i ,
10 3
the decimal expansion of :tri

3
contains every 3 digit sequence and since ~ >
10 3
:-k the decimal expansion of
-k
3
contains every l 00-digit sequence.
289
ARML Power Contest - January 1997 - Regular Closed Linkages
----------------------------------------------------------
----------------------------------------------------------
Regular Closed Linkages
The Definitions and Theorems
A linkage is a figure made up of line segments connected at their endpoints. The figures A, B, D, E, and F below
are examples of linkages. Figures C and G are not linkages but each consists of two overlayed linkages.
✓A . B/ \ /c\
- -. ' -- ~
~ E
C D co- --- - - ,,
A linkage is closed if each endpoint is shared by exactly two segments. Figures A, D, and E are closed linkages.
Figures Band Fare not closed linkages.
A closed linkage is regular if all segments in the linkage are crossed by other segments the same number of times.
In the figures above, only Figures A and D are regular closed linkages (RCL).
The symbol [n, k], where n and k are non-negative integers, can be used to represent a (RCL) made up of n segments
where each is crossed by other segments k times. Figure A is a [3, O] RCL and Figure D is a [5, 2] RCL.
In this set of problems, we will investigate for which values of n and k is it possible to construct a [n.. k] RCL.
The following theorems have evolved in the study ofRCL's and may aid in your investigation and used in proving
other theorems:
Theorem 1 Ifa [nJ, k] RCL and a [n2. k] RCL are constructible, then a [nJ + n2, k] RCL is also
constructible.
Theorem 2 If for some integer m, a [2m + 1, k] RCL is constructible, then a

[4m + 2, 2k + l] RCL is also constructible.
290
----------------------------------------------------------
The Problems:
1a. Constmct a [7. 2] RCL.
lb. Constmct a [7, 4] RCL.
2a. Construct a [6, I] RCL.
2b. Constmct a [8, 4] RCL.
3a. Construct a [ I 0, I] RCL.
3b. Construct a [ 12, I] RCL.
4a. Prove it is impossible to construct a [5, 1] RCL.
4b. Prove it is impossible to construct a [6, 2] RCL. (4b may be the most difficult problem of the set.)
5. Prove it is always possible to construct a [2n, 1] RCL for any integer. n, greater than or equal to 5.
6a. Construct a [8, 2] RCL.
6b. Prove it is always possible to construct a [n, 2] RCL for any integer. n. greater than or equal to 7.
7a. Prove Theorem 1.

Given two RCL's, [n1, k] and [n2, k], show how to constmct a [n1 + n2, k] RCL.
7b. Prove Theorem 2.
Given a [2m + 1, k] RCL, where m is a positive integer, show how to constmct a [4m + 2, 2k + I] RCL.
The Extensions (not part of the contest problem)

I. Does there exist a positive integer n for which a [n, 3] RCL can be constructed?
2a. Construct a [8, 1] RCL.
2b. If a [2m + 1, OJ RCL is constructible, then a [4(m+ 1), I] RCL is also constructible.
3. Ifa [n, jk] RCL is constructible, then a Un, k] RCL is also constructible for positive integers n, j, k.
4. If gcd(m, n) =I and n 2: 2m + I, then a [n, 2(m - 1)] RCL is always constmctible.
291
The Solutions
la. I b. In general, [2m + I, 2] is

constructible form '.:". 2.
In general, [2m + 3, 2m] is

constructible for m '.:". I.
2a. 2b.
3a. 3b.
4a. In a [n, k] RCL each of then segments is crossed k times. Therefore, there must be n times k divided by 2
crossing points (because each point is counted twice). Since the number of crossing points must be a whole
number, either nor k must be even. Therefore, [5, I] is impossible.
4b. Consider a graph called the dual of [6, 2]. In this graph each segment of [6, 2] , AB, BC, CD, DE, EF, and
FA, is represented by a point on a circle. Two points in this graph are connected if the segments they represent
in [6, 2] cross. Since in [6, 2] each segment gets crossed twice, each point in the dual of [6, 2] must be
connected to two (and only two) other points.
Lemma There are only two possible duals of [6, 2] (disregarding rotations of the labels).
AB BC
Proof: Since adjacent points in the dual represent segments with
common endpoints in [6, 2], they cannot be connected in the dual
• •
of [6, 2]. Therefore, AB must be connected with CD and DE or
AB must be connected with CD and EF. (Connecting AB to DE FA• • CD
and EF would be equivalent to AB to CD and DE and therefore
does not need to be considered.) • • DE
EF
292
Case 1 AB is connected to CD and DE.
Subcase la CD is connected to EF. Then EF must be connected CD

to BC and then BC would have to be connected to FA and
FA to DE.
EF DE
Figure la
Subcase 1b CD is connected to FA. Then FA must be connected

to BC and then BC would have to connected to EF. This
would force EF to be connected to DE, a contradiction.
Case 2 AB is connected to CD and EF

Figure I b
Then EF must be connected to CD or BC.
Subcase 2a EF is connected to CD. Then BC, DE and FA must

be connected forming two overlapping triangles. FA CD
Subcase 2b EF is connected to BC. Then BC can be connected to Figure 2a

FA or DE.
Subsubcase 2ba BC is connected to FA. Then FA must be

CD
connected to DE. This would force DE to be connected
to CD, a contradiction.
E
I
DE
Figure 2ba
Subsubcase 2bb BC is connected to DE. Then DE must be
connected to FA and FA to CD.
FA CD
Notice that the graph in Subcase 2bb is just a rotation of the labels of the graph
in Subcase I a. Therefore, they can be considered the same graph.
Therefore, [6, 2] has only two possible duals as shown in Figures la and 2a. Figure 2bb
293
Theorem It is impossible to construct a [6, 2] RCL.
The duals found in the lemma show that AB must be crossed by CD and DE or AB must be crossed by CD
and EF.
Case 1 AB is crossed by CD and DE. This can be done in two possible ways:
• E
'
B
A D
Subcase la Subcaselb
Now draw in BC in each of the two possible cases. (It cannot cross DE since according to the dual BC
crosses only EF and FA.)
~
A D A D
Subcase la Subcase lb
According to the dual EF must intersect CD and BC.
In subcase 1a, this is impossible because E is in the interior of Angle BCD and EF would have to intersect
both sides of this angle. This happens only at vertex C.
However, in subcaselb it is possible to draw EF intersecting both CD and BC if points C and E are
moved in the diagram.
A D
294
Finally, according to the dual, AF must cross DE and BC. This is also possible, but in doing so AF must
also cross CD. Would it be possible move points (and the accompanying segments) so this would not
happen? The answer this time is no. because in the original diagram of this case. A and E are on one
side of CD and F must be on the other side of CD since EF and CD cross.
Case 2 AB is crossed by CD and EF. Then either D and E are on the same side of AB or opposite sides of
AB.
Subcase 2a D and E are on the same side of AB. Then C and F are on the other side of AB. Segment
DE will be on one side of AB while FA and BC would be on the other and hence could not cross
DE. A contradiction.
Subcase 2b D and E are on opposite sides of AB. Then forming DE (without crossing AB) forces all
the other points to be on the same side of DE and the segments connecting these points could not
cross DE. Again a contradiction.
Therefore, it is impossible to construct [6, 2].
5. If k is odd (k 2:: 3), it is always possible to construct a [k, O] RCL as a convex polygon with k sides. By
replacing each of the sides with a pair of crossing segments, as shown in the diagram, a [2k, 1] will be
constructed.
/
•
/ /,A
Replace with
✓/
For example, [3, O] to [6, 1]
This still leaves the [2k, 1] RCLs where k is even. Each of these can be constructed by combining two previously
constructed [2k, 1] RC Ls using theorem 1.
6a.
295
6b. If n is odd, place the n points around a circle and

connect every other point. After the second trip
around the circle you will have come back to the •
starting point and each segment will have two
crossings. This works for any odd 11 2: 5.
• •
For example, [9. 2]:
If 11 is even, [8,2] is a special case and shown in part 6a. If 11 is even and greater than 8, it can always be expressed
as a sum of two odd integers, n I and 112, greater than or equal to 5. The construction above shows that [n I, 2]
and [112, 2] are constructible and theorem 1 states that [111 + n2, 2] will always be constructible when [nl, 2]
and [n2, 2] are constructible.
7a. Place [nl, k] and [n2, k] next to each other (not overlapping) so that one endpoint from each of the RCLs is
close to the other, one above the other. Move two of the four segments connected to these endpoints so that
the two upper segments go to the upper endpoint and the two lower segments go to the lower endpoint. (See
the diagram.)
Since no new points have been added and no crossings have been fonned or deleted, the figure would be
[nl + 112, k].
7b. Make a copy of the [2m + 1, k] RCL but make it a little larger. Overlay these two copies, lining up the parallel
sides. There are now 4m + 2 endpoints and each segment should now have 2k crossings. Replace each set of
parallel segments with two intersecting segments, producing the [4m + 2, 2k + 1] RCL. (See the diagram.)
Because n in the original RCL is odd, you are guaranteed to get a RCL using this method. If n had been even,
using this method would only produce two overlapping [n, k] RC Ls.
296
The Extensions
I. A general method has been found to generate [n, 3] RCLs. The following [16, 3] RCL should serve as a spring-
board for you investigations: It started with a [6. I] RCL and traced twice, going to the outside after almost
closing the linkage each time.
2a. 2b. From theorem 2, if [2m + I, OJ is constructible, then

[ 4m + 2, I] is constructible. You can always add two
more intersecting segments as indicated in the diagram:
\j
\/
==>
i
3. The following diagram should serve as a springboard to further investigation.
Changing [5, 2] to [IO, 1]:
==>
4. Put then points around a circle and connect every m-th point. This provides (2m - I) crossings and is closed
when gcd(m, n) = I.
Special thanks to Grisha Chelnokov from the Moscow Youth Science Center for the original idea behind the
problem and to Davide Cervone from the University of Minnesota Geometry Center for his assistance and
encouragement in the writing of this problem.
297
ARML Power Contest - November 1997 - Factorial Polynomials
Factorial Polynomials
The Definitions
Factorial polynomials are defined as follows:
Po(x) = I
P,(x) =x
P2(x) =x(x - 1)
PJ(x) =x(x- l)(x - 2), etc.
In general, Pn(x) =x(x - 1)(x- 2) ... (x - n + 1) , and recursively, Pn(x) = Pn -1(x)(x - n + I) and Po(x) = I .
Factorial polynomials can also be written in expanded form:
Po(x) = 1
P1(x) = lx+O
P2(x) =Ix 2 - 1x + 0
P3(X) = 1 x 3 - 3x 2 + 2x +O
P4(X) = 1x 4 - 6x1 + I lx 2 - 6x + 0
Ps(x) = I x 5 -10x4 + 35x 3 - 50x2 + 24x + 0 ,
creating the following triangle of coefficients:
0
-1 0
-3 2 0
-6 11 -6 0
-10 35 -50 24 0 ...
Let {z} = the coefficient of x" -k in the expansion of Pn(x) , with O :S k :Sn.
Forexample, {i}=ll, G}=-50,and {o}=l.
298
The Problems
la. Showthat P,(x+1)-P.,(x)=3P2(x).
1b. Show that P4(X + 1)- P4(x) = 4P,(x).
1c. Generalize the above two statements and prove your generalization is true.
P3(k+ 1)
2a. Show t hat P2(1) + A (2) + P2(3) + ... + P2(k) = 3
P4(k+ ])
2b. Show that P,(l) + Pi (2) + P,(3) + ... + P3(k) = 4 .
2c. Generalize the above two statements and prove your generalization is true.
3a i. Show that P1(x) +P2(x) = x 2 •
3a ii. Use the above statement to derive a simple, explicit formula for the sum 1 2 +2 2 + 3 2 + ... + k 2 in terms of k.
3b i. Show that P1(x) + 3 P2 (x) + P3(X) =x 3 .

3b ii. Use the above statement to derive a simple, explicit formula for the sum 13 +2 3 + 3 1 + ... + k 3 in terms of k.
3c. Express x 4 as the sum of factorial polynomials.
4a i. For n >
- 1' { 11 } = - - - -
11
4a ii. Explain this answer in terms of the roots of Pn(x).
4b i. For n 2: 2, { 1} = _ _ __
4b ii. Explain this answer in terms of the roots of Pn(x).
4c i. For n 2: 2, { 11-l
11 } =
----
4c ii. Explain this answer in terms of the roots of Pn(x).
4d i. For n 2: 2, ~ {i} = ____

4d ii. Explain this answer in terms of the roots of P,,(x).
5. In Pascal's Triangle, (k) = (k= D+ ( 11 , / ) • Likewise in this triangle, n

{z} can be derived from {z =
and { 11 k 1} . Complete this statement: Forn>2
- ' {n}=
k - - {n-1}+
k-1 - - {n-1} k .
299
----------------------------------------------------------
The Solutions
la. PJ(x+l)-PJ(x)=(x+l)(x)(x-1)-(x)(x-l)(x-2)
=(x)(x- l)((x+ l)-(x-2))
= (x)(x- 1)(3)
= 3P2(x)
1b. P4(X +1)- P.i(x) = (x + l)(x)(x- l)(x - 2) -(x)(x - l)(x - 2)(x - 3)
= (x)(x - I )(x-2)((x + I) - (x - 3 ))
= (x)(x - I )(x- 2)(4)
= 4P3(X)
le. P,,(x +I)- Pn(x) = (x + l)(x)(x-1) ... (x + I -11 + I) -(x)(x- l)(x-2) ... (x- 11 + I)
=(x)(x- l) ... (x+l-11+ l)((x+l)-(x-11+ I))
= (x)(x- l) ... (x-11 +2)(x +1-x + 11 -1))
= (x)(x - 1) ... (x - 11 + 2)( 11)
nPn- 1(x) = n(x)(x-1 ) ... (x- (11 - I)+ 1)
= 11(x)(x- l) ... (x-11 +2)

:.Pn(X +1)-P,,(x) =11Pn -1(x)
P3(k + 1)
2a. P2(1) + A (2) + P2(3) + ... + P2(k) = 3
1(1-1) +2(2-1) +3(3-1) + ... +k(k-1) (k+l)(k)(k-l)

3
Prove : For all k "?_ I, 0 + 2 + 6 + 12 + ... + k(k- 1) = (k + I) ~)(k- l) .
1) Show true fork= I.

0= (1 + 1)(1)(1- 1)
3
2) Assume O+ 2 + 6 + 12 + ... + JU - 1) =
U
+ i
1)(')(1'-1)
, prove
0 + 2 + 6 + 12 + ... + J·0-1) + U + l)(')=

J U+2)(j+l)U)
3 .
U+ l)~)U - 1) + U + I )(j) = (j + 2)~j+l)U)
U+ 1)())0 - 1) 30 + 1) U ) () + 2)()+ 1) U)
3 + 3 = 3
300
(j + l)(j)(j - l +3) (j + 2 X.i+ I )(j)

-------,---- = - - - , - - - -
3 3
(j + I )(j)(.i + 2) (j + 2) (j+ 1)(.i)
-'"---..a.....;---- = ...a....-...a....----''-
3 3
1(0)(-1) +2(1)(0) +3(2)(1) + ... +k(k-l)(k-2)= (k+ l)(k)(~- l)(k- 2)
Prove: 0+0+6+24+ ... +k(k- l)(k-2)= (k+ l)(k)(:-l)(k- 2).
1) Show trne for k = l.

0 = ( 2)( I )(0) ( - I)
4
2) Assume O+ 0 + 6 + 24 + ... + j(.i- l)(j 2) = (.i + !)(})(~- l )(.i - 2) , prove
0 + 0 + 6 + 24 + ... + (j + I )(.i Ki - 1) = (.i + 2)(.i +; )(j)(j - l) .
() + l X.i)(i-1)(}- 2) + (j+I)(.i XJ- l) = U+ 2)(}+ l)(j)(.i- I)

4 4
(j + I XJ)(j-1)(.i- 2) 4(.i +l)(j XJ-1) (j +2XJ + l)(j)(j -1)
4 + 4 = 4
(j + I X.i)(i - 1)( (j - 2) + 4) (j + 2)(} + l)(.i)(.i- 1)

4 = 4
(j + 1XJ)(j - l)((j +2) _ (j + 2)(} + l)(j)(.i -1)

4 - 4
2c. Prove for a given positive n, P,,(l) + R,(2) +P,,(3) + ... +Pn(k)= Pn+ ,(k
11 +
T1) for any k.
a) Show trne fork = 1.

Show for any n , Pn(l) = Pn + ,(12)
n+
1/ 11 =0, Po(l) = 1 and ~ 1 i2f = 2 (not true when 11 = O!)
If n = 1 P1(1) = 1 and PJ.( 2) = l

' I+ l
ff n ~2, Pn(l) =0 and Pn+ 1(2) =0

b) Assume P,,(l) + Pn(2) +P11(3) + ... + Pn(j) = Pn + ,(.i; l), prove
n+
301
----------------------------------------------------------
P11(1) + P,,(2) + ?11(3) + ... + P11(J) + P»U+ 1) = Pn+ i(j; 2) .
' 11+
p,,. 1() + 1) p ( . I)_ p,, + 1(} + 2)

11 + 1 + 11 J+ - -11---'+'--1-
(11 + I) P,,(j+I )= A+ 10 + 2)- P,, + 1(} + I)

=(}+2)(}+ l)(j)(j- l) ... (i+2-(11 + 1)+1)-(j+ l)(j)(j-1) ... (i+ 1-(n+ I)+ I)
= (j + 2)(} + l)(j)(j- l) ... (i-n + 2)- (j + l)(.i)(j- 1) ... (i-n + I)
= (j + l)(.i)(j- !) ... (}- n + 2)(.i + 2- (j-n+ 1))
= (j + 1)(.i)(j- I) ... (} n + 2)(11 + I)
= Pn(j+l)(n + I)
3a i. P1(x) +P2(x) =x +x(x-1)

=x +x 2 -x
=X2
3a ii. + + + ... +
P1( I) + P2(1) + P1(2) + P2(2) + P1(3) + P2(3) + ... + P1(k) + P1(k) =
P1( I)+ P1 (2) +P1(3) + ... +P1(k) + P2(1) + P2(2) +P2(3) + ... + P2(k) =
From problem 2 above,
P2(k+l) + Pi(k+I)
2 3
(k+ l)(k) + (k+ l)(k)(k-1)

2 3
3(k+ l)(k) + 2(k+l)(k)(k-l) _ k(k+ 1)(2k+])
6 - 6
3bi. P1(x)+3P2(x)+P3(x)= (x)+3(xXx-l)+x(x-l)(x-2)

= x(l + 3x - 3 + x 2 - 3x + 2)
=x(x2) =xi
3b ii. + + ... +
P1(1) + 3P2(1) + Pi(l) + P1(2) + 3P2(2) + Pi(2) + ... + Pi(k) + 3P2(k) + A(k) =
P1(l) + Pi(2) + ... +P1(k)+3P2(1) + 3P2(2)+ ... + 3P2(k) + P,(1) + Pi(2) + ... + PJ(k) =
P1(k +1) + 3(P3(k +I))+ P4(k +I) =

2 3 4
2(k+ l)(k) + 4(k+ I)(k)(k- 1) +(k + l)(k)(k- l)(k-2) = k4 +2k3 + k 2 = (k(k + 1)) 2
4 4 2
302
----------------------------------------------------------
----------------------------------------------------------
3c. x 4 = P1 (x) + 7 P2(x) + 6Pi(x) + H(x)
= P1(x) + aP2(x) + bP3(x) + P4(.t)
= x + a,(x- I)+ b(xXx - l)(x -2) + x(x - I )(x- 2)(x- 3)
=x(l +m-a +bx2 -3bx+ 2b+x 3 -6x 2 + l lx-6) ...
If (1 + a, - a +bx2 - 3bx + 2b + x 3 - 6x 2 + l Ix - 6) is to equal x 3 , then

I - a + 2b - 6 = 0, a - 3b + 11 = 0, and b - 6 = 0. Therefore, a = 7 and b = 6.
4a) {~} = 0 In any polynomial the product of the roots is equal to the constant term divided by the leading
coefficient. Since zero is a root of any factorial polynomial (when n 2'. I), this product will always be zero.
b) {7} =- (n)(i;-1) =T,, 1, then th Triangle Number. In any polynomial, the sum of the roots is equal to -b/a,
where a is the leading coefficient and b is the coefficient of x" - 1 • In any factorial polynomial, the roots are 0,
1, 2, 3, ... , n - I and the sum of these numbers is the (n - I )th Triangle Number.
c) {n ~ 1} = ( - I)" 1c 11 - 1) ! In any polynomial the sum of all the products of the roots where one of the roots is
left out of each product is equal to ( - I)" - I~ , where y is the coefficient of the linear term and a is the leading
coefficient. Since zero is always a root of a factorial polynomial (when n 2'. 2)., all the products will equal zero
except the product that has zero left out. This product will be 1(2)(3)(4) ... (n - 1) or (n -1)!.
d) t {'1} = 0 In any polynomial if one is a root then the sum of all the coefficients will always be zero. Since
one is a root of all factorial polynomials (where n 2'. 2), the coefficients must add up to zero.
5. For n 2'. 2, {k} = (1-n){k= j} + (1){ k1} 17 (Found by inspection and trial and error!)
303
ARML Power Contest - February 1998 - Integer Geometry
lnte2er Geometry
The Definitions and Theorems:
An integer polygon is a polygon whose sides (edges) and diagonals all have integer
lengths.
An integer polyhedron is a polyhedron whose fal:es are all integer polygons and interior diagonals also have integer
lengths.
The measure used to order such polygons and polyhedra is perimeter-plus which is
the sum of the lengths of the edges added to the sum of the lengths of all the
diagonals.
In the following problems the following theorems along with the Pythagorean Theorem and the Triangle Inequality
may be useful:
Theorem 1: All isosceles trapezoids are cyclic, that is they can be inscribed in a circle.
Theorem 2: (Ptolemy's Theorem) In any cyclic quadrilateral the sum of the products of the opposite sides
is equal to the product of the diagonals.
The Problems
la. Find all integer rectangles whose perimeter-plus is :S 40. Prove there are no more. (Draw each rectangle,
labeling the lengths of its sides and diagonals and next to each rectangle indicate its perimeter-plus.)
1b. Prove there are no integer squares.
le. Find the smallest integer rhombus and calculate its perimeter-plus number. (Draw the rhombus, labeling the
lengths of its sides and diagonals and next to it indicate its perimeter-plus.)
Id. Find all integer isosceles trapezoids whose perimeter-plus is :S 40. (Draw each trapezoid, labeling the lengths
of its sides and diagonals and next to each trapezoid indicate its perimeter-plus.)
le. There exist two small cyclic quadrilaterals with no parallel sides. Can you find them and their perimeter-plus
numbers? (Hint: Both have diagonals 7 and 8 units long.) (Draw each quadrilateral, labeling the lengths of
its sides and diagonals and next to each quadrilateral indicate its perimeter-plus.)
304
2. An integer pentagon can be formed by

overlapping two integer quadrilaterals:
Using the quadrilaterals you found in problems Id and 1e, find two integer pentagons and calculate their perimeter-
plus number. (Draw each pentagon, labeling the lengths of its sides and diagonals and next to each pentagon
indicate its perimeter-plus.)
3. Using a technique similar to the above, find the smallest integer hexagon. (Draw the hexagon, labeling the
lengths of its sides and diagonals and next to it indicate its perimeter-plus.)
4. Since all tetrahedrons ( 3-D objects with four faces) have triangular faces and
no interior diagonals, all tetrahedrons with integer length edges are integer
tetrahedrons. A pentahedron (3-0 objects with 5 faces) can be constructed
in two ways:
i) a pyramid with four triangular sides and a quadrilateral base.
ii) a frustum with three quadrilateral sides and two triangular bases.
l \. -------✓
Find the smallest integer pentahedron of each type and calculate each figure's perimeter-plus number.
5a. A "pyramidal" hexahedron can be formed from a pentagonal base with five
triangular sides. Find the smallest integer "pyramidal" hexahedron and
calculate its perimeter-plus number.
5b. A "frustumal" hexahedron can be formed from six quadrilateral faces. Show
how two 8 by 15 rectangles and four isosceles trapezoids with bases of 8
and 15 and sides of 7 can be arranged to form a "frustumal" hexahedron.
Find its perimeter-plus number.
305
ARML Power Contest- February 1998 - Integer Geometry
----------------------------------------------------------
----------------------------------------------------------
The Solutions
la. There is only one ... a 3 by 4 rectangle with a diagonal of 5.
Proof: The sides and diagonal will form a right triangle and therefore must be a Pythagorean triple. All
such triples can be written as (2mn, m 2 - n 2 , m2 + n 2 ) where m, n are integers with m > n > 0. Since the
perimeter-plus must be :S 40, 2mn + m 2 - n 2 + m2 + n 2 must be :S 20, and therefore, m(m + n) must be
less than or equal to I 0. The smallest value for (m, n) is (2, I) which is a solution to the inequality and
produces the triple (4, 3, 5). The next smallest value for (m, n) is (3, 2), which produces the triple
(12, 5, 13), is not a solution to the inequality and all larger values for (m, n) also fail to solve the
inequality. Hence, the 3 by 4 rectangle is the only solution.
I b. The diagonal of a square is always ✓2 times the length of the side and can never be an integer if the side is an
integer.
I c. The smallest rhombus has sides of 5 and diagonals of 6 and 8. It is important to know that the diagonals of a
rhombus are perpendicular and bisect each other to convince yourself that this is indeed the smallest integer
rhombus.
Id. Because current literature has two definitions for a trapezoid, the answer is either 8 or 9. If your definition
states a trapezoid is a quadrilateral with only one pair of parallel sides, the answer is 8; however. if your
definition states a trapezoid is a quadrilateral with a pair of parallel sides, the answer is 9 because all
rectangles would be considered isosceles trapezoids.
Since all isosceles trapezoids are cyclic, Ptolemy's Theorem holds. Their diagonals are also congruent.
Therefore, using the labeling in the figure, ab = d 2 - c 2 and a + c > d. In order to count them all,
consider the following table:
d
---~
C (a)(b) a,b d C (a)(b) a,b
3 2 5 1, 5 6 4 20 4, 5
c/
vd
I; ~ ~ // -;;;:\
a
///Y°',,,,"
d~
c
3
4 3
8
7
1, 8
2,4
1, 7
7
7
7
5
5
3
24
24
40
3, 8
4, 6
5, 8
b
4 2 12 1, 12 8 7 15 3, 5
2, 6 8 6 28 4. 7
3, 4 * 8 4 48 6, 8
5 4 9 3, 3 ** 9 6 45 5, 9
5 3 16 4, 4 ** 9 3 72 8, 9
10 6 64 8, 8
10 4 84 7, 12
306
The first six entries in this table can be eliminated because a + c :S d. But the seventh entry(*) produces a solution
where a= 3, b = 4, c = 2 , and d = 4. The next two solutions(**) are identical and produce the rectangle
solution. The last four entries and all subsequent entries produce integer isosceles trapezoids but their
perimeter-plus number is greater than 40. Here are the diagrams of these solutions:
4
P = 19 ~
4 6/ . 8
6
3
"-~- - - - -
~::."/// 3
I //"s , ~ i
L_ ~
4
//
P = 24 6
P = 34
5
// P
7
= 39
4
8
P = 38
8 5
le. Using the labeled figure below, ac + bd = 56 and u + b > 7, b + c > 8, a+ d > 7, and c + d > 8. Therefore,
c-a>l. This data produced this table :
ac + bd 56
(3)(5) (1)(41) <'
b ---------- 5 ----.
3g\
/.
------------
\
\ (3 )(6) (2)( 19) / \
~.~..
8 \c (3)(7) (5)(7)*
a/ 7 \ 4
(3)(8) (4)(8) 7
----------------- (4)(6) (4)(8)*
d 7~- ~'-, \\
(4)(7) (4)(7) -----~ 6----- --~\i
Of these six possibilities only the two starred entries satisfy all the inequalities (P = 37 in both figures)
307
2.
P = 63
5
Ii -~ J\ 3
i\~,, /
/:<"
' 5
/ :
/
I h P = 90
5 -"/
I
3. It is believed this hexagon was known to Euler. Note that 3
it has the asymmetrical pentagon as a subset.

5
4a. Select the smallest integer quadrilateral for the

base and place the smallest isosceles triangles on
its parallel sides. The vertex of each of these
=>
triangles will become the vertex of the pyramid.
4b. Use the smallest integer quadrilaterals for the P = 29

sides and equilateral triangles for the bases:
5a. Since the asymmetrical pentagon has the smallest
perimeter-plus number, use it for the base of the
pyramid and attach an isosceles triangle to each
side. Since the pentagon has at least one
diagonal of length 8, the congruent sides of all
these isosceles triangles must be 5 to satisfy the
Triangle Inequality. Therefore, P = 85
5b. Internal diagonals (like BG) are all diagonals on

interior isosceles trapezoids (like BAGF) B E
and will have length 17. Therefore the 8
perimeter-plus number will be 4(8) + 4( 15)
+ 4(7) + 4(17) + 4(13) + 4(17) = P = 308
For more information about integer geometry, see "Mathematical Gazette", vol 81, pp 18-28.
308
ARML Power Contest - November 1998 - Unit Fractions
Unit Fractions
The Definition
A unit fraction is a fraction whose numerator is I and whose denominator is a positive integer. In other
words, unit fractions are the reciprocals of the positive integers.
The Problems
1. As discovered in the Rhind Papyrus, dating from 1800 B.C., Egyptian mathematicians represented all positive
fractions less than I as sums of distinct (different) unit fractions. For example, ~ = ¼+ -is,
j = f +¼+TI· and o/ = ½+½+tr5· In 1202 A.D., Leonardo Fibonacci published a systematic way do
this:
If 0 < '-ff < I .then lff =¼+{representation of W- ~}. where q= l#i-1- This algorithm is repeated until the
{representation of W- ¼} is a unit faction.
Express each of the following as the sum of distinct unit fractions:
a) 1
8 b) IT c) Yt;
2. The unit fraction Ts can be represented as the sum of two unit fractions in five different ways:
T5 = :ffi + 3TT = to +m= ~ +:r5 = rn- + iTc) = TI; +-tlo ,but ¼can be represented as the sum of
two unit fractions in only three different ways· l
. 4
= l5 + J..
20
= 6l + J.. = l + l
12 8 8.
a) Express TI as the sum of two unit fractions in eight ways ignoring the order of the addends.
b) Express TT as the sum of two unit fractions in two ways ignoring the order of the addends.
c) In how many ways can ¼be expressed as the sum of two unit fractions? Prove your assertion.
3. Inthediagramattheright, AL, XM,and BN are parallel. If AL=a,
BN = h, and X M = x, prove ±= ¼+ -; .
M
309
4. In the diagram at the right, each circle is tangent to the line and
to the other two circles. If the radius of circle A is a, the
radius of circle B is h, and the radius of circle X is x, prove \
I
.l.=.L+..L A I
I✓
fx fa IF .
-" B
1,2(1,,
5a. Prove: ½+ -J; + TI + ... + k( k + l) + •••= I •
(Hint: Use the fact that k(k + I)= f- r±-I- and rewrite each term.)
Sb. Prove: f+n+fo+ ... +k(k+l)(k+2)+ ... =½

(Hint: Again rewrite k( k + 1)(k + 2) as the sum of two unit fractions.)
J' U-1)! U-1)!

5c. Prove: In general, k(k + l)(k+. 2) ... (k+ j) = k(k+ l)(k+ 2) ... (k + j-1) (k+ l)(k + 2) ... (k + j) .
6. Consider the following triangle of unit fractions known as the Harmonic Triangle of Liebniz:
diagonals
1
4
l .l. l
5 20 5
1 1 1 1
o ".ID !Ci o
For any n 2: k 2: 1, {k} refers to the kth fraction in the nth row of this triangle, e.g. {~} = 3~ .
a) Generate the next two rows of the Harmonic Triangle.
b) Complete this equation: For any n 2: k 2: 1, {k} = { ~ }+{ t }.

c) 2
{'{}=¼, { }=n(n-l)' {:;}=n(n-l)(n- 2),etc. Findanexplicitformulatocalculate {k}·
(Hint: Think factorials!)
d) Although there is no simple formula for the sum of the numbers in row n, kti {1} ,there is a formula to
determine the infinite sum of the numbers in any diagonal where k is held constant, "t {1} . (Note: The one
310
exception is the first diagonal when k = I. J,, {'{} = I + ½ + f + ¼+ ... , which is the harmonic series and is
known to diverge, i.e., increases without any bound.)
i) For any kc::: 2, find a fonnula for Jk {k:} .

ii) Prove the fomrnla is true.
The Extensions (Not part of the contest)
The following are interesting facts about unit fractions taken from The Penguin Dictionary of Curious and
Interesting Numbers by David Wells:
a) ~ is the only positive fraction less than I that cannot be expressed as the sum of two unique unit
fractions and therefore was a uniquely represented "Egyptian" fraction.
b) Erdos and Sierpinski have conjectured, respectively, that ~ and ~ are each expressible as the sum of
three unit fractions.
c) In 2 = I -½ + f- ¼+ ½-...
4 = I _ l3 + l5 - l7 + l9 + ... (Leibniz - 1673)
d) zt
e) e = iTI" + TI +T! +:TT+ if+ ... (Newton - 1665)
t) Ifn is an even number, y, + r,r + :jn + f, +?···=a multipleof rc". For example, if n = 2, the sum
equals f2 4
and if n = 4, the sum equals ~. (Euler - 1736)
g) Every number greater than 77 is the sum of integers, the sum of whose reciprocals is 1. For example,
78=2+6+8+10+12+40 and ½+i+½+fo+72+fo=l. (Graham-1983)
h) A neat connection between Pascal's Triangle and the harmonic series:
(lxl)=l
(1 X ])-(] x½)=½
(1 x 1) -(2 x l) + (1 x l) = l
2 3 3
(Ix l)-(3x½)+(3x½)-(lx¼)=¼
(lx l)-(4x½)+(6x½)-(4x})+(l x½)=½
311
The Solutions
la.
lb .5_-l+..1.-1+1+.l.
. II - 3 33 - 3 9 99
IC .L-l+..5..-1+.l.+...L
. 16 - 3 48 - 3 10 240
2b j_ - j_ + j_ - j_ + ...1..
. 17 - 34 34 - 18 306
2c. If ti = ¼+ r, ,then x ,y > n . Let x =n + j and y = n + k , then
ln = ___L. + _l_k ⇒ (n + j)(n + k) = n(n + k) + n(n + )) ⇒ jk= n2 • Therefore, j and

k are divisors of n 2 •
n+ J n+
Since a square has an odd number of divisors, d , add 1 to this number ( because j and k can both be n ) and
then divide by 2 since the order of j and k is insignificant. Therefore, if d is the number of divisors of n 2 ,
then ¼ can be written as the sum of two unit fractions in 4½1 ways.
3. Let LM = m and MN= n, by similar triangles, f = tll..:fjil and 'f' = ~ and so m= ag- . From the first
{!!l + 11
equation, x = ~ . Therefore, f = ~ and substituting for m, ! = ~ ⇒ f = ~ ⇒ ½= j; + ¼.
4. ln the following diagram, using the Pythagorean Theorem three times, you get:
(a-b) 2 +/=(a+b/ (a-x) 2 +z2=(a+x/ (b-x) 2 +w 2 =(b+x) 2
y=2/aF z=2fax w=2/fix
/
But y =z+ w \
~
Therefore, 2/lF = 2/Tix + 2(ax
Dividing each tenn by 2/afjx , 7i = Fa + 7ff .
I
5a. For any k , k(k + l) = f- r±-I- .

Using this fact, rewrite each term as the difference if
't
X
a~b/
◄
y
►
two unit fractions: ◄ ►◄►

z w
(1-½)+(½-½)+(½-¼)+ ... + (t-;zh)+ ...
Summing the opposites, results in: 1 - r±-r
As k ➔ 00 , r+"f ➔ 0 . Therefore, the infinite sum is I.
312
--- ? - - = _....__
Sb. For any k, it is easy to show :
k(k + I )(k + 2) k(k + I) (k+l)(k+2)
Using this fact, rewrite each term as the difference if two unit fractions:
(½- i)+ <t-h) +( i½--do)+ ... + (k(k+ I) - (k+ l)(k+ 2))+ ...
Summing the opposites, results in: ½ (k + 1)(k + 2) •
As k ➔ 00 , (k + l)(k + 2) ➔ 0 . Therefore, the infinite sum is ½.

(i - I)! (j- I)! ( k + j)(j - I)! - k(j- I)!
Sc. k(k+ l)(k+2) ... (k+j- I) ---"-----=
(k+ l)(k+2) ... (k+j) k(k+ l)(k+2) ...(k+.i)
k(j- I)! + j(j- 1)! - k(j - I)!
= --------,---"'--
k(k + I )(k+ 2) ... (k+ j)
j(j -1) !
= ----'-"-------
k( k + I)( k + 2) ... ( k + j )
= ___.........i!____
k(k + I )(k + 2) ... (k + j)
I I _I_ I I I
6a. I
i 42 105 T40 105 42 7
I I l I I I I
8 3'6 Tii8 '18a 280 Tii8 56 8
6b. Each term of the harmonic triangle is the sum of the two terms below it.
Therefore' {n} = {17 + I } + {n + I }
k k k+I .
6c. {11} I 0!
J = ii =n
{n}- I!
2 - 11(11- I)
{n}- 2!
3 - n(n - I )(11 - 2)
{11} 3!
4 = n(n -1)(11- 2)(n- 3)
I 1 n} _ (k-1) !
n genera' { k - n(n -1 Xn - 2)...(n- k + 1) or,
{ 17 } (k-1)! (11-k)(11-k- l)(n-k-2) ...(3)(2)(1)

k = n(n-l)(n-2) ... (n-k+l) · (11-k)(11-k- l)(11-k-2) ...(3)(2)(1)
{ n} = (k-1) !(n - k)!

k n!
313
ARML Power Contest- November 1998 - Unit Fractions
Also notice the Harmonic Triangle is related to Pascal's Triangle:
Therefore, using the binomial coefficient, { 11}

k
= ln + (nk-1
- l)
- l _,_ (11- 1)!
- n. (k-1)!(11-k)!
_ (k- l)!(n-k)!
- n(n- l)!
= (k - I) !(n - k)!
n!
6dii. From 6b, {z} = { ! I} + {k: l}. If k 2, can be replaced with n - 1 and k can be replaced with
11 ?_ 11
k-1 , to get {z =i} = {k: l} + {k} . Therefore, {z} = {z =i }-{k: I}.
For n?_k?_2, ,,t{k}=J,<{k=}}-{k:i})

= Jk {t~ n-
Jk {k ~ ]}
= ± {nk-1
- l} + f {nk-]
n=k
- I}- f {k-1
n="'r'+I
n} n=k
= {k-1}
k-1 +O
_....1_
- k- 1.
314
ARML Power Contest - February 1999 - Chromatic Polynomials
Chromatic Polynomials
A graph, G, is a finite non empty set VG of objects called vertices (also called points or nodes) and a (possibly
empty) set EG of two-element subsets of VG called edges (or lines). Every graph has a diagram associated with
it. For example. a graph G defined by VG= {vJ, v2, v3, v4} and EG = (v1v2, v1v3, v2v3, v3v4) has the
following diagram:
vz
In a null graph (designated N), the set EG is empty. In a complete graph (designated K), the set EG contains all
possible two-element subsets of VG- Subscripts are used to indicate the number of elements in VG· As
examples:
v2 0
v2
O VJ O V3
VJ
Other special graphs are cycles (C) and paths (P). As examples :
V3
Note that K3 = C3 and P2 = K2 and that C1 and C2 are undefined.
A graph can be "colored" by assigning a color to each vertex so that vertices connected with a edge are assigned
different colors. The minimum number of colors needed to color a graph G is called the chromatic number of
the graph, Xe;. For example, XK 3 = 3, XP1 = 2, XA 4 = 4, XP 4 = 2.
315
Blue Blue
/ \
I
Red
l /\
Red
._ _ _
Yellow
Red Red
). Red
Blue :,Blue
Two colorings of a graph are considered different if they assign different colors to the same vertex of the graph. For
example. using two colors, P3 can be colored in two different ways; however K3 cannot be colored with only
two colors. Using three colors, P3 can be colored in twelve ways and K3 can be colored in six ways.
A chromatic polynomial of graph G, PQ(x), is the number of different colorings of G using x colors. For example,
usin1t, the graphs:
I
Go =o o G1= 3
2 3 2
If you have x colors to color Go, you have x choices for v 1 and also x choices for v2 and v3. Therefore, Pc, 0 (x) =
(x)(x)(x) = x 3 .
If you have x colors to color G 1, you have x choices for v 1, (x - I) choices for v2 (it cannot be colored the same as
v 1), and x choices for v3.
Therefore, Pa 1 (x) = (x)(x - 1)(x) = x3 - x2.
If you have x colors to color G2, you have x choices for v 1, (x - I) choices for v2 (it cannot be the same as v 1), and
(x - I) choices for v3 (it cannot be the same as v2).
Therefore, Pc; 2 (x) = (x)(x - I )(x - I)= x3 - 2x2 + x.
If you have x colors to color G3, you have x choices for v 1, (x - I) choices for v2, and (x - 2) choices for v3 (it
cannot the same color as VJ or v2).
Therefore, Pc; 3 (x) = (x)(x - I )(x - 2) = x3 - 3x2 + 2x.
Notice that Pa 2 (2) = 2, PG 3 (2) = 0, PG 2 (3) = 12, and PG 3 (3) = 6 are the same results as stated in the paragraph
above.
316
Theorems about Chromatic Polynomials
I. For a null graph. Nn, with exactly n vertices and no edges, PN" (x) = xn.
2. For a complete graph, K11 , with exactly n vertices any two of which are connected with an edge,
PA- n (x) = x(x - 1)(x - 2) ... (x - n + 1).
3. For a path of n vertices, Pn, Pl',, (x) = x(x - 1)n - 1.
4. If G is a graph with n vertices and m edges, then

i) Po(x) has degree n. ii) the coefficient of xn in Po(x) is 1.
iii) the coefficient of xn -1 in Po(x) is -m. iv) the constant term in Po(x) is 0.
v) the coefficients of Po(x) alternate in sign. vi) if mi= 0, then the sum of the coefficients of Po(x) is 0.
As the graphs become more complex. counting the number of colorings and the choices of colors becomes very
difficult and so the following recursive algorithm is very helpful. Repeated use of the algorithm shows for any
graph, Po(x) is the sum the chromatic polynomials of related complete graphs and so Theorem 2 can be
applied.
5. Let G be a graph containing two vertices VJ and v2 not connected by an edge. From G,
form a new graph G' by connecting v1 and v2 and then form another graph G" by
replacing VJ and v2 by a vertex v (not already in G) and connecting v to any vertices
which v1 and v2 were connected to in G. Then Po(x) = Po•(x) + Po"(x). The
following diagram shows how this theorem can be used three times to find the chromatic
polynomial of this graph.
A'r\\
// I
/// /
VIO C
I
I/
Po(x) +
317
+ P(G')"(X) + + p (G")"(X)
Therefore, PG(x) = P(G')'(X) + P(G')"(x) + P(G")'(X) + P(G")"(X)

= PK6(x) + PK5(x) + PK5(x) + PK4(x)
= (x)(x - 1)(x - 2)(x - 3)(x - 4)(x - 5) + 2[(x)(x - 1)(x - 2)(x - 3 )(x - 4)] + (x)(x - 1)(x - 2)(x - 3)
= (x)(x - 1)(x - 2)(x - 3)[(x - 4)(x - 5) + 2(x - 4) + l]
= (x)(x - l)(x- 2)(x - 3)(x2 - 7x + 13)
= x6- J3x5 + 66x4- 16lx3 +185x2 - 78x.
318
The Problems
I. Find the chromatic number for each of these graphs:
a. b.
2a. Consider graph GI at the right.

i) Using two colors, show that G l can be colored in two different ways.
ii) Using three colors, show that G1 can be colored in eighteen different ways.
2b. Consider graph G2 at the right.

i) Using two colors, in how many different ways can G2 be colored?
ii) Using three colors, in how many different ways can G2 be colored?
/
3. Using the counting method, find PQ(x) for each of these graphs.
02
(You may leave your answer in factored form.)
a. b.
4. Determine a suitable graph for each of the following chromatic polynomials:

a. Pc(x) = x 4 - x 3 c. Pc(x) =x 4 - 4x 3 + 6x 2 - 3x
b. Pc(x) =x 4 - 3x3 + 3x 2 -x d. Pu(x) = x 4 - 2x 3 + x 2
5. a. Prove Theorem 1.
b. Prove Theorem 2.
c. Prove Theorem 4 part vi (Hint: How can you find the sum of the coefficients of any polynomial?)
319
6. Using theorem 5, find Pa(x) for each of these graphs. (You may leave your answer in factored form.)
a. b.
7. Prove Theorem 5.
The Extensions
Chromatic Polynomials were first used by Birkhoff in 1946 in an attempt to solve the famous Four Color Problem.
This problem conjectured only four colors were needed to color any map so countries or regions sharing a
common boundary were always colored a different color. It was first proposed in 1852 by Francis Guthrie, a
student at the University of London. A proof of the theorem was attempted by many eminent mathematicians,
including August de Morgan and Arthur Cayley. Several proofs were published in the l 880's but all were
found to contain errors. Interest in the problem grew and spawned many new techniques in mathematics. In
Birkhoffs attempt, if a planar graph (a graph which can be drawn in a plane with no edges crossing except at a
vertex) could be represented by a chromatic polynomial, then it was only necessary to show that Pa(4) is
always at least I. He was unsuccessful in completing the proof. The Four Color Problem was finally solved in
1976 by Appel and Haken; however, their proof required extensive use of a computer to check the validity of all
cases.
Some further problems to consider:
1. IfG is a cycle with n vertices, Cn, then Pc,,(x) = (x- l)" + (-l)"(x- I).
2. IfG is a tree with n vertices, Tn, then PrnCx)=x(x- l)"- 1•
3. In problems 3a and 5b, you found the chromatic polynomials for the graphs associated with the tetrahedron and
the octahedron. What are the chromatic polynomials for the graphs associated with the other Platonic solids -
the cube, icosahedron, and dodecahedron?
320
The Solutions
la. Because the graph contains K3 as a subgraph, at least three colors are needed. The coloring below shows this is
minimal.
1b. The outer pentagon requires at least three colors. The coloring below shows this is minimal.
le. Since the graph contains an edge, at least two colors are needed. The coloring below shows this is minimal.
G
G
2a i. R B B R 2a ii. R B R G R G R B B R B G B R B G
BR RB B G G B BR GR R G GR G B RB
GR G B GR G B RB RG BR BG GR GB
R B BR B G R G BR GR RB GB RG BG
2b i. It cannot be 2b ii. SIX: R B R G B R B G GR GB

colored in two colors G R B R G B R B B G R G
3a. P(x) = (x)(x- 1) 5 =x<,_ 5x5 + 10x 4 - l0x 3 + 5x2 -x
3b. P(x) = (x)(x- l}'(x-2) =x 5 -5x 4 + 9x 3 - 7x 1 + 2x
3c. P(x) =(x)(x- l)(x - 2) 2 =x 4 - 5x 3 + 8x 2 -4x
4c. Pc(x) =x 4 - 4x 3 + 6x 2 - 3x =x(x -1)(x 2 - 3x + 3)
4b. Pc(x) =x4 - 3x3 + 3x 2 -x =x(x- 1) 3
r p
or or [
u
Sa. For each of then vertices you have x choices of colors. Therefore, Pu(x) = (x)(x)(x) ... <x) =x" .
n times
321
5b. Pick a vertex, v I; you have x choices of colors for v I - Pick a second vertex, v2. Since it is connected to v J,
you only have x - I choices of colors for v2. Pick a third vertex, v3. Since it is connected to v I and v2, you
only have x - 2 choices of colors for v3. Continue this process until there is only one vertex not colored, vn.
Since vn is connected to all the other n - I vertices, it can be colored using x - (n - I) colors.
Therefore, PK. (x) = x(x - 1)(x - 2) ... (x - n + I).
5c. Since m / 0, G has at least one edge. The two vertices connected by this edge, must be colored differently.
Therefore, at least two colors are needed to color G and so PG( I) = 0. But in any polynomial, P( I) equals the
sum of the coefficients. Therefore, if m / 0, then the sum of the coefficients of Pc(x) is 0.
6.a.
= +
+ + +
Ks
Therefore, Pc(x) = (x)(x- l)(x -2)(x-3)(x -4) + 2(xXx- l)(x-2)(x -3) + (x)(x- l)(x -2)
Pc(x) = (x)(x- l)(x - 2)(x 2 - 7x + 12 + 2x- 6 + 1)
Pc(x) = (x) {t-- l)t-2) (t- 2 - Sx + 7)
Pc(x) = x 5 - 8x 4 + 2 4t- 3 - 3 Jx2 + 1 4t-
//
6b.
,,,- - - + -- -----',-----',
322
+ +
<c---------t------r-
~~ ' ~ ./~->Ii ---..... i~ i
,. ,-~ "
Ks Ks
Pc(x) = (x)(x - l)(x - 2)((x - 3)(x - 4)(x - 5) + 3(x - 3)(x - 4) + 3(x - 3) + l)

Pc(x) = (x)(x - l)(x-2)(x 3 - 9x 2 + 29x- 32)
Pc(x) = x 6 - 12x 5 + 5 8x4 - 137x3 + l 54r 2 - 64x
7. Consider graph G with two vertices, V] and v2, which are not connected. The number of different colorings ofG
using x colors is equal to the number of such colorings of G where the color of v] is different from v2 plus the
the number of such colorings of G where the colors of v J and v2 are the same. The number of colorings of G
where v J and v2 are different is the same as the number of colorings of G', the graph where v I and v2 are
connected. The number of colorings of G where v I and v2 are the same color is the same as the number of
colorings ofG", the graph where VJ and v2 are removed and replaced by a new vertex v with all vertices that
were connected to VJ and v2 now being connected to vertex v.
Therefore, PQ(x) = PG•(x) + PG"(x).
323
ARML Power Contest - November 1999 - XXV-Point Geometry
Twenty-five Point Affine Geometry
Definition 1: A point is any letter A through Y. There are twenty-five points in this geometry. The points are
arranged in blocks in the following three ways:
A B C D E A I L T w A X Q 0 H
F G H [ J s V E H K R K I B y
K L MN 0 G 0 R u D J C u s L
p Q R S T y C F N Q V T MF D
u V wx y MP X B J N G E WP
Definition 2: A line is any row or column in one of the three blocks above. Therefore, a line contains five distinct
points. There are only thirty lines and every point is on six different lines in this finite geometry. For example,
THUNB is a line and point Tis on lines PQRST, EJOTY, AILTW, THUNB, VTMFD, and XKCTG.
Theorem 1: Given any two distinct points, there is one and only one line containing both of them. For example,
given points T and K, only line XKCTG contains both of them.
Definition 3: Two lines are parallel if they have no points in common. The points are arranged in the blocks so
that parallel lines must lie in the same block and both be rows or both be columns.
Theorem 2: Given any two distinct lines, either they are parallel or they have only one point in common. Lines
FGHIJ and IVOCP both contain point I.
Theorem 3: Given a line and a point not on the line, there is one and only one line containing the point and
parallel to the given line. For example, given ASGYM and point R, line LERFX is the only line containing
R and parallel to ASGYM.
Definition 4: Two lines are perpendicular if one of them is a row and the other a column in the same block. For
example, lines FGI-1IJ and CHMR W are perpendicular and intersect at point H while lines FGI-1IJ and LERFX
intersect at point F but are not perpendicular.
324
----------------------------------------------------------
Theorem 4: Through any point, there is one and only one line perpendicular to a given line. For example, given
AFKPU and point N, line KLMNO is perpendicular to line AFKPU and contains N.
Definition 5: The distance between any two points is the least number of steps separating
the points on the line which contains them. Lines do not have end points so when
considering distances, it might be best to think of the five points of a line as being
cyclic: For example, on line AFKPU, distance AF~ 1, distance AK = 2, distance
AP= 2 (not 3), and distance AU= 1 (not 4). Row distances do not equal column
distances so row distances will be distinguished from column distances by using a
prime. Therefore, on line ABCDE, distance AB= 1', AC= 2' , AD= 2' , and
AE = 1' . Therefore, between two distinct points there are only four possible distances: 1, 2, l ', and 2'. ln
this geometry, row distances and column distances are not comparable.
Definition 6 Point x is the midpoint of points a and b if x is on the line containing points a and b and distance ax
equals distance xb. For example, on line {EJOTY} point J is the midpoint of points E and O while point T is
the midpoint of points E and J.
Two figures are considered different if they have a different set of points.
The Problems
Part A Triangles
Definition: A triangle is a set of three non-collinear points.
1. How many different triangles are there?
2. There are three types of triangles: equilateral, isosceles, and scalene. Each comes in four different sizes. What
type of triangle is each of the following: FAR, NHL, and SLO?
3. Prove or Disprove: All scalene triangles are right triangles. (A right triangle has two perpendicular sides.)
4. Find the circumcenter, the centroid, and the orthocenter of triangle ABN. Show these points are collinear. (This
line is called the Euler Line.)
Part B Quadrilaterals
Definition: A quadrilateral is a set of four points, no three of which are collinear.
325
5. How many different quadrilaterals are there?
6. There are eleven types of quadrilaterals. Match each of these quadrilaterals with its distinct name. (Vertices of a
quadrilateral are named in order.)
AVQK rectangle
AKCR rhombus
ACQG general parallelogram (adjacent sides are not = or ..l)
AKLQ right trapezoid with 3 congruent sides (!)
MIBF right trapezoid with 2 congruent sides
VBDW isosceles trapezoid
BWHL
AGQI
BFVH
general trapezoid (no
kite
dart
..l 's) ~
\~ ;
I
I\
_/ \
I /- \
y\< ' ~
\
AYDW general quadrilateral (no II sides) V

APNS general right trapezoid (no =sides) kite dart
Once the eleven types of quadrilaterals have been identified, the following theorems can be easily seen. This is
NOT part of this Power Contest Question:
The opposite sides of a parallelogram (rectangle, and rhombus) are congruent.
The diagonals of a parallelogram (rectangle, and rhombus) bisect each other.
The diagonals of a rectangle (and isosceles trapezoid) are congruent.
The diagonals of a rhombus (kite, and dart) are perpendicular.
The diagonals of a right trapezoid with three congruent sides are parallel !
7. Prove or disprove: If the opposite sides of a quadrilateral are parallel and congruent, then the quadrilateral is a
parallelogram.
8. Prove or disprove: There are no squares in this geometry.
Part C Circles
Definition: A circle is the set of six points a given distance (called the radius) from a given point (called the
center).
9a. Find a circle centered at point E with a radius of2'.
9b. What is the center and radius of the circle {K, J, Y, P, V, G}?
326
I 0. How many different circles are there?
11. In any triangle, the three midpoints of the sides, the three feet of the altitudes, and the midpoints of the
segments joining the vertices to the centroid all lie on a circle called the Feuerbach Circle. Consider again the
triangle ABN from problem 4 in Part A. What is the center of its Feuerbach Circle?
12. Prove or Disprove: A line through the center ofa circle always intersects the circle at two points.
Part D Conics
Definition: A parabola is the set of points equidistant from a given point (called the focus)
and a given line (called the directrix.)
13. How many different parabolas are there?
directrix
14. Determine the set of five points of the parabola whose directrix is line AFKPU and
focus is point M.
Definition: A line tangent to a parabola intersects the parabola at only one point and is
not perpendicular to the directrix. Of the five points of a parabola, one is the vertex and
the other four form two pairs of endpoints of a focal chord (a chord going through the
focus). focal chord
15a. Consider parabola {TIGED} whose focus is point X and directrix is line {YCFNQ}.
Determine the five tangent lines to this parabola.
15b. Make a conjecture regarding the tangent line through the vertex. Support your conjecture with evidence.
15c. Make a conjecture about a pair tangent lines that are tangent to the endpoints ofa focal chord. Your conjecture
should involve the directrix. Support your conjecture with evidence.
327
The Solutions
1 (25)(24)(20) = 2000
. 3!
2. FAR = equilateral; NHL = isosceles; SLO = scalene
3. For a triangle to be scalene all sides must have different lengths. Since there are only two row distances and
only two column distances, one side of the triangle must have a row distance length and another side must have
a column distance length. Since the lines containing these sides must intersect, they must lie in the same
block with one a row and the other a column. Therefore, these two sides are perpendicular and hence the
triangle must be a right triangle.
4a. The circumcenter is the intersection of the perpendicular bisectors of the

{JCUSL}
sides of the triangle. Therefore, the circumcenter is S.
A B
{DINSX}
4b. The centroid is the intersection of the medians of a triangle. Therefore
the centroid is X.
:AXQOH}
{MPXBJ}
4c. The orthocenter is the intersection of the altitudes of a triangle. A B

{DINSX}
Therefore, the orthocenter is I.
LTW}
Points S, X, and I all lie in line {DINSX}, the Euler Line of this
{RKIBY}
triangle.
f ·~ B
5 (25)(24)(20)(13) =6500 A {DINSX}
. 4!
6. A VQK: AK II VQ and AV = QK. Therefore, A VQK is a isosceles trapezoid.
AKCR: AK 11 RC, AR II KC, no perpendiculars. AKCR is a general parallelogram.
ACQG: AC = CQ, QG = GA, AC J_ QG at B, CQ J_ GA at Y. ACQG is a dart.
AKLQ: AK II QL, AK J_ KL, KL J_ QL, no opposite congruent sides. Therefore, AKQL is a general right
trapezoid.
328
MlBF: MI 11 BF, 18 11 MF, 18 .l IM. Therefore, MlBF is a rectangle.
AGVF: GV II FA, AG f. VF, no perpendiculars. Therefore, AGVF is a general trapezoid.
VBOW: VW II BD, VB .l WB, VB .l BD, no congruent sides. Therefore, VBDW is a general right trapezoid.
BWHL: BW II HL,WH II LB, BW = WH ~ HL = LB, no perpendiculars. Therefore, BWHL is a rhombus.
AGQI: no parallels, AG= GQ, QI f. IA. Therefore, AGQI is a general quadrilateral.
BFVH: no parallels, BF = HB, FY = VH, HB _l YB. YH _l HB. Therefore, BFVH is a kite.
A YOW: A Y I OW, OW _l WA _l A Y, AY =YD= OW f. WA. Therefore, A YOW is a right trapezoid with
three congruent sides.
APNS: AP II NS, no perpendiculars. Therefore, APNS is a general trapezoid.
7. False. Quadrilateral A YOW is a counter example. AY II OW and A Y = OW= 2 but

YD is not II WA. Therefore A YOW is not a parallelogram.
8. Trne. A square is a rectangle with congruent sides. Therefore, a square must have perpendicular and congrnent
adjacent sides. However, this is impossible since perpendicular sides must lie in the same block, one in a row
and the other in a column but row lengths are never equal to column lengths.
9a. {BKNCSP}
9b. center= A, radius = 2. Consider the midpoints of the five chords with endpoint K, the midpoints of the five
chords with endpoint J, and the midpoints of the five chords with endpoint Y. These sets have only point A in
common. (Or alternatively, the perpendicular bisector of chord KJ is {AIL TW} and the perpendicular bisector
of chord KY is {ARJVN}. They intersect at the center A.)
10. (25)(4) =JOO.
11. The midpoints of the sides are J, D, and H and the feet of the altitudes are R, T and D. The circle containing
these points is centered at N with a radius of 2.
12. False. Circle {HlBXWE} is centered at A with a radius of I '. However, line {AFKPU} goes through the
center A but does not contain any point on the circle.
13. (30)(20) = 600 .
329
14. I' from {AFKPU} and 1' from point M = {LJT}; 2' from {AFKPU} and 2' from point M = {DX}.
Therefore, the parabola is {LJTDX}.
15a. {PQRST}, {RKIBY}, {BGLQV}, {SVEHK}, and {HYLDP}
15b. The tangent line through the vertex is parallel to the directrix. E must be the vertex since T, X, and G lie on
line {XKCTG} and I, X, and D lie on line {DINSX}. {SVENK} is the tangent line through E. Both tangent
line {SVENK} and the directrix {YCFNQ} are rows in block 2 and hence they are parallel.
15c. Tangent lines through the endpoints of a focal chord are perpendicu-lar and intersect on the directrix. Points T
and Gare the endpoints ofa focal chord. {PQRST} and {BGLQV} are tangent lines through these points.
They are perpendicular in block I and intersect at Q, a point on the directrix {YCFNQ}. (Also, points I and D
are the endpoints of a focal chord. {RKIBY} and {HYLDP} are tangent lines through these points. They are
perpendicular in block III and intersect at Y, a point on the directrix {YCFNQ}.)
The Extensions
Many more theorems and investigations of this geometry are presented in Puzzles and Paradoxes by T. H.
O'Beime, including its association with a triangular lattice system. Of unusual interest is how this geometry is
used to solve a problem regarding the placement of photographic equipment around a circular race track!
330
ARML Power Contest - February 2000 - Square-Sum Partitions
Square-Sum Partitions
The Definitions
Consider a set A of distinct elements. A partition of set A is a set of disjoint subsets of A whose union equals set
A. For example, the set { { 1, 3}, {2, 4, 6}, {5}} is a partition of the set {I, 2, 3, 4, 5, 6}. A square-sum
partition of set A is a partition of A in which the elements in each of the subsets in the partition add up to a
square number. For example, the set { {0, I, 6, 9}, {2, 7}, {3, 5, 8], {4}} is a square-sum partition of
{O, I, 2, 3, 4, 5, 6, 7, 8, 9).
For ease in correcting please write all subsets in ascending order and all partitions in ascending according to the first
element of each subset as shown in the examples.
The Problems
Part I Square-Sum Pairs Consider the set { I, 2, 3, 4, 5, 6, 7, 8}. It can be partitioned into square-sum pairs:
( { I, 8}, {2, 7}, {3, 6}, i 4,5} } in which each subset contains a pair of elements that add up to a square
number. We will now investigate the partitioning of the set {0, I, 2. 3 .... , n} into square-sum pairs.
1. Partition the sets {0, I, 2. 3, ... , n} into square-sum pairs where:

a) n =1 b) n = 7 c) n = 9
2. Prove: If n is the square of an odd number, then {O, 1, 2, 3, .... 11} can be partitioned into square-sum pairs.
3. Prove the sets {O. I, 2. 3, ... , n} cannot be partitioned into square-sum pairs when:
a) n = 3 b) n = 5 c) ll = 11
4. Sometimes there exists a number, S, the square of an odd number less than 2n, that can be used to partition
{0,1,2,3, ... ,n} into two sets {0,1.2,3, ... ,S-n-l} and {S-n,S-n+l,S-n+2, ... ,n-1,n} where
the first set, {0, l, 2, 3, ... , S - n - 1}, is already known to be partitionable into square-sum pairs and the second
set, { S - n, S - n + 1, S - n + 2, .... n - 1,n}, is easily shown to be partitionable into square-sum pairs. For
example, given the set {O, I, 2. 3, ... ,15} and S = 25, then S - n = 10 . {0, 1, 2, 3, ... , 15} can be partitioned
into sets {0. I, 2, 3, .. ., 9} and { 10, 11, 12, 13, 14. 15} . In exercise I you demonstrated {0, 1, 2, 3, .... 9} is
partitionable into square-sum pairs and {10, 11, 12, 13,14, 15} can be partitioned into {IO. 15}, {11. 14}, and
{12, 13}, all of which are square-sum pairs. Use this technique to partition {0, 1, 2. 3, ... , n} into square-sum
pairs where:
a) n = 17 b) n = 23 c) n = 27 d) n = 29 e) n = 31 f) n = 33 g) n = 35
331
5. Sometimes the above technique does not work. Partition the set {0, 1, 2, 3, ... ,n} into square-sum pairs where:
a) n = 13 b) n = 19 c) 11 =21 d) 11 = 37 e) 11 = 39.
Theorem: For any odd number, 11 2: 41, there exists an odd square number between n + 13 and 2n.
6. Use the above theorem and the results so far to prove the set {0. I. 2. 3 ..... 11} can be partitioned into square-
sum pairs for any odd n except n = 3, 5, and 11.
Part II Square-Sum Triples Can the set {0. 1. 2. 3 .... , 11}. where n is one less than a multiple of three, be
partitioned into subsets of three numbers which add up to a square number? It is impossible for the set
{0, I. 2} .
7. Prove it is impossible to partition the set {0, I, 2, 3, 4, 5} into square-sum triples.
8. Show {0, 1, 2, ... ,8} can be partitioned into square-sum triples.
9. Show {0. 1. 2 .... , 11} can be partitioned into square-sum triples.
The Extensions (not part of the contest)

1. {0, 1, 2, ... ,17} can be partitioned into square-sum triples, {{0, 3, 6}, {l, 7, 8}, {2, 5, 9}, {4, 10, 11}} but
{0, 1, 2, ... ,14} cannot be partitioned into square-sum triples. Can you prove this?
2. Does there exist a number k, such that if n ~ k, it is always possible to partition {0, 1, 2, 3, ... ,3n-l} into
square-sum triples?
3. Can you prove the theorem used in problem 6?
4. Using the Bertrand-Cebysev Theorem [1850] which states for any n > 3, there exists a prime number between
n and 2 n, it can be proven, for any even number, 11, it is possible to partition the set {I, 2, 3, ... ,n} into
prime-sum pairs.
5. {0, 1, 2, ... , 15} can be partitioned into square-sum pairs in two different ways. Compare your answer to 4a
and the following set { { 0,4}, {1,8}, {2,14}, {3,6}, {5,11 }, {7,9}, {10,15}, {12,13} }. The set
{0, 1, 2, ... , 17} can be partitioned into both square-sum pairs and square-sum triples (See 4b and Ext. 1).
Can you write a computer programs to find all the square-sum partitions of the set {0, 1, 2, 3, ... ,n} for a
particular odd n?
332
The Solutions
1a. { {0, I} } lb. { { 0, l}, {2, 7}, {3. 6}, {4,5} le. { {0, 9}. { I, 8}, {2, 7}, {3, 6}, {4, 5} }
2. If n is the square of an odd number then {0, I, 2, 3, ... , n} can be partitioned as { (0, n}, {1, n - I},
{2, n - 2} ... {(n - 1)/2, (n + 1)/2}}, where each of the n/2 pairs add up ton, a square number.
3a. No other number of the set can be paired with 2 to add up to a square.
3b. The only number that can be paired with 5 is 4, leaving the set {0, I, 2, 3} which cannot be not partitioned
into square-sum pairs. Or again, no other number of the set can be paired with 2 to add up to a square.
3c. {11, 5} and { I 0, 6} must be pairs since a sum of 25 cannot be reached with this set. 9 can be paired with O or
7, but 2 must be paired with 7 and so 9 must be paired with 0. This leaves { I, 3, 4, 8} but 4 cannot be paired
with any of the remaining numbers to form a square.
4a. S = 25. (0, 1. 2, ... , 17} = {{17, 8}, {16, 9}, (15, 10), {14, II}, P3, 12}, {0, I, 2, ... , 7}} where {0, I,
2, ... , 7} was partitioned in problem I b.
4b. S = 25. {O, I, 2, ... , 23} = {{23, 2}, {22, 3}, (21, 4), (20, 5}, {19, 6}, {18, 7}, {17, 8), (16, 9).
{15, 10}, {14, 11}, {13, 12}, {O, l}} where {0,1} was partitioned in problem la.
4c. S = 49. {0, I, 2, ... , 27} = { {27, 22}, {26, 23}, {25, 24}, {0, 1, 2, ... , 23}} where {0, 1, 2, ... , 23} was
partitioned in problem 4b.
4d. S = 49. {0, I, 2, ... , 29} = { {29, 20}, {28, 21 }, (27, 22}, {26, 23}, {25, 24}, {0, 1, 2, ... , 19}} where
{0, 1,2, ... , I 9 J is partitioned in problem 5b.
4e. S = 49. {0, 1. 2, ... , 31} = { {31, 18}, {30, 19}, {29, 20}, {28, 21 }, {27, 22}, {26, 23}, {25, 24},
{0, 1, 2, ... , 17}} where {0, 1, 2, ... , 17} was partitioned in problem 4a.
4f. S = 49. {0, 1, 2, ... , 33} = { {33, 16}, {32, 17}, {31, 18}. {30, 19}, {29, 20}, {28, 21 }, {27, 22}, {26, 23},
{25, 24}, {O, I, 2, ... , 15}} where {O, I, 2, ... , 15} was partitioned in problem 4 example.
333
4g. S = 49. (0, I, 2, ... , 35} = { {35, 14}, (34, 15}, (33, 16}, (32, 17}, {31, 18}, {30, 19}, {29, 20}, {28,
21}, {27, 22}, {26, 23}, {25, 24}, (0, 1, 2, ... , 13}} where {0, 1, 2, ... , 13} is partitioned in problem 5a.
5a. {0, 1, 2, ... , 13} = {{0, 9}, { 1, 8}, (2, 7}, {3, 13}, {4, 12}, {5, 11 J, (6, 10}}
5b. Three possible answers: {0, 1, 2, ... , 19} = { {0, 1}, (2, 14}, {3, 13}, {4, 12}, {5, 11), (6, 19}, 17, 18},
{8, 17}, {9, 16}, {10, 15}} = {{0, 4}, {1, 3], {2, 14), {5, 11}, {6, 19}, {7, 18}, {8, 17}, {9, 16},
po, 15}, { 12, 13}} = { {O, 4}, { 1, 8}, (2, 14), {3, 6}, {5, 11 }, {7, 18}, {9, 16}, {10, 15], { 12, 13 },
{17,19}}
5c Four possible answers: {0, 1, 2, ... , 21} = {{0, 9}, {l, 3}, {2, 14}, {4, 21 }, {5, 11 }, {6, 19}, {7, 18},
{8, 17}, {10, 15}, {12, 13}, (16, 20}} = {{0, 9}, {l, 8}, {2, 14}, {3, 13}, {4, 12}, {5, 11}, {6, 10},
{7, 18}, {15, 21}, {16, 20}, {17, 19}} = { {0, 16}, {l, 8}, (2, 14}, {3, 13), {4, 12}, {5, 11}, {6, 10},
{7, 9}, (15, 21 }, {16, 20}, {17, 19}} = { {0, 9}, {l ,8}, {2, 14}, {3, 6}, {4, 21}, {5, 11}, {7, 18},
(10, 15}, {12, 13}, {16, 20}, {17, 19}}
5d. {0, 1, 2, ... , 37} = { {0, I}, {2, 7}, {3, 22}, {4, 21}, (5, 11 }, {6, 10}, {8, 28}, {9, 27}, { 12, 37},
{13, 36}, {14, 35}, {15, 34}, {16, 33}, {17, 32}, {18, 31}, {19, 30}, {20, 29}, {23, 26}, {24, 25}}
5e. {0, 1, 2, ... , 39} = { {0, 1 }, {2, 34}, {3, 22}, {4, 32}, {5, 31}, {6 ,30}, {7, 18}, {8, 28}, {9, 27},
{10, 39}, {11, 38}, {12, 37}, {13, 36}, {14, 35}, {15, 21}, {16, 33}, {17, 19}, {20, 29}, {23, 26},
{24,25}}
6. In questions 1, 4, and 5 above, it was demonstrated that for n = odd numbers 1, 7, 9, 11 through 39, the set
{0, I, 2, ... , n} can be partitioned into square-sum pairs. It must be now shown that for n c::: 41, the set
{O, 1, 2, ... , n} is also partitionable into square-sum pairs. Proof by Induction: Assume for some k 2'. 39
where k is odd, {0, 1, 2, ... , k} can be partitioned into square-sum pairs, prove {0, 1, 2, ... , k, k + 1, k + 2}
can also be partitioned into square-sum pairs. By the theorem above, for n c::: 41 there exists an odd square, s,
such that n + 13 < s < 2n. Replacing n with k + 2, implies that there exists an odd square, S, such that
k + 15 < S < 2k + 4.
Since k + 15 < S < 2k + 4, 13 < S - k- 2 < k + 2. Therefore, the set {0, 1, 2, ... , k, k + 1, k + 2} can be
partitioned into two sets: {0, I , 2, ... , S - k 3} and {S - k - 2, S - k - 1, .. . , k + 1, k + 2} . Since
334
S - k - 3 is an odd number 2: 13, the induction hypothesis states the set, {0, I, 2, ... , S - k - 3}, is
partitionable into square-sum pairs and the set [S k - 2, S k -1, ... , k + I, k + 2} = { (S - k - 2, k + 2},
{S-k-1,k+ !}, ... , {(S-1)/2,(S+ 1)/2}},whereallpairsadduptoS.
7. The numbers in the set sum to 15, so you need two squares that sum to 15. The smallest possible square triple
sum is 4 and the largest would be 9. However, using these square numbers, the only sums possible are 8, 13,
and 18. Therefore, since there do not exist two squares that add up to 15, {0, I, 2, 3, 4, 5} cannot be
partitioned into square-sum triples.
8. {0, I, 2, 3, 4, 5, 6. 7, 8} = { {0, I, 3}, {2, 6, 8}, i4. 5. 7)}
9. Twelve possible answers! : {0, I, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11} = { {0, 3, 6}, { I, 7, 8}, {2, 5, 9},
{4, 10, 1 I}}= { {0, 3, 6), { I, 7, 8}, (2, 4, 10}, {5 ,9, 11}} = { {0, 6, 10), {l, 7, 8}, {2, 3, 4}, {5, 9,
II}}= {{0, 7, 9}, {1, 3, 5}, {2 ,6, 8}, {4, 10, 11}} = { {0, 1, 8), (2, 5, 9}, {3 ,6, 7}, {4, 10, II}}=
{ i0, 2, 7}, {I, 4, 11 }, {3, 5, 8}, {6, 9, 10}} = { [0, 2, 7), { I, 5, 10}, {3, 4, 9}, {6, 8, 11}} = { [0. 7, 9},
(I, 3, 5}, (2, 4, 10), {6, 8, 11}} = {{0, 2, 7, [I, 6, 9}, {3, 5, 8l, {4, 10, 11}} = { i0, 5. II}, {I, 2, 6),
(3, 4, 9}, {7, 8, 10}} = { {0, 3, 6}, {l, 4, 11 }, {2, 5, 9), {7, 8, 10}} = { {0, 4, 5}, { I, 7, 8}, [2, 3, 11 },
(6, 9, 10)}
335
ARML Power Contest - November 2000 - Slides, Rolls, and Ro/ides
Slides, Rolls, and Rolides
The Definitions
In this problem you will be investigating various polygons and circles moving clockwise inside a rectangle always
staying in contact wilh the rectangle. Three different movements will be considered:
Slides - In this movement, the orientation of the polygon will not change. The u
polygon will move up, then right, down, and then left around the rectangle,
again always staying in contact with the rectangle.
'
GJ
Rolls - In this movement, the polygon will rotate about a vertex which is in contact
with the rectangle, "rolling around" the inside of the rectangle. The orientation
of the polygon will be constantly changing. The rotation of the polygon will
always be counter-clockwise, enabling the polygon to roll clockwise around the
rectangle. A circle will roll "without slipping."
Rolides - This movement is a combination of slides and rotations. In this

movement the polygon slides into the next comer and then rotates 90°
clockwise, always keeping two points of the polygon in contact with the
rectangle, and then slides into the next comer and rotates again. In this
movement, a polygon will make one complete (360°) rotation in one trip around
the rectangle.
In a trip around the rectangle, whether by sliding, rolling, or roliding, the polygon must come back to its original
position and original orientation.
The Problems
Part A - Slides
I. Consider a rectangle ABCD "sliding around" inside a rectangle, prove that the distance traveled by vertex A is
equivalent for both starting positions shown in the figures below.
C D
BA
fig. la fig. I b
336
2. Consider a right triangle ABC, with CA = 3, CB= 4 and AB= 5, "sliding around" inside an 9 by 12
rectangle. How much further does vertex C travel in one trip around the rectangle, if it starts in the position in
figure 2a versus if it starts in the position in figure 2b?
C
A1~ r~~
C ~B~~~
A B
fig.2a fig.2b
3. Consider right triangle ABC, with integer length sides and right angle at C, "sliding around" inside a rectangle.
When positioned as in figure 2a above, the distance traveled by point Cina trip around the rectangle is 12
units longer than the distance traveled by C when the triangle is positioned as in figure 26. Determine the
lengths of the sides of triangle ABC.
Part 8 - Rolls
Consider a unit square ABCD "rolling around" inside an 9 by 13 rectangle as in
figure 4. 6
,- I
12
fig.4
4a. How far does vertex A travel in one trip around the rectangle? 81 C
46. How far does the center of the square travel in one trip around the rectangle? A D
Consider rectangle ABCD, with AB= 1 and AD= 2, "rolling around" inside a 6 by 12 rectangle.
12
5a. How far does vertex A travel in one trip around the rectangle, if it starts in
the position in figure 5a? 6 L fig.5a
BL_,~ -
A D
5b. How far does vertex A travel in one trip around the rectangle, if it starts in
the position in figure 5b
6 fig.5b
ctf_
BA
Consider a circle whose diameter is I "rolling around" inside a (n + 1) by (3n+ I) rectangle.

3n+l
6a. How far does the center travel in one trip around the rectangle?
6b. How far does tangent point A travel in one trip around the rectangle, if it n+ J.
Al_
~ i
fig.6
starts in the position in figure 3? (Hint: "The length of a cycloid, from
cusp to cusp, is 4 times the diameter of the generating circle." Sir
Christopher Wren, 1658)
337
Part C - Rolides
12
7. Consider a circle whose diameter is I "roliding around" inside a 9 by 12
rectangle. How much farther does tangent point A travel than does the fig. 7
center travel in one trip around the rectangle?
Consider a unit square ABCD (with point A starting in the comer), "roliding around" inside a 9 by 12 rectangle.
8a. How far does point A travel in one trip around the rectangle?
8b. How far does the center of the square travel in one trip around the rectangle?
Theorem: Consider a right triangle ABC with right angle at C "roliding around" inside a rectangle. As
segment AB moves around any comer, from position I to position 3, the locus of points traced out by
vertex C is a line segment whose length is 2c - a - b.
l~~~
B C A
before the rotating during the rotating after the rotating
9. Consider a 3-4-5 right triangle with right angle at C "roliding around" inside a 9 by 12 rectangle,
starting as in position I above, how far does vertex C travel in one trip around the rectangle?
10. A right triangle, starting as in position 1 above, "rolides around" inside a rectangle. If b =a+ I and
vertex C travels 48 cm less than either vertex A or B, what are the lengths of the three sides of triangle
ABC?
11. Prove the above theorem.
338
The Solutions
1. Let W and L be the length and width of the large rectangle and w = AB and / = BC. The distance traveled
by point A in fig. la is 2( W - w) + 2(L- l) and the distance traveled by point A in fig. Ibis
2(W - /) + 2( L - w) . Both when expanded equal 2 W + 2L- 2w - 21 .
2. In fig.2a the distance traveled is 6 + 8 + 6 + 8 = 28 . In fig.2b, using similar triangles, the altitude from vertex
C is 2.4. Therefore, the distance traveled is 6.6 + 7 + 6.6 + 7 = 27 .2 . Therefore, the difference is 0.8
3. The distance traveled in Fig.2a is D1 = 2(W-a + L -b). The distance traveled in Fig.2b is
D 2 = 2( W - h + L - c) . The difference is 12 = D 1 - D2 = 2(h + c - a - b) . Therefore, 6 = h + c - a - b . Since
a, b, mu! c are integers, h must be an integer. Using the area formula for the triangle, ½(base Xheight),
½ab= ½he and so h =Of-, i.e., h must divide ah . If {a, b,c} is a primitive Pythagorean triple, i.e., a
Pythagorean triple with no common factors, h will not divide ab . To produce the desired results, multiply
the sides by c.
Looking at common primitive Pythagorean triples, the

following table is produced: C
a h C ac be c2 ab ab + c 2 - ac - be
3 4 5 15 20 25 12 2
5 12 13 65 156 169 60 8
8 15 17 136 255 289 120 18
7 24 25 175 600 625 168 18
20 21 29 580 609 841 420 72
Since in our problem ab + c 2 - a: - be = 6 , the only possible triangle is a multiple of the 15, 20, 25 right triangle.
To produce ab+ c 2 -a: -he= 6 , a triangle with side lengths 45. 60, 75 would be needed.
4a. On each roll the center will travel ~ units. There will be a total of 40 rolls and so the center will travel
10/2 n units.
4b. When the square "rolls" on vertex B, vertex A traces out a circular arc whose length ( ~ = ~ ) is equal to ~ ;
339
when the square "rolls" on vertex C, vertex A traces out a circular arc whose length is equal to ~ ; when the
square "rolls" on vertex D, vertex A traces out a circular arc whose length is equal to ~ ; and when the square
"rolls" on vertex A, vertex A does not move. Starting with A in the lower-left comer, vertex Bends up in the
upper-left comer, vertex C in the upper right, and D on the lower right corner. There will be a total of ten rolls
on each vertex. Therefore, vertex A will travel 10( ~ ) + 20( ~) = (5 /'I + 10)7t units.
5. Again using arc length = ~ = ~ , when the square "rolls" on vertex B, vertex A traces out a circular arc whose
length is equal to ~ ; when the square "rolls" on vertex C, vertex A traces out a circular arc whose length is
equal to ~ ; when the square "rolls" on vertex D, vertex A traces out a circular arc whose length is equal to
1t; and when the square "rolls" on vertex A, vertex A does not move. Therefore, the total distance vertex A
travels in fig.Sa is 6(~) + 6( ~ ) + 6( 7t) = (9 + 3/s)rr and the distance it travels in fig.Sb is
6a. Ans: 81t
6b. Ans: 32
7. Ans: 1t When the circle slides from corner to comer, the center and vertex A travel the same distance. When
turning in each corner the center is stationary while vertex A travels a quarter of the circumference or f.
8a. Ans: 42. When roliding around the rectangle, vertex A is always in contact with the rectangle and never needs
to backtrack.
8b. Ans: 46 - 4/2. When sliding inside the rectangle, the center of the square will travel on an 8 by 11 rectangle
and when the square rotates around a comer, it travels 2 - /2 units. Therefore, the total distance is
2( 8 + 11) +4(2 - /2) =38 + 8+ 4/2.
9. Ans: 34. Because side AB is always in contact with the rectangle when the triangle is sliding, vertex C travels
2(9 - 5 + 12 - 5) = 22 units. While the triangle rotates in each comer, vertex C travels 2(5) - 3 - 4 = 3 ,
Therefore, the total distance traveled is 22 + 4(3) = 34 units.
340
10. When MBC rolides around a rectangle with dimensions L hy W, vertex A travels 2(L + W) units, while
vertex C travels 2( W - c + L - c) + 4(2c- a - h) units.
Therefore, 2W + 2L- (2W- 2c +2L -2c + 8c -4a - 4h) =48, which simplifies to a+ b -c = 12. Since
b =a+ I , a+ b -c = 12 reduces to 2a - c = 11 or c = 2a - 11 . Substituting these equations into
a 2 + b 2 = c 2 yields a 2 +(a+ I/ =(2a - 11)2, which simplifies to a 2 - 23a + 60 = (a -20)(a- 3) = 0. And
so or. If a= 3 then MBC has sides of lengths 3. 4. and -5. an impossibility. Therefore, a= 20 and the
lengths of the sides are 20, 21, and 29 .
11. Part i) Consider right MBC roliding around a corner in the general position in figure 11 a. Since both
triangles ABC and ABO are right triangles with hypotenuse AB, ABCO is a cyclic quadrilateral (figure 11 b) .
...,. ,...
Draw ray OC (figure I le). L COA is always equal to L. B (of MBC) since both angles intercept arc AC.
Therefore, vertex C will always travel on a line when roliding around a corner.
Part ii) When right MBC rolides into a comer, vertex B is in the corner and vertex C is a units from the
corner (figure l ld). As MBC rolides it eventually gets to the position in figure I le, with the legs parallel to
the sides of the rectangle. Now vertex C is c units from the comer and hence has traveled c -a units in
going from figure l ld to figure I le. As MBCcontinues to rolide, eventually it gets to the position in figure
I lfand is ready to leave the comer. Vertex C is now b units from the comer and so in going from figure I le
to figure 11 f, vertex C has traveled c - b units. Therefore, vertex C will travel a total of 2c -a - b units
when roliding around a comer.
C C C
B B B
0
N A figure I la 0
N A figure 11 b figure 11 c
C B~i.C
~----~, -
C I
\
B A figure I Id A A figure 11 f
figure 11 e
341
ARML Power Contest - February 2001 - Pythagorean Triples
Pythagorean Triples
The Definition
A Pythagorean Triple {PT) is a set of three integers, {a, b. c} , in which a 2 + b 2 = c 2 • Therefore, a and b are the
lengths of the legs of a right triangle and c is the length of its hypotenuse. The ancient Babylonians were
familiar with these triples for their famous cuneifonn tablet, Plimpton 322 (c 2000 BC), lists fifteen
Pythagorean triples. The author of this tablet apparently knew for any pair of integers, 111 and n (with 111 > n ),
the set {211111,111 2 -11 2, 111 2 + n 2 } always produced a PT. (Ifm and n are relatively prime these formulas produce
a primitive PT, a PT with no common factors.)
The Problems
Part A - Generating Pythagorean Triples
Algorithm #1 For any pair ofrelatively prime integers, 111 and n (with 111 > n), {2mn,m 2 -n 2, m 2 + n 2 } is a PT.
la. Use the above algorithm to generate ten primitive PTs.
lb. Prove {211111,111 2 -n 2 , 111 2 + n 2 } is always a PT.
Algorithm #2 Take any two consecutive even or odd integers and add their reciprocals. The resulting numerator
and denominator are a and b of a PT.
2a. Show this algorithm works for the numbers 13 and 15.
2b. Prove this algorithm always works.
Algorithm #3 Take any two fractions whose product is 2. Add 2 to each number and rewrite each as an improper
fraction. Cross-multiply to produce two numbers a and b of a PT.
3a. Show this algorithm works for the numbers i and 1·

342
ARML Power Contest- February 2001 - Pythagorean Triples
Algorithm #4 Let a be any rational number greater than 1. Find two numbers, b and c, that differ by I and
whose sum is a 2 • The resulting set { a, b, c l is a rational Pythagorean triple. Multiplying each member
of the set by the common denominator of a, b, and c will produce a PT.
4a. Show this algorithm works for j.
4c. The PT you found in part 'a' of this problem is the same PT which would have resulted if you had started
with i! In addition, both ~ and ~ produce {28, 45, 53} and 1and 7 both produce (7. 24, 25). When
using this algorithm, for every fraction there exists a second fraction which produces the same PT. Looking al
patterns in the examples given, what other fraction will produce the same results as ~ ? Justify your claim.
4d. Using this algorithm, what rational number will produce the same PT as !ff ? Prove your conjecture.
Algorithm #5 Consider the Fibonacci numbers: 1, 1, 2, 3. 5, 8, 13, 21, 34, 55, 89, ... Take any four consecutive
Fibonacci numbers. The product of the outer two terms and twice the product of the inner two tenns will be a
and b of a PT. The value of c in the PT will also be a Fibonacci number and its subscript will be half the
sum of the subscripts of the original four numbers! (Charles Raine)
5a. Show this algorithm works using the numbers 5, 8, I 3, and 21.
Part B - Properties of Pythagorean Triples
6. From your list of primitive PTs, notice that
a) either a or b is a multiple of 3.
b) either a or b is a multiple of 4.
c) either a, b, or c is a multiple of 5.
Prove each of these results.
343
7. Pierre Fennat proved every prime of the form 411 + I (that is, the primes 5, 13, 17, 29, 37, ... ) is the sum of
two squares in exactly one way, while a prime of the form 411 + 3 (such as 3, 7, 11, I 9, 23, ... ) is never the
sum of two squares. Find two PTs that contain the prime number 929.
8. For any PT, { a, b, c }, the expressions, c· 2 +ab and c 2 -ab, can both be written as the sum of two square
numbers.
a. Show this works for the PT { 12, 35,37 }.
b. Prove this always works.
9. For any PT { a, b, c }, it is always possible to find positive integers r, s, and t, such that a= r + s and
b = s + t where r 2 + s 2 + t 2 is a square number.
a. ShowthisworksforthePT {8,15,17 }.
b. Prove it always is possible to find r, s, md t .
344
The Solutions
I. a. Here are 18 primitive PTs: 1b. Show (211111)" + (m 2 -n 2 ) 2 = (m 2 + 11 2) 2
4 3 5 4111 211 2 + 111 4 - 2111 211 2 +11 4 =

12 5 13 m 4 + 2m 2n 2 + 11 4 =
,., ? 2
8 15 17 (m- + ,r) =
24 7 25
20 21 29 2a. ..L + .l. = ..2B.. and 28 2 + 195 2 = 197 2 .

13 15 195
40 9 41
12 35 37 2b. l+ _j__ = ~
11 11 +2 n~ + 211
60 11 61
28 45 53 (211+ 2) 2 + (11 2 + 211) 2 =
56 33 65 411 2 + 811 + 4 + n 4 + 411 1 + 411 2 =

4 1 I ? 2
84 13 85 11 + 4,r + 811 - + 811 + 4 = (n- + 2n + 2)
16 63 65 3a. .l ~
2 3
48 55 73 .l + 2 ~ +2
2 3
80 39 89 1 l.Q
2 3
112 15 113 7 X 3 = 21.. .... 2 X 10 = 20
36 77 85 20 2 + 21 2 = 29 2 and so {20. 21. 29} JS a PT.

72 65 97
144 17 145
3b. Take any fraction, ~ . Divide 2 by it to produce the second fraction ~. Adding 2 to each fraction will result
in a...±fh- and ~ . Cross-multiplying produces a 2 + 2ab and 2ah + 2b 2 •
(
1
a~+2ab ) 2
+ ( 2ab+2b 2)2 = a 4 +4a 3b+8lrh-+8ab
1 ' ,
+4b 4 = ( a-+2ab+2b
1 2)2
.
4a. a= i 9 · { }· f. 7c;} = {21, 20, 29}
345
4b. Let a= lff . Then c - b = I and c + b = w:;.

n-
.
Solving this system, c = 21( ~) + 1

2= ~ and b = 12 (m;) _ 12 = W
2
-p 2 •
n 2n n 2n
') 2 -, ')
Therefore, 2n 2 · {'ff, m:;.=-.p-,
2n
~
211
} = {2mn,m 2 -11 2 , m 2 + n 2 } , a PT.
4c. Ans: ..J? . If 'ff is one fraction, the the other is ~ .
4d. In 4b. above it was shown that a= 'ff produces the PT, {2mn,m 2 -n 2 , m 2 + 11 2}, so it must be shown that
a= ~ produces the same PT.
- {Ul..±..Ll 4mn 2m 2 ± 211 2 }

- m-n· 2(m-n( 2(m-n) 2
2 2
- {Ul..±..Ll 2mn m ±u }
- m- n' (m- n( (m- n) 2
= {0i -n)(m + n), 2mn, m 2 + 11 2}
={m 2 -n2, 2mn, m2 +n 2},thesamePT(withaand b interchanged).
Sa. Fs=5 F6=8 F1=13 FR=21
½ (5 + 6 + 7 + 8) = 13 Fu = 233
Fs · Fs = 5 · 21 = 105 2 ·F6 · h = 2 · 8 · 13 = 208

105 + 208 = 11025 + 43264 = 54289 = 233
2 2 2
Sb. ½(k+ k +I+ k+ 2 +k + 3) = 2k+ 3. Therefore it must be shown that {h • Fk+3, 2 •Fk+I • h+2, F2k+3} is
a PT.
Let Fk=m-n and Fk+I =n, then Fk+2 =m and Fk+3 =m+n.
Fk·Fk+,=(m-n)(m+n)=m 2 -n 2 and 2·h+1·Fk+2=2mn.
Therefore, it must be shown that Fzk+ 3 = m 2 + n 2 •
346
Proof: F, = 111 - 11
F1+1=11
Fk+3 = m+ n
Fk + 4 = 2m + In = F., • m + F2 • 11
f\+s = 3m+ 211 = F4 · 111+ F, · 11
Fk + r, = Sm+ 311 = Fs · m + F4 · n ...
Fk + k = F2k = F, I · 111 + F, -2 · 11
F2k+3 = m · m + n · 11
F2k+3 = m 2 + 11 2
6a. Let a= 2mn and b = m 2 - ,/ • If m or n is a multiple of 3, then a is a multiple of 3. Otherwise,

111 =1 mod 3 or 111 =2 mod 3 . In either case, m 2 =1 mod 3 . The same is true for 11 , n2 =I mod 3 .
Therefore, 111 2 - 11 2 = 0 mod 3 and b must be a multiple of 3.
6b. Again, let a= 2m11 and b = m 2 -11 2 . If m or n is a multiple of 2, then a is a multiple of 4. Otherwise,
m =1 mod 4 or m =3 mod 4. In either case, 111 2 =1 mod4. The same is true for n, 11 2 =1 mod4.
Therefore, m 2 - n 2 = 0 mod 4 and b must be multiple of 4.
6c. Again, let a= 2m11 and b = m 2 - n 2 • If m or n is a multiple of 5, then a is a multiple if 5. Otherwise,

m =1 mod 5, m =2 mod 5, =3 mod 5, or m =4 mod 5 .111 In the outer two cases, m 2 =l mod 5 and, in the
inner two cases, 111 2 = 4 mod 5 . The same is true for 11 . There are four possible combinations of these cases:
m2 =1 mod 5 and n 2 =l mod 5 , m =I mod 5 and =4 mod 5 , m =4 mod 5 and 11 =I mod 5 , or
2 11 2 2 2
111 2 =4 mod 5 and 11 2 =4 mod 5 . In the outer two cases, =0 mod 5 and so b would be a multiple of
111 2 - 11 2
5. In the inner two cases, m 2 + 11 2 = 0 mod 5 and so c would be a multiple of 5.
7. Since 929 is prime and equal to 4(232) + 1 , it must equal to m 2 + 11 2 • The easiest way to find the two squares
that add up to 929 is to type Yl = ✓ (929 - x2 ) into your calculator and search the table for the integer solution.
Since 929 2 = 23 2 + 20 2 , m = 23 and n = 20 . So a= 2(23)(20) =920 and b = 23 2 - 20 2 = 129. Therefore,

{920,129,929} is a PT. 929 can also equal m 2 - 11 2 • Since m 2 - 11 2 =(m-n)(m + n)and 929 is prime,
m -n = 1 and m + n = 929, resulting in m = 465 and 11 = 464, producing the PT {929, 431520, 431521}.
347
Sa. 37°+(12)(35)=1789=5 2 +42 2 • (1789isa(4k+l)-typeprime!)
37 2 -(12)(35)=949=7 2 +30 2 . (Also, l8 2 +25 2 works.)
8b. Noticethat 12+35+37=84, 12 35+37=14, -12+35+37=60,and 12+35-37=10. (These

patterns are even more apparent if the theorem is applied to more PTs.) Therefore, it must be shown that
.2 cd -(a±h±c) 2 (a+h-c) 2 _ (a+b)2+2c(a+b)+c 2 (a+bf-2c(a+b)+c 2

c + )- 2 + 2 - 4 + 4
== + 4ab + -2b
2a- ,2........................ 2 + '.Jc 2
- -........;;;.;.... =
4ah + ..a.a...
4c 2 = ab +
..;.;;.a;.......... C
2
4 4
.'_ ib=(a-h+c) 2 (--a+h±c) 2 = ~1-hf +2c(a-b)+c 2 (-a+b)2+2c(-a+b)+c 2

cc 2 + 2 4 + 4
2a 2 - 4ab + 2b 2 + 2c 2 - -'=--=-
-- =--=--'-==---'-'::.;;.... 4ab - 4c 2 - ab - C
2
4 - 4 -
9a. 8=r+s 15=s+t

=1+7 =7+8 I 2 + 72 + 8 2 = 114
=2+6 =6+9 2 2 +6 2 +9 2 =121**
=3+5 =5+ IO 3 2 +5 2 +10 2 =134
=4+4 = 4 + 11 42 + 4 2 + 11 2 = 153
=5+3 = 3 + 12 5 2 +3"+1i=l78
=6+2 = 2 + 13 6 2 +l2+13 2 =209
=7+1 = 1 + 14 J2 + 12 + 14 2 = 246
Therefore, r = 2, s = 6, and t = 9.
9b. Looking at a few more examples:

{8.15.17} ➔ r=2,s=6,t=9
{3.4, 5} ➔ r =I, s = 2, t =2
{5,12,13} ➔ r= 1, s = 4, I= 8
{7.24.25} ➔ r= I, s=6, t= 18
Therefore.it seems r=c-band t=c-a and since a=r+s, s=a+b-c.
Hence, it must be shown that for any PT {a, b,c},

(c - b/ +(a+ b- c) 2 + (c -a/ is a square number.
348
c 2 -2bc + b2 +(a+ b/- 2c(a + b) + c 2 + c 2 - 2ac + a 2
3c 2 -4bc -4ac + 2a 2 + 2ah +2h 2
For this to be a perfect square, the coefficient of c' must be a square. Since c 2 =a 2 + b2 , the expression can be
rewritten as: 4c 2 -4c(b +a)+ a 2 + 2ab + t/ = 4c 2 -4c(a + b) +(a+ b) 2 = (2c-(a + b)) 2 •
The Extensions
l. In a PT, is either a. b, a- b.or a+ b always a multiple of 7.
2. Is c always of the form 411 + I ?
3. What numbers never occur in a primitive PT?
4. In an ordered primitive PT { a, b, c} with a < b < c , 1s c always equal to b + I , if a is odd, and equal to
b + 2 , if a is even?
5. Determine all triples {a, b, c} such that 3" + 4 b = 5 c •
6. Determine all triples {a, b,c} such that J,. + -1r = ~.

a b c
7. How do you generate all the Pythagorean Quadruples i.e. {a, b,c, d} such that a 2 + b 2 + c 2 = d 2 ?
8. Define D, the distance between two PTs, equal to lei - c2 I and H, the height of a PT, equal to c -b . Let
{ao, bo,c 0 } be a PT of height H, D be a positive integer, and ~ = ~. If ~ao and ~~ are integers, then
ak+ 1 = ak + D, bk+ 1 =[3th+ bk+ ~:, Ck +1 = [3ak +ck+ ~~ are recursive formulas that produce all PTs of
height Hand the distance between each pair is precisely D. In the case ofH =1, {ao, bo. c 0 } = {3, 4, 5} .
More about this theorem can be found in The College Mathematics Journal , Vol. 30, No. 2, March 2000 ..
349
ARML Power Contest- November 2001 - Cevians
Cevians
A cevian is a segment which joins the vertex of a triangle with a point on its opposite side ( or its extension.)
Throughout this problem set, unless it is stated to the contrary, the three cevians of !:,ABC will be labeled
AD, BE, and CF.
A
Some cevians have special names like altitudes, medians, and angle bisectors. Sometimes, AD , BE , and CF are
concurrent at point P, while other times they form an interior cevian triangle, called t:, PQR . When concurrent,
oftentimes, point P has a special name like orthocenter, centroid, incenter, or the Nagel, Gergonne, Lemoine, or
Miquel point.
Triangle !:, DEF is called the inscribed cevian triangle. The orthic and medial triangles are special inscribed cevian
triangles. The orthic triangle is formed when the cevians are altitudes while the medial triangle is formed by
medians. A
The following three theorems may be useful in solving this problem set:
C b
X
Stewart's Formula: The length of a cevian, x, and the sides of a triangle are
m n
related by the following formula: c 2n + b2m =x 2a + mna C
a
:)
B D C
Ceva's Theorem: The three cevians, AD, BE, and CF, are concurrent if and only if # •'1ft; -~ = I
350
ARML Power Contest - November 2001 - Ce vians
---------------------------------------------------------
---------------------------------------------------------
Routh's Theorem: If KABc is the area of fl.ABC, then the area of the interior cevian triangle PQR is
Kl'()R
-
= .
(st+ s +
t
I) s ' - Ii~ ( +
tr + t + rs r+
l) . KABC and the area of the inscribed cevian triangle DEF is
BD CE AF
vc=r, Ei'l"=s, FTJ=t.
The Problems
Part A - Cevian Length and Concurrency
. f'rom vertex C of uAABC.. x

1. Sh ow that th e 1ength of the me d ian = /2.a2+2hi-c2
2
2. Show that the length of the altitude from vertex C of fl.ABC,

x =/(a+ b- c)(b+ c-a)(c +a- b)(a + b+ c)
2c
3. Show that the length of the angle bisector from vertex C of A ABC,
,= /a·b-(a+b-c)(a+b+c)
- a+b
4. If the altitude, median and angle bisector from vertex C of a A ABC divide LAC B into four equal angles.
Find the degree measures of LA, LB, and L ACB.
5. In A ABC, the median AD, the altitude BE, and the angle bisector CF are concurrent. Show that.
6. In AABC, point D is on BC and halfway around the triangle from vertex A, i.e., AC+ CD= AB+ BD, and
E is on AC and halfway around the triangle from vertex B. and Fis on AB and halfway around the
triangle from C. Prove that the cevians AD, BE, and CF are concurrent.
7. Prove that the orthocenter of AABC is the incenter of its orthic triangle fl.DEF.
351
ARML Power Contest- November 2001 - Cevians
Part B - Areas of Cevian Triangles
8. The sides oft.ABC have lengths 14, 25.and /149. The three cevians divide each side in the ratio 3:5 and
fonn an interior cevian triangle t,. PQR. Find the area of t,. PQR.
9. The cevians of a triangle partition the sides into segments whose ratios are 1: I, 4: 1, and 7: I, consecutively. If
the area of Lhe inscribed cevian triangle is 28, what is the area of the interior cevian triangle?
10. The cevians of a triangle partition the sides into segments whose ratios are each 1:n. Find a simplified rational
KPQR .
function which expresses - K m tenns of n .
ABC
Part C - Symmedians
If two cevians from the same vertex of a triangle are symmetrical with respect to the angle bisector of that angle,
i.e. L D'A T: L DAT in figure below, they are said to be isogonal and fonn a pair of isogonal conjugates.
The Isogonal Theorem: If AD and AIY are cevians with point P on AD and point Q on AIY , then AD and
AIY are isogonal if and only if the ratio of distances from P to AB and AC are inversely proportional to the
- - -,,
ratio of the distances from point Q to AB and AC. (AT bisects angle BAC.)
j~
I \.\. \\
I
::,---o \ \ ----~.
B D' D C
\;r
11. Prove, if the cevians AD, BE, and CF are concurrent, then so are their isogonal conjugates, AIY, BE'. and
CF' .
A symmedian is a cevian which is isogonal to a median. Since medians are concurrent at the centroid, from problem
1 above, the symmedians of a triangle must also be concurrent. (Their point of concurrency is called the
Lemoine point of the triangle.)
352
A - ~
" AM is the median and AX is the angle bisector.
1
~\~',,,',,,',,,, ,, LDAXcLXAM
\ ~ ~ - So AD is a symmedian !
~ ~~~~
-- -- - - - ~ --~ - ~-
B ' D \ M' C
X
12. Prove P is a point on symmedian AD if and only if the ratio of the distances from P to AB and AC is
proportional to ~ .
- BD AB 2
13. Prove if AD is a symmedian, then DC= AC .
In /1 ABC let E be a point on AB and F a point on AC. If 11 AEF"" 11 ABC then EF is a parallel of side BC
of fiABC and if fiAFE"" 11 ABC then EF is an antiparallel of side BC of fiABC.
=c g C
14. In fiABC we know the median AM bisects every parallel of side BC. Prove the symmedian AD bisects
every antiparallel of side BC.
353
ARML Power Contest - November 2001 - Cevians
The Solutions
1. Using Stewart's Formula, a 2m+ h\1 =x 2c + mnc, let 111 = 11 =~-Then a 2 ~ + h 2 ~ =x 2c + ~ic ⇒
a2 +h2 =x2+c 2 ⇒ 2a 2 +2h 2 -c 2 =x2 ⇒ l2a 2 +2h 2 -c 2 =x

2 2 4 4 2
2. Area of the triangle, K = ½ex ⇒ x= 2f .

Area of the triangle, K = /s(s - a)(s -b)( s - c) where s = 0 ± b ± C
2
K = !(a+ b + cXb + c -a)(a + c -b)(a + b-c)

4
= (.f) !(a +b + cXb+ c-a)(a + c-b)(a + b-c)

X C 4
x =!(a+ b+ c)(b + c-a)(a + c-b)(a+ b-c)

2c
3. By the Angle Bisector Theorem, 1;f = ~ and m + n = c. Therefore, 111 = a ~-h and n = ~ . Plugging these
into Stewart's Formula yields: a 2 -1ii:....b + h 2 _a:_b = x 1c + _bi;_b_a:_bc ⇒ abc(a + b) = x 2c + ~ ⇒

a+ a+ a+ a+ a+ b (a+ h)-
,.,
⇒ w) -
abc 2 ,
_
-X
2
⇒
ah(a+b) 2 -ahc 2 _
-X
2
⇒
ah((a+b) 2 -c 2 ) _
-X
2
⇒
fab(a + b-c)(a +b + c) _
0 2 a +h -X ·
(a+br (a+b)- (a+b)
4. AM=MB and so AH+HM=HB-HM.

Therefore. Fig. 4
I
h tan x + h tan 2x = h tan 3x - h tan 2x
/ h
I _l~\
tan x + 2tan 2x = tan (x + 2x)
tan x + 2tan 2x = tan r ± tan 2 r

1-tanxtan 2x
A0 H I)
354
tan x - tan 2 x tan 2x + 2tan 2x - 2tan x tan 2 2x = tan x + tan 2x ⇒
- tan 2 x tan 2x + tan 2x - 2tan x tan 2 2x = 0 ⇒ - tan 2 x + 1 - 2tan x tan 2x = 0 ⇒
-tan 2 x+l- 4tan 2 x =0 ⇒ tan 4 x-6tan 2 x+l=0 ⇒ tanx={3±2n ⇒ x=tan· 1 (3±2n ⇒

1-tan-x
X = 22.5° or 67.5° ⇒ X = 22.5°. Therefore, LA= 67.5°. LB= 22.5°. and LC= 90°.
5. By Ceva's Theorem, '}Ji· 1/J!,; · Cjfj., = 1. But since M is a midpoint, 1/J!,; = 1 and since CD is an angle
bisector, 'PJi = *. Therefore, f/J. = ~. Let CH =at and HA= bt, then at+ ht =h, or t =-;;!;-;; .
By the Pythagorean Theorem, BH 2 = A B"l. - HA 2 , so

BH 2 = c 2 -b 2t 2 and BH 2 = BC 2 - CH 2 , so BH 2 = a 2 - a 2t2.
' "
Therefore. c 2 -h 2t 2 = a 2 -a 2t 2 and t 2 = a..:...=...e:,. . Substituting for C
a -b-
2
t.
_b_)2-
G b - ~
2_.2
and so
h2 _a+b
l .
""'T'""""'"! -
1+ a"-b a-c a-J A
A
6. Lets= the semi perimeter of 11ABC, then AE = s - c, EC= s a,

CD = s - b, DB = s - c, BF = s - a, and FA = s - b.
d..E.. CJ.2 .BE.-So...;;;;;.J;; - ~ - :i...=.il -1 B

~
--D
\ -- ·. -~~
C
EC DB FA - s -a s -c s -b - .
Therefore, the cevians are concurrent. This is known as Geronne's Point.
355
----------------------------------------------------------
----------------------------------------------------------
7. ~ DEF is the orthic triangle of
~ ABC. with P the ortho center,
the intersection of the altitudes.
~AFP==~CDP and so
LFAP= LDCP.
Since quadrilateral AEPF has two
right angles it must be cyclic. The
same is true for quadrilateral
CDPE.
Because they intercept the same

arc, L FAP = L FEP and
L DCP = L PED . Therefore,
L FEP = L PETJ . Fig. 7
....... --p .......

The same reasoning can be used to show L EDP= L FDP and L. DFP = L EFP and so DA, EB, and FC are
bisectors of the angles of ~ DEF and so P is the in center of ~ DEF .
8. ByHero'sFormulatheareaof ~ABC is KA 8 c=(s(s-14)(s-25)(s-/149) wheres= ~+32.
Routh's Theorem, the area of~ PQR is
9. Using Routh's Theorem for the inscribed cevian triangle,
28 = (0(4)( 7) MBC ⇒ 28 = 2.8.8o·MBC ⇒ 80 =MBC

(2) (5)( 8)
28
L et K =h
t eareao fh ....
t emtenorceviantnange, -( 28 + 4 +l)(<7 + -l)2
I K-
7 +!)(4 +l+l)
-80-2.l.!i
- ll ·
356
10.
(11 ' - l) 2
K PQR = -.---------.--------.---- . K AlJC
(n - + 11 + I)( 11 - + n + I )(11 - + n + I)
KPQR ((n-l)(n 2 +n+I))'
=--'---------'-
KAlJC (,/ + 17 + 1/
KPQR (11 - 1) 2 (11 2- 211+ l)

KAiie
=-.---__,.,..
(,r + n + 1)
= -(,r........--
+ 11 + I)
11. Let cevians AD, BE, and CE be concurrent al point
P and let Q be the intersection AIY co1f.i BE', the
isogonal conjugates of AD and BE . Let the

distances from P to sides BC. AC and BC be
a', b',and c', respectively, and the distances from Q
to the sides BC, AC, and BC be a", b", and c",
respectively. Using AD and AD' and The Isogonal
Theorem,
Ii, = .i:_b,,,, . Using BE
- and BE'
- and The
C
Isogonal Theorem, ½ = a..;;. Combining the two

a C
equations using multiplication yields
Let CF' be a cevian from C through point Q. By The Isogonal Theorem, CF and CF must be isogonal.
Therefore, isogonals of concurrent cevians are concurrent.
//1~
12. ⇒ Assume Pis a point on symmedian AD. By
the Isogonal Theorem, f = 1. Because Mis a 1 ,\ ~,."
l Fig. 12 .1 Y ~,
midpoint, KABM = KAcM and l2 ct = 2 bu. Therefore. I ·
/~ ~.
'-b = Li.t and so .I
Y = b' . ,,~"Ip \ u/~~
<= Assume f =t and M is the midpoint of BC.
Using the same area logic as above, ~=~and B

/
•· I NI
D M
~ .....
C
357
so. by The Isogonal Theorem, AP (AD) and AM are isogonal. Since AM is a median, AD must be a
symmedian.
13. Let AD be a symmedian and let BD = m and

DC= n . From problem 12 above, ~ =t .
Because they share the same altitude, KARn
KAm·
= W._
n
B D
14. Consider !:,.ABC with DE, a parallel to side

- .....
BC and AX, the angle bisector (Figure 14a).
- -
Let D'E' be the reflection of DE through AX.
.....
Using congruent triangles, it is easy to prove
that t:,.ADE=.t:,.ADE' and so
LADE= LADE'. Therefore, D'E' 1s an
antiparallel to side BC.
Using Figure 14b, since the median AM bisects DE

at G and the fact that the reflection transformation
.....
preserves distances, its reflection through AX will
- ......
bisect the reflection of DE through AX.
Therefore, the symmedian from A bisects any D
antiparallel to BC .
358
ARML Power Contest - February 2002 - Insane Tic-Tac-Toe
Insane Tic-Tac-Toe
The Definitions
In the game of Tic-Tac-Toe (also called Naughts and Crosses) two players, one using X's and the other using O's,
take turns placing their X's or O's in a 3 by 3 grid. The winner of the game is the first player to get three X's
or three O's in a row horizontally, vertically, or diagonally. The game oflnsane Tic-Tac-Toe (IT3 ) is played on
a 3 by 4 grid, but on any turn a player may place either an X or an O in the grid. Again the winner is the
first player to complete three X's or three O's in a row horizontally, vertically, or diagonally.
Consider the following game board (a partially played game):
It can be inverted (by interchanging all X's and O's), reflected through a horizontal line, reflected through a vertical
line, or rotated 180° to become the following boards.
Since the strategy for winning on any of these boards is equivalent, we will say these four boards are equivalent.
Therefore, any two boards are equivalent if one can be changed into the other by any combination of inversion,
horizontal or vertical reflection, or rotation.
A death cell on the IT3 board is a cell in which putting any mark, either an X or an 0, will allow the opponent to
win on his or her next move. The death cells have been shaded in the game boards above.
In analyzing the game ofIT3, a round will be considered a move (or tum) by player A followed by a move (or tum)
by player B. We will assume in all problems that each player is trying to win and playing smart. A person
playing smart will not provide the opponent a winning opportunity (2 of 3 X's or O's in a row) unless forced
to do so. Because of the play smart rnle, the following boards will never happen:
ID
(It would be impossible for two O's to be in a
X =--r-
0
X 0
row as on these boards, if the players are
playing smart.)
359
ARML Power Contest - February 2002 - Insane Tic- Tac-Toe
The Problems
In the following problems you can refer the a game board using a 3 by 4 grid and/or 1 2 3 4
refer to the individual cells using the following fonnat: Cutting and taping 5 6 7 8
grids from the grid sheets provided may make your work more readable. 9101112
Set A
1. Determine all the boards that are equivalent to each of the following game boards
a. b.
~
2. After the first round, if the players play smart, how many nonequivalent game boards are possible? Justify your
answer.
3. Determine five nonequivalent game boards that have seven death cells. Be sure to shade in the seven death cells.
Set B
When Ann told her sister Beth about Insane Tic-Tac-Toe. Beth was a little suspicious, especially when Ann said,
"You can always go first!" Beth said, ''I'll play, but you must go first." After a few games Beth realized a
simple non losing strategy. Every time it was her tum, unless she could win she would simply do the
opposite of what Ann did--in other words, rotate and invert Ann's last move. This is an example of how the
first rounds might go:
JJI
Ann Beth Ann Beth
Round 1 Round2
Throughout the problems in set B, Beth will use this copy-cat strategy.
4. After the first round, if both players are playing smart, how many nonequivalent game boards are there? Justify
your answer. Make a representation for each game board and shade in all death cells.
5. After the second round, if Beth has not already won and both players are playing smart, there are eight
nonequivalent game boards. Prove this and make a representation for each game board and shade in all death
cells.
6. For each of the game boards in problem 5 above determine how many rounds Ann will last if Beth continues
this strategy.
360
----------------------------------------------------------
Set C
After awhile Ann complained because she was never winning. So in an effort to make the game more fair, they
decided that Ann would be allowed to place two X's or two O's on the first tum. Beth's original strategy no
longer worked. Soon Ann was winning every game!
Throughout the problems in set C, Ann will make two X's on her first tum.
7. How many nonequivalent opening moves does Ann have, if Ann plays smart? Justify your answer.
8. If both players have played smart, how many nonequivalent game boards are there after the first round? Justify
your answer. Be careful of double counting!
Ann determined that if she always started with X's on cells 4 and 6, she could always win!
Assume that she always used this opening move answer the following questions to prove she
can always win regardless of how smart Beth plays. In each proot you may stop when all
the remaining cells are death cells. Careful!! After some moves, some death cells come back
to life!
9a. If Ann always started with the above opening move and both players played smart, how many nonequivalent
game boards are there after the first round? Justify your answer.
9b. If Beth's first move was an O in cell 1, then Ann's second move was an O in cell
11. (Likewise. if Beth's first move was an O in cell 11, Ann's second move was an
0 in cell I.) Prove that Ann can always win when starting off this way.
9c. If Beth's first move was an O in cell 3, then Ann's second move was an O in cell 9.
(Likewise, if Beth's first move was an O in cell 9, Ann's second move was an O in
cell 3.) Prove that Ann can always win when starting off this way.
9d. If Beth's first move was an O in cell 2, then Ann's second move was an O in cell 10.
(Likewise, if Beth's first move was an O in cell I 0, Ann's second move was an O in
-m~ ..... 1.- '.
cell 2.) Prove that Ann can always win when starting off this way. !
- - - I
9e. If Beth's first move was an O in cell 5, then Ann's second move was an O in cell 7.
(Likewise, if Beth's first move was an O in cell 7, Ann's second move was an O in
cell 5.) Prove that Ann can always win when starting off this way.
361
----------------------------------------------------------
9f. If Beth's first move was an O in cell 12. what must Ann's second move be to
guarantee a win? Show this move guarantees a win regardless of where Beth moves
and that Beth could win if Ann chooses any other second move.
9g. If Beth's first move was an O in cell 8 and Ann's second move is in cell 12, a win
can be guaranteed for either Ann or Belh depending on whether Ann plays an X or an
1~n-t
0. Show this is true.
Set D
Now on the losing end of the stick every time, Beth is no longer willing to give Ann two X's or O's on the first
move. Instead they return to the original rules but draw the game board on a toms so that the top of the grid is
connected to the bottom of the grid and the right side of the grid is connected to the left side. Now there are no
edges and every cell is adjacent to eight cells.
10. If Ann opens with an X in cell I, prove Beth must play an 0.
11. Prove Beth will always win on a toms!
362
----------------------------------------------------------
The Solutions
la. There are seven:
2. After the first round the game board must contain two different marks or two of the same marks.
Case 1-Two different marks. The X can be in one of four cells: a corner cell. a long side cell, a short side cell, or a
center cell.
If placed in a comer cell, there are eleven possible locations for the 0. [ 11]
If placed in a long side cell, there are eleven possible locations for the O but four of these are
inversions of a board already counted. [7]
O!X 0 0
If placed in a short side cell, there are eleven possible locations for the O but eight are inversions
0 0 0 0
0 of boards already counted. [3]
0 0 0
If placed in a center cell, there are eleven possible locations for the O but all but one are_inversions
o,olo o of boards already counted. [I] e--X--+-_
xlo QJ o
0 0 0 0
Case 2 - Two marks the same. Again the first X can be placed in one of four cells:
If placed in a corner cell, there are only five locations for the second X.
0 0 0 Oi Remember they play smart! [5]
o,x O CY
I ----+-----+-~
LQIO O 0 If placed in a long side cell, there are four possible locations for the second X
but one of these is an inversion of a board already counted. [3]
If placed in a short side cell, there are seven possible locations for the second X but all but
one of these is an inversion of a board already counted. [ 1]
If placed in a center cell, there are only two possible locations for the second X but both of
these are inversions of a boards already counted. [OJ
Therefore, there are 31 nonequivalent game boards after the first round.
363
ARML Power Contest - February 2002 - Insane Tic-Tac- Toe
----------------------------------------------------------
3. Obvious the board must contain an X and an O for there to be any death cells. Furthermore. it is easy lo see
that the board must contain a second X or 0. Consider a board with one O and two X's. There are only four
possible locations for the 0: a corner cell, a long side cell, a short side cell, or a center cell. If O is in a corner
cell, there are only six possible death cells and, if O is place in a short side cell, there are only four possible
death cells. If O is in a long side cell, there are eight possible death cells and, if O is place in a center cell,
there are nine possible death cells. Therefore, the O must be in a long side or center cell. If O is in a long side
cell, there are only two arrangements of the X's, while ifO is in a center cell, there are three arrangements for
the X's
4. Again the X can be in one of four cells: a corner cell, a long side cell, a short side cell, or a center cell. The 0
must be in the "opposite" comer cell, long side cell, short side cell, or center cell.
5. Again because of the symmetry of Beth's second move and the fact that it is the inversion of Ann's second
move, we only need to worry about where the next X goes. In each of the diagrams below, the light X's
represent the possible locations for the second X for each of the first round game boards.
lXX
~x
X
1~T~ X
The first board produces these four nonequivalent game boards:
The second board produces only two new nonequivalent

game boards (e and t): e. ;)~o
And the third produces two new nonequivalent game
=-msr
boards (g and h):
While the fourth produces no new nonequivalent game
boards.
364
6a. Game board (a) at the end of round three is at the right: Because only death cells remain,
Beth will win on the next (fourth) round.
6b. Game board (b 1) at the end of round three is to the right: The
board after round four (b2) is to its right. Because only death cells
remain, Beth will win in round five.
6c. The two possible third rounds of board ( c) are vertical reflections. Beth will win on the
next (fourth) round.
6d. Although there are four non death cells in

board (d), there are only three possible pairs
of moves for the third round. In each of
these, only death cells remain after the third
round. Beth will win in the fourth round.
e.
6e. In board (e) rounds three and four can be played in any order but the results are the same.
Beth will win in the fifth round.
6f., 6g. Only death cells remain in boards (e) and (f). Therefore, Beth will win in round three on both of these
boards.
6h. Game board (h) at the end of round three is at the right. Again, all of the remaining cells h.
are death cells so Beth will win in round four. f--+.,,-;,,,.,.,..'-"+.
Therefore, Beth will always win if she uses this strategy!
7. As was shown in problem A2, case 2, there are nine ways for Ann to put two X's on the game board while
playing smart:
a. b. C. d. e. f.
g. h. I.
365
ARML Power Contest - February 2002 - Insane Tic- Tac-Toe
----------------------------------------------------------
8. In board (a), Beth could put a O in cells 2, 5, 6, 9, or IO to produce non-equivalent boards. [5]
In board (b ), Beth could put a O in any of the open cells or an X in cell I 0. [ 11]
In board (c), Beth could put a O in cells 2, 3, 4, 7, or 8. [5]
In board (d), Beth could put a O in any of the open cells. [IO]
In board (e), Beth could put a O in any of the open cells or an X in cell 8. However, having an X's in cells I,
8, and 10 has already been counted in board (b) [ 10]
In board (f), Beth could put a O in any of the open cells or an X in cells 8 or 11. [ 12]
In board (g), Beth could put a O in any of the open cells or an X in cells 5, 9, or 11. However, having X's in
cells 2, 8, and 5 or having X's in cells 2, 8, and 9 has already been counted in board (f) and having X's in cells
2, 8, and I I was already counted in board (b ). [IO]
In board (h), Beth could put a O in cells I, 5, 6, 9, 10 or an X in cell 5. However, having X's in cells 2, 11,
and 5 has already been counted in board (f) [5]
In board (i), Beth could put a O in cells 1, 2, or 6. or an X in cell 2. However, having an X's in cells 5, 8,
and 2 has already been counted in board (f) [3]
The total number of nonequivalent game boards after the first round is 71.
9a. As shown in board (d) above, Beth can only use O's in any of the ten open cells.
9b. On her second move, Beth must put an O in cell 8 and Ann can put a X in cell 12,
leaving only death cells.
9c. On her second move, Beth must put an O in either cell 8 or 12. In either case, Ann
responds by putting an O in the other one, leaving only death cells.
9d. On her second move, Beth must put an O in either cell 5 or cell 8. Whichever one she
choses Ann will put an O in the other cell, leaving two non death cells, 7 and 12, which
must be filled with X's. Whichever one Beth choses for her X, Ann will put an X in the
other, leaving only death cells.
366
ARML Power Contest- February 2002- Insane Tic-Tac-Toe
9e. Two cells. 8 and I 0, are available. Beth must put an O in one of them and Ann counters
with an O in the other. This frees up cell 12 in which Beth must put an X. This in turn
frees up cell 2 in which Ann puts an 0, leaving only death cells remaining open on the
board.
9f. Although there are five cells available, cells 1, 3, 5, 8, and 9, only cell 9 guarantees Ann
a win!
Case I Show cell 9 guarantees a win. If Ann puts an O in cell 9, only cells 3 and 8 are
available. Beth must put an O in one of them; Ann will put an O in the other, leaving
only death cells.
Case 2 Show the other four starts could force Ann to lose.
a) If Ann puts an O in cell 1, Beth could counter with an O in cell 8,which would
force Ann to put an X in cell 11. Beth could counter with an O in cell 10, leaving only
death cells for Ann.
b) If Ann puts an O in cell 3, Beth counters with an X in cell 9, leaving only cells
5 and 8 available. Ann must put an O in one of them and Beth counters with an O in the
other, leaving only death cells for Ann.
c) If Ann puts an O in cell 5, Beth counters with an O in cell 8, forcing Ann to put
an O in cell 3. Beth counters with an X in cell 9, leaving only death cells for Ann.
d) If Ann puts an O in cell 8, Beth counters with an O in cell 5, leaving only death
cells for Ann
9g. Case 1 Ann places an X in cell 12.

Beth must counter with an O in one of the eight remaining circular cells.
Ann counters with an O in the "opposite" cell of the circle. Beth now must
put an O in one of the two remaining safe cell; Ann counters with an O in
the "opposite" safe cell, leaving only death cells. Ann wins.
Case 2 Ann place an O in cell 12.

Beth counters with an O in cell I. Ann must play an X in cell 11 and Beth counters
with an O in cell 10, leaving only death cells. Beth wins.
367
I 0. The easiest way to think of the game board on a torus is to

imagine that the board repeats itself on each side and each
comer. From the original board, if Ann plays an X in cell I,
Beth cannot play an X in cells 2, 3, 5, 9, 6, or 11. Looking
at the expanded game board, cells 7, 12, and 1 form a
downward diagonal of three cells and cells 8, 1 and IO form an
upward diagonal of three cells. Therefore, Beth cannot play an
X in cells 7, 8, 10, or 12. Cell 4 is on the left of cell 1 and
hence is not a smart play for an X. Therefore, Beth must play
an 0.
11. In problem D1 above, it was shown that every cell is in a row of three with every other cell. Prior to Ann's
second tum, the board contains an X and an 0. Since every cell is in a row of three with the cell containing the
X and every cell is in a row of three with the cell containing the 0, all remaining cells are death cells and Beth
will win.
368
ARML Power Contest - November 2002 - Three Addition Problems
Three Addition Problems
The Definitions
Consider the following alphameric addition problem:
ABC
+DEF
GHJ
where A BC. DEF, and G HI represent three three-digit numbers with A BC+ D E F =G H I. In this
problem A, B, C. D, E. F, G, H, and I are distinct digits from the set {I, 2, 3, 4,5. 6, 7, 8. 9}.
2 35
+746 is a solution to this problem.
981
- i-
2 45 352 2 15
Likewise, +736 +4 67 +478 are also solutions to this problem. Notice that in each of these
981 819 69 3
- ii - - iii - - IV -
solutions two of the addend digits total more than ten and so a carry digit must be added to the column on its
left.
Although each of these three new solutions is unique from the original solution i, only the last one iv will be
considered a solution independent of the original solution. In this problem, two unique solutions,
a 1, b 1, c1 ,d1, e1, fi, g 1, h 1, ii and a2, b 2, c2, dz, e2, fi, g2, h2, ii , are considered independent if and only if the
set is insignificant.) For example, in the solutions cited, solutions i, ii, and iii are not independent because
the set {{2. 7}. {3, 4}, {5,6}} = the set {{2. 7}, {4. 3}. {5,6}} = the set {{3, 4}. {5, 6}. {2. 7}}, but
solution i is independent of solution iv because the set {{2. 7}, {3,4}, {5.6}} -f. the set
{{2, 4}, {I. 1}. {5,8}}.
So how many unique solutions are there to this problem? In this problem,
A + B + C + D + E + F + G + H +I= 45 and by considering whether there are zero, one, or two carry digits, it
can be shown that G + H +I= 18. (Can you verify this?) Therefore, {G, H, f} must be one of the following:
{1,8,9}. {2.7,9}, {3,6,9}, {3,7,8}, {4,5.9}, {4.6,8}, or {5,6,1}. (Canyouverifythis?)
Each of these produces one, two or three independent solutions, totaling 21 independent solutions. (This also could
369
----------------------------------------------------------
be verified by examining each of the above seven cases. For example, ! I, 8, 9} produces these three
independent solutions: 234 + 657 = 891, 235 + 746 = 981, and 324 + 657 = 981. Notice that in these
solutions, {{2, 6}. {3. 5}. {4. 7}} -f- {{2. 7}. {3.4}. {5.6}} -f- {{3. 6}. {2. 5}. {4. 7}}. )
Furthermore, each of these twenty-one independent solutions produces eight unique solutions since the order of each
of the three pairs of addend digits can be permuted in (2) 3 = 8 ways (Compare solutions i and ii on page 1).
In addition, in each of these eight unique solutions, the first column can always be interchanged with the last
pair of columns (Compare solutions i and iii on page I), producing another eight unique solutions.
Therefore, there is a total of 21 (8)(2) = 336 unique solutions to this problem!
Although alphameric problems usually have a solver trying to find a single unique solution. the alphamerics in this
ARML Power Contest have many unique solutions where the solvers must determine a method to find the total
number of unique solutions.
370
The Problems
Since the order in which elements are listed in a set is irrelevant, for ease of correcting and checking independence,
please always list the elements of a set in numerical order, smallest to largest.
1. Consider the following alphameric addition problem:

ABC
+DEF
GHJJ
where A BC and DEF represent two three-digit numbers with A BC + DEF = G HI .I, a four-digit
number and A, B. C. D, E, F, G, H, /, and .l are distinct digits from the set {O, 1, 2. 3,4, 5, 6. 7. 8, 9}. (Note:
A. D. or G cannot be zero.) In this problem. two unique solutions, a 1,b 1, c1 ,d1, e1,./i. g1. h 1, i1, )1 and
1a. Make two quick observations about the digits O and 1. Justify your conjectures.
1b. It is obvious that there is one carry digit from the hundreds column to the thousands column. Prove there can
be either one carry digit or three carry digits, but not two carry digits, in this problem. Then show that the
only possible sets equal to {G, H, I, J} are {O, 1, 2, 6}, {O, I, 3, 5}, and {O, 1, 8, 9}.
I c. For each of the three sets above, find all independent solutions, i.e., find the possible sets for { {A, D}, { B, E},
{C, F}}.
1d. Determine the total number of unique solutions to this problem. Justify your answer.

A
BC
+DEF
GHJJ
where A is a one-digit number, BC a two-digit number, and DEF a three-digit number with
A + BC + DEF = G HI .I, a four-digit number and A, B. C, D, E, F, G, H, I, and .I are distinct digits from
the set {O. I, 2. 3,4. 5, 6, 7, 8, 9}. (Note: B. D, or G cannot be zero.) In this problem. two unique solutions,
a 1, b 1, c 1 ,d1, e1. Ji. g 1, h 1, ii, )1 and a2, b 2, c2. d2, e2, f;.. g 2. /12, i2, ft , are considered independent if
2a. Determine all the possible four-element sets that are equal to {G, H, I, J}. Show your work.
371
2b. For each of the sets from 2a above, determine all the independent solutions, i.e., find the sets which would
equal { :A, C, Fl, {B,E) }.
2c. Detennine the number of unique solutions to this alphameric problem. Justify your answer.
B,wt'fre
C Dhfff1·c
FG11n,1.·1·
+HI J,,,,,1;,,
KL M1wefre
where A and B are one-digit base twelve numbers, CD and F G are two-digit base twelve numbers, and HI J
and KL M are three-digit base twelve numbers with A + B + CD + F G + HI J = KL M and
A, B, C, D, F, G, H. I. J. K,L. and Mare distinct digits from the set {0. 1, 2, 3,4. 5, 6. 7, 8, 9, T, E}, where T
represents the digit ten in the base twelve system and E represents the digit eleven. (Note: C. F, H, or K
cannot be zero.) In this problem two unique solutions, a 1, b 1, c1 ,di, f1, g1, h 1, ii, }1. k 1, Ii, m 1 and
a2. b 2. c2. d2,f2, g2, h2. i2, }2,kz, h,m2 are considered independent if {{a 1, b 1, d1, g1. }1}. {c1, f1, i1}. {h 1}} f.
3a. It can easily be shown that the sum of the carry digits must be even in this problem. Using this fact, find the
twenty possible sets equal to {K, L, M}. Show your work.
3b. Pick one of the {K, L, M} sets from 3a and list all the independent solutions for that set. Then determine the
number of unique solutions for this set of independent solutions.
BONUS (Not part of this competition) Prove there are I 08,960 solutions to this problem!
372
The Solutions
I a. G must be 1. It must be produced as a carry digit and A + B is at most 17, causing a carry digit of 1.
C or F cannot be 0. IfC ~ 0, then F = J and if F = 0, then C = J. A or D cannot be 0. If A = 0, then D
must be 9 and H would have to be at least 2, implying that B + E > 30, an impossibility. There is a similar
argument if D = 0.
lb. Case I Only I carry digit: G = 0. C + F = 1. B + E =I.and A+ D= 10 + H. Since the carry digit is
produced in lhe hundreds column, H = 0. A + B + C + D + E + F + G +I+ 1 = 45. Substituting produces:
10+!+1+0+1+!+1=4S ⇒ 21+2!+11=4S ⇒ 1+1=17. Therefore, {f,~={8,9} and
{G. H, I,J} = {O. I. 8, 9}

Case 2 Exactly two carry digits: Whether the second carry is produced in the units or the tens column makes
no difference. G = I , C + F = IO + 1, B + E + I = I. and A + D = 10 + H. Since
A + B + C + D + E + F + G + I + 1 = 45 ⇒ 10 + H + I - l + 10 + J + l + H + I = 4S ⇒
2H + 21 + 21 + 20 = 45 ⇒ H +I+ J = l 2.5 ..... impossible!

Case3 Three carry digits: G=l,C+F=lO+l, B+E+I =10+/, andA+D+l=lO+H. Since
A+B+C+D+E+F+G+/+1=45 ⇒ 9+H +9+!+ IO+J+H+/=45 ⇒ 2H+2/ +2/+29=45 ⇒
H +I+ 1 = 8. One of these three variables must be 0 because 2 + 3 + 4 > 8 . Therefore, {G, H, I,J} =
{0, 1, 2, 6} or {G, H,l, J} = {0. 1, 3, 5}.
le. Case 1 {G, H, I,J} = {O, I, 8, 9}
This case produces two possible values for the four-digit number GHIJ , either 1089 or I 098 with
{A. B, C,D, E, f} = {2. 3,4,5, 6, 7}. If GH/1 = 1089, then C + F = 9, B+ E = 8, and A+ D= 10.
Therefore, {C, f} = {2. 7}, {3, 6}, or {4, 5}.
Case Ia: {C, r) = {2. 7}, then {B. E} = {3, 5} and {A, D} = {4, 6}. producing this solution:
432 + 657 = 1089 and {{A, D}, {B. E}, {C, F}} ={{4. 6}, {3, 5}, {2, 7}}.
Case lb: {C, f} = {3. 6}. then none of the remaining numbers add up to 8. Therefore, {C, f} =t= {3, 6}.
Case le: {C. f} = {4. 5}, then {B. E} = {2. 6} and {A, D} = {3. 7}. producing this solution:
324 + 765 = 1089 and {{A. D}. {B, E}. {C, f)} ={{3. 7}, {2. 6}. {4. 5}}.
All solutions for 1098 can be found by interchanging sets {C, f} and {B. E}.
Case 2 {G. H, I,J} = {0. I. 2. 6}

This case produces six possible values for the four-digit number GHLJ, either 1026, 1062, 1206, 1260, 1602 or
1620 with {A, B, C, D, E, f} = {3, 4,5, 7, 8, 9}.
373
Case 2a: If GHJJ = 1026, then {A. B, C. D,E. r} = {3. 4.5, 7, 8, 9}. C+ F = 16,B + E = 15. and A+ D= 9,
producing this solution: 437 + 589 = 1026 and {{A. D}. {B, E}. {C. rj} ={{4. 5}, {3. 8}, {7, 9}}.
All solutions for 1206 can be found by interchanging sets {A. D} and {B. E}.
Case 2b: If GH/J = 1062, then C+F= 12,B+E= IS.and A +D=9. Therefore, {C.f}= {3.9}, {4,8}, or
{5. 7}.
Case i: {C. r} = {3. 9} , then {B. I:.,}= {7. 8} and {4. D} = {4, 5}, producing this solution:
473 + 589 = 1062 and {{A. D}, {B. E}. {CJ}} ={{4. 5}. {7. 8}. {3, 9}}.
Case ii: {C, r} = {4.8} or {5, 1j, then none of the remaining numbers add up to 15. Therefore,
{C, f} ::t {4,8} or {5. 1j.

All solutions for 1602 can be found by interchanging sets {A. D} and {B, t..}.
Case 2c: If GHIJ = 1260, then C + F = 10,B + E = 15. and A+ D = 11. Therefore, {C, r} = {3, 1j but none
of the remaining pairs of digits add up to 11 or 15. Likewise, there are no solutions for 1620.
Case 3 {G. H, I.J} = {O, I. 3, 5}

This case produces six possible values for the four-digit number GH/J , either 1035, I 053, 1305, 1350, 1503,
1530 with {A. B. C, D, E. f} = {2. 4,6. 7. 8, 9}.
Case 3a: If GHIJ = 1035, then C + F = 15,B + E = 12, and A+ D= 9. Therefore, {A, D} = {2. 7} and so
{B, E} = {4, 8} and {C, r) = {6, 9}, producing this solution: 246 + 789 = 1035 and
{{A, D}, {B, E}, {C, F}} ={{2.1j, {4, 8}, {6, 9}} ..
All solutions for 1305 can be found by interchanging sets {A, D} and {B, t..}.
Case 3b: If CHI.I= 1053, then C + F = 13,B + E = 14, and A+ D= 9. Therefore, {A. D} = {2, 7} and so
{B. t..} = {6, 8} and {C, Pj = {4, 9}. producing this solution: 264 + 789 = 1053 and
{{A. D}, {B. E}. {C, F}} ={{2, 1j. {6. 8}, {4. 9}}.
All solutions for 1503 can be found by interchanging sets {A, D} and {B, E}.
Case 3c: If CHI.I = 1350, then C + F = 10, B + E = 14, and A + D = 12. Therefore, {A. D} = {4. 8} but none
of the remaining pairs of digits add up to 10 or I 4. Likewise, there are no solutions for 1530.
Id. In le six independent solutions were found:

432 324 437 473 264 246
657 765 589 589 789 789
1089 1098 1206 1062 1053 1035
374
In each solution two columns can be interchanged to produce another set of solutions:
423 342 347 743 624 426
675 756 859 859 879 879
1098 1098 1206 1602 1503 1305
In each of these ten solutions the pairs of addend digits in any column can be permuted to produce 23 = 8 unique
solutions. Therefore, there are 6 x 2 x 8 = 96 unique solutions.
2a. It is easy to verify that the following must be true: G = 1,H = 0, and D = 9. As in problem 1, the carry digits
must total to an odd number. A+ C + F:2'. 2 + 3 + 4 =9, but J * 9. Therefore, A+ C + F:2'. 10 and
B + E + carry digit :2: 10. So A+ C + F= J + 10, B + E + I =I+ 10, G =I, H = 0, and D = 9. Since
A + B + C + D + E + F + G +I+ J = 45 ⇒ J + IO+/ + 9 + 9 + 1 + 0 + / + J = 45 ⇒ 2J + 2/ + 29 = 45 and
so J + 1 = 8.
Therefore, {G. H, I.J} = {0. I. 2, 6} or {0, 1,3. 5}.
2b. Case 1 {G, H, 1.1} = {0. L 2, 6}

This case produces two possible values for the four-digit number GHJJ • either 1026 or I 062 with
{A.B,C,E.r) ={3,4,5,7,8} and D=9.
case la: If GHJJ = 1026, then A+ C + F= 16 and B + E= 11. Therefore, {B. E} = {3. 8} or {4, 1).
If {B. E} = {3. 8}. then {A, C, F} = {4,5, 1}, producing the solution, 4 + 35 + 987 = 1026, with
{{A. C. r). {B. E}} = {{4. 5, 1}, {3,8}}.

If {B, E} = {4, 1}. then {A, C, F} = {3, 5, 8}, producing the solution, 3 + 45 + 987 = 1026, with
{{A. C, f}. {B, E}} = {{3, 5, 8}. {4, 7}}.

case b: If GH/J = 1062, then A+ C + F= 12 and B + E= 15. In this case {B. E} = {7, 8} and
{A, C, F} = {3 .4. 5}, producing the solution, 3 + 74 + 985 = 1062, with
{{A, C, f}, {B, E}} = {{3.4, 5}, {7.8}}.
Case 2 {G. H, l,.f} = {0, I, 3, 5}

This case produces two possible values for the four-digit number GHJJ , either 1035 or I 053 with
{A, B. C, E, f} = {2, 4, 6, 7, 8} and D = 9.
case 2a: If GHIJ = 1026, then A + C + F = 15 and B + E = 12. Therefore, {B, E} = {4, 8} and
{A, C, F} = {2, 6, 1}, producing the solution, 2 + 46 + 987 = 1035, with

{{A, C, r), {B. E}} = {{2, 6, 7}. {4,8}}.
375
case 2b: If GH/J = I 062, then A + C + F = 13 and B + E = 14. Therefore, {B, E} = {6. 8} and
{A. C. F} = {2,4, 7}, producing the solution, 2 + 64 + 987 = 1053. with
2c. In each of the five independent solutions found in 2b., the digits A, C, and F can be permuted in 3! = 6 ways
while the digits B and E can be permuted in 2! = 2 ways with each pennutation producing a unique solution.
Therefore, there are 5 x 6 x 2 = 60 unique solutions.
3a. Case 1: If there are two unit carry digits, then A+ B + C + D + F +G +I+ J + K + L + M = 66 with
A+B+D+G+1=12+M, C+F+/+l=L+I2. and H+l=K. Substituting,produces:
12 + M + 11 + L + K - I + K + L + M = 66 ⇒ 2M + 2L + 2K + 22 = 66 ⇒ M + L + K = 22. Therefore,
{K. L.M} = {I. T, E}. {2. 9, E}, {3, 8, !:), {3. 9, 7), {4, 7. E}. {4, 8, 7). {5. 6, !:), {5, 7, 7), {5, 8, 9}. or
{6,7,9}.
Case 2: If there is a carry digit of I and a carry digit of 3, then A + B + C + D + F + G +I+ J + K + L + M = 66

with A + B + D + G + J = 36 + M, C + F +I+ 3 = L + 12, and H + 1 = K. Substituting, produces:
36 + M + 9 + L + K - I + K + L + M =66 ⇒ 2M + 2L + 2K + 44 = 66 ⇒ M + L + K = 11. Therefore,
{K, L,M} = {0. I. 7), {0, 2, 9}, {0. 3, 8}, {0, 4, 7}. {0, 5, 6), {1. 2, 8}, {l, 3, 7}, {l. 4, 6). {2, 3, 6), or
{2.4, 5}.
3b. (The solution to this problem is long because I have included all ten cases.)
{I, T, E}: five independent solutions:

7 8 8 8 8
6 6 6 7 7
85 54 35 25 36
43 73 24 43 42
902 902 970 960 950
TlE TlE TlE TlE TlE
The digits in the units column can be permuted in 5! ways while the digits in the twelves column can be permuted
2 ! ways in the first two solutions and 3 ! ways in the last three solutions. Therefore, there are
2 x 2! x 5! + 3 x 3! x 5! = 2640 unique solutions.
376
----------------------------------------------------------
{2, 9, El: sixteen independent solutions:
8 7 7 T 6 7 8 7 T
5 6 6 5 4 6 6 6 6
74 84 TS 74 83 85 T4 TS 84
63 53 34 63 71 43 73 83 73
T08 TOI filli 801 TSO TIO _l2Q 140 150
E29 E29 92E 92E E92 E29 2E9 2E9 29E
6 T 8 8 T 8 8
4 6 6 7 7 6 7
T3 74 TS TS 35 74 65
71 53 74 63 41 53 41
850 810 130 140 860 TIO no
9E2 92E 29E 29E 92E E29 E29
2! ways in the first four solutions and 3! ways in the last twelve solutions. Therefore, there are
4 x 2! x 5! + 12 x 3! x 5! =9600 unique solutions.
{3, 8, E}: twelve independent solutions:

7 7 9 T 7 7 9 9
6 6 6 6 5 6 5 6
94 TS TS 94 92 95 74 74
52 94 42 52 61 42 62 51
TOI 201 701 701 T40 TIO TIO T20
E38 38E 83E 83E E83 E38 E38 E38
9 T 9 T
6 5 7 6
T4 94 T6 95
71 71 51 42
250 260 240 710
3E8 3E8 38E 83E
2! ways in the first four solutions and 3 ! ways in the last eight solutions. Therefore, there are
377
----------------------------------------------------------
{4, 7, E:: fourteen independent solutions:
8 9 6 8 8 9 9 T
5 6 5 5 6 5 6 6
93 T5 93 92 93 83 83 95
62 82 82 61 52 62 51 82
TOI 301 TIO T30 TIO TIO T20 310
E47 47E E74 E74 E47 E47 E47 47E
9 T T T 8 T
8 6 8 9 5 3
T5 92 93 83 T2 92
61 81 52 51 91 81
320 350 610 620 630 650
47E 4E7 74E 74E 7E4 7E4
2! ways in the first two solutions and 3! ways in the last twelve solutions. Therefore, there are
2 x 2! x 5! + 12 x3! x 5! =9120 unique solutions.
{5, 6, E}: eleven independent solutions:

7 8 T 9 8 9 8 9
4 4 7 8 4 4 7 4
93 93 93 T3 93 83 92 83
82 72 82 72 72 71 41 72
TOI TOI 401 401 TIO T20 no TIO
E65 E56 56E 56E E65 E65 E56 E56
T T 8
8 9 7
93 83 T2
72 71 91
410 420 430
56E 56E 5E6
2! ways in the first four solutions and 3 ! ways in the last seven solutions. Therefore, there are
378
{3, 9, T): nine independent solutions:

7 7 E E 8 E 7 8 E
5 5 5 6 7 5 6 7 6
E2 E4 74 74 ES 84 ES E6 84
61 26 62 51 61 71 84 51 71
840 810 fil.Q 820 240 260 210 240 250
9T3 93T 93T 93T 3T9 3T9 39T 39T 39T
3! ways. Therefore, there are 9 x 3! x 5! = 6480 unique solutions.
{4, 8, T}: thirteen independent solutions:

E 6 7 7 7 E 6 9
5 5 5 6 6 5 5 6
93 E3 E3 E2 ES 73 E3 ES
62 72 61 51 32 61 92 32
701 910 920 930 2lQ 920 710 lLQ
84T T84 T84 T84 T48 T48 8T4 84T
9 9 E 7 E
6 7 6 6 6
15 25 13 15 52
72 61 52 92 71
3E0 3E0 790 3E0 390
48T 48T 84T 4T8 4T8
2! ways in the first solution and 3! ways in the last twelve solutions. Therefore, there are
Ix 2! x 5! + 12 x3! x 5! =8880 unique solutions.
379
----------------------------------------------------------
{5, 7, T}: seventeen independent solutions:
8 8 E 8 8 E 8 9
4 6 3 6 6 4 4 4
E3 E2 82 E3 E4 83 E3 E3
62 41 61 42 31 61 92 81
910 930 940 910 920 920 610 620
T75 T75 T75 T57 T57 T57 7T5 7T5
E 9 9 E 8 9 9 E E
3 8 8 8 6 6 8 6 8
92 E3 E4 92 E3 E3 E3 93 92
81 42 31 41 92 81 62 82 61
640 610 620 630 410 420 410 410 430
7T5 75T 75T 75T 5T7 5T7 57T 57T 57T
3! ways. Therefore, there are 17 x 3! x 5! = 12240 unique solutions.
{5, 8, 9}: twelve independent solutions:

E T T T E T T E
4 6 6 4 3 6 7 6
T3 E3 E4 E2 T2 E3 E2 T2
62 42 31 61 61 71 61 71
701 710 720 730 740 420 430 430
859 859 859 895 895 598 598 598
T T E E
6 7 6 7
E3 E3 T3 T2
72 61 71 61
410 420 420 430
589 589 589 589
2! ways in the first solution and 3! ways in the last eleven solutions. Therefore, there are
380
{6, 7, 9}: eleven independent solutions:

E T T E T T E E
4 4 5 4 4 5 4 5
T3 E3 E2 T2 E3 E3 T3 T2
82 51 41 51 52 41 51 41
501 820 830 830 810 820 820 830
679 976 976 976 967 967 967 967
T E T
4 4 8
E3 T3 E2
82 81 41
510 520 530
697 697 679
2! ways in the first solution and 3! ways in the last ten solutions. Therefore, there are
{0, 1, T}: two independent solutions:
4 5
6 6
27 27
38 38
95E 94E
TIO TOI
2! ways. There are 2 x 3! x 5! = 1440 unique solutions.
381
{0, 2, 9}: seven independent solutions:
5 4 4 3 4 3 4
6 6 5 5 5 5 6
37 37 36 47 16 17 17
48 58 7T 6T 3T 4T 3T
!ET !TE !SE !SE 87E 86E 85E
290 290 290 290 920 920 902
{0, 3, 8}: ten independent solutions:
4 4 4 1 4 4 4 2 4
6 5 5 5 6 5 5 4 5 5
17 17 16 49 57 67 16 19 29 19
59 69 7T 6T 19 19 2T 5T 4T 2T
2ET 2TE 29E 27E 2ET 2TE 79E 76E 76E 76E
380 380 380 380 380 380 830 830 830 803
{0, 4, 7}: nine independent solutions:
5 2 2 I 2 3 2
8 6 5 6 5 8 5 5
19 18 18 28 29 19 18 18
2T 59 6T 5T 6T 3T 29 3T
36E 3TE 39E 39E 38E 65E 6TE 69E
407 470 470 470 470 704 740 740
382
{0, 5. 6l: two independent solutions:
2
7 7
18 28
39 39
4ET 4TE
560 560
{0, 5, 6}: two independent solutions:
2
7 7
18 28
39 39
4ET 4TE
560 560
3 ! ways. Therefore, there are 2 x 3 ! x 5 ! = 1440 unique solutions.
{l, 2, 8}: two independent solutions:
3 3
5 4
49 59
6T 6T
70E 70E
812 821
in only 2! ways ( I must be 0). There are 2 x 2! x 5! = 480 unique solutions.
383
{I, 3, 7 l: six independent solutions:
5 5 4 4 4 4
8 6 6 5 5 5
49 48 58 69 29 29
6T 79 9T 8T 8T 8T
20E 20E 20E 20E 60E 60E
317 371 371 371 713 731
The digits in the units column can be permuted in 5 ! ways while the digits in the twelves column can be permuted
{I, 4, 6}: nine independent solutions:
5 2 2 3 2 3 2 7 7
7 7 5 7 8 7 4 9 9
29 58 79 29 39 28 37 2T 2T
8T T9 8T 8T 7T E9 T8 5E 3E
30E 30E 30E 50E 50E 50E 50E 380 580
416 461 461 614 614 641 641 461 641
In the first seven solutions, the digits in the units column can be permuted in 5! ways while the digits in the
twelves column can be permuted only 2! ways. In the last two solutions, the digits in the units column can be
permuted in 5! ways while the digits in the twelves column can be permuted 3! ways. There are
{2, 3, 6}: ten independent solutions:

5 4 5 4 4 4 1 4 0
7 8 7 7 5 7 8 7 7 8
49 59 48 T8 79 18 49 E8 49 19
8T 7T E9 59 8T T9 7T 19 8T 4T
JOE lOE JOT lOE lOE 50E 50E 50T 50R 57E
236 236 263 263 263 623 623 632 632 632
In the first nine solutions, the digits in the units column can be permuted in 5 ! ways while the digits in the twelves
column can be permuted only 2! ways. In the last solution, the digits in the units column can be permuted in
5! ways while the digits in the twelves column can be permuted 3! ways. There are
384
----------------------------------------------------------
(2, 4, 5 l: six independent solutions:
6 3 6 3 6
7 8 7 7 7 7
38 69 38 69 18 69
T9 7T E9 8T T9 8T
JOE JOE !OT !OE 30E 30E
245 245 254 254 425 452
There are 6 x 2! x 5! = 1440 unique solutions. WHEW!!
385
ARML Power Contest - February 2003 - Number Theoretic Functions
Number Theoretic Functions
The Definitions. Symbolism. and Theorems
Number theory is a branch of mathematics that deals with the properties of positive integers. This problem will
deal with six functions that are used to state some of these properties. They are defined as follows:
't(11) = the number of positive integral divisors of 11. For example, 't( 12) = 6 because 12 has six divisors,
namely L 2, 3. 4, 6, and 12.
0(11) = the sum of the positive integral divisors of 11. For example, <J( 12) = 28 because
I + 2 + 3 + 4 + 6 + 12 = 28 .
q>(11) = the number of positive integers less than n and relatively prime ton. <p(l2) = 4 because I, 5, 7, 11
are relatively prime to 12. Two integers, a and b, are relatively prime if and only if their greatest common
factor is I (or symbolically gcf(a, b) =1 or simply (a, b) = I).
The fundamental theorem of arithmetic states that any positive integer, n, can be written as a product of primes in
onlyoneway,i.e., n=p'; 1 •p?-p; 3 - ••• -p;k ,where p,'s areprimesand als are positive integers. For
example, 12 = 2 2 · 3 1 •
A.(11) = ( - 1) a 1 +a 2 + ... +ak and A())= 1. Therefore, A.(12) = (-1)( 2+l) = -1.
µ(11) = 0 if any ak is greater than I, otherwise µ(n) = (-1/ and µ(l) = I. Therefore, µ(12) = 0.
8(11) = number of ordered pairs (a,b) such that (a,b) = 1 and ab = n. 8(12) = 4 for the pairs (I, 12), (3, 4),
(4, 3), and (12, I) have no common factors. The pairs (2, 6) and (6, 2) are not relatively prime.
It is amazing that such functions with such distinctive properties can be related in interesting ways but later in this
problem it will be shown that for any prime number, n., cr(n) + <p(n) + A.(n) + µ(n) + 8(11) = n • 't(n). But first,
here are some symbols and fundamental theorems regarding these functions.
Symbols:
d I 11 means d is a divisor of n.
386
----------------------------------------------------------
----------------------------------------------------------
L j(cl) means sum up all the values off(cl) where d In. For example,
din
1(11) = L
dl11
I and cr(n) = L cl.
din
L
(d.111::::.1
L I I.
f(d) means sum up all the values off(d) where dis relatively prime ton. For example, cp(n) = (c/.11)-;;c,.
Q j(cl) means multiply together all the values off(d) where d I 11.
Useful Theorems:
I. A function, f; is called multiplicative if for any positive integer, n, where n =a · b and (a.b) =I .
f(n) = f (a· b) = f(a) · f(b). All six functions are multiplicative. For example, cp(a·b) = cp(a) · cp(b) .
2. If a is a non-negative integer and p is a prime number, then ...

d) A(pa) = (-1 )a.
b) cr(p 0 ) = p
a+ I
p-1
-
l
e
) (pa) =
µ
{-1, if a =
0, otherwise
I} ·
a) t(n) = (aJ +l)(a2 +J) ... (ak + 1).

nPa+I_J
b)cr(n)= pinp-1
c) cp(n) = nn (1 -~'.
pinr~
In math class you probably learned about the composition of two functions, f (g(x)) , sometimes denoted as
.f g(x) . In number theory two functions can be combined by a method called the convolution of two
O
functions, often denoted as f*g(n) . It is defined as .f*g(n) = L j(d) · g(lld)

1/'in
.
The following chart may be useful in solving the problems of the contest.
387
Number Theoretic Function Table

prime
11 factorization r(n) CJ(n) <p(n) A(n) µ(n) 0(n)
2 2 2 3 -I -1 2
3 3 2 4 2 -I -1 2
4 22 3 7 2 0 2
5 5 2 6 4 -1 -I 2
6 (2)(3) 4 12 2 4
7 7 2 8 6 -1 -1 2
8 23 4 15 4 -1 0 2
9 32 3 13 6 0 2
10 (2)(5) 4 18 4 4
11 11 2 12 10 -1 -1 2
12 (2 2 )(3) 6 28 4 -1 0 4
13 13 2 14 12 -1 -1 2
14 (2)(7) 4 24 6 4
15 (3)(5) 4 24 8 4
16 24 5 31 8 0 2
17 17 2 18 16 -1 -I 2
18 (2)(3 2 ) 6 39 6 -1 0 4
19 19 2 20 18 -1 -1 2
20 (2 2 )(5) 6 42 8 -1 0 4
21 (3)(7) 4 32 12 4
22 (2)(11) 4 36 10 4
23 23 2 24 22 -1 -1 2
24 (2 3 )(3) 8 60 8 0 4
388
The Problems
Part 1
I. Find the value for each of the six functions when n = 72.
(/ + t _ l
2. Prove, if a is a non-negative integer and pis a prime number, then cr(p") = pp- 1
3. For what values of n is cp(n) =~ ? Prove your conjecture.
4. For what values of n does cp(n) = cp(211) ? Prove your result.

Part 2
5. For each of the following expressions:
i) Find its value when n = 24. Show your work.
ii) Experiment with enough other values of n (No need to show your work here.) which will allow you to
make a conjecture about the expression that is true for all n. (All conjectures will involve a single
number theoretic function or a constant.)
a. cp * 't(11) .
b. A* 8(11).
c. ~ 't(d 2 )µ(j) .
d. cp * cr(11) .
e. fen (T(d)) 3 •
6. For what values of 11 does cp(11) + cr(n) = 211 ?
~ A(d) = I ?
7. For what values of n does tr,:
Part 3
8. Find all values of n for which <p(n) = 24 . Show your work.
9. If cr(n) = 2n , then n is called a perfect number.

a. Show 6, 28, 496, 8128 are perfect.
b. Show if 2" - I and 11 are prime, then 2"- 1 (2" - I) is perfect.
c. Show ~ J= 2 if and only if 11 is perfect.
10. Prove, if n is prime, then cr(n) + cp(n) + A(n) + µ(n) + 8(11) = n · 't(n)
389
The Solutions
Part I
I. 72=2 3 -3 2 andthefactorsof72are 1,2,3,4,6,8,9, 12, 18,24,36, 72.
i-(72) = (3 + I )(2 + I) = 1 2, the number of factors of 72.
cr(72) = ( 2 4 - 1) ( 3 3 - 1) = 15 · I 3 = 19 5 or
2-1 3-1
cr(72) = I + 2 + 3 + 4 + 6 + 8 + 9 + 12 + 18 + 24 + 36 + 72 = 195 .
<p(72) = 2' (1 -½)3 2 (1 -½) = 8(½)9G) = 4 · 6 = 2 4, the number of elements in the set
{I. 5, 7, II, 13, 17, 19, 23, 25, 29, 31, 35, 37, 41, 43, 47, 49, 53, 55, 59, 61, 65, 67, 71}.
A(72)=(-1)( 3 +2 i=(-I/=-1.
µ(72)= 0.
0(72) = 4, the number of elements in the set {(I, 72), (8, 9), (9, 8), (72, I)}.
2. The only factors of p" are I, p. p2, p3, ... ,p".
an(l - r")
Therefore, cr(p") = I + p + p 2 + p 3+ ... + p" • an infinite geometric sequence, whose sum is -----'----'-
1- r
J(J a+]\ a+I J

Therefore, cr(p") = - p 1 =p -
1-p p-l
3. <p(n)=~ iffnisapowerof2.
Proof: <== If n is a power of 2, then n = 2 a • <p(2j =2 a(l - ½) =2 "(½) = ~.

==> 2<p(n) = n. Therefore, n has a factor of 2. Assume there are other prime factors, i.e.,
2(2" 1(½} p/2 (! --k}p/3 (1 _ JJ .... p/k(] _ -k)) = 2" I. p/2. p/3 . .... Pk ak.
2 a 1• (1 - ~J (1 -i}· ... · (1 - -k) = 2" 1 • If P2, Pi, ... ,Pk exist then the left side is less than the right
side ... a contradiction. Therefore, n must have no prime factors other then 2.
4. <p(n) = <p(2n) iff n is odd.
Proof: <== If n is odd then (2, n) = 1 and so <p(2n) = <p(2) · <p(n) = 1 · <p(n) = <p(n) .
==> Suppose cp(n) = cp(2n) and n is even, i.e., n = 2" 1 • p/' 2 p/ 3 • ••• • Pkak and
<p(n) = 2"1(½)· p2a2 (1 -i}p/ 3 (1 - fJ ... ·p/k(I -fJ. Then 2n = 2a +11 • f)]." 2 •p/' 3 • ... · p/k and
<p(2n) = 2"+ 11 (½)· /'2." 2 (1 -j-Jp/ 3 (1 - kJ· ... ·p,/{1-;k), which equals
390
contradiction.
Part 2
I: cp(d) · tta)
5a. <p * 1(24) = a124 ~ . The divisors of 24 are L 2, 3, 4, 6, 8, 12, and 24.
So <p * 1(24) = <p(l) · 1(24) + cp(2) · 1(12) + cp(3) · 1(8) + <p(4) · 1(6) + cp(6) · 1(4) +
cp(8) · 1(3) + <p(l2) · t(2) + <p(24) · 1(1).
= 1(8) + 1(6) + 2(4) + 2(4) + 2(3) + 4(2) + 4(2) + 8(1) .
= 8 + 6 + 8 + 8 + 6 + 8 + 8 + 8 = 6 0.
Therefore, q> * 't(n) = a(n) . Other examples should verify this.
5b. ')-.. * 8(24) = ')-...(1) · 8(24) + ')-...(2) · 8(12) + ')-...(3) · 8(8) + ')-...(4) · 8(6) + ')-...(6) · 8(4) +
')-...(8) · 8(3) + ')-...(12) · 8(2) + ')-...(24) · 8(1).
= 1(4) + (-1)(4) + (-1)(2) + 1(4) + (-1)(2) + (-1)(2) + (-1)(2) + 1(1).
=4-4-2+4-2+2-2+1=1.
Therefore, A. * 8(n) = 1 . Other examples should verify this.
5c. );4 1(d 2 ) • µ(';?) =1(1) · µ(24) + 1(4) · µ(12) + 1(9) · µ(8) + 1(16) · µ(6) +
1(36) · µ(4) + 1(64) · µ(3) + 1(144) · µ(2) + 1(576) · µ(l).
= 1(0) + 3(0) + 3(0) + 5(1) +9(0) + 7(-1) + 15(-1) + 21(1).
= 0 + 0 + 0 + 5 + 0 - 7 - 15 + 21 = 4.
Therefore, );4 t(d 2 ) • µ(~ = 8(n) . Other examples should verify this.
5d. cp * cr(24) = cp(l) • cr(24) + cp(2) • cr(l2) + cp(3) • cr(8) + cp(4) • cr(6) +cp(6) • cr(4) +
<p(8) • cr(3) + <p(l2) • 8(2) + cp(24) • cr(l)
= 1(60) + 1(28) + 2(15) + 2(12) + 2(7) + 4(4) + 4(3) + 8(1)
= 60 + 28 + 30 + 24 + 14 + I 6 + 12 + 8 = 19 2.
Probably not enough to make a conjecture yet and so more examples are needed:
cp* cr(l) =I, <p* cr(2) = 4, <p* cr(3) = 6, <p* cr(4) = 12, cp* cr(5) = 10, and
Therefore, q> * <J(n) = Wt(n) .
~ [ ]3
5e. a"r!
4 1(d) = 1 3 + 2 3 + 2 3 + 3·3 + 4 3 + 4 ·3 + 6 3 + 8 3 = 900. Agam
• •
probably not enough to make a conJecture and
so more examples are needed: ,~ [1(d)] 3 = 1 3 + 2 3 = 9, ~ [1(d)] 3 = 1 3 + 2 3 = 9,
391
All the answers are squares ... but squares of what? Observe that ~" ('t(d)) 3 = ~ 't(d)) 2 .
6. Examples: cp(2) +cr(2) = 4, cp(3) +cr(3) = 6, <p(4) +cr(4) = 9 t:- 8, <p(5) +cr(5) = 10, cp(6) +cr(6) = 14 t:- 12 .. .If n
is prime then <p(n) +cr(11) = 211.

7. Examples: ~ Al._d) =I , ~ Al._d) = 0, ~ Al._d) = 0, ~ Al._d) = I ... If 11 is a square then ~ Al._d) = I .
Part 3
8. cp(n) = 24 =nIT (1- pl)

f! In
= plal . P2a2 . p3"3 . ···.Pk"k(] - p1 1)(·1 - [)21)(I - /J3I) . · · ·. () - /JI-1)
Therefore, if p is a prime factor of 11 , then (p - I) must be a divisor of 24 and so (p - I) could be 1, 2, 3, 4, 6, 8,

12, or 24. Therefore, the only possible prime factors on could be 2, 3, 5, 7, or 13.
<p(l3) = 12,<p(7) = 6. <p(5) = 4. cp(3) = 2. <p(i) = 6, <p(2) = 1, cp(2 2 ) = 2.<p(2 3 ) = 4. and cp(2 4 ) = 8.
All other powers of these primes produce q>--values which are not divisors of 24. Since <p(n) is multiplicative
and equal to 24, the following are the only possible values for n :
24=(12)(2): <p(l3)·<p(3) ⇒ n=39
cp(l 3) • cp(2 2) ⇒n= 52
24 =( 12)(2)( 1): cp(l3) · <p(3) · <p(2) ⇒ n=78
24 = (8)(3): not possible ... all q>--values are even except <p(l) and <p(2) .
24 =(6)(4): <p(7) · <p( 5) ⇒ n=35
<p(7) · <p(2 3 ) ⇒ n=56

,,
cp(3~) ·cp(5) ⇒ n=45
<p(32) ·<p(22) ⇒ n=72
24 = (6)(4)(1): <p(7) · cp(5) · <p(2) ⇒ n=70
<()(3 2 ) · <p(5)) · <p(2) ⇒ n=90
24 = (6)(2)(2): <p(7) · <p(3) · <p( 2 2) ⇒ n=84
9a. 6 = 2 • 3 and (2,3) = 1, so cr(6) =cr(2)cr(3) = 3 • 4 = 12 = 2(6).

28 = 4 -7 and (4,7) = 1,so cr(28) = cr(4)cr(7) = 7 · 8 = 56 = 2(28).
392
496 = 16 · 31 llild ( 16.31) =I.so cr(496) = cr(l6)cr(3I) = 31 • 32 = 992 = 2(496).

8128 = 64 · 127 and (64,127)= I. so cr(8128) = cr(64)cr( 127) = 127 • 128 = 2(8128).
9b. 2"- I is odd and 2"- 1 is even. Therefore, (2" - I, 2"- 1) = I and so cr(2''- 1 (2" - 1)) =cr(2"- 1) • cr(2" -1).
By Theorem 2b, cr(2" - 1) = ~, ~ l . Since the only factors of a prime, p, are 1 and p, cr(p) = p + I and so
cr(2" -1) = 2 11 • Therefore, cr(2"- 1 (2 11 - 1)) = (2" - I) (2') = 2(2" 1)(211 -1) =211.
9c. <== If n is perfect, cr(n) = 211 . But cr(11) = L d =

din
L Ill= n L ld . Therefore, 211 = n L ld and 2 = L ld .
din { ,1111 din d!n
==> L ld = 2 and multiplying both sides by

din
11 produces: n L ld = 211 , which equals
din
L l .d = 211 and so
din
cr(11) = 211 .
10. If 11 is prime , it has only two divisors, I and 11, so 't(11) = 2 and cr(11) = 11 + 1 . All the numbers less than n
are relatively prime ton, and so <p(11) = 11 - I and A(n) = - I. µ(n) = - I, and 8(11) = 2. Therefore,
cr(11) + <p(n) + A-(11) + µ(11) + 8(11) = 11 • 't(11) is equivalent to (n +I)+ (11- I)+ (-1) + (-1) + 2 = n · 2 which
is true. (Note: The converse is also true but more difficult to prove.)
Further Notes
• Euler went on to prove that all perfect numbers are of the form 2"- 1(2" - 1), where 2" - I is prime. Primes
of the form 2" - 1 are known as Mersenne primes. If 2" - I is prime then n must be prime but the
converse is not true for 2 6 7 - 1 is composite. As of this writing there are only 39 known perfect numbers.
In November 2001 it was proven that 2 1346 69 17 - I is prime. Its corresponding perfect number has
8, I 07 ,892 digits! No one has yet found an odd perfect number but much is known about them I Euler
proved that any prime factor of an odd perfect number must be of the form 4n + 1 and more recently it has
been shown that an odd perfect number must have at least eight prime factors, one being greater than IO 18.
• An n - gon can be drawn using a compass and straightedge if and only if <p(n) is a power of two.
• In Lure of the Integers, Joe Roberts states that cr(n) < 6 '?
1c
for all values of 11 , except 2, 3, 4, 6, 8, and 12;
't(n) < 3R except when n is 2, 4, 6, 12; and cp(n) > (ii except for 2 and 6.
• More investigation and discussions about number theoretic functions can be found in An Introduction to
Arithmetical Functions, by P. J. McCarthy.
393
ARML Power Contest - November 2003 - Errors in Math Reasoning
Errors in Mathematical Reasonine
In this contest problem you will be exploring errors in mathematical reasoning that ultimately lead to correct
solutions. Do the errant methods work some of the time, all of the time, or just in one isolated case?
The Problems
1. Sometimes fractions can be reduced by canceling out digits. For example,
Find the other two fractions that can be reduced in a similar manner, i.e. f =1i[ =~, where ~ is a reduced
fraction with a< c .
2. While the distributive property of multiplication over addition, a· (b + c) =a· b +a• c, always works with any
set of real numbers, the "other" distributive property of addition over multiplication,
a+ (b · c) =(a+ b) · (a+ c), is less successful. The "other" distributive is valid for the following example:
.5 + (2 • .3) = '5 + .2) • (.5 + .3) = (.7) • (.8) = .56 and .5 + .2 • .3= .5 + .06 = .56. Under what conditions
for a, b. and c, does a+ (b • c) = (a+ b) • (a+ c) '?
3. Anna incorrectly solved the following inequality but came up with the correct answer:
lxl + Ix- 1I< 2

I2x - 1 I < 2 (In general, Ia + b I -:t Ia I + Ib I )
-2 < 2x- I< 2
-I <2x<3
=t<x<~
Underwhatconditionsfor a,b,andc with a<b and c>O. islx-al+lx-bl<c the same as
I2x - (a+ b) I < c '?

4. Ning was asked to solve the following problem:
Solve for x: fx + fx - a= 4, where a 2:0.
Remembering the difference of squares identity, he proceeded as follows:
(ft'+ ,x - a,C(x- ,:x=a) = x - (x -a)= a.
Therefore, (fx - , x - a) = (fx ) .

x + ,x-a
And so, ((x- ,x-a) = 2.
394
Adding (fx+fx-a)=4 and (fx-fx-a)=~- hegot 2fx=4+~
2(x = 1.6...±a
4
(x = 1.6...±a =2 + {l
8 8
X= (2+ ~r
When a= 4, x = f. a correct solution. When a= 8. x = 9. and when a= 16, x = 16. Both of these are also
correct solutions to this problem. However, when a= 32. x = 36 and this is not a correct solution to the
problem! Where. if any place, is the flaw in Ning's solution' 1
5. Oftentimes, due to carelessness, log (b") and (logb)" get interchanged in a problem. For what positive values
ofb and positive integer values ofn will this interchange not create a wrong answer to the problem, i.e., under
these conditions, when does log (b ") = (log b)" ?
3~
6. Simplify: 3"7
j
Method I: Change to improper fractions and divide:

32 17
~=~=1/+¥=1/
j j
·h= ~
Method 2: Eliminate fractions by errantly multiplying by 1 (in this case f! ):
Under what conditions does

at
de reduce to the same fraction using both methods?
7
m:r:; ~m
7. To add fractions. sometimes Siu Hin errs and adds the numerators together and then adds the denominators
together, which generally produces a wrong answer. However. he finds ~+ = = ,-rs~m

produces a correct solution! Under what conditions for a. b. c, and d, does ~ + ~ = g+S'?
8. Paul used the following algorithm to factor a quadratic: ax 1 + bx+ c.
395
Remove the leading coefficient and multiply the constant term by it .

x 2 + bx+ a·c
Factor the resulting quadratic into linear factors.
(x- r)(i: -s)
Divide both r and s by a.
(x-&)(x-t)
Reduce the fractions, if possible.
Remove the fractions by multiplying the factors by a' and a", respectively.
(a'x - r') (a"x - s ''

For example.
6x 2 -7x-3 ➔ x 2 -7x-18 ➔ (x-9)(x+2) ➔ (x-iXx+i) ➔ (x-~Xx+}) ➔ (2x-3)(3x+ I)
Although this method is seemingly full of algebraic errors, it worked in this example! As a matter of fact, any
factorable quadratic can be factored using this method!! Prove this is true.
9. Mr. Kilkelly wrote on the board a quadratic equation in the form x 2 -Ax+ B = 0 and asked the students to
quickly solve it. In copying the problem, Sindhuja erroneously transposed the two digits of B as well as the
plus and minus signs. However, she still got one of the correct roots. If A and B are integers, what was this
root? What were the two possible equations Mr. Kilkelly wrote on the board?
I0. Using the formula for the sum of an infinite geometric series, Hwa-Sheng produced:
x +x 2 +x 3 +x 4 + ... = .....L....l-x and I+¼+~+~+ ... = -L1 = .....L....1 . Adding these two equations
X X I-- x-
x
resulted in: I +x + l +x 2 + -1, +x 3 + ~ + ... = .....L + .....L

x x x 1-x x-l
= .....L + ...;;;;.L
I- X I -x
However, if x is positive, the left side sums to a positive number! Where is the error in his reasoning?
11. Although each step of the following proof seems justified, the final result would cause Peano to rise from his
grave! Where is the flaw?
(x+lf=x 2 +2x+I ➔ (x+I)2-(2x+l)=x 2 ➔ (i:+1)2-(2x+l)-x(2x+l)=x 2 -x(2x+1) ➔
(x+If-(2x+l)(l+x)+(2.l:f-1)2=x 2 -x(2x+l)+(2.l:f-1)2 ➔ (x+1-~)2=(x-2.If-1)2 ➔
x+l- ~ =x - 2.l:f-1 · Therefore, x + 1 =x !!
396
The Solutions
I lO.a...±.1z = a ➔ Wac + be= lOab + a: ➔ 9ar +be = lOab ➔ c = .lilah... .

· !Ob + c c 9a + b
Therefore, 9a + b must evenly divide lOab . This is always tme when a= b = c and in four other situations:
a= I. b = 6,c = 4; a= I, b = 9. c= 5; a= 2.b = 6, c= 5; a= 4. b = 9, c= 8, producing
~ = ¼, ~ = }, i = ·l ~ = ~- (The last one is not acceptable because f is not reduced.)

2. a+(bc)=(a+b)(a+c) ➔ a+(bc)=a 2 +a,:+ab+bc ➔ a-a 2 -ab-a.· =0 ➔ a(l -a-b-c)=O
:. a = 0 or a + b + c = 1 .
3. With a< b, let j(x) =Ix-a I + Ix - b I and g(x) = l2x- a - b I. The graphs below show that the functions are
identical when y > b - a. Therefore, c > b - a
y = -2x +a+ b
y = -2x +a+ b
y = 2x-a-b
y = 2x- a b
y = b-a
f(x) g(x)
4. Ning's method of solution is correct. However, since the square root of a number is always positive, the greatest
x can be is 16. Therefore, all values ofx between 4 and 16, inclusively, solve the equation. (a must be
between O and 16, inclusively.)
5. log(b")=nlog(b). Letx=log(b). Therefore, nx=x" ➔ O=x"-nx ➔ O=x(x"- 1 -n)
:. x = 0 or x"- 1 =n . This implies log(b) = 0 or {log b)"- 1 = n

Case I: log(b) = 0, implies b = 1 and n is any positive integer.
Case 2 : n = 1 implies x0 = 1 , which is true for any positive x. So, if n = 1, b is any positive number and
n -lf;,
if n > 1 then b = 1 0 ". (The exponent of IO can be positive or negative if n is odd.)
397
h a·+ h
a+c_ . _<uj. + hf Method 2: a+ a+ hf
b
c
1-~- cdf +ce
6. Method 1:
d+ d+y=d+ce·
Qlf +bf a+hf

Therefore, df
c +ce
= -d-
+ce
➔ Q('c!l + hdf + ac 2ef + hcef = Q('{ff + Q('e + bed r, + hcef ➔
ac 2ef-Ql'l' =lx¥i/2-bdf ➔ Ql'e(cf- I)= bdf(cf- l) ➔ ace= bdf.
7. ~ + ~ ➔ ~ . Therefore, ~ = m. And so, afxf +lx¥i = afxf + ad 2+ b 2c + lx¥i ➔
8. Start with a quadratic that is factorable: (ax+ b)(cx+d) = a:x 2 + (al + bc)x+ bd and follow the algorithm:
a:x 2 +(al +bc)x+bd ➔ x 2 +(al +bc)x+acbd ➔ (x+al)(x+bc) ➔ (x+'{f.Xx+'lf.) ➔
(ex+ d)(ax + b) .
9. Board'sequation: x 2 -Ax+IOt+u=O; Sindhuja'sequation: x 2 +Ax-I0t-t=0 Let r1andr2 bethe

roots of the first equation and r2 and r3 be the roots of the second. Then r1 + r2 = A and r2 + r3 = -A ,
implying r1+r3=-2r2. Also r1 ·r2=l0t+u and r2·r3=-10u-t, implying r1·r2+r2·r3=9t-9u.
Therefore, rz(r1 +r~=9~-u) ➔ r2(-2rz)=9(t-u) ➔ -2rl=9(t-u). Therefore, r2=3 and
t - u = - 2 . Therefore, the possible values for the constant term of the original equation are 13, 24, 35, 46,
57, 68, and 79. But the only ones that are divisible by 3 are 24 and 57, implying r1 is either 8 or 19.
Therefore, the original equation was either x 2 - 1lx + 24 = 0 or x 2 - 22x + 57 = 0 .
10. a+ <r + ar2 + ar3 + ... = _.o,_l , provided

-r
Ir I < 1 . Only one of the two equations can be correct because either
lxl < 1 or ltl <I, but not both.

11. If a 2= h 2 • then a= h or a= -b . Therefore, the penultimate line should read:
(x + I - 2.l.:..±.1)
2
2
= (x - 2.r..±..1.)
2
2
➔ .x + I - 2..i:.±..1
2= .x - 2.r..±..1.
2 2 = -x + 2..i:.±.l
or x + I - 2..i:.±..1 2 · The left
solution implies x = x + I, an impossibility; while the right solution implies O = 0, affirming that the original
equation is always true!
398
ARML Power Contest - February 2004 - Mathematical Strings
Mathematical Strings
The Definitions
A mathematical string is any ordered list of symbols, where a symbol can occur more than once. While strings
occur in probability theory, topology, combinatorics, set theory, and other branches of mathematics, they are
also very commonplace in today's world. Words, phone numbers, license plate codes, UPC symbols are all
examples of strings. Using the letters from a set, such as {a, b,c}, this problem set will look at strings such
as t:lHI , ababba , and bbccbccbb. The length of a string, fl, is equal to the number of symbols in the string.
The Problems
1a. Using the set of symbols, {a, b, c. d, e}, how many strings of length fl can be made? Justify your answer.
lb. Using the set of symbols, {a, b,c, d, e}, how many palindromic strings of length n can be made? (A
palindromic string is the same when read forwards or backwards, like abba, ccc, and bacab.) Justify your
answer.
le. Using the set of symbols, {a, b,c, d, e}, how many strings of length fl (fl~ 2) can be made that contain
exactly 2 a's ?
Using an ordered set of symbols, {a1, a2,. a3, ... ,ak}, where a1 < a2 < a3 <... <ak, many nondecreasing strings of
length n can be made. (A string is considered nondecreasing if for each pair of adjacent symbols in the string,
the symbol on the left side of the pair is less than or equal to the symbol on the right.) For example, using an
ordered set of symbols, {a, b}, where a < b, there are three nondecreasing strings of length two ( m , ab, bb )
and four nondecreasing strings of length three ( an , aab , abb. bbb ). In general, using an ordered set of
symbols,{a,b}, where a<b,thenumberofnondecreasingstringsoflength n thatcanbemadeis n+ 1.
2a. Using an ordered set of symbols, {a, b,c}, where a< b < c, how many nondecreasing strings of length fl can
be made? Justify your answer.
2b. Using an ordered set of symbols, {a, b,c, d}, where a< b < c < d, how many nondecreasing strings of length
n can be made? Justify your answer.
2c. Using an ordered set of symbols, {a1, a2 .. a3, ... ,ak}, where a1 < a2 < a3 <... <ak, how many nondecreasing
strings of length n can be made? Justify your answer.
The string abc has six substrings: a, b.c.ab,bc, and alx.·. Using the symbols in the set {a, b}, abbaa is a
string with four substrings of length two: ab. bb. ba, and cu. With a length of five, aabba is an example of
the longest string with no repeated substrings of length two and aaabbbalxw is an example of the longest
string with no repeated substrings of length three.
399
3a. Using the symbols in the set {a, h}, how long is the longest possible string with no repeated substrings of
length four? Give an example of such a string.
3b. Using the symbols in the set {a, h,c}, give an example of the longest possible string with no repeated
substrings of length two. How long is your example?
3c. Using the symbols in the set {a, h,c, d}, give an example of the longest possible string with no repeated
substrings of length two. How long is your example?
3d. Using the set of distinct symbols {a1, a2,. a,, ... ,ak}, how long is the longest possible string with no repeated
substrings of length m? Justify your answer.
In many games, such as tennis, volleyball, ping-pong, and ultimate frisbee, you must win by two points. For
example, in a high school volleyball game, generally the first team to reach 15 points is the winner. However,
the game can not end with a score of 15 to 14. Therefore, the score must have been tied at 14 all and
"overtime" must be played. If a represents Team A winning a point following the 14-14 tie, and h represents
Team B winning a point, a string of a's and h's could represent the "overtime" portion of the game. The
strings hh and abbaaa could be strings representing the overtime, while abbaa and amaaa could not
represent an overtime.
4a. List all the "overtime" strings of length 6. How many are there?
4b. Determine a formula for the number of"overtime" strings oflength n, where 11 is any positive even integer.
4c. If two teams are evenly matched, what is the average or expected length of an overtime string? Show your
work, justifying your answer.
A string consisting of two a' s and three h's could represent the counting of ballots in an election where candidate
A receives two votes and candidate B receives three votes. The ten different arrangements of these a's and h's
(aabhh , ababh , abbab , abhba , baahh , babab , babba , hbaab, hbaba, hhbaa) ) could represent the ten different
ways of the counting the five ballots. Notice that in only two of the ten countings, namely bbaba and
hhhaa, candidate Bis always ahead of candidate A at each step in the count. If N(A). the number of votes for
candidate A, equals 2 and N(B), the number of votes for candidate B, equals 3, the probability that candidate B
is always ahead in the count is 2/1 o or Ifs.
Sa. If N(A) = 3 and N(B) = 4, how many different arrangements of a's and h's are there?
Sb. If N( A) = 3 and N( B) = 4, make a list of the strings where candidate B is always ahead of candidate A
throughout the counting of the ballots.
Sc. In general, if N(A) represents the number of votes for candidate A and N(B), the number of votes for
candidate B, with N(B) > N(A), what is the probability, in terms of N(A) and N(B), that candidate B will
always be ahead of candidate A throughout the counting of the ballots? Justify your answer.
400
The Solutions
la. 5" There are five choices for each of then spots.
n n+I
1b. If n is even, 5 z and, if n is a.Id, 5 --r-. There are five choices for only half of the spots.
1c. The two spots to place the a 's can be selected in (:0 ways. There are 4 choices for each of the remaining n - 2
spots. Therefore, there are ( 2)(4n-, strings with exactly two a's.
~1 + l)(n + 2) n k=3
2a. 2 I 3 - a, b, c
2 6 - aa, ab, ac, bb, be, cc
3 10 - aaa, aab, aac, abb, abc, bbb, bbc, ace, bee, ccc
4 15 -1(3)+3(2)+6(1)
... The triangle numbers!
H~
11
(n + I)(n + 2)(11 + 3) 1 - a, b, c, d
2b. 2 - aa, ab, ac, ad, bb, be, bd, cc, cd, dd
6
3 - aaa, aab, aac, aad, abb, abc, abd, ace, acd, add, bbb,
bbc, bbd, bee, bed, bdd, ccc, ccd, cdd, ddd
4 35 - 1(10) + 2(6) + 3(3) + 4(1) (From n=3, fl can preceed
all strings beginning with g_, both fl and /J. can
preceed all strings beginning with /J., etc.)
... The tetrahedral numbers!!
2c. k- 1
D; (n+ i) {a, b} 2 n+l
(k-1)!
{a, b, c} 3
(n + l)(n +2)
2
or
{a, b, c, d} 4
(µ + l)(n + 2)(µ + 3)
6
(µ + l)(n + 2)(n + 3)(n + 4)

{a, b, c, d. e} 5
24
3a. aaaabbbbabbaababalJ
length=l9
3b. aabbaccbc1
length= l 0
401
3c. aabbccddacbdcadb
length= l 7
aaa aaa a
3d. km + m - 1 (There are k"' strings of length m using k different aab aab a
abb abb a
symbols ( k choices for each of the m spots) and, with proper bbb bbb b
bba bba b
arrangement and staggered alignment, the letters in each column are bah bah b
identical. The first symbol in each row is used to form the string, aba aha a
baa baa b aa
along with the remaining m - 1 symbols of the last string. (See
diagram at the right and discussion about this problem following the solution to problem 5.)
4a. There are eight: ababaa, ababbb, abbaaa, abbabb, baabaa, baabbb, babaaa, bababb
4b. To form all the overtime strings of length n, tack an ab or ba onto the beginning of all the overtime strings
of length n - 2 .
Therefore, Tn =Tn-2 * 2, where T2 =2 or t(n) = 2% for all positivtl!ven n.
4c. If two teams are evenly matched half of the the overtimes will be represented by strings oflength two, a fourth
by strings of length four, an eighth by strings of length six, and so on. The expected length of a overtime
period would be the sum of all possible overtimes, each multiplied by the probability that the overtime goes
that long: 2( 1h) + 4(14) + 6(1/s) + 8(116) + 10(1'32) + l 2(Yr,4) + . . . =I+ I + /4 + 12 + 11" + <fo + ...
Separate this sum into some partial sums:
= 1 + 1/c + J/4 + 1/H + 1/10 + 1'32 + .. .
+ 1h + 1/4 + 1'8 + 1/16 + 1'32 + .. .
+ ¼+ 1'8 + 1/16+ 1132 + .. .
+ 1/s + 1/16 + 1h2 + .. .
+ 1/10 + 1'32 + .. .
+ 'h2+ .. .
+ .. .
Since each sum is an infinite geometric series, each has a finite sum:
= 1 + 1h + ¼ + 1/x + 1/16 + 1132 + ... = l ~ th= 2
+ '/16+ 1132 + ... = 1 ~'712 =½

- ___:m__ - .l_
+ 1/30- + ·---~-16
+ ...
402
The sum of these finite sums again forms an infinite geometric series with a finite sum:
= 2+ I+ 1/2+ 11➔ + 1/K+ l/16+ ... = 13½ =4.
Sa. From the seven possible spots, select 3 spots to be represented by a vote for A. This can be done in a) ways.
Fill in the remaining four spots with votes for B. CT) = 3 5.

Sb. There are five: bbbbaaa, bbbabaa, bbbaaba, bbababa, bbabbaa
Sc. Collect more data:
A string of N(A) a's and N(B) b 's can be thought of as a
N (A) N(B) p I
I
"northeast path" on a lattice grid from (0.0) to
(N(B), N(A)), where a means "go up" and b means "go
2 113
right". The path representing the string bbabbaa from
1 3 2/-1 = 1/2
part Sb is shown below:
2 3 2/10 = Ys
2 4 S/15 = 1r, N(A)! c, c,
2 s 9'21 = 17
3 4 5/35 = Y1 4a
14156 = 2/8 3a
s
C
3
2a _, <v,' .'
la i
' ',
I h 2h 3h 4h Sh 6b 7b Sb 9b ... N(B)
Notice that: P = N(B)- N(A)
N(B) + N(A)
Since there are (~J)~:C'iJ;? ! arrangements of N(A) a's and N(B) h's, this also represents the number of
"northeast paths" from (0, 0) to (N(B), N(A)).

For candidate B to always be ahead of candidate A in the ballot counting, the sting of a's and h's must start with a
band N{B) must be greater than N(A). All the strings starting with ab can be represented by a path going to
the right first. The number of such paths is equivalent to the number of paths going from (lb, 0) to
~(A)+ (N(B) - 1)) 1
(N{B), N(A)). There are N(A)! (N(B) -1) ! such paths.
However, for candidate B to always be ahead of candidate A in the ballot counting, these paths cannot cross the
diagonal line from (0, 0) to (1 b, Ia) to (2b, 2a) to (3h. 3a), etc. For each of these paths, starting at (lb, 0) and
crossing the diagonal, there exists a dual path formed by taking the portion of the path from (lb, 0) to (mb,ma),
the point where the "northeast path" first crosses this diagonal, and reflecting this portion across the diagonal.
The following illustrations show an example ofa path and its dual for strings ofa's and h's:
403
N(A)' N(A)
4a
;~3a .,
·[.· 3a
2a 2a 1--....-..--.....
Ia Ia
// lh 2h 3h 4h Sb 6b 7b 8b 9h ... N(BJ lb 2b 3b 4b Sh 6b 7h 8h 9h ... N(B)

bbaabaabbb it's dual
These duals represent all the northeast paths from (0, 0) to (N(B), N(A)) going through ( 1a, 0). The number of such
((N(A)-1) + N(B))! f h
paths is equal to the number of paths from (la, 0) to (N(B), N(A)) and there are (N(A)- l)!N(B)! o t ese
paths.
Therefore, in a string ofN(A) a's and N(B) b's representing ballot counting, the probability that candidate B will
always be ahead of candidate A would be: 01(A) + (N(B)- I))! ({N(A)-1) +N(B))'
N(A)!tJ(B)-1) 1 - (N(A)-l)!N(B)!
(N(A) + N(B))!
N(Aj!N(B)!
(N(A) +N(B)- I)'N(B)- (N(A) +N(B)-l)!N(A) _ 01(A) +N(B)- I)!(N(B)-N(A)) _ 01(A) + N(B)- 1) 1{N(B) N(A)) _
(N(A)+N(B))! - (N(A)+N(B) 1 - (N(A)+N(B»(N(A)+N(B)-1)!-
N(B)-N(A)
N(A) + N (B) ·
Solution to #3 continued.
Although many teams may have solved #3 using a "guess and check" method, the following methods may appear
useful and instructive. Given a set of k different letters or symbols, the length of the longest string containing
no substrings of length m is km+ m - 1 . This is because there are km strings of length m using the S
different symbols. They overlap each other, with the initial letter of each substring being forming the
consecutive letters of the desired string. However, the last substring has m - 1 more letters to be attached to
the end of the desired string.
The longest string using {a, b} with no repeated substrings of length 5:

aaaaababbbababaabaaabbabbaabbbbbaaaa
404
The 32 substrings of length 5 from {a, b} showing how they overlap (Decimal equivalents of the binary numbers
represented by the strings with a = 0 and h = I):
aaaaa 0
aaaah
aaaba 2
aabah 5
ahahh 11
hahhb 23
abbha 14
bbhah 29
hhaha 26
babab 21
abaha 10
babaa 20
abaab 9
baaba 18
aabaa 4
abaaa 8
baaah 17
aaabb 3
aabba 6
abbab 13
bbabh 27
babha 22
abbaa 12
bhaab 25
baabh 19
aahhh 7
abbbb 15
hhbhb 31
bbbba 30
bbhaa 28
hbaaa 24
haaaa 16
Therefore, the length of the longest possible string with no repeated substrings of length m is known, but how does
one arrange all the substrings to form this string and is it always possible to do so? Euler circuits and graph
theory give us the answer!
405
----------------------------------------------------------
(aaa)
In problem 3, you were given the longest string of u's and h's with no repeated \-- -
substrings of length 2 or 3. Here is a method from graph theory for finding the ◄
second one when you know the first one. Let the vertices of the graph be all the (b~a) aa~aah)
strings of a's. and h's of length 2. (There are four of them.) Vertex ah is (bah)
connected with an arrow to vertices oo and bb, forming directed edges aha and ha ◄~ ► ab---~
ahb, respectively. (These are the only substrings of length 3 that start with ab . ' (aha) /4,
Notice that they end with oo and bb .) Now do this to the other three vertices: (hha)~ JJf,. (abb)
m is connected to m and ah, forming directed edges <m and aab ; oo is bb
connected to m and ah, forming directed edges lxK1 and bah; bb is connected
I ~
(hhh)
to ha and hh, forming directed edges lxKi and hhb. Since this is a connected
graph and very vertex has an even degree of 4, Euler proved that this graph has an
(aa)
Euler circuit, i.e., starting at any vertex you can travel each of the edges without
0
having to ever retrace your path. Here are the graphs for some of the other strings
in problem 3: (aaaa) (aa) (ab),, (ba)
~/ \t/7 b
J'a ►
(ac) (bb)
(ha)//
. 1(ab) (ca)
/ /
//,: (be)
I -►
-;
C
(aabb) ;~◄ (ch) ~

(bb) (cc)
(3b)
(3a)
406

Arml: 1995-2003

Uploaded by

Copyright:

Available Formats

Arml: 1995-2003

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Arml: 1995-2003

Uploaded by

Copyright:

Available Formats

What are the six number theoretic functions defined in the text?

What are the six number theoretic functions defined in the text?

What does the fundamental theorem of arithmetic state?

What does the fundamental theorem of arithmetic state?

AMERICAN REGIONS

by Donald Barry and Thomas Kilkelly

Copyright 2004 by ARML. All rights reserved.

Published by ARML. Website: htqJ:!!www.arml.com

Printed in the United States ofAmerica.

Preface: American Regions Math League·--------···--------------········-·--------····-···-•···----------------------------- 1v

Preface: ARML Power Contest __ ··-----------········---------·----·--·---------------·-------------------------------------------- vi

Acknowledgements: Problem writers and reviewers _____________________________ ·······---------------------------------- vm

AMERICAN REGIONS MA TH LEAGUE: 1995 - 2003

Team questions and solutions __________________________ ···--------------------------------------------------------- 3

Team questions and solutions _____________________________________________________________________________________ _ 27

Team questions and solutions ___________________________________________________ ._________________________________ _ 57

Team questions and solutions _____________________________________ ···---------------···-·-------------------------- 89

1999 ARML 115

Team questions and solutions______________________________________________________________________________________ 117

2000 ARML 143

2001 ARML 175

Team questions and solutions______________________________________________________________________________________ 177

2002 ARML 207

Team questions and solutions______________________________________________________________________________________ 209

2003 ARML 237

Team questions and solutions______________________________________________________________________________________ 239

1994-95 Color Transformations ___________________________________________________________________________________ _ 269

1995-96 Induction _____________________________________________________________________________________________________ _ 274

Rook Polynomials__________________________________________________________________________________________ 280

1996-97 Rotating Decimals__________________________________________________________________________________________ 286

Regular Closed Linkages_________________________________________________________________________________ 290

1997-98 Factorial Polynomials ____________________________________________________________________________________ _ 298

Integer Geometry___________________________________________________________________________________________ 304

1998-99 Unit Fractions________________________________________________________________________________________ . ______ _ 309

Chromatic Polynomials___________________________________________________________________________________ 315

1999-00 Twenty-five Point Affine Geometry _________________________________________________________________ _ 324

Square Sum Partitions_____________________________________________________________________________________ 331

2000-01 Slides, G!ides, and Rolides _____________________________________________________________________________ _ 336

Pythagorean Trip Jes_________________________________________________________________________________ .. _____ 342

2001-02 Cevians________________________________________________________________________________________________________ _ 350

Insane Tic-Tac-Toe _______________________________________________________________________________________ _ 359

2002-03 Three Addition Problems-------------------------------------------------------------------------------- 369

Number Theoretic Functions--------------------------------------------------------------------------- 386

2003-04 Errors in Mathematical Reasoning____________________________________________________________________ 394

Mathematical Strings _______________________________________________________________________________ .______ 399

Thomas Kilkelly, Wayzata High School

Problem Writers for the ARML Power Contest

1. New York City A___________________________________ 126 1. Chicago B ___________________________________ 82

1995 Individual Competition

1. Daniel Stronger_____________________________________ New York City A ________________________________________________________ 7

1996 Team Competition

1996 Individual Competition

1. Nathan Curtis---------------------------------------- Thomas Jefferson A______________________________________________________ 8

1997 Team Competition

1. Minnesota Gold_____________________________________ 125 1. Suffolk A____________________________________ 90

1997 Individual Competition

1. Davesh Maulik ---------------------------------------------- Nassau A-------------------------------------------------------------- 7

1998 Team Competition

1. Massachusetts A____________________________________ I 71 1. San Francisco Bay Area B___________ 138

1998 Individual Competition

Team questions and solutions ___________________ ._ _ 57

1998-99 Unit Fractions__________________________________________________________________________________ . _ 309

Pythagorean Trip Jes_____________________________________________________________________________ .. _ 342

Mathematical Strings _________________________________________________________________________ . 399

1. New York City A_ 126 1. Chicago B _ 82

1. Daniel Stronger_ New York City A ____________________ 7

1. Minnesota Gold_ 125 1. Suffolk A 90

1. Massachusetts A__________________________ I 71 1. San Francisco Bay Area B_ 138

1. Chicago A__________________ 172 1. Connecticut A_ 127