the ART of PROBLEM SOLVING Volume 2: and BEYOND Richard Rusczyk Sandor Lehoczky Art of Problem Solving Books ¢ Online Classes e Videos e Interactive Resources Bae www.artofproblemsolving.com Bee iCopyright © 1994, 1996, 2003, 2004, 2006, 2011, 2013, 2015, 2017 Sandor Lehoczky and Richard Rusezyk. All Rights Reserved. Reproduction of any part of this book without expressed permission of the authors and publisher is. strictly forbidden. For use outside the classroom of the problems contained herein, permission must be acquired from the cited sources. ISBN #: 978-0-9773045-8-5 Published by: AoPS Incorporated 10865 Rancho Bernardo Rd Ste 100 San Diego, CA 92127 (858) 675-4555, booksGartofproblemsolving.com Visit the Art of Problem Solving website at http: //www.artofproblemsolving.com [EEtEIIR] Scan this code with your mobile device to visit the Art of Problem Solving web- if, site, to view our other books, our free videos and interactive resources, our online community, and our online school. Printed in the United States of America. Seventh Edition; printed in 2017. Editor: David Patrick Cover image designed by Vanessa Rusczyk using KaleidoTile software. Cover Image: “Niagara Falls” by Vanessa Rusczyk, This book was produced using the ATX document processing system.www.artofproblemsolving.com The Art of Problem Solving (AoPS) is: > Books For over 24 years, the classic Art of Problem Solving books have been used by students as a resource for the American Mathematics Competitions and other national and local math events. Every school should have this in their math library. ~Paul Zeitz, past coach of the U.S. International Mathematical Olympiad team The Art of Problem Solving Introduction and Intermediate texts form a complete curriculum for outstanding math students in grades 6-12. The new book [Introduction to Counting & Probability] is great. I have started to use it in my classes on a regular basis. I can see the improvement in my kids over just a short period. ~ Jeff Boyd, 4-time MATHCOUNTS National Competition winning coach > Classes The Art of Problem Solving offers online classes on topics such as number theory, count- ing, geometry, algebra, precalculus, calculus, computer programming, and problem solving at beginning, intermediate, and Olympiad levels. Over 12,000 students will participate in an AOPS online class in 2017, All the children were very engaged. It’s the best use of technology I have ever seen. ~ Mary Fay-Zenk, coach of National Champion California MATHCOUNTS teams > Forum As of September 2017, the Art of Problem Solving Forum has over 295,000 members who have posted over 6,000,000 messages on our discussion board. Members can also participate in any of our free “Math Jams.” I'd just like to thank the coordinators of this site for taking the time to set it up... think this is a great site, and Ibet just about anyone else here would say the same. . - AoPS Community Member » Videos, tutorials, interactive resources, and much, much more! Membership is FREE! Come join the Art of Problem Solving community today!Yalli: I know that somewhere deep down you believe in me, and that makes this worthwhile. I flower dogs. SL Vanessa, thank you for your dedication, support, and patience. Without you, these books would still just be an idea. Thank you for making all my dreams real. —RRThanks A large number of individuals and organizations have helped make the ART of PROBLEM SOLVING possible. All of the following people and groups made very significant contributions, and we offer our deepest gratitude to them all. Samuel Vandervelde. Sam collaborated with us in creating the Mandelbrot Competition; he continues producing the contest to this day. His work in developing innovative and challenging problems astounds us. In addition to writing these tests, Sam has also contributed problems to the U.S.A. Mathematical Olympiad and created the Stanford Math Circle. Sam is a 1993 graduate of Swarthmore College and earned his Ph.D. in mathematics from the University of Chicago. He was a member of the 1989 U.S. International Mathematics Olympiad team, and was a grader for three years at the Math Olympiad Program, a seminar that determines and prepares that team. He is now the Head of School at Proof School (www .proofschool org). Many times when trying to find a proof for some theorem, we'd call on Sam and he’d give us three or four. We owe Sam many thanks for his contributions as a mathematician, our partner, and our friend. MATHCOUNTS is the premier extracurricular middle school mathematics program in the United States. MATHCOUNTS produces educational problem solving materials and conducts a nationwide contest consisting of school, chapter, state, and national levels. MATHCOUNTS was the starting point in mathematics for one of the authors, and is a great entry into mathematics for seventh and eighth graders. To Barbara Xhajanka we offer an extra thank you for her help. For more information, visit www.mathcounts org. The Mandelbrot Competition was started in 1990 by Sam Vandervelde and the authors. It is a four round high school competition designed to teach students not only the common stibjects of geometry and algebra, but also subjects that don’t appear in high school classes, like number theory and proof techniques. Each round of the Competition consists of an Individual Test and a Team Test. The Individual Test is a short answer test while the Team Test is a series of proofs designed to enhance participants’ knowledge of a particular subject area. There are two divisions of the competition, one for beginners and one for more advanced problem solvers. For more information regarding the Mandelbrot Competition, visit www.mandelbrot .org. Dr. George Berzsenyi. We could go on for pages about Dr. Berzsenyi’s many contributions to mathematics education through his involvement in competitions and summer programs. He hasbeen involved in writing the AHSME, AIME, and USAMOas well as other independent competitions. He also created the U.S.A. Mathematical Talent Search and its international counterpart; participating students in each round are given a month to prepare full solutions to five problems. These solutions are graded by mathematicians and comments on the papers are returned to the students. The USAMTS is an excellent way for students to learn how to write proofs. The USAMTS is now administered by the Art of Problem Solving Initiative (ww.artofproblemsolving.org), and is funded primarily by the National Security Agency. For more information on the USAMTS, visit www usants.org Dr. Berzsenyi was also an editor and contributor to the Mathematics and Informatics Quarterly (M&IQ). In addition to many practice problems, M&IQ contains articles written (in English) by people all over the world on various subjects of interest to the high school mathematician. While entirely within the reach of the average student, the articles are fascinating and have shown the authors many new approaches to various fields of mathematics. American Mathematics Competitions. The AMC produces the series of tests that determine the United States mathematics team. The tests are currently called the AMC 10, the AMC 12, the Ameri- can Invitational Mathematics Exam (AIME), and the U.S.A. Mathematical Olympiad (USAMO). The ‘AMC 12 used to be called the American High Schoo! Mathematics Exam (AHSME). Top performers in the contests are invited to the Math Olympiad Summer Program (MOP). For mote information on the contests and the MOP, visit amc.maa.org. There are a handful of problems in this book that appeared on tests at the MOP. These were kindly provided by Professor Cecil Rousseau, who instructed both of the authors of this text at the Math Olympiad Program in 1989. The American Regions Mathematics League (ARML) is an annual competition in which 15- member teams representing schools, cities, and states compete in short answer, proof, and relay contests. The contest is held concurrently at multiple sites. The authors of this text were teammates on the Alabama team at ARML in 1988 and 1989. We highly recommend this experience to students, as they will learn not only about mathematics but also about teamwork. ARML's primary question writers for the tests from which we have drawn are Gilbert Kessler and Lawrence Zimmerman. For more information on ARML, visit www. arml .com. David Patrick, Amanda Jones, and Naoki Sato. The original the ART of PROBLEM SOLVING texts were written in 1993 and 1994 on old Macintosh PCs that have less computing power than most watches now have. To produce the current edition, Amanda Jones recovered these ancient files from our old Macs. Unfortunately, recovering the files was not enough. David Patrick reformatted and edited the book, using his IATpX expertise to convert our decade-old code to modern ISIEX standards. Finally, nearly all of the diagrams of the book were re-created by Richard Rusczyk, Naoki Sato, and Amanda Jones. This text also contains questions from the Mu Alpha Theta (MA@) National Convention. Mu Alpha Theta is a national high school math honor society. For more information, visit their website at www mualphatheta.org. We gathered some problems from a few international sources in order to offer a wealth of challenging problems on some advanced topics. We collected problems from the national olympiads of Bulgaria (provided by Borislav Lazarov) and Canada (provided by Graham Wright). Both of these sources provide excellent practice for problem solvers. We also include problems that were either used in or proposed for the International Mathematical Olympiad (IMO). Each year many of the countries in the world send a six person team to the IMO to participate in the Olympiad. Theproblems in this text come from the 1989 Olympiad in Germany (provided by Paul Jainta), and the 1986 and 1985 Olympiads in Poland and Finland respectively (provided by Dr. George Berzsenyi). Key Curriculum Press produces The Geometer’s Sketchpad, which was used to generate most of the diagrams in the first edition of this text. The Sketchpad is an amazing program that forces students to learn geometry while producing fascinating visual output. The Sketchpad can be used to do everything from teaching simple geometric principles in an interactive way to generating complex fractals. For more information on the Geometer’s Sketchpad, visit www. keypress.com. Special thanks to Vanessa Rusczyk and Vladmir Vukicevic for their help in proofreading this book and to Kai Huang, Joon Pahk, Lauren Williams, and many members of the online Art of Problem Solving Community at wiw.artofproblensolving. com, and particularly Hussain Zahid Sheikh, Ravi and Meena Boppana, and Justin Venezuela, for their corrections for this seventh edition.To Students Unless you have been much more fortunate than we were, this book is unlike anything you have used before (except Volume 1!). The information in this book cannot be learned by osmosis. What the book teaches is not facts, but approaches. To learn from a section, you have to read, and comprehend, the text. You will not gain from just looking for the key formulas. Important ideas may be in seemingly out-of-the-way places, where someone skimming might miss them, since things are ordered by topic, not by importance. Don’t expect to find a uniform difficulty level. Read slowly, spending minutes on a single line or equation when you need to. Fly when you can. There will be times for both, so don’t get impatient. Some very important concepts are introduced only in examples and exercises. Even when they are simply meant to increase your comfort with the idea at hand, the examples and exercises are the key to understanding the material. Read the examples with even more attention than you pay to the rest of the text, and, no matter what kind of hurry you are in, take the time to do the exercises thoroughly. This book is about methods. If you find yourself memorizing formulas, you are missing the point. The formulas should become obvious to you as you read, without need of memorization. This is another function of the examples and exercises: to make the methods part of the way you think, not just some process you can remember. The subjects in this volume cover a much broader range of difficulty than those in Volume 1; therefore, you may wish to do a lot of skipping around. If you hit a subject you simply don’t understand, move on and return later when you've had more practice problem solving. Don’t give up; learning takes time. Most of all, this book is about problems. We have gone to great lengths to compile the end-of- chapter problems and other problems in the book. Do them, as many as you possibly can. Don’t overload on a single subject, though, or you'll forget everything in a week. Return to each subject every now and then, to keep your understanding current, and to see how much you've grown since you last thought about that subject. If you have trouble with the problems, don’t get neurotic, GET HELP! Consult other students, consult your teachers and, as a last resort, consult the Solution Manual. Don’t give up too quicklyand begin using the Solution Manual like a text. It should be referred to only after you've made a serious effort on your own. Don’t get discouraged. Just as importantly, if these last sentences don’t apply to you, you should be the one other students can come to for help. The book thus comes with one warning: you will not learn if you don’tdo the problems. Cultivate a creative understanding of the thought processes which go into solving the problems, and before too long you will find you can do them. At that same instant you'll discover that you enjoy them!To Teachers the ART of PROBLEM SOLVING is our conception of what a motivated student's instruction in high school non-calculus mathematics should be. We strongly feel that a student should learn all subjects simultaneously. There are two reasons for this. First, it is better to convey the interconnectedness of it all; how geometry naturally leads to coordinates and how those coordinates make it easy to define conic sections and the complex plane; how counting leads to probability, the Binomial Theorem, and number theoretical ideas. Second, it all sinks in better. Overloading on a single subject can cause students to acquire a surface understanding which doesn’t connect to any deeper comprehension, and is thus rapidly lost. There are many subjects in this text which your students have likely not seen before. We feel it is very unfortunate that students aren’t introduced to such subjects as collinearity, inequalities, and number theory. Again in this volume we put an emphasis on geometry, which we feel is the most neglected subject in many curricula: students take a year of geometry, then don’t ever see it again. We also warn the teacher that the difficulty level of the subjects in this book vary much more greatly than in the first volume. Some of the text may be too advanced for your beginners, while other portions are likely too elementary for your advanced students. Thus, take care in the chapters or sections you assign your students. Our notation sometimes diverges from the accepted notation. In these cases, however, our decisions have been made with full deliberation. We strive to use symbols which evoke their meanings, as in the use of the less-popular | | to denote the greatest integer function instead of the usual { J Each chapter of the text is meant to feel like the discussion of a subject with a friend. In one aspect of such a discussion, the text must fail: the answering of questions. This weakness must be repaired by teachers or strong students who are able to assume a leadership role. Teachers are crucial to the process of the book, whether teaching the material directly or simply being available for explanation. We urge teachers using the ART of PROBLEM SOLVING in a classroom or club setting to encourage students who understand certain areas to explain the subjects to the rest of the class, or perhaps rotate such responsibility among a large group of willing students. This will not only give the other students a different view, perhaps closer to their own thought process, but it also greatlyenhances the teaching student's understanding of the subject. Furthermore, the instructing student will have a chance to see the rewards that come from teaching. We also suggest that after covering each subject, students attempt to write problems using the principles they have learned. In writing a problem, one does much more math than in solving one. This further inspires the creative drive which is so essential to problem solving in math and beyond, and if students are asked to take a crack at each others’ creations, the competitive urge will also be tickled. In closing, this book is about methods, not memory. The formulas we prove are important ones, but we intend for our explanations to be such that memorization is not necessary. If a student truly understands why a formula is true, then the formula can be internalized without memorization However you choose to use this book, we hope that the focus remains that students understand why formulas work. Only in this way can they understand the full range of the formulas’ applications and the full beauty of the mathematics they are learning.Justify Your Love Throughout high school and even middle school, the authors of this text participated in a lot of math contests. After high school we then produced our own contest, The Mandelbrot Competition, along with Sam Vandervelde. One question has persisted from the skeptics: why bother? They argue that the math involved in competitions is largely useless for the rest of participants’ lives. While correct (It won't be often that your boss says, “Tell me $(45) or you're fired!”), this argument is misguided, because math is by far not the most important aspect of the contests. Through math competitions and projects, students learn how to attack problems. Unlike specific techniques, this skill is crucial to virtually any area of life. Successful problemists go on to be successful not only in mathematics, but also in every other field (not just technical ones!) that you can think of. The authors’ math training didn’t just make us able to write this text, but it taught us the rewards of hard work, gave us confidence, and—most importantly—developed our ability to solve problems. Good problemists are very creative people. Knowing alll the tools at your disposal will not always guarantee finding a solution; the key to solving problems is cleverly choosing the right method of attack. A great way to ‘train’ for problem solving is to do various brainstorming and other creative ideas. Not only will these help you open your eyes to new ideas, but they can often be a lot of fun. This is not to say that the mathematics itself is useless. Hopefully through this text and other work, you'll develop the same interest in mathematics we have. While some people might think we're nuts, we view an elegant mathematical concept or a neat proof with the same admiration as others view a Rembrandt painting or a Beethoven symphony. This is the reason for our choices of the covers of our texts. The beauty of nature is dictated by a mathematics which we humans are still struggling to understand. The last and, for many, most important aspect of math contests is the people. The authors of this text met each other and Sam, as well as many other friends, through math. When your days in contests are over, you'll cherish the memories far more than you will the contests themselves.The eye will be found looking at especially important areas of the text. When you see it, pay extra attention ‘The threaded needle indicates particularly difficult problems or concepts. If your hands are too shaky, you may need help from someone else. The bomb signals a warning. If you see it, tread lightly through the material it marks, making sure you won't make the mistakes we warn against.Contents 0 Prove It! 1 Logarithms 2 Not Just For Right Triangles 21 Trigonometric Functions : 22 Graphing Trigonometric Functions . 2.3 Going Backwards 24 Tying It All Together 2.5. Solving Problems Using Trigonometric Identities 3. More Triangles! 3.1 Triangle Laws 3.2 Areas, Areas, Areas . 3.3 More Important Lines 4 Cyclic Quadrilaterals 4.1 Properties of Cyclic Quadrilaterals . . 4.2 Finding Cyclic Quadrilaterals 43 Ptolemy's Theorem 5 Conics and Polar Coordinates 5.1 Parabolas 5.2 Ellipses .. . 53 Hyperbolas . 54 Polar Coordinates Revisited . 5.5 That Pesky xy Term . 6 Polynomials 6.1 Whatisa Polynomial? . 62 Multiplying and Dividing Polynomials 63 Finding Roots of Polynomials . . . . 64 Coefficients and Roots . 10 10 12 14 15 20 24 24 7 294 CONTENTS 65 Transforming Polynomials 62 66 Newton’s Sums 65 7 Functions 69 71 Thelnverse ofa Function ©... 6.0 ee ee eee o) 7.2 Functional Identities. 70 7.3. Solving Functional Identities 7 731 Isolation .........-+ oes 71 7.3.2 Substituting in Values . . at ewmer sire 2 73.3 Using Cyclic Functions 2 7.34 Arbitrary Functions 2 8 Taking it to the Limit 76 8.1 Whatisa Limit? 76 8.2 Tricky : 7 83 Working with Limits . 7 84 Continuity 80 85 Asymptotes . 81 86 Trig Limits 83, 87 ¢ 84 9 Complex Numbers 88 9.1 Drawing the Complex Numbers... 2.0605 88 9.2. The Complex Absolute Value : 89 9.3 Complex Multiplication and Coordinates a1 9.4 Complex Powers and Geometry... .. +. 91 9.5 DeMoivre's Theorem . 92 9.6 Exponential Form 94 9.7 Twofor One . ot 96 9.8 The Roots of Unity ........- 96 10 Vectors and Matrices 100 10.1. What is a Vector? - 100 10.2 The Dot Product . . .. 101 10.3 Coordinate Representation of Vectors - 102 10.4 What isa Matrix? . 103 10.5 Matrix Multiplication ©... eee 104 10.6 Matrices in Higher Dimensions . 107 107 Better Matrix Notation ........ 108 11 Cross Products and Determinants m2the ART of PROBLEM SOLVING: Volume 2 23 111 The Cross Product... . 0... 12 11.2 The Cross Product in Coordinates - 113 113 The Determinant : 113 11.4 Determinants in Higher Dimensions 115 115 Minors 7 +. 16 11.6 Row and Columa Operations eee ee 118 AL? The Inverse of Matrix... eee eee eee - 120 12 Analytic Geometry 14 12.1 Lines, Angles, and Distances... 2.0.0.2... fee tee eee ee 124 12.2 Parameters - 125 123 Vectors ...... . 17 124 Points, Lines, and Planes +e rome gs 2180 125 Curved Surfaces... eee vee e eens .. 134 12.6 Using Analytic Geometry . . «+ 135 12,7 Vectors and Geometry Problems . 136 13 Equations and Expressions 143 13.1 Linear Equations 43 13.2 Convenient Systems sees . 148 133 Symmetric Expressions and Advanced Factorizations - «150 13.4 More Polynomials. . . . .. 152 13.5 Squares and Cubes .. 154 13.6 Using Graphing . . 155 14 Inequalities 159 14.1 Trivial Inequality Revisited 1 $ HOSHE TEE Sore se emcee + me we om 159 142 Arithmetic Mean-Geometric Mean Inequality . +. 160 143 Cauchy's Inequality . . +. 162 144 Maximization and Minimization » 164 165 145 Geometry and Inequalities 14.6 Wrap-Up and Parting Hints 166 15 Combinatorics 170 15.1 Identities 170 15.2 Pascal's Identity 170 15.3 More Identities +. 172 15.4 Block Walking . . . - 173 15.5 The Binomial Theorem . . . - 175 16 Sequences and Series 180Ww 18 19 20 2 2 16.1 Fractions in Other Bases . 16.2 Some Fpecial Deries . 16.3 The Fibonacci Numbers 16.4 Dealing with Recurrences 165 Dealing with Sums 166 ‘The Binomial Theorem Revisited 16.7 Harmonic Sequences Counting in the Twilight Zone 17.1 One to One . 17.2 Clever Correspondences 173 Easy as. ‘ 174 Generating Functions 175 Partitions . . 17.6 Counting on Graphs 17.7 Counting Infinite Sets . . ‘Again and Again 18.1 Repeats... . 18.2 Off to Infinity 18.3 Rational Continued Fractions . 18.4 Real Continued Fractions Probability 19.1 Review, Definitions, and Notation 19.2 Going a Step Further... .. . 19.3 Geometry and Probability . . 19.4 Conditional Probability Find It and Make It 20.1 Locus : 20.2 Construction . . Collinearity and Concurrency 21.1 Three Points and a Line 21.2 Three Lines and a Point Geometry Tidbits 22.1 Projections 22.2 Inversion 22.3 Homothecy... . 22.4 Geometric Continuity . CONTENTS - 181 180 182 .. 184 . 186 188 - 191 196 196 . 196 .. 198 - 200 . 202 204 - 205 211 -. 211 21 212 213 216 216 217 - 217 219 224 - 224 207 233 233, . 236 241 . 241 243, 245 . 246the ART of PROBLEM SOLVING: Volume 2 22.5 Given a Finite Number of. 23 Number Theory 23.1 Divisibility . . 23.2 Division in Congruences . 23.3 Solving Linear Congruences . . 23.4 Solving Quadratic Congruences . . 23.5 The Sum of the Divisors 23.6 Fermat's Theorem . . . . 23.7 The $ Function iwi 23.8 Wilson's Theorem... ... . 24 Diophantine Equations 241 axtby=c .. 24.1.1 c=0 24.1.2 c#0. 24.2 P+ y= 2 243 eye... 24.4 The Pell Equation 24.5 General Methods .. . . 25 Graph Theory 25.1 PointsandLines....... 25.2 PlanarGraphs ..... . : 25.3 Example: The Platonic Solid: 25.4 Walking Around on Graphs . . 255 Buler Trails : 25.6 Colorings 26 Parting Shots - 267 247 252 252 253 - 254 = + 256 . 257 258 = + 260 = 262 266 266 266 269 270 2.271 272 275 - 275 276 277 - 278 279 . 280 284Chapter 0 Prove It! Unfortunately, proofs in the standard school curriculum are either overlooked or confined to geom- etry classes. Proofs are absolutely essential to mathematical understanding, because if you don’t know why a tool works, you can’t use it to its full capacity. Don’tignore the proofs in this text! While they don’t occur in most classes of competitions, they do occur in the most challenging contests. In Volume 1 we dedicated an entire chapter to proof techniques and the language of proofs. ‘We'll review certain methods here because they are the most common and have the widest variety of uses, The many other less common techniques are scattered throughout the text, included among the subjects where they are most commonly used. * Contradiction. Suppose we wish to prove some statement A. We can use contradiction by showing that if A were false, then some impossible statement would have to be true. * Mathematical Induction. Induction is generally used to prove statements which are true for all positive or all nonnegative integers. Suppose we wish to show that some statement B is true for all integers n. We show that B is true for n = 1, then we show that if B is true for n = k (commonly called the inductive hypothesis) then it is also true for n = k +1. Hence, since B is true for n= 1, it is true form = 141 = 2. Therefore it is true for 2+ 1 = 3, then 4, and so on. * Pigeonhole Principle. The Pigeonhole Principle states that given kn +1 objects which are placed in 1 boxes, there must be some box with at least k + 1 objects. The principle is clearly most useful in problems where there is something that can be divided into categories. EXAMPLE 0-1 Prove that there are infinitely many prime numbers. Proof: Suppose there aren't infinitely many primes and look for a contradiction. Let there be k prime numbers, namely p1, p2,...., pk. Consider the number Z = pipops:**PkaPe + 1. Clearly Z isn’t divisible by any of our k prime numbers, so it must divisible by some other prime. This contradicts our assumption that there are only k prime numbers, so there cannot be a finite number of primes. EXAMPLE 0-2 Let Fy = Fy-1 + Fr-2, Fo = 0 and F; = 1, Prove that for all n, Fot Fy +Fate+-+Fy = Fy. <12 & CHAPTER 0. PROVE IT! Proof: The recursion in the problem generates the Fibonacci numbers. There are many identities like the one in this problem which are true for Fibonacci numbers. Most of these can be proven with mathematical induction. For this one, we prove the statement first for n = 0, for which we have Fy = Fx—1, whichis obviously true, since Fy = 0 and Fz = F +Fo = 1. Now we assume the statement is true for n =k: Fot Fy + Fg tee + Fy = Fig 1 This is our inductive hypothesis. We then wish to show the identity holds for n = k +1, or Fo+ Fy +Fote-+ Fat Fest = Fea From our inductive hypothesis, the sum Fy + Fi + +++ + Fy equals F,,2 ~ 1. Hence we have FotFitiyte+ Ft Fea = Foo-1+Fea = Fea-1 as desired. EXAMPLE 0-3. Prove that if we select 5 points within the boundaries of a unit square, then some pair of them are no more than V2/2 apart. Proof: We can apply the Pigeonhole Principle by dividing the square into four squares with side length 1/2. By the Pigeonhole Principle at least two of the points must fall in the same square. The farthest apart two points can be ina 1/2 inch square is ¥2/2, where the points are on opposite ends of a diagonal. Since there must be two points in one of the little squares, there must somewhere be a pair of the five points at most V2/2 apart.Chapter 1 Logarithms As we mentioned in the BIG PICTURE in Volume 1, logarithms were originally devised to turn multiplication and division problems into addition and subtraction ones. Let's take a closer look at how this works. Suppose we are asked to find (1234)(5678). Normal multiplication would be quite tedious. Instead, we note that for some x and y, we can write 10* = 1234 and 10” = 5678, so that log 1234 =x and log 5678 = y. Hence, (1234)(5678) = 10°10Y = 10+. Taking logarithms of this last relation (remember that a logarithm with no base indicated is assumed to be base 10), we have 1og(1010") = log(10"*¥) = x + y = log 10* + log 10%. In other words, log(1234)(5678) = log 1234 + log 5678. Neat! The logarithm of a product of two numbers is the sum of the logarithms of the two numbers. Think about why this must be so. Recall that the value of a logarithm is an exponent. We add exponents when we multiply two numbers with the same base. As logarithms are these exponents (xand y above), their sum must be the exponent of the product (log(1234)(5678) = x + y above). Now to find the product, we merely look up log 1234 and log 5678 in logarithm tables, find the sum of the two values, then find the number z from the tables such that log z = log 1234 + log 5678. If you try this, you may find that your logarithm table only goes from 1 to 10. How can you find log 1234? Use scientific notation, so that log 1234 = log(1.234)(10°) = log(1.234) + log(10%) = 3 + log 1.234. This relationship between multiplication and addition is not the only useful property of loga- rithms. Using the same logic as above, division becomes subtraction: 1234 log sa7g = 108 1234 - log 5678, and exponentiation becomes multiplication: log 1234°° = 5678 log 1234. <34 CHAPTER 1. LOGARITHMS ‘These are by no means proofs, nor are these manipulations confined to base 10 logarithms. In the following pages, we'll formalize these rules and introduce a few more, as well as show you how to prove them. Properties of Logarithms 1. log, b" = nlog,b 2, log, b + log, c = log, be 3. log, b— log, c = log, b/c 4, (log, 6) (log, 4) = (log, d) (log, ®) log.b _ 5. gg =log.b 6. logy. b" = log, b WARNING: Note that in Properties 2, 3, and 5, the bases of the logarithms added, subtracted, or divided are the same. This is very important to understand; we can’t simplify logy x + log, y° with, Property 2 for the same reason we can’t add exponents to evaluate the product 2 3°, as we would for 22°. You should try to prove these properties on your own, as the proofs are fairly simple. Some are proven on page 5, and the proofs of the others are left as exercises. EXAMPLE 1-1 Evaluate each of the following in terms of x and y given x = log, 3 and y = log, 5. i. log, 15 Solution: Since 15 = 3(6) we think of Property 2: og, 15 = log, 3(6) = log, 3 + log, 5 = x + y. ii, log,(7.5) Solution: Since we already know log, 15, we note that 7.5 = 15/2 and think of Property 3. This is abit tricky, but remember that in addition to log, 3 and log, 5, we also know log, 2 = 1: 10g9(7-5) = log,(15/2) = log, 15 - log, 2 =x +y-1. ii, logs 2 Solution: Since we have a different base in this than in the given quantities x and y, we look for a property which allows us to change the base. Thus, we use Property 5: In general, it is always true that log,,z = 1/log, w. (Can you prove it?) Remember this; you'll probably see it again.the ART of PROBLEM SOLVING: Volume 2 <5 iv. log, 15 Solution: First, 15 = 3(5), and we know log, 3, s0 we use Property 2 to get log, 15 = log, 3+log,5 = 1+ log, 5. Now we must find a logarithm with base 3, but we only know base 2 logarithms. This leads us to Property 5: Jog, log,3 1 * x" log, 15 =1+log,5=1+ ¥. log,9 Solution: Ourbase is the square ofthe base we are given in our information, so we look to Property 6. When working problems, always try to manipulate the bases so they are the same, or as close as. possible, throughout the problem. When working with various powers of the same number, like 2. and 4, use Property 6 like this: log, 9 = logy: 3* = log, 3 = x. vi. log, 6 Solution: Seeing a different base that is not a power of , we look to Property 5. Noting that 6=2(3), we also apply Property 2: _ log,6 _ log,3+log,2 10856 = og, y xt > ‘We'll now prove three of the six properties; the proofs of the other three are left as exercises. The first step for the proofs, since we can’t do anything with the expressions as they are written, is to write the logarithms in exponential notation. Thus, we let x=log,b, y=log,c, and z= log, c, from which we have a@=b, a =c, and # =c, ‘These relationships will be used in the first two proofs below. EXAMPLE 1-2. Prove Properties 1, 2, and 4. i, Property 1: log, b" = nlog, b Proof: Let w = log, b”. We want to show that w = log, b = nx. Make sure you understand why this will complete the proof. Putting our expression for w in exponential notation, we have a” = b’. Since a* = b, we find a” = b" = (@*)" = a, so xn = w. Thus, nlog, b = log, bt ii, Property 2: log, b + log, c = log, be. Proof: We wish to show that log, be = x + y. Since a = b and a¥ = c, we can get the quantity x+y by multiplying a* and a: a*aY = a'*¥ = be. Putting this last equality in logarithmic notation gives us log, be = x + y = log, b + log, . (Notice how this proof is similar to our discussion of evaluating log(1234)(5678).) iii, Property 4: (log, b) (log, d) = (log, d) (log, b). Proof: We let x=log,b, y=log.d, w=log,d, and z=log.b.6 > CHAPTER 1. LOGARITHMS We wish to show that xy = wz. As before, we write the above logarithmic equations exponentially. We find bat=ac d= aac g= cw), oy cols) = la), Thus we have from which we have the desired xy = wz. Using this relation we can show that (log, ) (log,¢) = log, ¢, a frequently occurring identity sometimes called the chain rule for logarithms. Itis important that you realize that these proofs are not just pulled out of thin air. They involve methods that you should learn, namely, the practice of changing logarithmic notation to exponential notation and manipulating the exponential expressions. Make sure you understand this method before proceeding to the exercises. After writing logarithmic expressions in exponential notation, ask yourself what you wish to prove in terms of the exponents (x, y, etc. above). Then, manipulate the exponential equations to complete the proof. EXERCISE 1-1 Prove Properties 3, 5, and 6 without using Properties 1, 2, and 4. EXERCISE 1-2 Prove the chain rule for logarithms using Property 4. OS WARNING: Don’t overlook the fact that the base and the argument of all logarithms must be positive, for sometimes devious, or careless, test writers will create problems in which some seemingly correct solutions violate one of these rules. EXAMPLE 1-3. Find all x such that log, (x + 2) + log, (x + 3) = 1. Solution: Seeing the sum of two logarithms with the same base, we think of Property 2, which yields loge (x + 2) + log (x + 3) = log, (x? + 5x +6) = 1. Putting this equation in exponential notation gives x° + 5x +6 = 6, or 2? + 5x = 0, so our solutions are x = -5 and x = 0. You may be tempted to stop here and claim that these are both valid solutions, ut your last step in all problems involving logarithms must be checking that each solution makes the argument and the base of all logarithms positive. In the given problem the arguments of the initial logarithms are negative when x = —5, so this is not a valid solution. The only valid solution isx=0. EXAMPLE 1-4 Find the sum 99 1 2,, 3 log 5 + log 5 + 10g 5 +--+ 108 355the ART of PROBLEM SOLVING: Volume 2 <7 Solution: Seeing the sum of logarithms we think of log x +log y = log xy. Calling our given sum S, this identity gives 1 S=log G 3 Notice that in the product every number from 2 to 99 appears once in the numerator and once in the denominator, so they all cancel. EXERCISE 1-3 Find log, 10 and log, 1.2 in terms of x = logy 4 and y = log, 3. EXERCISE 1-4 Iwant to use my calculator to evaluate log, 3, but my calculator only does logarithms inbase 10, Should I go find a better calculator, or should Ibe able to find a way to make my calculator tell me log, 3? EXERCISE 1-5 Show that x!&¥ = y, 7, 28 By at a 98°99 100) ~°°8 705 Problems to Solve for Chapter 1 1. Evaluate the product (log, 3)(log, 4)(log, 5)(logs 6)(log, 7)(log, 8). 2. If log 36 = a and log 125 = b, express log(1/12) in terms of a and b. (MAO 1992) 3. In how many points do the graphs of y = 2log x and y = log 2x intersect? (AHSME 1961) 4, Find all the solutions of 3 x log = 2 7 100° (AHSME 1962) logy (X 5. Ifa>1,b>1,andp= ate, then find a” in simplest form. (AHSME 1962) 5b 6. If one uses only the information 10° = 1000, 10* = 10000, 2" = 1024, 241 = 2048, 22 = 4096, 2° = 8192, what are the largest a and smallest b such that one can prove a < log, 2 < b? (AHSME 1967) 7. For all positive numbers x # 1, simplify 1 1 1 + +, Togs * log,x * Tog,x (AHSME 1978) 8. Given that log, 2 ~ 0.3010, how many digits are in 5“? (MAe 1991) 9. If logs 3 = P and log, 5 = Q, express log,, 5 in terms of P and Q. (MAG 1990) 10. Suppose that p and q are positive numbers for which 1085p = loging = logis(p + 4). What is the value of q/p? (AHSME 1988)8 > CHAPTER 1. LOGARITHMS S—_11. Given that log, 40 V3 = log, 45, find 1°, (MAO 1991) $~__ 12. Suppose @ and b are positive numbers for which Logg = logs b = logys(a + 2b). What is the value of b/a? (MA® 1992) S~_ 13. If 60" = 3 and 60# = 5, then find 12{0--H/2-291, (auSME 1983)the ART of PROBLEM SOLVING: Volume 2 <9 >the BIG PICTURE One area in which logarithms play a surprisingly large role is music. Musical sound is created by something vibrating—a string on a violin, a column of air ina flute. The rate of vibration translates to a pitch; the faster the vibration, the higher the pitch. For instance, top Cona flute is 2048 Hz (Hz, or Hertz, means “cycles per second,” so this is 2048 vibrations per second), a violin’s low G is 192 Hz, and bottom A on a piano is 27.5 Hz. Notes played together either “sound good’ or they don’t. This sounding good corresponds to the frequency of one tone being a nice multiple of another. Two tones an octave apart have frequencies differing by a factor of two, like middle C (256 Hz) and the next C up (512 Hz); two tones a major fifth apart have frequencies in the ratio 3/2, as C (256) and the next G up (384). On the other hand, tones with nasty frequency ratios (say 31/17) sound displeasing, in part because the ear hears not only the two frequencies, but an artificial beat frequency resulting from the times when the two vibrations are in sync. Scales were originally formed on the basis of frequency ratios described, but such scales were found to be lacking. A scale in which every note was the right frequency multiple of C would no longer work when Af was the central note. The resolution of this problem came with the discovery of even tempering in the early 1700's, in which the octave was divided up into 12 pieces such that each frequency was the right multiple of the last. To see how this works, let the octave go from frequency F to 2F. For some multiplier m, the scale would be F, Fm, Fm2, Fin’, ---, Fm? = 2P. Solving this last equation, we find that the multiplier is 2", Why is this scale special? Suppose we wanted to start four notes up, at Fm; the scale would then be Fm3, Fm, Fm, ..., all notes in the original scale. The scale works in any key. We can find out about what note a tone at 1.5F is by solving 1.5F = 24/2P for k as k = 12log, 1.5 = 7. Thus our note is seven notes up, so it’s a G. Even with an even-tempered scale, we'd still like to get, as closely as possible, nice frequency ratios; otherwise our mathematically perfect scale will contain no worthwhile harmonies. But it does. For example, we found above that the tone 1.5F = 3F/2 is almost exactly seven notes up the scale. Check for yourself where other notes which harmonize well with F, like 4F/3 or 5F/4, end up in the new scale; it turns out the new scale does very well musically as well as mathematically. J.S. Bach proved this explicitly in his Well-Tempered Clavier, which contained pieces in every major and minor key.Chapter 2 Not Just For Right Triangles 2.1 Trigonometric Functions In Volume 1 we introduced sine, cosine, and other trigonometric quantities as ratios of sides of a right triangle. In fact, these quantities are used for much more than just right triangle geometry. ‘To be able to use them as widely as possible we must first understand what the values of sine and cosine are for non-acute angles. Consider the unit circle; that is, the circle with radius 1 centered at the origin. Any point on the circle can be described by the polar coordinates (r, 6) = (1,0). For example, the point (0, 1) in rectangular coordinates can be described by (1,90°) in polar coordinates. It could also, however, be described by (1,450°), since we could go around the circle once before continuing 90° more to our final point, for a total of 360° + 90° = 450°. Conversely, given the coordinates (1,450°), we could convert to the more manageable (1,90°). By adding multiples of 360° (or 27), we can find an equivalent angle between 0° and 360° for any angle, even negative angles (which mean going around the circle clockwise rather than counterclockwise). EXAMPLE 2-1 Find angles between 0° and 360° which are the same as ~7/2, 1180°, and 91/4. Solution: The first is negative, so we add 2n: —n/2+2n = 3n/2. The last are both over 360°, so we subtract, sometimes repeatedly: 1180° = 1180° - 360° = 820° = 460° = 100° 9n/4 = 9n/4—-2n = 1/4. Why are we fussing with the unit circle when we are supposed to be dis- y cussing sines and cosines? First consider an acute angle as shown in the diagram. The point described by (1, @) is in the first quadrant. If we draw an altitude from A to B on the x axis and see that ZAOB = @, we can find the oa (x,y) coordinates of point A from trigonometric relations applied to AAOB. For example, since cos @ = OB/AO = OB (since OA = 1), we have x = OB = cos 0. 7 Similarly, y = AB = sin. Thus the rectangular coordinates of the polar point A= (1,8) are (cos 6,sin 8). < 10>the ART of PROBLEM SOLVING: Volume 2 s 11 Impressed by this success, we could nafvely try to apply it to (1, 8) for an obtuse angle @. But we don’t know what cos @ and sin @ are for obtuse 6! To solve this problem, we simply define cos 6 and sin to be the Cartesian coordinates of the polar point (1, 8). ¥ Now matters are a little trickier. We'll consider @ = 150°. Once again, we draw an altitude from A to point B on the x axis. Since ZAOC = 150°, we have ZAOB = 30°. Thus we find x = -OB = — cos 30°, where we have a negative sign since B is clearly on the negative x axis. Hence, cos 150° = ~ cos 30° = ~ 3/2. Similarly, we see that sin 150° = AB = sin 30° = 1/2, where we have a positive result since A is above the x axis. This gives a general method to find the sine and cosine of any angle, whether acute, obtuse, or worse. First we determine which quadrant the angle is in, then we draw the picture and make a right triangle by drawing an altitude to the x axis. We use basic trigonometry to get the rectangular coordinates (2, y) of the point (1, 6), remembering that points to the left of the y axis have a negative x and those below the x axis have a negative y. We then set cos @ = x and sin @ = y. EXAMPLE 2-2 Find tan77/4. Solution: First we determine the: ‘quadrant of 77/4. Since 7/4 is greater than 37/2 and less than 2m, the angle is in the fourth quadrant, as shown. We draw (| y our altitude and find ZBOA. Let C be the point (1,0). Since the are from C to A counterclockwise has measure 77/4 radians and an entire circle has 27 radians, Th the remaining arc AC has 2n ~ 71/4 = 7/4 radians. Thus OB = cosn/4= 2/2 AY and AB = sin /4 = 2/2. Since A is in the fourth quadrant, x is positive and “A y is negative, so cos 7n/4 = OB = 2/2 and sin7n/4 = ~AB = — y2/2. Hence we have sin?n/4 cos 7/4 ~ (Notice that we found the sine and cosine of the angle and then used these to determine the tangent; this is how we will almost always determine trigonometric functions.) Do not be too intimidated by this example; finding sines and cosines is very easy once you've had practice. Eventually you'll be able to do all this reasoning in your head quite quickly. EXERCISE 2-1. In what quadrants are 30°, 700°, 57/3, and ~370/5? EXERCISE 2-2 How would we find sin @ and cos @ if the point (1, 6) is on the x axis? the y axis? For what angles @ does this occur? EXERCISE 2-3. Evaluate sin 300°, cos 225°, csc 150°, cot 57/3, tan x, and sec 57/6 EXERCISE 2-4 Evaluate more trig (shorthand for trigonometric or trigonometry) functions using angles which are multiples of 30° or 45°. Then use a calculator to check your work. tan7n/4 C EXERCISE 2-5 Always make sure your sign (positive or negative) is correct when evaluating trig (> functions. How can we tell what the signs of cos 0 and sin @ are given an angle 0? Its awfully easy to mix up the cosine and sine values for multiples of 30°. If you're ever confused, draw the angle on the unit circle; using the resulting geometry you should be able to reason which is which.> CHAPTER 2. NOT JUST FOR RIGHT TRIANGLES ‘EXAMPLE 23 Use our geometric approach to show that for any obtuse angle 6, sin @ = sin(n ~ 8). Solution: In our figure we have AB = sin ZAOB from right triangle AOB (with AO = 1). Since A is in the second quadrant, we have AB = sin LAOC = sin 8. Equating these expressions for AB, we find sin 6 = sin LAOB = sin(n - 6) as desired. EXERCISE 2-6 Use the same reasoning as above to show that cos @ = — cos( ~ 8). 2.2 Graphing Trigonometric Functions Knowing what a function looks like is often quite useful in understanding problems involving the function. In this section we'll discuss the graphs of the basic trigonometric functions sin x, cos x, and tanx. For cosine, we know thatcos0 = 1, cos 7/2 = cos 37/2 = ww 0, and cost = ~1. We also know that outside the range 0 < x < 2n, the values of cosine repeat those inside the range (since the angles outside the range (0,272) are equiv- alent to those inside). Thus, we sketch cos x for the range 0 < x < 2n and continue this pattern indefinitely in both directions as above. This graphical form is called a sinusoid. Using the same analysis as for cosine we can generate the graph of sinx, shown at left. Make sure you see how this graph describes sine. Look for particular points you know, like sin 7/2 = 1, and look for where you know sine is positive. Graphing tanx is a little trickier, and we do so by noting that tanx = sinx/cosx. Since sin0 = 0 and cos0 = 1, the graph of tanx must pass through the origin. Since sine and cosine are both positive in the first quadrant, tan. is positive for 0 < x < n/2. Finally, since cos x gets closer and closer to zero as we approach x = 7/2, tanx gets larger and larger. Try dividing 1 by 0.1, 0.01, and 0.001 to see what happens to tan.x as we approach x = 7/2. On the other side of the origin, where the angle is in quadrant IV for —r/2 < x < 0 (make sure you see why), sine is negative and cosine is positive, so tan. is negative. ‘As before, as we approach x = ~7/2, cosine approaches 0, so the magnitude of tanx grows and grows, producing the graph on the right. (It continues past y = 4 upward and y = -4 downward.) Unlike the graphs of sinx and cos x, the graph of tan x repeats after every interval of length 7 rather than 2re. (Why?) Use your knowledge of secx, csc, and cotx to graph these functions.the ART of PROBLEM SOLVING: Volume 2 « 13 EXERCISE 2-7 Before proceeding, make sure you are satisfied that the above graphs do indeed represent sine, cosine and tangent. Now that we understand the trigonometric functions, let’s try applying the functional transfor- mations we discussed in Volume 1 to trig functions. We'll start by letting f(x) = sin x. The values of this function range from —1 to 1. It follows that the function 2f(2) (a vertical stretch of f(x)) varies from ~2 to 2. This length of this range can be used to help describe trigonometric functions. The amplitude of a graph is half the difference between its largest and smallest values. Thus, for f(x) = sin x, the amplitude is 1 and that of 2f(x) = 2sin x is 2. A simple extension of this reveals that for any trigonometric function, the amplitude is dictated by the coefficient of the function. (Remember, amplitude measures a distance so it is never negative.) Let’s look at f(2x) = sin2x, At right are the graphs of f(z) (solid line) and f(2x) (dashed line). The transformation (2x) is a horizontal shrink by a factor of 2. Thus, while f(z) goes from 0 to 1 as x goes from 0 to 7/2, f(2x) goes from 0 to 1 in half this interval. ‘We now are ready to define the period as the amount of the graph (in terms of x) we can draw before we must start repeating. For example, to graph J (2x) we draw the graph from x = 0 tox = 7 then repeat this range indefinitely. Thus, the period of f(x) = sin2xis 7. Similarly, the period of sin x is 2r. Clearly, the coefficient of x in our trig functions determines the period, since this coefficient is responsible for the horizontal shrinking or stretching of the graph. From this analysis of period, we see that the period of f(kx) = sin kx is 2r/k for all k since f(x) is shrunk by a factor of k by the transformation f(kz). Related to the period of a function is the frequency, or how often the graph of the function repeats. For example, the graph of sin x repeats every 2r, so the frequency is one per 2n, ot 1/(2r). Since a graph of a function repeats every period of the graph, the frequency is always 1/(period). y Having looked at a horizontal shrink, we move on to the horizontal slide f(x—n/2) = sin(x—n/2). The graphs of f(x) (solid line) and f(—7/2) (dashed line) are shown at left. As discussed in Volume 1, the transformation f(x 7/2) results ina 7/2 slide to the right of f(x). We define the phase shift ofa function f(x—K) as the amount the graphis shifted from the ‘parent’ function f(x). If a direction is not given for the phase shift, a positive phase shift is to the right and a negative to the left. WARNING: What about the phase shift of sin(2x ~ 7/2)? Referring to our above discussion of phase shift, the ‘parent’ function is g(x) = sin 2x, since sin(2x ~ 7/2) is a shift of sin 2x, not a shift of sinx. (Graph them and see!) The desired function sin(2x — 1/2) is g(x - 77/4) (make sure you see this), so the phase shift is 7/4 to the right, not 7/2. yh EXAMPLE 2-4 Determine the period, amplitude, phase shift, and frequency of f(x) = 3sin(4x+7) +7. Solution: Since the period of sinx is 2n, the period of f(x) is 2n/4 = m/2. Since 3sin(4x + 7) ranges from 3 to ~3, f(x) ranges from 3 +7 = 10 to -3 +7 = 4, so the amplitude is (10 - 4)/2 = 3. Notice that the amplitude is still the coefficient of the trigonometric function despite the +7"So @ m 14> CHAPTER 2. NOT JUST FOR RIGHT TRIANGLES For the phase shift, we see that the parent function is g(x) = sin4x +7 and f(x) = g(x + 7/4). Hence the phase shift is x = —1/4 (or 71/4 to the left). Finally, the frequency is just the reciprocal of the period, or 2/r. EXERCISE 2-8 Prove that the phase shift of f(x) = sin(ax + 6) is ~b/a. EXAMPLE 2-5 Find the frequency, period, and amplitude of the function at right. Solution: The important thing here is that trigonometric functions are not the only periodic functions; there are many, many functions which repeat over and over. Draw some yourself! Since the given gtaph repeats every 3 units, its period is 3 and its frequency is 1/3. Since it varies from ~2 to 5, its amplitude is [5 - (-2)]/2 = 3.5. EXERCISE 2-9 Whats the period of tan 2x? (Not mt!) 2.3 Going Backwards Suppose we are 1000 feet away from a 500 foot tall tower and we wish to aim a laser at a mirror mounted on top of the tower. At what angle 6 should we aim the laser? Although we don’t have the angle immediately, we do have enough information to figure it out. The tangent of the desired angle is 500/1000 = 1/2. Looking back at our graph of tan.x, we see there is only one acute angle 0 for which tan @ = 1/2. This is our desired angle, and we write it as @ = tan“1(1/2). This is also often written as @ = arctan(1/2). WARNING: The expression tan“ y does not mean 1/(tan y), and the same is true of the other trigonometric functions. ‘To write the reciprocal of sinx, we must write (sinx)*. (Unlike sin? x, which does mean (sin x)’. Sorry for this discrepancy, but we didn’t make the rules!) Even a casual glance at the graph of tanx will show that there are actually many values of x at which tanx = 1/2. How do we know which one is intended by the expression tan™'(1/2)? We don’t. There are infinitely many values of tan“!(1/2). To show that we want the acute angle as in the diagram above (rather than one of the other values), we write @ = Tan”! 1/2, where the capital T shows that we are interested only in the principal value, or the value that lies in the first period of tanx. Since this period ranges from —7/2 to /2, we will always have -n/2 < Tan7!x < 7/2. Similarly, Arctanx implies the principal value as well. If we apply this concept to Sin“! 1/2, we will note that there are still two values of x in the first period of sinx for which sinx = 1/2. Thus we need to restrict ourselves to the first half-period in which sin x ranges from —1 to 1. So, we have —/2 < Sin" y < 7/2. (Note that Sin“ y can equal 7/2 or —71/2, while Tan“ y cannot.) After seeing that the principal values of inverse sine and inverse tangent are both between —7/2 and 11/2, we may suspect that the principal values of inverse cosine are also in that range. So what is Cos *(-1/2)? For all x in the range from ~m/2 to m/2, cos. is positive. This is clearly not the right range. We should use 0 < Arcos < 7 instead. Convince yourself by looking at the graph of cosx that this is a sufficient range.the ART of PROBLEM SOLVING: Volume 2 s15 EXAMPLE 2-6 Find sin“! 0 and Arcsec2. 10 =n, for Solution: Since sinx = 0 at all x which are integral multiples of 7, we can write si +7 -2,-1,0,1,2,.... Since secx = 1/cosx, for Arcsec2 we seek the angle in the range [0,72) such that cosx = 1/2. Thus, x = 12/3. Is it clear why the principal values of inverse secant are in the same range as those of inverse cosine? EXERCISE 2-10 What is wrong with the statement sin“ 1 — Sin“! 1 = 0? EXERCISE 2-11 Evaluate arccsc—1, Cos"! -Y2/2, and Arctan(— 3/3). 24 Tying It All Together Since the graphs of sinx and cosx are very similar, you may suspect that they can be easily related. You're right. First, since sinx is the same as cosx shifted to the right by 7/2, we can say sinx = cos(x — 7/2). (Look at the graphs and see!) Inthe graph at right, we have plotted sin x and sin(x—r) on the same graph. We see that sin(x ~ n) is the reflection of sinx in the x axis! Thus, sin(x — x) is everywhere the negative of sinx, or sin(x ~ m) = ~sinz. Try finding the relationship between cos x and cos(x — n). Now consider reflections in the y axis. Just as —f(x) is the reflection of f(x) in the x axis, f(-x) is the reflection of f(x) in the y axis. Choose a few functions {(2) and plot the functions and the respective f(~x) to see their relationships to each other. Applying this principle to cosx, we see that the reflection of cosx in the y axis gives the same graph back. Thus, cos(-x) = cos x, so we find that cos x is an even function. y The reflection of sin x in the y axis is another matter. As with sin(x ~ 7), sin(—x) is everywhere the negative of sin x, so sinx = ~ sin(—x) and sin is an odd function. When working with expressions involving negatives and multiples of 90° as we have above, it is often useful to look at the graphs of the resulting functions to determine their connections to sinx, cosx, or tanx. While we will later examine faster methods to do this, it is very important to learn how the trigonometric functions are related and to understand their graphical representations. 1 EXAMPLE 2-7 Use graphical analysis as above to show sin x = cos(90° — x), which we showed in Volume 1 using the geometry of right triangles.16 > CHAPTER 2. NOT JUST FOR RIGHT TRIANGLES Solution: We'll draw cos(7/2—x) in steps. First we draw cos(~x) by reflecting cos. in the y axis (the dashed lines in the graph). Then we draw cos(n/2~ x) = cos(-x + 1/2) by shifting the graph of cos(-x) to the right by 7/2 (the solid lines). Make sure you see that why this is a shift to the right, not the left. The resulting graph of cos(—x + 7/2) is the same as sin x, as you can verify. EXERCISE 2-12 Which of the following functions are odd and which are even: sin x, cosx, tanx, escx, secx, and cot x? EXERCISE 2-13 Find each of the following as trigonometric functions of x: sec(270° + x), cos(m + x), tan(450° + x), and sin(37 — x). We've figured out how to handle trigonometric functions of sums and differences of angles where one of the angles is a multiple of 90°. How about other angles? To answer this, we will use a method proposed by Masakazu Nihei of Japan in Mathematics & Informatics Quarterly (Vol. 3, No. 2) A Consider the triangle ABC at left. We'll find the area of this triangle in two ways, both as }(AB)(AC)(sin BAC) and also as [ABH] + [ACH]. We let AH = 1 and express the sides of right triangles ACH and ABH in terms of a and p. (For @ d example, cosa = AH/AB = 1/AB so AB = 1/ cosa.) Hence, we have H [ABC] = [ABH] +[ACH] 3(AC)(AB)(sin BAC) 3(AH)(BH) + }(AH)(CH) 4(1/ cos B)(1/ cosa)(sin CHAPTER 2. NOT JUST FOR RIGHT TRIANGLES Use the sum formulas for cosine and tangent to prove these. Using sin?x + cos*x = 1, which we proved in the first volume, we can write cos 2x in a couple other, equally useful, ways: 2cos*x-1 = 1-2sin?x, cos 2x " These two are often used to evaluate integrals involving cosines and sines, so when you learn integral calculus, you'll be seeing them again. Related to sin? x + cos?x = 1 are the identities 1 + tan? x = sec? x and 1 + cot? x = esc? x, which were also discussed in the first volume. (And which you should be able to prove quickly.) These three identities are clearly most useful when working with squares of trigonometric functions. EXAMPLE 2-11 Use the above formula for cos 2x to create formulas for sin x/2 and cos x/2. Solution: Applying our double angle formulas to cos x, we have = coe? sin? 2 = 2 in? ~ cosx = cos? 5 ~ sin® 5 = 2cos" 5 2sin? 5. The first of the three expressions for cos: isn’t terribly useful, but the second and third are, as we can solve for the desired expressions in terms of cos: x _ , flFcosx css = 5 x sing = 4 The + signs are a result of taking square roots. How do we know which to use? We use our knowledge of the signs of sine and cosine. If x/2 is in the first quadrant, we use + for both, and so on. EXERCISE 2-19 Use the above formula for sin x/2 to determine sin 15° EXERCISE 2-20 When | use sin(60° — 45°) to evaluate sin 15°, I get (V6 - V2)/4, but when I use sin(30°/2),1 get (V2 - V8) /2. Have I done something wrong? Which method is easier? EXERCISE 2-21 Find two expressions which contain no square roots for tanx/2 in terms of sinx and cos x. EXAMPLE 2-12 Find the amplitude of f(x) = 3sinx + cos x. Solution: We might be tempted to say the answer is (3 + 1) = 4, but this is not right. (Can you find an x for which this function equals 42) If we can express this sum as a single sine or cosine, we can find the amplitude easily. Recall our formula for sin(x + y), sin(x + y) = sinxcosy + sin ycosx. Comparing this to f(x), we can write the function as a single sine if we find an angle y for which cosy = 3and siny = 1. However, cosy = 3 is ridiculous for real numbers y. A little better isthe ART of PROBLEM SOLVING: Volume 2 a i S(2)/4 = (8/4)sinx + (1/4) cosx. Now we need an angle for which cos y = 3/4 and siny = 1/4. Still no such angle exists, because these values violate sin” y + cos* y = 1. We're not lost yet, though. Let’s try a generic scaling A, so we can write Af(x) = 3Asinx + Acosx, where cosy = 3A and siny = A. We thus have sin? y + cos? y = 104? = 1, or A = 1/ V10. We then have £2) “= = cosysinx + cosxsin y = sin(x + y). -Vi0 ¥ y ¥y) Thus f(x) = V10sin(x + y). Remembering that y is a constant angle, the amplitude of f(x) is VI0. EXERCISE 2-22 Notice that the answer to the previous problem is V3? + 12, where the 3 and the 1 are the coefficients of sine and cosine. Is this true in general, i. is the amplitude of asin x + bcosx always Va? +0? EXAMPLE 2-13 Find the period of f(x) = sin 2x + cos3x. Solution: The period of sin 2x (light dashed line) is and that of cos 3x (light dotted line) is 271/3. In one full period of length T of f(x), both sin 2x and cos3x must go through an integral number of periods. (Why?) Hence, for some positive integers m and n, we have T = mn = 2nn/3. Writing this as n = 3m/2, the smallest possible solutions are (m,n) = (2,3). (Why do we want the smallest solutions?) The period of f(x) is then T = 2m. EXERCISE 2-23 Extend the above argument to find the period of sin(ax/b) + cos(cx/d), where a, &, S~_ ¢, and d are integers and a/b and c/d are in lowest terms. To derive a final set of trig identities, notice what happens when we add cos(x + y) and cos(x-y). The products of sines cancel and we are left with 2cos x cos y! Thus cos(x + y) + cos(x - y) = 2cosxcos y. Letting a = x + yand f = x - y, we can write this as +B cosas cosp=200s( 22) os (258),20 > CHAPTER 2. NOT JUST FOR RIGHT TRIANGLES Similarly, we can find the following expressions: cosa —cosp = sina + sinB sina -sinB 2.5 Solving Problems Using Trigonometric Identities Look closely at each of the following examples; each one exhibits a common technique in attacking problems using the trig identities of the previous sections, Some methods of solving these trig identity problems are discussed below, but the best tool is experience, so take the time to work through all the problems yourself. » Look for angles whose sum or difference is a multiple of 90°. If these exist, we can often use our relations like sin(180° ~ x) = sin x and sin(90° ~ x) = cosx. » When you see squares of trigonometric relations, try using sin? x + cos*x = 1 or the related identities. > Look for pairs of angles whose ratio is a power of 2. For example, sin20° _ 2sin 10° cos 10° cosi0° cos 10° = 2sin 10° Using the double angle formulas as above will often simplify such expressions. » When working with the trigonometric functions besides sine and cosine, it is often helpful to. write the problem in terms of just sine and cosine. > Don't work with inverse trig functions. Apply trigonometric functions to equations involving inverse trig functions to get rid of them. EXAMPLE 2-14 Evaluate tan 10° tan 20° tan 30° --- tan. 80°. Solution: Writing this in terms of sines and cosines, we have sin 10° sin 20° sin 30° ---sin 80° cos 10° cos 20° cos30° ---cos 80°" “Applying sin x = cos(90° ~ x) to each term in the numerator, we get 0s 80° cos 70° cos 60° -+-cos 10° 08 10° cos 20° cos 30° --cos 80° EXAMPLE 2-15 Write (sin 13° + sin 167° + cos 13° + cos 167°)(sin 13° — sin 167° + cos 13° — cos 167°) in the form asinx°.the ART of PROBLEM SOLVING: Volume 2 < 21 Solution: Instead of multiplying out the product, we note that 13° + 167° = 180° and use the rela- tions sin sin(180° —x) and cos x = — cos(180° ~x). Thus sin 13° = sin 167° and cos 13° = — cos 167°, and our product is (2sin 13°)(2.cos 13°) = 4 sin 13° cos 13° = 2 sin 26°. EXAMPLE 2-16 Find xif Tan“ x = Tan“!4+ Tan! 6, Solution: Working with inverse functions is pretty difficult, so how do we get rid of them? Simply use tan(Tan™ y) = y. Taking tangents of both sides of the given equation, we have x = tan(Tan’4+Tan“6) tan(Tan™ 4) + tan(Tan™ 6) 1 = tan(Tan™ 4) tan(Tan=! 6) (4+ 6)/(1- 4-6) = -10/23. u EXAMPLE 2-17 Find secx in terms of y if x = Tan“ y. Solution: Taking tangents of the given equation, we have tanx = y. Using 1 + tan?x = sec? x, we et sec*x= 147, orsecx = + /14 y2. 8 y EXERCISE 2-24 Simplify 4/7 in2x secx fOr * Such that 0° Given a value for sinx + cos. or sinx ~ cos x, we can find sin 2x, and hence x, by squaring the given relation. Try it and see! Problems to Solve for Chapter 2 14. Find the value of sin? 10° + sin? 20° + sin? 30° + -.- + sin290°. (mae 1992) 18, Evaluate csc (Arcsin } ~ Arccos }). (MA@ 1991) 16, Given that Arcsin x = y, find tan y in terms of x. 17, Given a positive integer m and a number ¢, ~1 < c < 1, for how many values of q in [0,2n) is sin ng = c? (MA@ 1992) 18. Given the triangle ABC with side a opposite 2A, side b opposite 2B, and side c opposite ZC, find sinA + sin2B + sin3C if@ = 3, b = 4, and c = 5. (MAG 1991)22> CHAPTER 2. NOT JUST FOR RIGHT TRIANGLES 19, Solve for x: Arctan 3 + Arctan . (MAO 1991) 20, Find the period of 2sin(4nx + 11/2) + 3cos(Srex). (MAO 1992) 21. Write sin‘ x in terms of cos 2x and cos 4x. (MA@ 1991) 22, Which of the following equals cot 10 + tan5: cse5, esc 10, sec5, sec 10, or sin 15? (AHSME 1989) 23, If sinx = cos2x and 0 the BIG PICTURE In the 1700's, Daniel Bernoulli, one of a large family of brilliant mathematicians and physi- cists, was studying the vibration of guitar strings. Supposing the string is of length 1, the fundamental vibration of the string can be represented as sin zx. With this function, the two ends are fixed at zero, while the middle vibrates freely. (Plot the function for 0 < x < 1 and see.) Bernoulli knew that when a string is plucked, you hear more than just the fundamental vibration, however; you hear a series of overtones with vibrational speed two times higher, three times higher, and so on. Extending the argument used to justify sin nx, Bernoulli argued that the overtones could be represented as sin 2rcx, sin 3x, and so on. (In fact, if you vibrate the ends of, say, a phone cord, you can see these vibrational modes yourself.) Bernoulli asserted that the general vibration would be some combination of the fundamental tones and overtones, 4 sin(rex) + ap sin(2nx) + a3 sin(Brex) + + Other mathematicians of the time also used such series, including Leonhard Euler, who came up with a method to find the coefficients aj. In 1822 the French engineer Joseph Fourier published a treatise on heat transfer, looking at situations like, what would be the temperature anywhere on a rectangular plate if one side was held at a temperature of 100 degrees and the other three sides at 0°. Fourier used infinite series of sines and cosines in his solutions, and tried to prove the controversial assertion that any periodic function f(x) could be written as a sum £02) = San cos(2nnx/T) + by sin(2nnx/T)], m= where T is the period. (Do you see why the sum of all these sines and cosines have period T?) This surprising fact was proven in 1829 by Dirichlet, but such series continued to be called Fourier series. Fourier series are now crucially important in most branches of mathematics and physics, as they break complicated functions down into manageable sines and cosines. 23Chapter 3 More Triangles! 3.1. Triangle Laws 3-74 __ How many non-congruent triangles ABC are there such that AC = 4, BC = 3, and ZACB = 30°? Recall that by SAS congruency all triangles which satisfy the above criteria are congruent; thus, AB can only have one length. (Try drawing 2 non-congruent triangles which satisfy the above specifications!) At left is the described triangle. Since c there is only one possible value of AB, we should be able to find it, but how? Draw a circle with center A and AC as the radius. Let F be the point E D where side BC (extended if necessary) meets the circle. Since ACFD is right (why?), CF = CD.cos30° = 4-V3, so point B must be on segment CF c (since CB < CF) and hence is inside the circle. We continue the sides of the triangle to meet the circle at the points shown. Applying the Power of a Point Theorem to point B, we have (BG)(BE) = (BF)(BC). E Now we relate these lengths to sides and angles of our original AABC. First we have AG = AC = AE = bas radii of the circle, so BE = AE ~ AB = C AC ~ AB = b ~c. Furthermore, we find BG = BA+ AG = BA+AC=c +b and BF = CF—BC = CD cos C—a = 2bcos Ca, where we have used right triangle FCD to determine CF = CDcosC = 2bcosC. Putting these expressions into our Power of a Point equation, we get (BG)(BE) = (BF)(BC) (b+0@-0) = (2bcosC~a)(a) B-2 = ~a+2abcosC, or the more common =a +b — 2abcosC This is called the law of cosines and can always be used to determine the third side of a triangle ‘when given two sides (a and b) and the angle included between them (ZC). EXAMPLE 3-1 Use the law of cosines to find the answer to our introductory example. Solution: Applying the law directly, we have AB? = 3? + 4? — 2(3)(4)(cos 30°), «24>the ART of PROBLEM SOLVING: Volume 2 < 25 80 AB = 25-12 V3. EXAMPLE 3-2 Use the law of cosines to find 2B if a = 5,b = V21, and c = 4 in AABC. Solution: Using the law of cosines, we have 2 cosB =» Thus, 2B = 60°. Hence, we can find any angle of a triangle if given the three sides of the triangle EXERCISE 3-1 In our proof of the law of cosines, point B was inside the circle centered at A with radius AC. Complete the proof of the law of cosines by addressing the cases in which B is on the circle and outside the circle. In any field involving geometry, such as land surveying or architecture, the law of cosines is a necessity in determining lengths and angles. However, the law of cosines is sometimes not enough. For example, suppose we are given two angles 2A and CHAPTER 3. MORE TRIANGLES! S\__ EXERCISE 3-3. Complete the proof of the law of sines by showing that a/ sin A = 2R when 4A is obtuse and when ZA is right. EXAMPLE 3-4 Given ZA = 75°, AB = 4, and AC = 2 V6, find 2B, C, and BC. Solution: First we use the law of cosines to find BC? = AB? + AC? -2(AB)(AC)cos A = 16 +24 —2(4)(2 Vé)[( V6 - V2)/4] = 16 + 8. V3, so BC = 16 +83 = 2/4+2V3 = 2(1 + V3) = 2 +2-V3. We can use either the law of sines or the law of cosines to find the remaining angles. The law of sines is much easier to use, so we write BC/sinA = AB/sinC, and ABsinA_ 4 V6+¥2_ VO+V2_ V2 BC 24+2¥3 4 -242V8 2” Hence £C = 45° (why not 135°?), and 2B = 180° — 75° — 45° = 60° sinC = These examples should have given you a grasp of how and why the law of sines and the law of cosines are used in basic problems and in many fields. In problem solving, the law of cosines is used in problems involving squares of sides (for obvious reasons). These problems are usually pretty easy to spot, as expressions such as (c? - a ~ b?)/(—2ab) are hard to miss. Make sure you recognize this expression as cos C! The use of the law of cosines in a problem can often get very algebraic. Try not to resort to the law of cosines immediately in problems besides those already discussed; chances are there is a more elegant approach. The law of sines is much less complicated algebraically than the law of cosines and hence can be used creatively in a broader range of problems. Problems involving sines of angles and the circumradius of a triangle are often ripe for the use of the law of sines. The presence of the circumradius or @/ sin A terms is often a giveaway that the law of sines will be useful. The law of sines is also useful when circles are present; equal inscribed angles allow us to make clever manipulations like those in the proof of the law of sines. $S__ EXAMPLE3-5 Prove the law of tangents, which states that in triangle ABC witha = BC and b = AC, a-b_ a+b” Proof: First we write the latter fraction in terms of sines and cosines: tan 432 sin (438) cos (452) tan 48 ~ sin (4f2) cos (458) Recalling our expressions for the sum and difference of sines from page 20, we recognize the numerator as (sin A ~ sin B)/2 and the denominator as (sin A + sin B)/2, so tan 4the ART of PROBLEM SOLVING: Volume 2 < 27 We're clearly very close now. We need a way to relate the sines to the sides a and b; looks like a good place for the law of sines. Since sin A = a/2R and sin B = b/2R, we have tan 458 _ a/2R—b/2R _ a-b © a/2R+b/2R~ a+b Ze tan" EXERCISE 3-4 Use the law of sines to show that in ABC, 2A > 2B if and only if BC > AC. EXAMPLE 3-6 Use the law of cosines to justify the statement that if a, b, and c are the sides of triangle AABC and a < b < c, then AABC is acute if a? + b? > c? and obtuse if a? + b? < 2. Solution: From the law of cosines, ee cosC = SE . Ifc? a? + L?, cos C is negative and ZC is obtuse. EXERCISE 3-5 In AABC, let D be a point on BC such that AD bisects ZA. If AD = 6, BD = 4, and DC = 3, then find AB. (MA 1991) 3.2 Areas, Areas, Areas In the first volume we investigated three methods of finding the area of ABC, namely [ABC] = oo sinc =15, where 7 is the inradius, s the semiperimeter, and hy the altitude to side BC. There are far more formulas for the area of a triangle than are useful to remember; in this section we will explore a few more useful methods. If we are told the three side lengths of a triangle, we can draw the triangle in only one way. By this we mean that all triangles we draw with these three side lengths will be exactly alike (by $55 congruence). Hence, we should be able to find the area of a triangle from just its three sides, but how? Our three formulas all involve either angles or other lengths. Hence, we look for a way to modify one formula to be in terms of only the sides of a triangle. Since we know how to relate the angles of a triangle to the sides, we'll use [ABC] = }absin C. The law of sines isn’t directly useful because it introduces the circumradius, a length we don’t know. The law of cosineshas only the triangle side lengths, so we'll try that by writing sinC = Vi —cos®C. From the law of cosines and lots of algebra (mostly recognizing perfect squares and factoring differences of squares), we have [ABC] = e Vi = cos? ab f (@-a@-P 2 ee<< 28 > CHAPTER 3. MORE TRIANGLES! 0 (atte — (2 By) /16 V2ab = 2 + 0? + P)Qab + 2 — a — 7/16 VIG + bP - elle - @- bY I/16 Varb-atb+ abr oj(-atb+o/l6 vals = (5 BYE 0), where s is the semiperimeter of the triangle. The last step can be seen by noting that (a +b ~ )/2 = (a+b +c—20)/2 = s—c. This formula for the area of a triangle is commonly called Heron’s formula. EXAMPLE 3-7 Find the area of the incircle of a triangle with side lengths 13, 14, and 15. Solution: The only relation we know so far involving the inradius of AABC is [ABC] = rs. We do know s = (13 + 14 + 15)/2 = 21, so if we find [ABC], we can find r. Since we also know [ABC] = 506 —a)(s — BY(S =), we can find r from rs = -Vs(6— a(S — B= 6), or We thus find r = 4, and the desired area is 167. What if we used the law of sines to evaluate the area? Writing sin C = c/2R, we find (aBc] = $ sinc = we, which is another very important relation. ‘As we saw in the previous example, the importance of these area formulas is not only in finding the area of a triangle, but they also can be used to relate various important triangle lengths and angles. In the following examples, you will have some practice with such formulas and manipulations. ‘Among these problems and those at the end of the chapter you will see many more formulas for the area of a triangle and for various lengths and angles. Don’t memorize them all; learn how they are derived. EXAMPLE 3-8 Find the circumradius of a triangle whose sides are 13, 14, and 15. Solution: In the previous example we saw that the inradius of such a triangle is 4. Since 13 = [ABC] = abc/4R, we have R=abc/4rs = 65/8. EXAMPLE 3-9 Given two angles and side a of a triangle, how would you find the area of the triangle without determining any more side lengths? Solution: Notice that it doesn’t matter which two angles we know, as we can determine the third from the two given. Starting from the formula [ABC] = (ab/2) sin C, we can express b in terms of a from the law of sines, or b = (asin B)/sinA, 50 asin BsinC 2sinA ‘After finding the third angle from the two given we can use this expression to find the area [ABC] =the ART of PROBLEM SOLVING: Volume 2 < 29 EXERCISE 3-6 Show that [ABC] = 2R*sin Asin B sin C. Once again, these formulas aren’t always most important for the determination of areas. You will see them again in our discussion of geometric inequalities (page 165) and you will often use them as intermediate steps to solutions. The information given in a problem usuaily indicates which methods are useful: if we are given the sides of a triangle, Heron’s formula is appropriate; given two sides and the angle between them, we use (ab/2) sin C; for problems involving altitudes, inradii, or circumradii, we use the corresponding methods. Finally, problems involving perimeter often call for [ABC] = rs, and keep an eye out for the product abc and remember that it equals 4R[ABC]. 3.3 More Important Lines A segment drawn from a vertex of a triangle to the opposite side is called a cevian. In the first volume we discussed a few important cevians, namely angle bisectors, medians, and altitudes. In this section we will discuss methods of finding the lengths of these cevians. Altitudes are the easiest, because we can use the area of a triangle to find the altitude from [ABC] = al /2. Since we have many ways to find the area, we have many ways to find the altitude. EXERCISE 3-7 Find the length of the shortest altitude of a triangle with side lengths 13, 14, and 15. Angle bisectors and medians are significantly more difficult than altitudes; we will first reiterate a method we mentioned in the first volume. EXERCISE 3-8 Use the law of cosines to show that the sum of the squares of the diagonals of a parallelogram equals the sum of the squares of the sides. EXAMPLE 3-10 Use the previous exercise to find the length of median AD of SABC if AB BC =7, and AC =8. Solution: Extend median AD beyond D to E such that AD = DE. Since D is the midpoint of both AE and BC, the diagonals of ABEC bisect each other. Thus, ABEC is a parallelogram, so we can apply the previous exercise: AE? + BC? = AB? + BE? + EC? + AC?. Since AC = BE, AE = 2(AD), and AB = EC, we have 4AD? + BC? = 2(AB? + AC2), 80 AD = yX05 + 64) = 49/2 = ¥129/2. EXERCISE 3-9 Use the previous example to find a general formula for the length of median AD of ABC,30 > CHAPTER 3. MORE TRIANGLES! We've now figured out medians. Before addressing angle bisectors, we will introduce and prove Stewart's Theorem. In the figure, cevian AD with length d cuts BC (length a) into segments of A length m and n as shown. Since cos ADB = ~ cos ADC (because the angles are supplementary), we can relate all the given lengths using the law of cosines: 2-P-met _ P-&- we B et oe T al oh cos ADC, DT or after multiplying by -2mnd, cos ZADB n= a@n— nm? = Wm + dm +? Rearranging this, we have cn + #m = d?(m +n) + mn(m +n), or (since m +n = BC =a) the easier to remember cnc + bmb = dad + man. You will usually only use Stewart's Theorem to find the lengths of angle bisectors and medians. To find the length of an angle bisector, we recall the Angle Bisector Theorem, which gives n/m = b/c. We can use this relation and Stewart's Theorem (and a bit of algebra) to find If you don’t see how to get this, solve Stewart's Theorem for d and use the Angle Bisector Theorem in the form m/(m +1) = c/(b +0). EXERCISE 3-10 Don’t take our word for it; work through the algebra to obtain the above formula, EXAMPLE 3-11 Show that if AD is an angle bisector of AABC, then AD? + (BD)(CD) = (AB)(AC). Solution: Dividing both sides of Stewart's Theorem by 4, with d = AD, m = BD, and n = CD, we have Bam mb sene From the Angle Bisector Theorem, bm = cr, 80, cnb+bme _ bo(m+n) _ @ + mn= be, which is the desired result. EXERCISE 3-11 Use Stewart’s Theorem to find an expression for the length of median AD in terms of the sides of ABC. Does your formula agree with your earlier result for median AD? EXAMPLE 3-12 Segment CX divides AB such that AX = 2BX = 4. If AC = 7 and BC = 5, find CX. Solution: Applying Stewart's Theorem, we have 2(6)(4) + 6 CX? = 5(4)(5) + 7(2)(7), so CX =5.the ART of PROBLEM SOLVING: Volume 2 < 31 A The most important lesson to learn from Stewart’s Theorem is the method of using the law of cosines on two equal angles or two supplementary angles. For example, if AD is an angle bisector in AABC, we could use the law of é g cosines to determine cos BAD and cos /CAD and set these equal. This will D achieve the same result as cos ADC = - cos ADB (Stewart's Theorem) Problems to Solve for Chapter 3 36. Is a triangle whose side lengths are in the ratio 6 : 8 : 9 acute, right, or obtuse? (AHSME 1952) 37. Find the length of the altitude to the 14 inch side of a triangle whose two other sides have lengths of 13 inches and 15 inches. (MA@ 1990) 38. In AADC, angle bisector DB is drawn. If AB = 3, AD = 6, and CD = 8, find BD, (MA@ 1987) 39. If the sides of a triangle are in the ratio 4 : 6 : 8, then find the cosine of the smallest angle. (MAQ 1991) 40. Triangle ABC is such that AB = 4 and AC = 8. If Mis the midpoint of BC and AM = 3, what is the length of BC? (AHSME 1975) 41. Use the law of sines to prove the Angle Bisector Theorem. 42. Find tan C, where Cis the angle opposite side c of a triangle whose side lengths a, b, and c satisfy C++ atbte (MA@ 1991) 43. Point O is the center of the circle circumscribed about isosceles AABC. If AB = AC = 7 and BC = 2, find AO. (Mandelbrot #3) 44, Prove that if the sides of trapezoid ABCD (AB || CD) satisfy the condition AC? + BD? = (AB+CDY?, then ABCD is orthodiagonal. (M&iQ 1991) @+P +3 = =. = 7, compute the maximum value of @. (ARML 1984) sin? A + sin’ B+ sin°C P u 45. In ABC, a > b> c, If 46. Show that in AABC, 47. Use the last problem to find a similar expression for cos(A/2) without using the law of cosines. 48. Show that A BoC [ABC] = Poot > cot 5 cot 5.32 > CHAPTER 3. MORE TRIANGLES! 49. Points A, B, C, and D are on a circle of diameter 1, and X is on diameter AD. If BX = CX and 3 O anda +B +y = 7, then sin2a + sin26 + sin2y = 4sinasinB sin y. (M&IQ 2) 7"Chapter 4 Cyclic Quadrilaterals 4.1 Properties of Cyclic Quadrilaterals Unlike triangles, not all quadrilaterals can be inscribed in a circle. Those which can be are called cyclic quadrilaterals, Such quadrilaterals have the following two special properties. » The sum of the opposite angles in a cyclic quadrilateral is always 180°. Thus, Doe LA+ LC = (B+ LD = 180° To prove this, we note that angles A and C are inscribed angles, so a B BCD BAD _BCD+BAD _ 360° ° CA = EE = 100", » When we draw the diagonals of a cyclic quadrilateral, we form four pairs of equal angles like the pair shown below. D. ‘These angles are equal because they are inscribed angles which subtend the same arc. We have such a pair of equal angles for each side of the quadrilateral: ZABD = ACD —_ZACB = ADB 34 > CHAPTER 4. CYCLIC QUADRILATERALS is cyclic, we draw the circumcircle of AABC and show that if /B + ¢D = 180°, the circle must pass through D as well. We do this by proving that D cannot be inside or outside the circle of ABC, so D must be on the circumference. In both of the diagrams below, we let £F=0, AEC =a, and ABC=8. Since AEC and ABC make up the whole circle, their sum is 360°. For the case where D is inside the circle, pA _a@ +0 Bra 0 12, 6 (K \) wB+ = 4 PF ERS a 180+ 5 ° , ‘Thus the sum is greater than 180” if D is inside the circle, violating the given fact \ 4 that ZB + D = 180°. c For the case of D being outside the circle, we do the same thing: D be _a@,B-@_atB 8 2, 6 KON, eB = 54 E82 EE 7 = 180° - 5. \ B If Dis outside the circle, 2B + £D is less than 180°, which is again a contradiction. ZI Since D cannot be outside or inside the circumcircle of ABC when /B + £D = ¢ 180°, it must be on the circle. oS ‘The other method for proving that four points ate concyclic, meaning they all lie on the same circle, is as follows. If points C and D lie on the A 3 _ same side of segment AB such that ACB = ZADB, then the four points AB A,B,C, and Dare concyclic. By ‘the same side’ we mean the points are as. \, inthe diagram on the leftand not the right. We can prove this just like we proved the prior method for showing a quadrilateral is cyclic. We draw the circumcircle of ABC and show that if D is inside this circle, then ZADB > ZACB, a contradiction to the two being equal. Similarly, if D is outside this circle, ADB < ACB, a contradiction. Thus, point D is on the circumcircle of AABC. Since this proof is almost exactly like the one we did above, we'll let you do it yourself. Now that we know how to find cyclic quadrilaterals and how we can use them once we find them, we must discuss when to look for them and how they are useful. Cyclic quadrilaterals are most useful for proving that angles are supplementary or equal. In fact, usually problems involving cyclic quadrilaterals involve first showing that a quadrilateral is cyclic, A B then using the equal angles formed by the diagonals and sides to show that a pair of angles are equal. ‘Although cyclic quadrilaterals can be used to show that angles are supplemen- tary, this is more commonly the way we show that a quadrilateral is cyclic. We then use the cyclic quadrilateral to show that angles are equal. We can use cyclic quadrilaterals to show that angles are equal when the angles are situated as in the top diagram at right. Angles A and B have sides which intersect at C and D, so we C D can show that ZA = CHAPTER 4. CYCLIC QUADRILATERALS Rearranging these, we have the equations (BC)(AD) (AB)(CD) (AC)(ED) (AC)(BE). We then add these to get (AB)(CD) + (BC)(AD) = (AC)(ED + BE) = (AC)(BD). This, finally, is Ptolemy’s Theorem, which states that in cyclic quadrilateral ABCD, with a = AB, b = BC,c = CD, d = DA, and diagonals e and f, we have ac +bd = ef. ‘The use of Ptolemy is generally quite straightforward once we have shown that a quadrilateral is cyclic. EXAMPLE 4-3 Find the diagonal length of an isosceles trapezoid with bases of lengths 8 and 20 and legs of length 10. Solution: Since the opposite angles of an isosceles trapezoid are supplementary, the trape- zoid is cyclic. Since the trapezoid is isosceles, its diagonals are equal. Letting these diagonals have length x, we apply Ptolemy's Theorem and find x2 = (8)(20) + (10)(10) = 260, 80x = 2-¥65. Problems to Solve for Chapter 4 58. Prove that a trapezoid is cyclic if and only if itis isosceles. 59. In cyclic quadrilateral ABCD with diagonals intersecting at E, we have AB = 5, BC = 10, BE = 7, and CD = 6. Find CE. 60. In the diagram, let ADM = ZACD and ZABM = CACB. Prove that AB = AD. A (Mandelbrot #2) 61. Quadrilateral ABCD with consecutive sides of 8, 15, and 12 is inscribed in a circle with circumference 177. Given that AC is a diameter of the circle, whatis 7 the length of the other diagonal of the quadrilateral? (MA 1987) 62. Inscribed ina circle is a quadrilateral having sides of lengths 25, 39, 52, and c 60 taken consecutively. What is the diameter of this circle? (AHSME 1972)the ART of PROBLEM SOLVING: Volume 2 < 37 A a 63. Suppose an angle BAC is rotated about point O to a new angle B‘A’C’ such that AB and A’B’ intersect at M and AC and A’C’ intersect at N, as y shown. Prove that A, A’, M, N, and O all lie ona circle. (Mandelbrot #3) 64. A parallelogram ABCD with an acute angle BAD is given. The bisector of BAD intersects CD at point L, and the line BC at point K, Let O be the circumcenter of ALCK. Prove that the quadrilateral DBCO is inscribed in a circle. (Bulgaria 1993) oO B c 65. Side AB of the square ABCD is also the hypotenuse of right triangle ABP (ABP lies outside ABCD). Prove that the angle bisector of APB bisects the area of ABCD, (M&IQ 1992) 66. Prove that the midpoints of the sides of a quadrilateral lie ona circle if and only if the quadrilateral is orthodiagonal. (Ma&IQ 1991) 67. In the figure, ABCD is a quadrilateral with right angles at A and C. Points B Eand F are on AC, and DE and BF are perpendicular to AC. If AE = 3, DE = 5, and CE = 7, then find BF. (AHSME 1990) c A 68. In quadrilateral ABCD with diagonals AC and BD intersecting at O, BO = 4, OD = 6, AO = 8, OC = 3, and AB = 6. Find AD. (AHSME 1967) 69. Prove that if in ABCD we let a = AB, b = BC, c = CD, and d = DA, we have a B +P) [ABCDF? = (~a)(s ~B)(s— os ~ 4) ~ abet cos? ( where s is the semiperimeter of the quadrilateral. What does this expression, called Brahmagupta’s formula, yield for a cyclic quadrilateral?Chapter 5 Conics and Polar Coordinates In Volume 1 we examined methods of graphing lines and circles. In this chapter we develop methods of describing and graphing other curves as well as using polar coordinates to describe curves. Be forewarned that while there are seemingly many formulas to be memorized in this chapter, if you take the time to understand the forms of the conics and the derivations thereof, you will need no memorization, only logic, to determine the formulas. Take the time to understand the proofs and the lessons and intuition they offer. 5.1 Parabolas We already know how to solve quadratic expressions like x? + 2x +4 = 0, but how do we graph the quadratic y = x2 + 2x + 4? The answer to this question is the parabola. Given a line | and a point P ina plane, a parabola is the set of points $ in the plane such that the length SP equals the distance from S to I. The point P is called the focus and the line | is called the directrix. Using this definition of a parabola, we can make the rough sketch shown, /s where | is a horizontal line. The minimum point on the curve is called the vertex, and we label it X = (HK). If we let the distance from X to P be a, we have P = (h,k +a). Similarly, lis a below X (since X is equidistant from Pand!) _1__¥4 and thus can be described by y = k~a. (Remember, | is a horizontal line.) If we choose any point S = (x,y) on the parabola, we have SP = y/(x— h)? + (y— ka)? and the distance from S to lis merely y-(k—a) = y—k-+a. Hence, from our definition of a parabola we have Squaring both sides and rearranging, we have (e-nP+(y-k-aP = (yk +a? (e-nP = (y-k+a)—-(y-k-a? (xh ly-k+a-ytk+a]ly-k+a+y—k-a] 2 0 WY = [a2 -B)- Dividing by 4a we have the general form of a parabola with a horizontal directrix: 1 y-k= pe hy. < 38 >the ART of PROBLEM SOLVING: Volume 2 < 39 Such parabolas always open either upward or downward (the one in this example opens upward). Similarly, if the directrix is vertical, the equation =1y xoh= Eh) describes the parabola (which then points either to the right or the left). In this case, the vertex is still (1,k), but now the focus is +a to the right of the vertex, or (h +a,k), and the directrix is a vertical line ~a from the vertex, or x = h—a. EXERCISE 5-1 In the two general equations for the parabola above, what effect does the 1/4a term have on the graph of the parabola? What does negative a mean? Large a? EXERCISE 5-2 The axis of symmetry is the line through the focus and the vertex of a parabola. The axis thus divides the parabola precisely in half. Find the equation of the axis of a parabola which opens upward and that of a parabola which opens to the right. EXAMPLE 5-1 Graph and find the vertex, focus, and directrix of the parabola x= -2y? + 12y- 15. Solution: First we complete the square to get the parabola in one of our general forms: 2? - 6y) - 15 x+(-2)(9) = -2(y? -6y +9) -15. x Hence our parabola is described by x - 3 = ~2(y—3)2, Our vertex then is (i,k) = (3,3). To determine the focus and directrix, we find a by noting 1/4a = -2, so a = -1/8. Hence, the directrix is x = 3 — (-1/8) = 25/8 (remember, the y term is squared, so the directrix is vertical) and the focus is (3 ~ 1/8,3) = (23/8,3). To plot the parabola, we first plot the vertex, then find a few more points on the parabola by selecting values for y and finding the corresponding x values. We then plot the parabola as shown. EXERCISE 5-3 Draw a line through the focus parallel to the directrix. Suppose this line intersects the parabola at A and B. The latus rectum is the segment AB. Prove for our above described parabolas that AB = |a|. How can this fact be used to sketch parabolas easily? EXERCISE 5-4 Why do you think the line | in our definition of a parabola is called the directrix? EXERCISE 5-5 Find the focus, the vertex, the directrix, and the length of the latus rectum of the parabola y=2/243x44, Now that we can find all the significant points and lines of a parabola, we should be able to find the equation of a parabola given some of these points or lines. We do so by first determining the direction of the parabola (right, left, up, or down), then using the given information to determine h, Kanda.40 > CHAPTER 5. CONICS AND POLAR COORDINATES EXAMPLE 5-2 Find the equation of a parabola with focus (3,2) and vertex (3,4). Solution: Since the vertex and focus lie on a vertical line (x = 3) and the vertex is above the focus, the parabola points downward. Hence, we are dealing with the form y ~ k = (1/4a)(x ~ h?. From the vertex we know that h = 3 and k = 4 and we expect a to be negative (since the parabola opens downward). We know the focus is always a away from the vertex. In this case, the focus is 2 units below the vertex, so a = -2. Hence, our parabola is y ~ 4 = ~(1/8)(« ~ 3) EXERCISE 5-6 Find the equation of a parabola with directrix x = 3/2 and focus (5/2, 4). EXAMPLE 5-3 A box with two dimensions of 10 feet and 4 feet is to be slid through a parabolic arch which is 5 feet tall at the center and 6 feet wide at the base. If the side facing the ground is 4 feet by 10 feet, what is the largest the other dimension can be and still slide through the arch? A Solution: Draw the segment AB from the top of the arch to the midpoint x y of the base as shown. If we let B be the origin, we can determine A = (0,5), C= G,0), and D = (-3,0), since the arch is 5 feet high and 6 feet wide. Hence, D. C__ the vertex of our parabola is (0,5) and it passes through (3, 0). Since the parabola WBZ points downward, we have y—5 = (1/4a)x as its equation. Using the point (3,0), we have —5 = 9/4a, or 1/4a = -5/9. Hence, the equation of the parabola is Now we draw the box WXYZ in the arch as shown. Since WZ = 4, point Z has coordinates (2,0) Point Y then has coordinates (2, z), where zis the desired third dimension. Since Y is on the parabola, we have z~ 5 = (-5/9)(22). Hence z = 25/9 and our largest possible third dimension is 25/9 feet Make sure you understand this problem. Variations of it using elliptical and hyperbolic arches (figures which are discussed in the next few sections) are very common. Don’t let them trip you up! 5.2 Ellipses ‘The general equation for a circle is (e- hy + (y-bP = RB Dividing by R?, we can write this as (=h? , Y-h? ete 1 In our discussion of distortion in Volume 1, we noted that we can stretch a ; circle to form an ellipse. Namely, we could stretch the radius in the x direction so that it differs from that in the y direction. The resulting curve is at right. We could then associate two radii, a and b, with our curve and write it as Game, yk? 1the ART of PROBLEM SOLVING: Volume 2 < 41 From our diagram, we see that we have two different ‘diameters’ in the x and y direction. These have length 2a and 2b, respectively. These are called the major axis and the minor axis of the ellipse, where the major axis is the longer of the two. EXERCISE 5-7 What if a = b above? Do you see why circles are ellipses? Taking two points F; and F2, we can define an ellipse as the set of points Z such that ZF; + ZF2 has some constant value. We call F; and F, the foci of the ellipse. To see that this new definition of an ellipse, (*: E B satisfies our equation form, we apply the distance formula. First we 5 draw the axes of the ellipse and note that CD = 2b and AB = 22. Now we can find the constant sum ZF; + ZF». Letting Z be A we have ZF, + ZF; = AF, + AF2 = BF, + AF; = AB = 2a, where we note that AF, = BF). (Why?) Hence our constant sum is 2a, or the length of the major axis. Since F\C = CF) and F\C + CF) = 2a, from right triangle CEF; we have 2a = 2CF, = 2 CE? + EF? = 2/6? + EF}. EF; be the distance from each focus to the center, we square a = \/#? + EF? to find c= a?—}?. (Why are the foci equidistant from the center?) Now we are ready to apply our constant sum principle to find the equation of an ellipse. We let the center of the ellipse be the origin, so that F, = (-c,0) and Fz = (c,0). Taking a general point Z = (x, y) on the ellipse, we have 2a = ZF, +ZFp=Ve+cPry+ Ve-e+y. We move one radical to the other side of the equation and square, so (22- Jerry)” = (very) 4? — da (e+ eP+ P+ ercPry = «P+ 4g? —4afQ+cPryt2rr2c+e+y = 2-2+2+y Again we rearrange the equation to isolate the square root. Then we divide by 4 and square both sides of the resulting equation, yielding 2 (-aVeroRr x) = (-a?— xc)? Plxtor+yP] = ab + 20x04 22 P+ wreta@r+ay = ab +2axc+x22, Simplifying where possible and applying the relation c? = a ~ b?, we have Pre reay = a+r Pre @-P+ey = at+2@-P),42 > CHAPTER 5. CONICS AND POLAR COORDINATES Putting all terms involving x and y on the left and the constants on the right we find Petey? = ab or ‘The simple translation from (x, y) to (xh, y—K) gives us the equation for an ellipse with center (Hk) rather than (0,0), or Gh? , y-wF eR For this ellipse, the foci are (/ + c,k) since they are c to the right and c to the left of the center. ‘We measure the amount an ellipse is stretched away from a circle by its eccentricity, which is c/a. Finally, the area enclosed in an ellipse is ab, which is proven on page 242. EXAMPLE 5-4 What happens to our ellipse equation if the major axis is parallel to the y axis rather than the x axis as above? Solution: Again we let the major axis have length 2a and the minor axis length 2b, so that through the same discussion as above, the equation of the ellipse is ‘The foci are now c above and c below the center (at (h,k ¢)) and c still equals Va? — 7. EXERCISE 5-8 Why is the quantity c/a called the eccentricity? Now we have a way to describe all important points and lengths of an ellipse whose axes are parallel to the coordinate axes. WARNING: In describing an ellipse, we associate the letter a with the larger of the ‘radii’ in the x and y directions. Always make sure your value of a is greater than that of b; otherwise, your value of c = Va? — & will be nonsense. Furthermore, make sure you know which direction the major axis points. This will help you determine where the foci are. EXAMPLE 5-5 Graph and find the center, foci, area, and lengths of the axes of the ellipse given by 9x? — 36x + dy? — 24y + 36 = 0. Solution: We attack this just as we did parabolas that weren’t written in the nice general form; we complete the square: 9(x? — 4x) + 4(y? - 6y) = -36. To make perfect squares, we add (—4/2)* = 4 inside the x parentheses and (-6/2)? = 9 inside the y parentheses. Hence, we have 0 9(x? — 4x + 4) + 4(y? - 6y + 9) —36 + 9(4) + 4(9) Wx-27 +4(y-3) = 36.the ART of PROBLEM SOLVING: Volume 2 < 43 Dividing by 36, we get our form for the ellipse, Y= 3P _ 7 ot 9 Hence the center is (2,3). Since the number under the y is greater than that under the x, the major axis is parallel to the y axis. From our equation we have a = y9 = 3 and b = V4 = 2. Thus, the major and minor axes have lengths 6 and 4, respectively. From a and b we find that the area is 6m and c = ¥9—4 = Y5. Since the major axis is parallel to the y axis, the foci are then found by adding and subtracting c from the y coordinate of the center, or (2,3 + V5). To graph the ellipse, we locate the endpoints of the axes. The major axis endpoints are 3 above and below the center (since the major axis has length 6) and thus are (2, 6) and (2,0). Similarly the endpoints of the minor axis are (0,3) and (4, 3). Plotting these points, we can draw the ellipse as shown. ¥ EXAMPLE 5-6 What are a and b for the ellipse (x =1 (y-2)2 sa=1) .w 2) 9 Solution: Neither anor bis V9 = 3! Notice the 4 before the (x~1)2, We put this in the denominator as (= IP, (y-2? 9/4 8 Our a and b values are then V8 = 2 V2 and 9/4 = 3/2, respectively. EXERCISE 5-9 Find the center, foci, and length of the axes of the ellipse 3x? + 4y? - 6x + 8y +3. =0. EXERCISE 5-10 What if I complete the square for a problem like the the previous exercise and find = (> (x= 2P (y+? 4773 as my equation of the ellipse? (Notice there is a 0 on the right, nota 1.) How many solutions (x,y) are there to this equation? What if I get a negative number on the right? These are cases of degenerate ellipses, or ellipse equations with either no solution or only one solution. EXERCISE 5-11 As with the parabola, we can define the latus recti of an ellipse as the segments through the foci parallel to the minor axis with endpoints on the ellipse. Find the length of each of these segments in terms of a and b. Just as with parabolas, we are often asked to find the equation of an ellipse given certain information about the ellipse. Again the first step is to determine which direction the ellipse points (ie. the direction of the major axis). We then use the given information to determine the center as well as @ and b.44> CHAPTER 5. CONICS AND POLAR COORDINATES EXAMPLE 5-7 Find the ellipse with major axis length 8, center (2, 1), and one focus at (2,3). Solution: Since the given focus is directly above the center, the major axis is parallel to the yy axis. From the information we can also deduce ht = 2, k = 1, and a = 8/2 = 4. Since the given focus is 2 away from the center, we have c = 2. Thus, from c? = a” ~ b?, we find =a? -c? = 12 and our ellipse is, (=H? y= KP _ B 2 i 16 Make sure you see why the 12 is under (x — 2)? rather than (y ~ 1). EXERCISE 5-12 Find the equation of the ellipse with foci at (3,1) and (5,1) and minor axis with length 4. With the ability to find the equation of an ellipse given some information about it, we move on to less obvious applications of ellipses. EXAMPLE 5-8 To give my dog some space to run, I drive two stakes in my lawn 10 feet apart. T tie the ends of a 30 foot rope to the stakes (one end to each stake) and loop my dog's collar loosely around the rope, so she is free to move along the rope. Over how many square feet is my dog free to roam? Solution: If my dog walks until the rope pulls taut, she will get to the boundary of her roaming area. This boundary is an ellipse since the sum of the distances from any point on the boundary to the stakes is the length of the rope, 30 feet. The stakes correspond to the foci and the rope to the constant sum of distances. Hence, the major axis has length 30. If we let the stakes lie on the x axis and the midpoint of the line connecting the stakes be the origin, the equation of the ellipse is (since a = 30/2 = 15) We then find b by noting that the distance from a focus to the center is c = 5 since the stakes are 10 feet apart, so b = Va2 —@ = 10-V2. The area of the roaming region is the area of the ellipse, or abrt = 150n-Y2 square feet. Keep an eye out for these slick applications of the constant sum of distances property of an ellipse to problems. 5.3 Hyperbolas Suppose we change the + in the general ellipse equation to a -, resulting in =n? y-k? 2 & ‘The graph of this equation is called a hyperbola. As you may have guessed, with each hyperbola we can associate a pair of foci F; and F2 so that the hyperbola is the set of all points $ where FS ~ F2S| 1.the ART of PROBLEM SOLVING: Volume 2 < 45 has some constant value. We can slug through the same exact algebra as we did with the ellipse to show the equivalence of the constant difference condition to the above equation. If you don’t trust us, try it yourself. Find this common difference of distances and use the fact that c = a? + 2 (where cis the distance from the focus to the center). If you find a bunch of points on the hyperbola, you will eventually find a curve like the bold curve shown below. The center O has coordinates (i, ), just like the ellipse. Points I and J are the foci and have coordinates (h + c,) since they are c to the right and left of the center. The points C and Gare the vertices. They have the same y coordinate as the center, so letting y = k in the equation for a hyperbola, we have (x ~ hi)*/a? = 1, so x = ha and the vertices are (h +a,k). The segment CG is called the transverse axis, and like the major axis of an ellipse it has length 2, Similar to the minor axis we define the conjugate axis as the segment with endpoints (hk +). Suppose we sketch the rectangle, as we have in dashed lines, with center O and sides equal in length to our two axes. The lines | and m which are the extensions of the diagonals of this rectangle are called the asymptotes of the hyperbola. These are lines which the curve approaches but never actually meets. We can find the equation of line I by noting that it passes through the center O = (h,k) and has slope HG/OG = b/a. Hence, the equation of line is yk = (b/a)(x—h). Similarly, the equation for line m is y—k = (b/a)(x—1). As the following example will show you, using the asymptotes helps graph the hyperbola Please do not memorize all of these formulas; understand them instead. EXERCISE 5-13 Why are the lines y — k = +#(x — h) asymptotes of the above hyperbola? What happens in our hyperbola equation if we let y- k = +£5(x— h)? EXAMPLE 5-9 Find the asymptotes, vertices, center, foci, and lengths of the axes of 9x2—4y? + 18x + 16y ~ 43 = 0 and graph the hyperbola, Solution: Grouping our x and y terms we have 9(x? + 2x) - 4(y? — 4y) = 43. Completing the square then dividing by 36 we get (+I? _(y-2P ee a1 4 9 (Always make sure you have positive 1 on the right.) The center is (-1,2), and we have a = 2 and b= 3, Hence, the transverse axis has length 2(2) = 4, the conjugate axis has length 6, and the vertices are (-3,2) and (1,2). Since c = Va? + # = V13, the foci of the hyperbola are (-1 + 13,2). Finally, the asymptotes are y— 2 = +3(x +1).7 46 > CHAPTER 5. CONICS AND POLAR COORDINATES ‘To graph the hyperbola, we plot its center and vertices. We then draw our asymptotes, which we can do most easily by sketching the rectangle with center (-1,2) and sides of length 2a and 2b as in our introduction. The lines through the opposite corners of the rectangle are the shown asymp- totes. Now we can draw the curve through the vertices approaching the lines asymptotically as shown. EXERCISE 5-14 Find the center, vertices, foci, asymptotes, and the lengths of the axes of the hyperbola (y-k? _@=-h? @ e Note that with a hyperbola we always associate a with the positive term rather than the one with the larger denominator. EXERCISE 5-15 We can define the latus recti of a hyperbola as the segments through the foci parallel to the conjugate axis with endpoints on the hyperbola. We can also define the eccentricity as we did for the ellipse, c/a, Find the length of the latus recti of a hyperbola. EXAMPLE 5-10 What if completing the square for the hyperbola results in (x~1)?/4~(y+2)"/9 = 0, rather than equalling 1 like usual? Solution: This is a degenerate hyperbola. We can write the equation as (x ~1)*/4 = (y+ 2/9. ‘Taking the square root of both sides, we find (x= 1/2 = #(y +2)/3, so the degenerate hyperbola is just a pair of lines. the graph we find the curve shown. This looks very much like a hyperbola, and it is! The coordinate axes are the asymptotes and the origin is the center. We can find the vertices of the hyperbola by noting that x = y at the vertex. (Why?) Hence the vertices are (v6, V6) and (- V6,— V6). Similar to this example, any \ Consider the graph of the curve xy = 6. Graphing several of the points on \ curve of the form xy = cis a hyperbola. EXERCISE 5-16 Why are the coordinate axes asymptotes of the hyperbola xy = 6? Once again, we can determine the equation of a hyperbola given various information about the hyperbola. EXERCISE 5-17 Find the equation of a hyperbola with vertices (-2, ~1 2 V2) and conjugate axis of length 4.the ART of PROBLEM SOLVING: Volume 2 < 47 Now that we've finished introducing parabolas, ellipses, and hyperbolas, we can discuss why we call them conic sections. Take a pair of congruent cones and hold them tip to tip so they have the same vertex and same axis but open in opposite ditections. Consider the various cross-sections that occur when you cut the resulting solid with a plane. Cutting completely through one cone forms an ellipse. Cutting with a plane parallel to the axis will form a hyperbola, and a plane intersecting one cone but not the other (but not passing all the way through the first cone) forms a parabola. 5.4 Polar Coordinates Revisited As we saw in the first volume, we can identify any point P in the plane by its distance from the origin (OP) and the angle @ which OP forms with the . positive x axis. Calling the distance from the point to the origin r, we relate the WAN polar point P = (r, 8) to the rectangular coordinate point (x, y) by x = rcos @ and y=rsin@. Hence, we have P+yY=Pos'6+rsin?a =r, and we relate the angle @ to x and y by Zane x cos6’ or @ = tan"! 7. WARNING: Make sure when determining 6 that it properly corresponds to the quadrant in which (x, y) lies. EXAMPLE 5-11 Convert the rectangular point (3,3) to polar coordinates and the polar point (6,30°) to rectangular coordinates. Solution: For the point (3,-3), we have r = x+y? = VI8 = 3-V2. We find the angle as @ = tan“1(-1). Since (3, -3) is in the fourth quadrant, @ = 315°. The point (3, ~3) in polar coordinates is then (3V2,315°). Notice that we could use 315° + n(360°) for any 1 as the angle to determine the point in polar coordinates as well. For the polar point (6, 30°), we have x = 6cos30° = 3-V3 and y = 6sin30° = 3. Our point thus is ¥3,3). EXERCISE 5-18 Express (6,—6 V3) in polar coordinates and (~2, 405°) in rectangular coordinates. Polar coordinates are often useful in describing curves. For example, the equation r = 3reptesents a circle with center (0,0) and radius 3. Using the expressions x = r cos @ and y = rsin 8, we can easily turn any curve in rectangular coordinates into a polar equation. EXAMPLE 5-12 Express the equation x” — y* = 9 in polar coordinates. Solution: Using x = rcos@ and y = rsin@, we have x? ~ y2 = 1°(cos? @ — sin? @) = rcos26. Hence, our polar form is 7? cos 20 = 9. EXERCISE 5-19 Express 6xy = 8 in polar coordinates.48 > CHAPTER 5. CONICS AND POLAR COORDINATES Going from polar coordinates to rectangular is generally a little bit tougher. To do so, we replace any 7? with x2 + y?, rcos 6 with x, and rsin @ with y. Sometimes we have to manipulate the equation a bit first, as you will see. EXAMPLE 5-13 Write r= = in rectangular coordinates. Solution: First we multiply by 3 - cos 8, yielding 3r-rcos@ =5. Since rcos 6 left) yields 97 ., we have 37 — x = 5. Isolating the r we find 3r = 5 + x. Squaring this (to get r? on the 25 +10x + x2. Since r? = x2 + ¥?, we have (x2 + y?) = 25 + 10x + x7, or 8x7 + 9y? — 10x = 25. You should now be able to recognize this as an ellipse. EXAMPLE 5-14 Identify the curve r = 30s. Solution: To identify a curve in polar coordinates, it is often best to convert the equation to rectangular coordinates and name the resulting curve. For this equation, we multiply by r to force an r? on one side and r cos @ on the other, namely r? = 3rcos @. Hence, x? + y* = 3x and our curve is acircle. EXERCISE 5-20 Express r = 4sec @ and r = 3sin @ in rectangular coordinates. EXERCISE 5-21 How would you express vertical or horizontal lines in polar coordinates? How about a line through the origin? EXERCISE 5-22 Describe as specifically as possible the class of curves described by r = asin@ + bos. Finally, for you trivia buffs, there are a few more families of curves which have simple polar forms. A limagon has the one of the forms: a+bsin@ +bcosd a—bsin® r r - bcos 6. ‘Try choosing some pairs (a,b) (where a,b > 0) and sketching the resulting graphs by choosing different values for 6, then computing r. Ifa limacon has a/b < 1, it will have a loop. If a/b = 1, the curve, which is shown at left, is called a cardioid. If 1 < a/b < 2, the limacon is ‘dimpled,’ and otherwise, it is ‘convex.’ Curves of the form 7? = £2cos20 or 1? = +a*sin26 are called lemniscates, which look like infinity symbols. The curve r = a6, where the radius increases with the angle, is called the spiral of Archimedes, Graph it and see why.the ART of PROBLEM SOLVING: Volume 2 < 49 Finally, r = asinn@ and r = acosn@ represent roses. Shown is a graph with n = 5, which has 5 ‘petals.’ Try choosing other values of 1 and plotting the results. Can you develop a rule for the number of petals in a rose? As weinitially stated, these curves aren’ t terribly useful for problem solving outside of trivial pursuits, however through using computer graphics you may be able to generate quite artistic results based on these simple polar forms. 5.5 That Pesky xy Term A general conic has the form Ax? + Bry + Cy’ + Dx + Ey + F = 0. In previous sections, the only time we have seen an xy term is when we discussed hyperbolas of the form xy = c for some constant c. Otherwise, we have avoided xy as much as possible because conics without xy terms are easy to analyze. Those with an xy term have axes which are no longer parallel to the coordinate axes and these are much more difficult to resolve. We call these conics oblique. Now that we've mastered polar coordinates we can analyze oblique conics by rotating them so that their axes are parallel to the coordinate axes. But how do we perform the rotation? Rather than rotate the conic, we rotate the coordinate axes to be yy parallel to the axes of the conic. In the diagram is the rotation of the coordinate axes through an angle a counterclockwise about the origin. Point P, which originally has coordinates (x, y) before rotation is at (x’, y') with respect to our new axes x’ and y’. Letting the angle OP forms with the positive x’ axis be B, we can relate the two pairs of rectangular coordinates of P to a and f through polar coordinates (where r = OP) as x=rcos(a+B) and y=rsin(a+6), and x =rcosp and y'=rsing. (Make sure you understand these.) Expanding our expressions for x and y and using the ones for x” and y’ we find x = rcosacosf—rsinasing = x’ cosa ~ y'sina y = rsinacosp + rcosasin§ =x’ sina + y' cosa, where we have used our expressions for x’ and y’ to express (1, y) in terms of (v’,y’) and the angle of rotation. EXERCISE 5-23 In our above rotation, find x’ and y’ in terms of x, y, and @. Now we return to our conic Ax? + Bry + Cy? + Dx + Ey + F = 0, where B # 0. Our problem now is to rotate the axes through some angle @ such that the resulting conic has no x'y/ term in (2, i’) coordinates. Using the above equations for x and y in terms of x’, y’, and a, we have AG cosa — y' sina)? + B(x’ cosa ~ y'sina)(x’ sina + y' cosa) + C(x’ sina + y’ cosa)? +D(x' cosa ~ y' sina) + E(x’ sina + y' cosa) +F =0.50 > CHAPTER 5. CONICS AND POLAR COORDINATES Since we want to get rid of the 2’y’ term, we only consider those terms which produce x’y/ terms. Combining these terms and setting their sum equal to 0, we have -2Ax'y' cosasina + Bx’ y'(cos” a - sin?a) +2Czx'y' cosasina = 0 xy (C - A)(2cosasina) + Bx'y'(cos*a-sin?a) = 0 x'y(C- A)(sin2a) + Br’y'(cos2a) = 0. Dividing by x'y’ and rearranging a bit, we find B cos 2a = (A—C) sin 2a, so cot2a = (A—C)/B. Hence, aas given by cot2a = (A —C)/B is the angle through which we must rotate the axes to eliminate the xy term of Ax? + Bry + Cy? + Dx + Ey+F =0. EXAMPLE 5-15 Through what acute angle(s) can the conic 3x” + 4xy ~ 4y* - 6 = 0 be rotated in order to remove the xy term? Solution: From the above discussion, we have cot2a = 7/4 for the angle a through which we must rotate the axes counterclockwise; hence, we must rotate the conic clockwise a to get rid of the xy term. Thus, one such angle is @ = (1/2) Cot™"(7/4) clockwise. We can also rotate the conic through an angle 90° ~ a counterclockwise to get rid of the xy term. The axis which becomes parallel to the x axis under a rotation of a clockwise will be parallel to the y axis upon a rotation of 90° - a counterclockwise. Make sure you see why! How can we tell if the general conic Ax? + Buy + Cy? + Dx + Ey + F = 0 is an ellipse, parabola, ora hyperbola? We already know how if B = 0. The conic is a parabola if A or C is 0; it’s an ellipse if AC is positive, and it’s a hyperbola if AC is negative. (Why?) If B # 0, we rotate the conic so that there is no xy term, just as above. For those of you with a yen for algebra, use our rotation method above to prove that if we rotate Ax? + Bry + Cy + Dx + Ey+F=0 Ax? 4 Bixty + Cy? + Dix! +E'y' + F =0, then B? — 4A’C’ = B? — 4AC, no matter what the angle of rotation is. This value, B? - 4AC, is called the discriminant of the conic. Suppose the rotated conic is such that B’ = 0. Hence, BP — 4AC = B? — 44’C’ = -4A’C’, Since our new conic has no xy term, it is an ellipse if A’C’ > 0, a parabola if A’C’ = 0, or a hyperbola if A’C’ < 0, Thus, if B? - 44C = -4’C’ = 0, the original conic is a parabola, if the discriminant is negative (so that A’C’ > 0), the conic is an ellipse, and if the discriminant is positive, the conic is a hyperbola. EXERCISE 5-24 Can every circle be described without an xy term?the ART of PROBLEM SOLVING: Volume 2 < 51 Problems to Solve for Chapter 5 70. Find the equation of a hyperbola with asymptotes y — 1 = +$(x - 4) and one vertex at 1). x + bx +c has vertex (4,2), If (2,0) is on the parabola, then find abc. (AHSME 1966) 71. A parabola y 72. Find the radius of the smallest circle whose interior and boundary completely contain the two circles with centers (0, 0) and (24,7) and radii 3, 4, respectively. (Mandelbrot #2) 73. A tiny bug starts at a point (x, y) on the graph of x2/9 + y2/4 = 1. It walks ina straight line to the point (— V5,0), then in a straight line to (-V5,0), and then in a straight line to its initial point. How far has the bug walked? (MA@ 1990) 74, Ifeach of two intersecting lines intersects a hyperbola and neither line is tangent to the hyperbola, then what are the possible numbers of places where the lines can intersect the hyperbola? (AHSME 1956) 75, Points A and B are selected on the graph of y = ~x7/2 so that triangle ABO is equilateral, where Cis the origin. Find the length of one of the sides of ABO. (MATHCOUNTS 1991) 76. A parabolic arch has a span of 24 feet. Its height is 18 feet at a point 8 feet from the center of the span. What is the height, in feet, of the arch? (MA@ 1992) 77. If the line y = mx + 1 intersects the ellipse x2 + 4y? = 1 exactly once, then find m?. (AHSME 1971) 78, Find the equation in rectangular coordinates of the curve whose polar equation is r = 2sec@ + cos 0. (MA@ 1987) 79. A circle has the same center as an ellipse and passes through the foci Fy and F of the ellipse. The two curves intersect in four points. Let P be any point of intersection. If the major axis of the ellipse has length 15 and the area of triangle PF;F) is 26, compute the distance between the foci. (ARML 1984) 80. A point P lies in the same plane as a given square of side 1. Let the vertices of the square, taken counterclockwise, be A, B,C, and D. Also, let the distances from P to A, B, and C, respectively, be u, v, and w. What is the greatest distance that P can be from D if u2 + 02 = w2? (AHSME 1983) 81. An ellipse is drawn with major and minor axes of lengths 10 and 8 respectively. Using one focus as the center, a circle is drawn that is tangent to the ellipse, with no part of the circle being outside the ellipse. Compute the radius of the circle. (ARML 1986) 82. The points of intersection of xy = 12 and x? + y? = 25 are joined in succession. What is the resulting figure? (AHSME 1956) 33. A circle rests in the interior of the parabola with equation y = x? so that it is tangent to the parabola at two points. How much higher is the center of the circle than the points of tangency? Mandelbrot #42) 7,Chapter 6 Polynomials 6.1 What is a Polynomial? In Volume 1, the equations of one variable we saw were usually no more complicated than quadratic equations. What happens when we introduce terms with higher powers than 2? This brings us to the general subject of polynomials. A polynomial is a function of the form F(R) = nx" + Oy-ax" +++ + Moy where the a; are called coefficients (any of these except a, can be 0) and n, the highest power of x in the polynomial, is the degree, written deg f. As the form suggests, 1 is always a nonnegative integer. Examples of polynomials in x are: O42x+5 At -284562-V2xt4 7-5. ‘The expressions below are not polynomials in x: x-1 werd Ve-7 ar-t log, x + sinx. Throughout this chapter, unless we specifically state otherwise, we are considering only polyno” mials with all rational coefficients. 6.2 Multiplying and Dividing Polynomials In working with polynomials we sometimes encounter expressions like 43x 44x44 2 2_ (2 +3r4 YQ? -3x+4) and —Z IF We can expand this first expression, the product of polynomials, using the distributive property Usually it is easiest to set up the multiplication just like when we multiply large numbers. A: example of this is shown below. «52>the ART of PROBLEM SOLVING: Volume 2 < 53 e+ Bxt 1 e- 3x4 4 42 +12x+ 4 (1) = 39-92 — 3x (2) x44 334 2 (3) A+ 0-4 + Ox 4 Here we have multiplied x? + 3x + 1 and x? ~ 3x + 4 by multiplying (x? + 3x + 1) first by 4 (line (1)), then by -3x (line (2)), then finally by 2? (line (3). Last, we add the results as shown; grouping the common terms in columns makes this easy. Multiplying the two given quadratics yields x* - 4x? + 9x +4. (There is no need to keep the 0x3 term). One pretty obvious result of polynomial multiplication is that for all polynomials f and g, deg(fg) = deg f + deg g. The proof of this is straightforward. If the degree of f is n and that of g is m, then the product will contain only one x"*" term and no terms of higher degree. Can you show that if deg f > deg g, then deg(f +g) < deg f? Just like multiplication of polynomials, division of polynomials can be done very much like long division. The best way to describe this is by example. x +1 @ Pse+1 | P+ 3b are = -22- x (2) e+ Bx 44 = 2-2-1 (3) x+3 (4) Let g(x) = x? + 2x41 and f(x) = x3 + 3x2 + 4x44, We divide the first term, x, of g(x) into the first term of f(x), 2°, yielding the x on line (1). We then multiply g(x) by x and subtract this product from F(x) as shown on line (2). Again we divide the first term of g(2) into the first term of the result of the subtraction, yielding 1. Finally, we multiply this quotient, 1, by g(x) and subtract the result as in line (). The result is line (4). Since we can’t evenly divide the leading term of g(x) into that of line (4), we are done. The quotient is x + 1 and the remainder, line (4), is x +3. Compare this process to long division of large numbers—it’s exactly the same, Don’t memorize the steps; understand the process of division. If'we let the remainder be r(x) and the quotient be q(x), we can write the above division as BB = 001+ 23, ox fa) = ate) +109, It is very important to note that the quotient and the remainder above are unique. That is, given g(2) and f(x), there is only one pair of polynomials (q(x), 7(x)) such that deg r < deg g and LC) = q(x) g(x) + r(x).54> CHAPTER 6. POLYNOMIALS EXERCISE 6-1 Prove that the quotient and remainder (q(x) and r(x)) are unique for each pair (fF), 8@)- ‘There is a special shorthand method called synthetic division for dividing polynomials by expressions of the form (x). To introduce synthetic division, we'll take you step by step through a problem which will be solved with both long division and synthetic division. Pay close attention not only to how to perform synthetic division, but also why it works. 2+ 3x + 2 @) x-1[ 2 4+22- x+3 (2) = +1? (3) 3x2 -— x +3 (4) = 3x2 + 3x 6) 2x +3 (6) =2r+2 (1) 5 @) Above we did the long division of x~ 1 into f(x) = x3 + 2x? -x+3. For synthetic division (shown below), we don’t write any xs. The 1 from the constant term of x ~ 1 goes to the left of the vertical line on line (9), The coefficients of f(x) are then copied into the remainder of that line, Line (11) represents the coefficients of the quotient. Clearly the first such coefficient is the first coefficient of *flx) (since the leading coefficient in (x ~1) is 1). Hence, we copy the first coefficient of f(x) in line (9) into line (11). Now we have to figure out how to get the rest of line (11). 12-13 177338 13 25 (11) Line (10) represents the subtractions at lines (3), (5), and (7) in the long division. In the long division, we get these by subtracting the product of the quotient and x ~ 1. Since the first term in the long divisions on these three lines always cancel, we are only interested in the second terms (the boldface coefficients). These results are from multiplying ~(~1) by the quotient (Line (1)). (The first negative comes from the fact that we are subtracting the products of the quotient and x ~ 1 on lines (3), (5), and (7),) The coefficients of the quotient are on line (11), so we get line (10) from multiplying line (11) and the 1 at the left of our vertical line. Finally, how do we determine line (11), the quotient? Since the leading coefficient of (x ~ 1) is one, the coefficients of the quotient are the coefficients of the leading terms resulting from the combinations of lines (2) and (3), lines (4) and (5), and lines (6) and (7). Note that these are just the ‘sums of the boldface numbers and the coefficients of the original f(x)! Hence, we get line (11) from just adding lines (9) and (10). Here’s how synthetic division works in action. We'll divide x ~ 2 into x° ~ 3x? + 7x + 4. First we copy the 2 from the constant term of x ~2 and the coefficients of x3 - 3x? +7x +4 into our table. Then, we copy the first coefficient into line (3): 1-37 4B 1 @)the ART of PROBLEM SOLVING: Volume 2 «55 We now get the first number in line (2) by multiplying the 2 to the left of the vertical line and the 1 in line (3). After this, we add the number in line (2) to the number above it to get the next coefficient of the quotient in line (3): 2)1-374 3 2 (2) 1-1 @) We continue by multiplying our 2 and the next term in line (3), ~1, to get the next term in line (2): E4tg 1-1 5 (3) Now we can finish off the problem by getting our last terms in lines (2) and (3): LAAB 1-1 514 @) So what's the answer? The last line gives us the coefficients of x2 — x +5, but what's the 14 for? Compare synthetic division to long division and you'll find that 14 is the remainder, so the above synthetic division tells us that 2-37 +744 =Y—x454 4, x-2 x-2 There are a couple of important points to remember when doing synthetic division. First, it only works when we are dividing by a linear polynomial (x ~ a). Second, the leading coefficient of this linear term must be 1. (Look at our development of synthetic division to see why we can’t use synthetic division with a linear coefficient other than 1.) Finally, in synthetic division the term to the left of the vertical line is the negative of the constant term of the linear divisor. For example, in the above problem where we divided x - 2 into x3 - 3x? + 7x + 4, we put 2, not -2, at the left of the vertical line. EXAMPLE 6-1 Use synthetic division to determine (8x4 — 12x? + 2x + 1)/(2x + 1). Solution: First, we must make the coefficient of x in the divisor 1. Hence, we divide the numerator and denominator by 2 to get 4x4 63 +x 41/2 x+1/2 -1/2| 4 0 11/2 4-2 1/2 Now we do our synthetic division: ©56 > CHAPTER 6. POLYNOMIALS (Why is there a 0 in the first line above?) Thus, we find at - 68 +x 41/2 3 1 3417 = 4 81 + 4x-1+ py Bxt- 123 +2041 _ 2 2 ea 8H +ar-1+ EXERCISE 6-2. Use synthetic division to divide x +3 into x° + 3x4 + 2x9 - x? +x—7. 6.3 Finding Roots of Polynomials Suppose we are given the polynomial f(x) and asked to find the solutions to f(x) = 0. We call these solutions roots of the polynomial. Unfortunately, no quick and easy method like the quadratic formula exists to solve general polynomials. Instead we must go searching for the roots. Does this mean that we just have to keep guessing values for x until we find one for which f(x) = 0? And how will we know if we've found all such x? Fortunately, we are not completely consigned to guessing. We do have some helpful hints to guide our way. First, if is a root, then (x—a) divides f(x) evenly; thatis, there is no remainder when we perform the division. To see this we write £02) = (= a)q(x) + 12). Since deg r(x) < deg(x ~ a)=1, deg r(x) = 0, and r(x) is some constant c. Letting x = 4 gives fla) = (@-a)h(a) +e = If f(@) = 0, we have c = 0, and thus there is no remainder when we divide f(x) by («~ 4). EXAMPLE 6-2. Prove that the remainder upon dividing f(x) by x — ais f(a). Solution: As above, we write f(x) = (@— @)h() + 1). Since degr(x) < deg(x ~ a), r(2) is a constant r. Letting x = a gives f(a) = 1, so the remainder upon dividing f(x) by xa is f(@). Therefore, we can use synthetic division to determine f(a) by finding the remainder when f(x) is divided by (x ~ a). EXAMPLE 6-3 P(x) is a polynomial with real coefficients. When P(x) is divided by x ~ 1, the remainder is 3. When P(2) is divided by x — 2, the remainder is 5. Find the remainder when P(2) is divided by x? - 3x +2. (MA@ 1990) Solution: We write P(x) = (x? — 3x + 2)q(x) + 702),the ART of PROBLEM SOLVING: Volume 2 < 57 where r(x) is the desired remainder. Since deg r(x) < deg(x? - 3x + 2), we can write r(x) = ax +b for some constants a and b. From the given information, we know P(1) = 3 and P(2) = 5. Since x? 8x42 =O for x= 2 and x= 1, we put these values in our equation for P(2), yielding ()q(l) +1(1) = a+b = PA) =3 (O)q(2) + r(2) = 20 +b = P(2) =5. Solving this system, we find (a,b) = (2,1), so the remainder is 2x+1. Remember this method of cleverly choosing values for x in polynomial equations; it can be very useful! The Fundamental Theorem of Algebra states that every nonconstant polynomial has at least one root. Thus, there is at least one value a such that f(a) = 0. This a may be real, imaginary, rational, or irrational, but the Fundamental Theorem of Algebra assures us that at least one such root exists, Unfortunately the proof is a bit too complex for this text, but we shall put the theorem to good use by showing that any degree m polynomial has exactly 1 roots. This means we can write any polynomial f(x) as F(X) = yx" + agg) + +X + 9 = An(x — 1y)(X = 12) +++ (X= TH). The rare the roots of the polynomial and they are not necessarily real or rational. It should be clear why f(r) = 0. To show that all polynomials can be written in such a fashion we invoke the Fundamental Theorem of Algebra. By this theorem, we know that for some number r; we can write FQ) = &~r1)n@). Since deg f = n = degl(~rs)qu(x)] = deg(x ~r) + deg qi, we find deg q1 = n— 1. Now we apply the Fundamental Theorem to i(x) to get SQ) = (&-n)Ye-r2)q2e), where degqz = n- 2. Thus, we can continue applying the Fundamental Theorem until finally we have the desired factorization F(R) = an x — Ye 12) + Showing the roots exist is one thing; finding them is another thing altogether. Rather than provide a recipe-like formula, the best we can do is give a batch of methods to guide us to the roots, For the rational roots of a polynomial, there is a method we can use to narrow the search. Although there are infinitely many rational numbers we could guess as roots of f(x), the only ones which have a chance of being roots are given by the Rational Root Theorem. For any polynomial F%) = px" + px) + 49 with integer coefficients, all rational roots are of the form p/q, where |p| and ql are relatively prime ntegers, p divides ag evenly, and q divides a, evenly. The Rational Root Theorem will be proven as in example on page 58,58 > CHAPTER 6. POLYNOMIALS EXAMPLE 6-4 Find all the roots of x9 — 6x? + 11x - 6. Solution: From the Rational Root Theorem, we know that all possible roots are of the form p/q, where p divides -6 and q divides 1. Thus the possible roots are (#1, +2, +3, +6). If we substitute these in the polynomial, we find that {1,2,3) all satisfy f(x) = 0, so these are the three roots of the polynomial. (How do we know there aren’t any more?) EXAMPLE 6-5 Find all the roots of 2x? - 5x? + 4x - 1. Solution: Once again we apply the Rational Root Theorem and determine that the possible roots are (1/2, #1}. Trying these, we find that both 1 and 1/2 are roots of the polynomial. We know that there must be one more (why?), but we also know that no other rationals could possibly be roots, We might think that the third root is irrational or perhaps imaginary, but as we will see, no polynomial with rational coefficients can have just one irrational or one imaginary root. Thus, we come to the conclusion that this polynomial must have a double root, just like quadratic expressions which are perfect squares, such as x° + 2x +1. Indeed, in this problem, we can use synthetic division to find (2x° - 5x2 + 4x — 1)/(x - 1) = 2x7 - 3x + 1. Factoring this quadratic, we find =(@-1P@x-2), so that the root x = 1 is a double root, meaning the factor (x ~ 1) occurs twice. 2x8 - 5x? + dx — We have already come across two shortcomings of using the Rational Root Theorem alone. One is that we will miss multiple roots. Another is that it could still end up taking a very long time, as there are many numbers for polynomials like 12x* ~ x ~ 60 which satisfy the Rational Root Theorem criteria. ‘To avoid missing multiple roots and to shorten our search for the roots, when we find a root r1 of the polynomial, we divide (x — rn) into f(x), as, Fe) = @— rag). Then, we continue our search for roots with (x), because all roots of 4(x) are also roots of f(x). As we saw in the previous section, synthetic division provides a swift method for performing the division EXAMPLE 6-6 Prove the Rational Root Theorem. Proof: Let p/q be a rational root of the polynomial f(x), where p and q are relatively prime positive integers. The case where the root is ~p/q is virtually the same. Since p/q is a root, we have " wt f (pia) =«,(2) tana () tea = 0. Multiplying by q” gives agp" + Oq-ap" tq +++ + apg" * + oq" = 0. Now look at this equation modulo p. The first terms on the left will become 0 since they an multiples of p, so we have agp" + ayap"g +--+ apa? + aod 4 +++ +0 +99” (mod p) = 0 (mod p)the ART of PROBLEM SOLVING: Volume 2 < 59 Thus, aoq" = 0 (mod p), so plaoq”. Since p and g are relatively prime, it follows that plao. By the same argument, we can evaluate the sum mod q to show that gla,p". Thus gla, and the proof is complete. There are a few more guides to tell us where to look for roots. The first is Descartes’ Rule of Signs, which gives us a method to count how many positive and how many negative roots there are. We do this by counting sign changes. The number of sign changes in the coefficients of f(x) (meaning we list the coefficients from first to last and count how many times they change from positive to negative) tells us the maximum number of positive roots the polynomial has, and the number of sign changes in the coefficients of f(—-x) gives us the maximum number of negative roots the polynomial has. Hence, for f(x) = 3x8 + 2x4 - 3x? + 2x = 1, there are at most 3 positive roots and at most 2 negative roots (since f(-#) = -3x8 + 2x4 32 -2x-1), Furthermore, the actual number of positive or negative roots will always differ by an even number from the aforementioned maximum, so our above f(x) has 1 or 3 positive roots and 0 or 2 negative roots. Another root location method is finding upper and lower bounds. Suppose we use synthetic division to find f(x)/(x — c) where f(x) has a positive leading coefficient and c > 0 as below: 3] 1-12 3.6 2 28 1 3K SRRa If all the resulting coefficients in the quotient are positive (including the remainder), as in the example above, then no roots are greater than c. (Why?) This c is called an upper bound on the solutions since no roots can be higher. Similarly, if c < 0 and the coefficients of the quotient and remainder alternate in sign, then there is no root smaller than c (which we then call a lower bound for the roots). Locating upper and lower bounds will often help you shorten your search for roots. Lastly, recall from our discussion of quadratic equations in Volume 1 that complex roots and toots of the form a + b yc come in pairs if the coefficients of the quadratic are rational. This is also true of any polynomial with rational coefficients. For example, if the complex number z = a + biis a root of f(x), we have S] = Oz” + yz" + +++ + 012 +09 = 0. Now we use some of our useful properties of complex numbers, such as OFZ = 0 +2, = @)*, and w =z implies @ = z. Applying these principles to f(z), we have Fo baz tap te AE + mH OZ" + Oyarzo aE tag = eca Hence, if f(z) = 0, then (2) = 0, so Zs also a root. This proof, with slight modifications, can be used to show that if z = a +b y¢ is a root, then z = a ~b Y¢is also a root.60 > CHAPTER 6. POLYNOMIALS EXAMPLE 6-7 Find all of the solutions to the equation x! ~ 10x? + 35x? - 50x + 24 = 0. Solution: Since the signs of the coefficients of f(~x) are all positive, none of the roots are negative ‘This cuts our search in half. Now we use the Rational Root Theorem to deduce that the roots are all factors of 24. (Why?) We'll start with 1 (usually the best place to start). Synthetic division yields 0 35-50 24 1-9 26-24 1 -9 26-24 0 Since there is no remainder, x ~ 1 is a factor. Now we continue our search, not with x* ~ 10x + 35x? — 50x + 24, but with the quotient above, since of = 10x? + 35x2 ~ 50x + 24 = (x - 1)(0? - 92? + 26x ~ 24), Continuing in this manner we find that x = 2 is also a root and we have of —10x3 + 35x? - 50x + 24 = (x — 1)(@ ~ 2)(0? - 7x + 12). Factoring the quadratic yields (x ~ 1)(2 - 2)( ~3)(x - 4) = O and our solutions are 1, 2,3, and 4 EXAMPLE 6-8 I'm trying to find the roots of (2) = 2x4 — 15x? + 15x? + 20x ~ 12. Istart with x = 1 After finding f(1) = 10, what should I try next? Solution: Since f(0) = ~12 and f(1) = 10, there must be some number c between 0 and 1 such that f(c) = 0, because f(0) and (1) have opposite signs. (Graph y = f(2), noting that the points (0,12) and (1, 10) are on the graph, to see why there's a root between x = 0 and x = 1.) From the Rational Root Theorem, the only possible rational root between 0 and 1 is 1/2. Using synthetic division we find that this indeed works. Using this ‘location principle’ we can zero in on roots. Namely, if f(a) and f(b) have opposite signs, then there is a root between a and b. EXERCISE 6-3. Find the roots of x4 + x3 + 2x? + 17x - 21. EXERCISE 6-4 Given a quartic polynomial with rational coefficients and roots 3 éand 4 + V2, find the other two roots. 6.4 Coefficients and Roots Suppose we are asked to find the sum of the roots of a polynomial. We could just find the roots and add them all, but that may not be easy to do and could take a long time. Of course, there is a better way. For example, remember from Volume 1 that the coefficients of a quadratic are directly related to the sum and product of the roots.the ART of PROBLEM SOLVING: Volume 2 < 61 ‘The coefficients of a polynomial tell us much more than just the sum of the roots. To see this, let the roots of f(x) = x3 + apx? + ayx + a9 be 71, 7, and r3, so we can factor f(x) as f(2) = (=) 12)(a 13). If we multiply this out we get (check and see) FQ) = (1 +1 + 15)2? + (are + ors + TB)x NNW. By comparing this with f(x) = x° + azx? + 1x + ao, we see that the coefficients not only give us the sum of the roots (a2), but the product of the roots (~aq) and the sum of the products of the roots taken two ata time (a; = ryrp + r2r3 + T1753). Now, what if the leading coefficient of the cubic is something besides 1? Like we did with quadratics, we change the problem to one involving a monic polynomial, ie. a polynomial with leading coefficient 1. The roots of F(X) = 333 + agx? + ayx + a9 =0 are the same as those of ge) = 2 Br My Mg 3 a3” * a3" * as since if f(z) = 0, then g(x) = f(x)/as = 0. Thus, the sum of the roots of g(x), and therefore f(x), is @/a3, the product of the roots is —a9/a3, and so on. Now a quick definition and we will be ready to use these results on any polynomial, not just cubics. Suppose our polynomial has 1 roots. We define the sum of all products of the roots taken k at a time, or the kth symmetric sum, as the sum of all products formed by multiplying k of the n roots. Thus, if we have 4 roots which are 1, 1, 2, and 3, the second symmetric sum is 1-141-241-341-241-34+2-3=17, For the general polynomial f(x) = a,x"+----+ap, the kth symmetric sum of the roots is (—1)ay_4/an. We can prove this through algebra much like our n = 3 case above. The proof is made rigorous through induction, EXERCISE 6-5 Use induction to prove the assertion that the kth symmetric sum is (—1)kay_¢/ap. EXAMPLE 6-9 Find the constant term of a monic quartic polynomial with rational coefficients that has two roots equal to 2—iand 2+ v3. Solution: The other two roots are 2 + i and 2 ~ V3, so the product of the roots is (2 + i)(2 — (2 + VB)2~ V3) = (6)(1) = 5. The constant term of a quartic is equal to (~1)* = 1 times the product of the roots, so the constant term is 5. EXAMPLE 6-10 If three roots of x* + Ax? + Bx + C = 0 are -1, 2, and 3, then what is the value of 2C = AB? (Mae 1992)62 > CHAPTER 6. POLYNOMIALS Solution: Since the coefficient of x} is 0, the sum of the roots is 0. Thus the fourth root is —4, Hence (12) + 1B) + HY(-4) + 2B) + 2-4) + GY(-4) = ~15 B= -[(-1)2)G) + (-YQ(-4) + NEY-4) + EVA] = 10 C= (-1)2)8)(-4) = 24, so 2C - AB = 198. EXERCISE 6-6 Find the largest solution of x° - 27x? + 242x ~ 720 = 0 given that one root equals the average of the other two roots. (MA 1990) 6.5 Transforming Polynomials Through the following examples, we will examine how to transform polynomials in various ways. EXAMPLE 6-11 Find the polynomial whose roots are the reciprocals of the roots of x* 3x? + x —9. Solution: Once again, let the given polynomial be f(x) and the roots be r:, 72, 73, and rs. One equation whose solutions are the reciprocals of these is just f(1/x) = 0 because 1 I (qa) =f =0 ‘Thus, the solutions of f(1/x) = 0 are the reciprocals of the roots of f(x), as claimed. Unfortunately, {f(1/2) is not a polynomial. On the other hand, the function given by g(x) = x4 f(1/x), which also has roots 1/r;, is a polynomial: 9) = ot 429-32 +1. This g(x) is our desired polynomial. Now compare g(2) to f(x). Look closely and you'll see that the coefficients of g(x) are the same as those of f(x) in reverse order! We can prove that this is true in general in the same way. Let the general polynomial f(x) be f(x) = ay" + dy-1x""* +> + ao with roots r,...,Tm The solutions to f(1/x) = O are x = 1/n, 1/ta,-- +, 1/n, because again we have I (Gg) fe ‘Thus, the solutions of f(1/x) = 0 are the reciprocals of the roots of f(x). The desired polynomial is then 0. g(x) = x"f(1/2) =x" (@ + | ot * +a) = age" + ax" 1 + Oy or the original polynomial with the coefficients reversed.the ART of PROBLEM SOLVING: Volume 2 < 63 EXAMPLE 6-12. Find a polynomial whose roots are twice those of f(x) = x4 — 3x2 +x -9. Solution: If we are given a polynomial f(x) = ayx” + dy-ax"~1 +--+ + ao, a polynomial with roots which are k times the roots of f(x) is f(x/X). (Let the roots of f(x) be ri, 12, ..., Ta; then for = kn, kr2,..., kty, we have f(x/k) = f(r) = 0.) Hence, the desired polynomial is 4, vt a Ger tt Eta, Slajk) = + Multiplying both sides by k" to simplify the expression (i.e. to get rid of the fractions), we have BOR) = RF (XR) = gx” + Kay gx + tae + Rag as a polynomial with roots kr;. We form g(x) by multiplying the coefficients of f(x) in turn by 1, k, , ..., K*, Hence, one answer to our problem is B(x) = x4 — (2°)(3)x? + (23)x — (24)(9) = x* - 122 + Bx - 144. EXAMPLE 6-13 Find the polynomial whose roots are half the reciprocals of the roots of 5x4 + 123 + 8x? — 6x -1. Solution: Let the roots of this polynomial be a, b, c, and d. We seek the polynomial whose roots are 1/2a, 1/2b, 1/2c, and 1/24. The polynomial whose roots are 2a, 2b, 2c, and 2d is B(x) = 5x4 + 12(2)x3 + 8(2?)x? — 6(23)x - 1(24) = 5x4 + 24r3 + 32x — 48x - 16, The polynomial we desire is the one whose roots are reciprocals of the roots of g(x), or h(x) = -16x4 ~ 48x3 + 32x? + 247 +5. EXAMPLE 6-14 Find an equation whose roots are 3 greater than those of x4 ~ 3x3 — 3x2 + 4x - 6. Solution: Let the given polynomial be f(x) and the roots be rj, r2, 73, and r4. In the spirit of the examples involving reciprocals and multiples of roots, consider the polynomial g(x) = f(x ~ 3). We have 8%; +3) = fri +3-3) = f(r) = so the roots of g(x) are 3 greater than those of f(x). Hence, g(2) is the desired polynomial and our answer is 8(2) = fle—3) = (x-3)4-3@-3)? — 3-3) + 4(x-3)-6. Similarly, we can show that the polynomial whose roots are k greater than those of a general polynomial h(x) is h(x — k). However, the above expression for g(x) will take quite a bit of time to evaluate, so it is useful to find a swifter method if possible. We will determine f(x — 3) term by term. First we find the constant term, which is just the polynomial evaluated at x = 0, or f(-3). (Why?) Now we must find the coefficient of x. This is64 > CHAPTER 6. POLYNOMIALS tricky. If we subtract the constant term from a polynomial and divide the result by x, the constant term of the new polynomial is the coefficient of x in the original. This is shown for f(x) below: [Lf@) - fO]/x = [xt 3x9 3x? + 4x - 6 - (-6)] x = 9 - 3x? - 3x + 4. Now we need a swift way to find (f(x — 3) — f(-3)) /xin order to get the coefficient of x in f(x ~ 3). Remember that we can write f(x) as F(R) = (& + 3)qu(x) + f(-3) for some polynomial q:(x). This leads us to our short cut: f(a -=3) = (X34 3)qu(x—3) + f(-3) = aque ~ 3) + f(-3), or f@=3)- fC3) x Since we want the constant term of qi(x — 3), we want q:(-3) (since setting x = 0 eliminates all terms except the constant term). Now our problem is determining qi(x)- Synthetic division of f(x) by (x +3) gives us this polynomial, and synthetic division of q:(x) by (x + 3) gives us the desired remainder q;(~3). By the same argument as above, to find the coefficient of x? we divide q(x ~ 3) — qu(~3) by x and find the constant term of the resulting polynomial. Once again, synthetic division of qu(2) by x + 3 can be used to find this polynomial and constant term. Since after each synthetic division, the resulting quotient is used for the next synthetic division, we can just ‘stack’ our divisions as below: u(x ~ 3). -3| 1-3-3 4 -6 =3_18 -45 123 -3) 1-6 15 27 = a2 36 -3 78 -3 Our desired coefficients, then, are the boldface remainders above, so the polynomial f(x ~ 3) is af — 15x3 + 78x? — 167x + 117. What we have done here to describe this method is not a complete proof, but we hope it gives you a clear idea why this ‘trick’ works. While this method is somewhat quicker and more reliable for higher degree polynomials, it is also easy to forget. It is most important to remember that the polynomial whose roots are k more than those of polynomial f(x) is always given by f(x — f) EXERCISE 6-7 The roots of f(x) = 3x3 - 14x? + x + 62 = O area, b, and c. Find the value of (Mae 1991)the ART of PROBLEM SOLVING: Volume 2 < 65 WARNING: While you could just memorize the three methods to solve the three general types of problems described above, it is much more important to understand why these methods work, because when you forget the ‘trick’ you'll be able to arrive at the solution without it. Don’t waste too much time memorizing; once you've done a few problems with these methods, you'll have committed them to memory anyway. 6.6 Newton’s Sums Givenx+yand xy, how would we find x°+4?? As we saw in Volume 1, we write x2+y? = (x+y)? '—2xy. Let's try a tougher one. Write x° +? in terms of x? +y?, x+y, and xy without squaring or cubing any of these expressions. We can only get x° + y° from the product (x + y)(x2 + y2) = 29 +P +2y2 + wy. Hence, we have PHP alee Merry) —xyP— xy = (x4 ers P)— rye +9) Now let x and y be the roots of the quadratic a2? +a)2+a) = 0 and ; = x*+y*. Our above expression can then be written as = My _ M0, 83 = ——8)- 5) a a ~a) /az and xy = ag/az. Rearranging this, we can write since x+y = 283 + 152 + 098; = 0. This nice form suggests that other similar relationships may be true as well. EXAMPLE 6-15 Show that if 5, is the sum of the kth powers of the roots of a3x° + a,x? + ax + a9, then 4382 + 495; + 2m = 0. Proof: Let § = a3(s2 + a281/a3 + 2m /a3). We wish to show that $ = 0. Using our relationships between the roots of the polynomial, which we calll r,s, and t, and its coefficients, we have S = w((P+%+P)+((+s+Hr+s+8) +2(r5 + rt +st)) a3 ((7 +5? +P) ~(P +3? +P + Drs + Ort + 2st) + (2rs + Qrt + 2st)) 0. Perhaps you see where we're going with this. The family of equations which relates the sum of the mth powers of the roots of a polynomial to the coefficients of the polynomial as we've done above iscalled Newton’s sums. If we let sn be the sum of the mth powers of the roots of f(x) = aqx"-+----+a9, then the Newton’s sums can be written as yS1 + dy-1 = 0 yS2 + Oy-181 + 2q-2 = 0 983 + dy-182 + On-281 + 3-3 = 0 Ay + On 153 + On282. + On-381 + Ady4 = 0 and so on.66 > CHAPTER 6. POLYNOMIALS EXERCISE 6-8 What happens to the Newton's sum ays + 4n-15k—1 + ‘1* + Ky = 0 when n < k? EXAMPLE 6-16 Find the sum of the cubes of the solutions of x* ~ 3x +3 = 0. Solution: We use Newton's sums: 51 + (-3) = 0, 08 = 3; 52 + (-3)s1 +2) = 0, 5052 =3. Now, in the next Newton sum, we have a term 32.1, but there is no a1, $0 this term is just 0. We find 53 + (-3)s +351 = 0, 8053 = 0. ‘This is much easier than cubing the solutions to the quadratic. EXERCISE 6-9 Find the sum of the cubes of the roots of 2x4 + 3x3 + x? - 4x ~ 4. ‘As we've seen, Newton's sums are just a result of algebraically manipulating expressions in- volving the roots of polynomials. The Newton's sums equations can be proven in general using the same algebraic techniques as above. Those of you very comfortable with summation notation and manipulation should try to do so. The leading term in every Newton's sum is a,5,. We present here a less algebraic proof for all Newton’s sum equations in which k 2 » because it involves a very important problem solving technique. Let the roots of the polynomial f(x) be r1, 72,..-, fu» Since these are solutions of the equation f(x) = 0, for each r; we have F(t) = aur + an ar? +--+ + ay = 0. Multiplying each of these equations by ri” yields ath + ay a oo haar Aah dy gt fP ooo ag Oath + Oya7kD + + gr We can add all of these equations, which gives us fig (Ab) ages (PEE oes Doo bag (PM oe tet) = 0. Thus, we find BpSk + An—18¢-1 + °° + AOSk-n In the special case where k = n, we have Spon 90= M+ Bt ehaleltetlan, so the Newton sum is nS Oy-aSq-1 ++ fdg = 0.the ART of PROBLEM SOLVING: Volume 2 4 67 1000, + rite EXAMPLE 6-17 If ri, 72,..., Y1009 are the roots of x! — 10x + 10 = 0, find 71000 + 71000 4... Solution: Since only aio99, a1, and ag are nonzero, we can write the 1000th Newton sum as 1000 ~ 10s; + 1000(10) = 0. Since the coefficient of x is 0, 3; = 0 and syoq9 = -10000. Problems to Solve for Chapter 6 84. Find the remainder when 219 + 1 is divided by x1. (AHSME 1950) 85. Find all the roots of 2y* — 9y° + 14y? + 6y - 63 = 0. 86. Find all values of m which will make x + 2a factor of x3 + 3mx? + mx +4. (MA@ 1991) 87. Find the product of the nth roots of 1. (MA@ 1991) 88. The equation x* ~ 16x° + 94x? + px +q = 0 has two double roots. Find p +9. (MAO 1991) 89. Let f(x) = ax’ + bx® + cx — 5, where a, b, and care constants. If f(-7) = 7, then find f(7). (AHSME 1982) 90. For nonzero constants ¢ and d, the equation 4x3 — 12x? + cx +d = 0 has two real roots which add to give 0. Find d/c. (MA@ 1991) 91. An equation with roots 3+ V2, 3~-V2, ~3+i V2, and -3—i y2is in the form x4+Ax3+Bx2+Cx+D =0. Find A +B + C+ D. (Mae 1991) 92. Polynomial P(x) contains only terms of odd degree. When P(2) is divided by x3, the remainder is 6. What is the remainder when P(2) is divided by x2 - 9? (Mae 1991) 93. Let PUA) = gx dy a1 4 tay + ag, where the coefficients a; are integers. If p(0) and p(1) are both odd, show that p(x) has no integral Toots. (Canada 1971) 94. If.x4 + 4x9 + 6px? + gx + ris exactly divisible by x3 + 3x2 + 9x +3, then find (p-+q)r. (AHSME 1965) 95. Let r, s, and t be the roots of x3 - 6x? + 5x—7 = 0. Find 11,1 prate (Mae 1991) 96. Suppose x = a + biis a solution of the polynomial equation caz* + icg2? + cz” + icyz + cp = 0,68 > CHAPTER 6. POLYNOMIALS where co, C1, C2, ¢3, C4, a, and b are real constants and # = —1. Show that -a + bi is also a solution. (AHSME 1982) 97. If q;(x) and 7; are the quotient and remainder, respectively, when the polynomial x* is divided by x + 4, and if qa(x) and rz are the quotient and remainder, respectively, when g(x) is divided by x+ 4, then find rz. (AHSME 1979) 98, Solve the equation (x + 1)(x-+2)(x+3)(x +4) = 1. (M&IQ 3) SA __ 99, Let (1+ x + 22)" = ay + myx + agx? +--+ day be an identity in x. Find ag + a2 + a4 +--+ dan in terms of n. (AHSME 1966) 100. Give the remainder when x2 — 1 is divided by x4 — 1. (MA@ 1991) ‘S~___ 101. Given the equation (2 -3x-2)' - 30? -3x-2)-2-x=0, prove that the roots of the equation x? - 4x ~ 2 = 0 are roots of the initial equation and find all real roots of the given equation. (Bulgaria 1993) S—__ 102. Let bea positive integer. Find all polynomials with real coefficients which satisfy the equation P(P@)) = [PI (Canada 1975) {_ 108. Ifa, c,d are the solutions of the equation x* ~mx—3 = 0, then find the polynomial with leacling coefficient 3 whose roots are atbtc atbtd atctd 4 bterd pe @ (AHSME 1981) 104. For n > 1 let ai, az,..., dy be n distinct integers. Prove that the polynomial f(x) = ( ~ a1)( = a) an) — 1 cannot be written as g(x)h(x) where g and ht are nonconstant polynomials with integer coefficients. (MOP)Chapter 7 Functions 7.1 The Inverse of a Function A function is a machine. It takes one thing in, and outputs something else. But what if we ran this in reverse? If we cram something into the output slot, is the machine flexible enough to give us back the input which would create that output when run forwards? The inverse function toa function f(x) is anew function g(x) which “undoes” f, so that g(f(x)) = x. In other words, if you put an input x into f, then put the output, f(x), into g, you will get back x—the original input. EXAMPLE 7-1 Prove that if g is the inverse of f, then f is the inverse of . Solution: Consider some x in the domain of f, so that f(x) = y for some y. By the defini- tion of the inverse, we have g(y) = x. Substituting this for x, we have f(g(y)) = y. Since the range of g is the domain of f, we don’t have to worry about g(y) not being in the domain of f; thus, since F(g(y)) = y holds for all y in the domain of g, f is the inverse of g. The inverse of the inverse is the original function. EXAMPLE 7-2 Let's find the inverse g of the function f(x) = x/(1 + x). Since g is the inverse of f, we have from Example 7-1 that f is the inverse of g, so that F(g@)) = g@)/[1 + gx)] = x. Solving the second equality for g(x), we obtain (x) = x/(1- 2). EXERCISE 7-1 Find the inverse function of f(x) = ¥&. ‘The method of Example 7-2 can be used to find the inverse of many functions. The inverse of a function f(x) is denoted by f-1(x). To be really perverse, we can iterate the inverse function, as we iterated functions in Volume 1: f-1(f-(x)) = f-2(x), and so on. EXAMPLE 7-3. The “composition exponents” can be manipulated in some of the same ways as normal exponents. For example, (Po fe) = FFL E EKKO = £70): < 69>70 > CHAPTER 7. FUNCTIONS that is, we can add the “exponents.” As we warned in Volume 1, though, don’t let these similarities confuse function composition with exponentiation. EXERCISE 7-2. To what should f(x) correspond? Is it equal to [f()]"? Does every function have an inverse function? The inverse, if it exists, must itself be a function. Consider the function f(x) = x2. The inverse is found by setting x = (g(x))?, so that g(x) = + Vx. But this is not a function! Why? Consider the input x = 4; (4) could be either 2 or ~2, but a function can have only one output. Thus x2 does not have an inverse function. In general, no function can have an inverse function if it assigns two different x values to the same y, because in the inverse function, that value y won't know which of the two 2's to go to. A function which has an inverse function takes different x values to different y values, and is thus called a one to one function, which is often written 1:1. One way to see if a function has an inverse function is to graph it. If any horizontal line crosses through the function at more than one point, then there is a y which can be generated by two different x's, and the function cannot be one to one. EXERCISE 7-3 Which are 1:1? i. f= ii, g(x) = bel iii, A(x) = Lx] iv. jl) = 2/2 ‘The most interesting thing about inverses comes last. What happens when we draw the graphs of a function and its inverse on the same axes? Because the one takes x to y, and the other takes y to x, the graph of the inverse is, exactly the original graph with the axes reversed. In practice, this means that the graph is flipped over the line x = y to form the graph of the inverse. Examine the picture at right to see this graphically, then try graphing some yourself to get a feel. 7.2 Functional Identities ‘An important thing to consider in some problems involving functions is the identities they might satisfy. For example, the logarithm f(x) = log x always satisfies fley) = fe) + FY), since log(xy) = log x + log y for any positive x and y. In our study of trigonometry we have already encountered some other functional identities, though they were not identified as such. If we let f(x) = sinx and g(x) = cosx, then two important trig identities can be expressed as Fe + 9) = FH) + FB) and [F@P + [g@P =1.the ART of PROBLEM SOLVING: Volume 2 < 71 EXERCISE 7-4 Verify these two trigonometric identities. EXERCISE 7-5 Which of the following identities are satisfied by f(x) = [xl? i fey) = fofy) ii, fe+y) = f@) +f) ii, f(f@) =x EXERCISE 7-6 Does the floor function (greatest integer function) f(x) satisfy f(nx) = nf(x) i, ifboth m and x are integers? i. if m is an integer but x is any real? iii, if m and x are any reals? EXERCISE 7-7 Find some identities which are satisfied by f(x) 7.3 Solving Functional Identities We have seen how some functions satisfy interesting identities, but the real trick is to go backwards — given only the identity, to find the functions which satisfy it. There are many general techniques for this. 7.3.1 Isolation The method of isolation is exemplified in solving an identity like f(z) = xfly). In cases like this, we can bring all expressions involving x to one side and all those with y to the other, converting the given expression to £@) _ fo x yo We can now define a new function g(t) = f(é)/t; then we have 8(x) = sy) for any x and y. Clearly this can only happen for alll pairs (x, y) if g(x) is a constant, say c. Thus mi so f(x) = cx, for any constant c, is the family of solutions. 7 WARNING: Once we have shown that all solutions must be of the form f(x) = cx, we also need oe to test to show that every function of this form is a solution. To do this we go back to the defining identity yf(x) = xf(y) and substitute in the functional form. Then f(x) becomes cx and f(y) becomes cy, making our relation ycx = xcy, which is always satisfied. Thus any function f(x) = cx does the job.72 > CHAPTER 7. FUNCTIONS 7.3.2 Substituting in Values A surprising amount of information can often be obtained by substituting in values for the variables For example, consider the general functional identity Fry) = xf(y). Substituting in the value y = 1 yields f(x) = xf(1). Letting f(1) = c since f(1) is a constant, all the solutions to the identity are given by f(x) = cx. EXERCISE 7-8 Are all functions f(x) = cx solutions of f(xy) = xf)? ‘As another example of the power of substitution, consider the seemingly complicated fla + y) + fe—y) = 22 + 2y7, Substituting y = 0 immediately gives 2f(x) = 2x2, or f(x) Does this work? We have f(x+y)+ f(x—y) = (x+y)?+(x-y) as desired, so f(x) = x is a solution. It is unique. as the only candidate for a solution. PH2xy tyr tx?—Dayty? = 2x7 +2y", EXAMPLE 7-4 The previous example differs from earlier ones in that there is no loose constant; f(x) = x2 is the only solution. In an earlier example where f(x) = cx, there was instead an infinite family of solutions: some examples include f(x) = x, f(x) = 100x, f(x) = ~mx, and so on. EXERCISE 7-9 Find all solutions to the equation fe+y) + fle y) = 20 ~2y. 7.3.3 Using Cyclic Functions A cyclic function is a function g(x) such that Bg gH) =x (71) fot some number of nested g's. For example, g(x) = 1/x is cyclic because (g(x) = g(1/x) = 1/(/2) = x. The number of nested g’s in (7-1) is called the order of g; for example, the order of 1/x is 2. EXERCISE 7-10 Which of the following are cyclic? Of what order?the ART of PROBLEM SOLVING: Volume 2 <1 73 How do cyclic functions help with solving functional identities? Consider one like F@) + 2f(U/a) If we substitute 1/x for x, we get a new equation, S(/x) + 2f() = 1/x. Subtracting the first equation from twice the second to eliminate f(1/2) yields 3f(2) = 2/x—x, so the only possible solution is, 2 x fG)= 3-3 Substituting this into the original equation shows that this is in fact a solution. We have used the fact that 1/ is cyclic to help us find the solutions. 7.34 Arbitrary Functions ‘We have seen solutions to functional equations which were unique, and some which depended on an arbitrary constant. However, the solutions to some functional equations are much more general. Consider the functional equation fs *) _ ( (4) =r If we notice that the equation becomes Thus, if we create a new function g(x) = f(x) ~x, we have the simpler equation lta 1-#@ a( 2 ) =e( 2 ) To simplify stl further, we'll create a third function h, such that g(x + 3) = h(). Then l+a lia a s(5) = (5 +5) ="). and similarly ¢(*5*) = n(=*). Our equation for g thus becomes =a W(Z). But this last equation is satisfied as Jong as h is even! (Recall that if h is even, then A(-x) = h(x) for any x) Thus for any even funetion M(), we can construct g(x) = h(x~4) and fle) = g(a) + = a(x 3) +x, and we'll have a solution to our equation. Rather than just having an arbitrary constant, our solution has an arbitrary function, because h can be chosen any way we like (as long as it’s even),74 > CHAPTER 7. FUNCTIONS EXERCISE 7-11 We have claimed that the function f(x) = nx - 3) +x will solve our functional equation for any even function h. For the particular even function h(x) = 2°, show that it does. Problems to Solve for Chapter 7 105. If f(2x) = 53; for all x > 0, then find 2f(x). (AHSME 1993) 106. If f(x) = <4 and g(x) = 2x, then find all x such that f(g(x)) = g(f(@))- (Mae 1991) 107. Given that f(ax) = af(x) for all real a, and f(2) = 5, find f(17). (MA® 1992) 4108, Find all solutions to the functional equation f(x) + f(x + y) = y +2. (M&Q 1991) 109, Find all solutions to the functional equation f(x)/ f(y) = y/x. (M&IQ 1991) 110. Given g(x) = 2x + 8 and f(x) aia: find g © f-"(—2). (MA 1990) 111. Let f(t) = 74, £# 1. Ify = f(@), then x can be expressed as: A. faly) B. -fly) Cc. -fCy) D. fy) E. fly) (AHSME 1967) 112. How many of the following sets of functions have the property that, given any two elements (8) and g(x) of the set, the composition f(g(x)) is in the set? 1, functions of the form ax +b 2. functions of the form ax? + bx +c 3. polynomial functions 4, polynomial functions with 12 as a root (MA@ 1992) 113. Given f(ax) = log, x, find f(x). (MAG 1992) 114, Find all solutions to the functional equation 21(x) — 7; (2) = 12x, (Ma&iQ 1991) 115. If g(x) = 1- x? and f(g(x)) = GF when x # 0, then find f(1/2). (AHSME 1974) 116. If, for all x, f(x) = f(2a)¥ and f(x +2) = 27f(x), then find a. (MAG 1992)the ART of PROBLEM SOLVING: Volume 2 s75 117. Suppose f(x) is defined for all real numbers x; f(x) > 0 for all x; and f(a) f(b) = f(a +b) for alla and b. Which of the following statements are true? (AHSME 1975) L f(Q)=1 TL f(-a) = 1/f(@) for alla ML. f(a) = 9/F@a) for alla IV. f@)> faifb>a 118. If f (47) = } for allx #0,1and 0 < @ < 3, then find f(sec? @). (AHSME 1991) ei) = 3 U9. If f(x) = x7 +x 1 for x > -2 and g(x) = x? ~ x forx <5, then what is the domain of go f? (MAO 1991) 120. Solve the functional equation f(x + £) — f(x — #) = 4xt. (Ma&iQ 1991) 121. Given f(x) such that f(1 - x) + (1~x)f(@) = 5, find (5). (MA@ 1992) 122. Consider a family of functions f,(x) such that f(0) = band f(x) = 2 fi(x—a). Find an expression for fe(2x) in terms of fy(x). (MA® 1992) 123. If f(x) = log (85) for -1 C, w € Ca € C, @ = 1, and w # 1. Show that there is one and only one function . fC Csuch that f@) + fez +a) = g@, z€C, and find the function f. (IMO 1989)Chapter 8 Taking it to the Limit 8.1 What is a Limit? Consider the sequence absaba~ A particularly interesting question that one can ask about such a sequence is to what value does it tend? The value of the sequence for n = 17 is obviously 17/18, but as n goes to ce, what is the limiting value? To be rigorous about this concept is fairly difficult, so we will try to examine the limit in a commonsense way. In this case, as n gets larger and larger, the fraction gets closer and closer to 1. EXERCISE 8-1 Does the last sentence make sense to you? If not, you should talk to someone before you go on. If you understand that one sentence, the rest of this chapter should be a breeze. Really. ‘To express the concept that the sequence — y approaches 1 as 1 gets larger, we say that the sequence tends to 1, or that the limit of the sequence as n tends to oo is 1. To express this concept in symbols, we write taal ‘EXAMPLE 8-1 One important example of a limit is lim 4 rn As m increases, the sequence decreases: mie Nie wie It is clear that the sequence tends to 0. EXERCISE 8-2. Whatis lim, e lim, 4? Generalize. anethe ART of PROBLEM SOLVING: Volume 2 < 77 The previous examples can be used to evaluate a great many limits. For example, they immedi- ately solve the entire category of rational functions, functions like 38 +9042 be — 1d + x41 which are the ratio of one polynomial to another. What happens as x tends smoothly to 00? To analyze a rational function like (8.1), we divide the top and bottom by the highest power of x present. In this case, the power is x°, so the result is 34543 5-B+3+3. When we take the limit x + co, everything with x in the denominator goes to 0 and we are left with 3/5, the final limit. (8.1) EXERCISE 8-3 Evaluate i lim 2xt - 722 +1 8 dod aa ed — 6x + 17 li 28-72 +1 i MG asta ee Ti 28-722 +1 iii, Jim soe dat — 43 + 42 — 6x +17 8.2. Tricky Though the limits we dealt with in the previous section were quite simple, limits can actually be a very tricky business. For many sequences, the limit may not even exist! The simple example of a function for which this is a sequence like 1, 2, 3,..., for which there is simply no limiting value. A similar case was the third part of Exercise 8-3; there, as x increased, the function increased without bound. A sequence or function for which there is no upper limit on the values is called unbounded; we write lim f(x) = co. (WARNING: This “co” is just a symbolic shorthand for saying that the limit diverges in the positive direction. DO NOT treat it as a regular number.) EXAMPLE 8-2 Let's construct a rigorous definition of our terms. A sequence {a} or function f(x) is unbounded if and only if it gets as big as we want at some point; that is, if for every number N, there is some choice of n or x such that lay| > N or [f(x)| > N. Note that by using the absolute value we have allowed for functions to be unbounded toward the negative as well, like -1, -2, ~3, .... We've also taken care of others that might try to evade the definition, like 1, -2, 3, -4,... or similar miscreants. EXERCISE 8-4 Rigorously define what it means for a sequence or function to be bounded, the opposite of unbounded. EXERCISE 8-5 Is every sequence either bounded or unbounded?78> CHAPTER 8. TAKING IT TO THE LIMIT This boundless increase (or decrease or alternation) is only the simplest way in which a sequence or function can fail to have a limit. We can come up with many other devious ways as well. For example, consider the sequence 0,1,0,1,0,1,..., or alternatively the function sin x. Although the sequence and the function are both bounded, there is again no limiting value. (Why?) So being bounded is not enough for a function to have a limit. ‘A function which has no limit, regardless of the particular way in which it fails, is called divergent EXERCISE 8-6 Think about how you might rigorously define a convergent sequence or function. Before you get too confident about this task, think about this: your definition should distinguish between a sequence like 0, 1, 0, 1,0, 1,0,1,..., which has no limit, and one like 0, 1, 0, 1, 0,1, 0,0,0, 0,0,..., which has limit 0. Only the long term behavior should matter. EXERCISE 8-7 Which rational functions are convergent? Which are convergent to nonzero values? 8.3 Working with Limits In Section 8.1, we analyzed the limits of rational functions. But we implicitly assumed several things. For example, we assumed that the limit of a ratio is ‘equal to the ratio of the individual limits; that is, that fe) lim, f(x) #55 g(x) ~ lim g(@) Without this being true, we would have been unable to make the crucial last step, where we said that since the limit of the numerator of was 3 and that of the denominator 5, the final limit must be 3/5. We made another assumption as well, that the limit of a sum equals the sum of the individual limits. This assumption is what allows us to say that the limit of the top, tm (2+ 5+ 5) must equal ling 3 + lim 2, + Jim 4 =3+0+0=3. Ams+ Bat BS In fact, these two assumptions are generally true (and similar results hold for multiplication and exponentiation), as long as all the limits exist and no divisions by zero occur. A proof of this is a little too complicated to deal with here. WARNING: When divisions by zero occur, interesting—and dangerous—things occur. Suppose we have two functions f(x) and g(t) and we wish to evaluate im £0) 10)the ART of PROBLEM SOLVING: Volume 2 < 79 If Jim g(x) # 0, then everything is fine; we just write lim f(x Tim £2) _ BB fo J 9 g(x) lim, g(x) Even if Jim g(x) = 0, we can make an immediate conclusion if lim f(x) # 0. We then have fo 2B ea = (Why?) But if both Jim g(x) = 0 and Jim f(x) = to zero much faster than the bottom, resulting in a limit of 0 (for example, try f(x) = 4, g(x) = 3). ‘The top could go to zero much slower than the bottom, resulting in a limit of oo, (Can you find an example of this situation?) Or, in the most interesting case, the two can go to zero at comparable speeds, resulting in some finite, nonzero limit. A simple example of this last case is f(x) = 3, g(x) = 8, in which lim f(2)/g(x) = 3/5. A more complicated example is f(x) = sin (2), g(x) = }. EXERCISE 8-8 Ona calculator, evaluate sin(1/1000)/(1/1000). (Make sure you are using radians for the angle measure!) Do you have a guess as to what lim sin(1/x)/(1/x) is? Be very careful that you don’t see a limit in which the top and bottom both go to 0 and automat- ically assume that the limit is 1. As we have shown, a 0/0 limit can have any value whatsoever. EXAMPLE 8-3 Certain simple transformations are possible when working with limits. For example, suppose we know that lim f(x) = L and we want to know Jim f(5x). As x goes off to infinity, so does 5x, so we can write Jim, fx) = Jim, f(6x) = lim fly) = L, where we have made the substitution y = 5x. EXAMPLE 8-4 A more useful transformation can be used to convert a limit as x tends to infinity into a limit as x tends to 0, or vice versa. We simply define y = 1/x and note that as x goes to infinity ¥ goes to 0, and as x goes to 0, y goes to infinity. Thus inf) = lim f(1/y) and lim fx) = lim FQ/y) EXAMPLE 8-5 Find im X22 10- Ver 73 mt x-3 . Solution: For x = 3, we find that our limit is 0/0. A general technique in dealing with limits involving square roots is multiplying top and bottom by a conjugate expression, or V2x + 10+-Vx + 13 in this case: N2x+TO~VEFTS _ (VOR TO~ Ve +S) IRF 10+ Ver 1B) lig “7 mo 3 23 (x —3)(V2x + 10+ V4 13) him x-3 295 (x= 3)(V2e+ 10+ Ve +13) 1 1 lim ———__ = -. 23 V2x+10+Vx+13 8 ©80 > CHAPTER 8. TAKING IT TO THE LIMIT Remember this use of multiplication by the conjugate of a radical expression, it is often the key to simplifying limits. 8.4 Continuity We have so far only considered the limit of a function as x — ov. This is because other limits usually aren’ all that interesting! For example, lim 2? +2is just 2? +2 = 6. In general, saying that the limit of a function f(x) as x goes to some finite a is equal to f(a) is the same as saying the function is continuous at the point a. In Volume 1, we defined a continuous function as being one which could be drawn without picking up the pen. However, we didn’t make the distinction that a function is often continuous in most places with isolated discontinuities. For example, consider a function g(x), defined as ey a {% fora #2 BO)" 3, forx=-2. We have plotted g(x) at right. Clearly g(x) is continuous everywhere except at x = -2. But consider the behavior near that point. The limit lim, g(x) is the value that is being approached as we get closer and closer to ~2, not the value at -2. Thus lim, g(x) = ~2, while of course, g(-2) = 3. The limit is not equal to the value! This means that the function is not continuous at x = -2. Everywhere else, the limit equals the value, so % the function is continuous everywhere else WARNING: Always remember that the limit is the value which the function approaches as we get closer and closer to a point, not the value of the function at the point. The two are equal only for continuous functions. EXAMPLE 8-6 The discontinuity we just saw can be completely removed by changing the behavior of the function at a single point: set g(-2) = -2 and it is completely removed. This “nice” type of discontinuity is called a removable discontinuity. Such discontinuities often appear when dealing with rational functions. For example, take the function 2 f@=5 Since the top factors into (x ~ 1)(x -3) and the bottom into (x ~ 1)(x- 2), the function is always equal to (x ~3)/(¢ — 2)... EXCEPT when x ~ 1 equals zero. Then the function is not defined. Thus there is a “hole” in the function at x = 1, where it is not defined at all. Why do we say this discontinuity is removable? Because we can set the value of the original function at x = 1 to its limit, ea Be tig 23 Bx+2 0 x1x—2 which will immediately remove the discontinuity.the ART of PROBLEM SOLVING: Volume 2 < 81 Other types of discontinuities aren’t so easy to remove. Consider the step function 0, forx <0 f= { forx>0. Here there is no way to remove the discontinuity at x = 0, since the limit of the function does not exist at x = 0! We can define separate limits from the left- and right-hand sides to see why this is so; the left-hand limit (limit approaching 0 from the negative side) Jim fc) is 0, while the right-hand limit lim f(z) ro is 1. Clearly, the limit of a function at a point only exists if the left and right limits are equal to one another at that point. Since there is no limit for the step function at 0, we certainly can’t “fix” the discontinuity just by changing the value at one point. Thus this is an essential discontinuity. EXERCISE 8-9 Evaluate x im lim > and - xo [xl x10 [x] Does lim — 0 [x] exist? 8.5 Asymptotes Limits are especially useful for functions, because the limits give vital information as to a function’s structure. For example, let’s try to plot 3x2 6x43 $0) = See + 20° Plotting point-by-point would be very tedious for such a function, and important structural details might be missed altogether. In seeking a different way to plot the function, we note that 3 inn, fe) = lim, f(2) This immediately tells us something that we wouldn’t find using the plug-in method of plotting: for very large and very small x, the function 2 must get closer and closer to the horizontal line y = 3/5. This line, shown 5554 at right, is called an asymptote, and will be a useful guide; try to identify it now in the final graph of the function, shown below. (Can you see how to find the horizontal asymptotes of any rational function?)82 > CHAPTER 8. TAKING IT TO THE LIMIT Factoring the top and bottom of the equation yields f(x) = 3(x-1)?/5(x~4)(x—1), which presents more guidelines. The function is not defined for x = 4 and x = 1, since the bottom of the fraction becomes 0 at those points. But the two points don’t behave the same. The limit of f(x) as x approaches 4is undefined, since the top does not go to 0 and the bottom does. As x approaches 4 from the right, (2) will diverge to positive co; as x approaches 4 from the left, f(x) will diverge to -o0, (To see ‘what the signs should be, imagine, but don’t actually calculate, what would happen for x = 4.1 and x= 3,9.) As x gets closer and closer to 4, the graph gets closer and closer to the line x = 4, so this line is a vertical asymptote. Identify the line in the graph below, and figure out why it is also called an asymptote. (How can we find vertical asymptotes of a general rational function?) On the other hand, the limit of the function as x approaches 1 can easily be evaluated: 3x12 3¢r-1) 0 lim yet ~ Maa 4 ~ EMTs Unlike the unruly behavior near x = 4, the behavior near x = 1 is nice: the curve is smooth except for the removable discontinuity exactly at x = 1 As a final clue to the shape of the graph, let’s see where the function Le crosses the x axis. The numerator of the fraction must be 0, so x = 1 is the 2 only point, but this point is the removable discontinuity shown with the open circle. Combining this with our knowledge of the function as it nears both the vertical and the horizontal asymptotes generates the graph at right. 4 28 - 3x2 +x-6 EXAMPLE 8-7 Plot the function 3x42 Solution: Factoring the bottom yields (x ~ 2)(x — 1), so we see if either x ~ 2 or x ~ 1 also divides the top. The first does, and the top factors into (x ~ 2)(2x? + x +3). Since the top and bottom share a factor x — 2, there is a removable discontinuity at x = 2, with value = 2QP +443) jing CE FE+9) _ lim = 13. ty e@-De-1t)) rt e-D On the other hand, x ~ 1 does not divide the top, so there is a vertical asymptote at x = 1. As x approaches 1 from the positive side the function soars to 02, and as x comes in from the negative side the function dives to -co. There is no horizontal asymptote since the numerator has greater degree than the denominator; thus you might think we have gleaned all the clues we can. On the contrary, getting rid of the (x~2) which is common to the numerator and denominator, we are left with (2x2 +x+3)/(x-1), which upon polynomial divison becomes Qx+3+ x For x — 00 or x + ~o», the fraction tends to zero, so the function approaches the line 2x + 3! This slant asymptote is an important graphing tool, as you can see. We now sketch the graph noticing its behavior at x = 1, and that it approaches the line 2x +3 as x gets large and as x gets small. The open circle on the curve represents the removable discontinuity. 10:the ART of PROBLEM SOLVING: Volume 2 < 83 8.6 Trig Limits Many, if not most, limits require the machinery of calculus to be done with any efficiency. However, certain basic trigonometric limits can be done with simple geometry, and are nevertheless very important. The simplest trigonometric limits one can think of might be things like Jim sin or lim cos x. These aren't too interesting —the first diverges, with sin x oscillating between 1 and —1; the second is equal to cos0 = 1 since cosine is continuous. A more challenging limit is lim sin x/x. Here the limits of the top and bottom are both 0, so the usual methods don’t apply. Consider the diagram at right to get an intuitive feeling for the limit in P question. If the radius of the circle is 1 and ZPOB = x, where x is in radians, then elementary trig shows that PA = sinx, where CHAPTER 8. TAKING IT TO THE LIMIT or 1 cosx Now lim cos = 1, so lim(1/ cos) = 1. Taking limits, we thus have sinx cosx << sinx 1 0 is the same as a limit as 3x > 0 and then substituted y = 34. 8.7 e IfTputa dollar in the bank at 100% interest per year, after one year I will have $2. Suppose, however, that the interest is compounded once during the year. This means that after six months I receive the interest so far, 1/2 dollar, and for the second six months I receive interest on all the money I have, 1+1/2 dollars. Hence my total at the end of the year is (1 + 1/2) +1/2(1 + 1/2) = (1+ 1/2). (Figure out how much this is.) Thave more with the compound interest, because I am paid interest on the first six months’ interest during the second six months.the ART of PROBLEM SOLVING: Volume 2 < 85 My interest can be compounded more than once. If my interest is compounded twice, at the end of 1/3 year [have 1 + 1/3, at the end of 2/3 year Ihave (1 + 1/3) + 1/3(1 + 1/3) = (1+1/3)2, and at the end of the year Ihave (1 + 1/3)? + 1/3(1 + 1/3)? = (1 + 1/3). Similarly, if my interest is compounded 1 times during the year, at the end of the year Ihave (1 + 1/n)" dollars. EXERCISE 8-11 Evaluate the expression (1+ 1/n)" on your calculator for n = 10 and n = 100. As my interest is compounded more and more often, does my yield diverge or approach a fixed limit? As you should have seen in Exercise 8-11, compounding interest more and more often does not lead to an infinite amount of money. Rather, it approaches the fixed limit 2.71.... We define this limit to be the constant e: . it Jim (1+ 4)" =e. dim (1+; The importance of the constant ¢ really comes out in calculus, but it also has some importance in our discussion. For example, the so-called natural logarithm Inx denotes the logarithm base e, log, x. Moreover, the exponential function e* appears in many contexts. 2\3" EXAMPLE 8-9 Evaluate lim (a + 2) : Solution: We fist use the substitution x = 2u to write the given limit as Jim (1+ then use the fact that lim y* = (lim y)* to write a(t) - (eed) Problems to Solve for Chapter 8 131, For what value of k is the following function continuous at x = 2? (MAS 1991) _ { BYEZ forx 42 ro={; = fore =2 132. Evaluate the following limits. sind Be . 48 ii, lim eh ae? in NEW 4 Pam i. 133. The graph of f(x) = (x* — x ~ 2)/(x + 2) has an oblique (slant) asymptote. Find the equation of this asymptote. (MAO 1990)86 CHAPTER 8. TAKING IT TO THE LIMIT 134, 136. 137. . Evaluate lim( Vax? + 5x - V4x? + x). 135, Evaluate lim sin? x/x. . Evaluate lim @ cot 0. a0 . Use the identity 2 = cosx __sin’x © x(1+ osx) to prove that lim(1 ~ cos.x)/x = 0. 138. Find all asymptotes of the function x9/(x* - 1). . Evaluate lim 6x/ V9x? + 17s. 139, 140, . Evaluate lim, 3x2 + 17xthe ART of PROBLEM SOLVING: Volume 2 “ >the BIG PICTURE Centrally important in the field of computer science is the study of algorithms, repetitive Procedures used to accomplish tasks on a computer. For example, suppose I was given a list (a1 @2 43 +++dn) of numbers and asked to find the largest element L. A simple algorithm would be as follows: first take L = a}. Then go down the list, comparing each a; to the current value of L; if aj > L, then set L to be a;. Thus if the list were (2 1 3 4 2), after each step, the current largest element would be 2, 2, 3, 4, 4. An algorithm is only useful if it can be run in a reasonable length of time. For this reason, computer scientists have a way to classify algorithms based on their running time; this notation depends on several concepts of limits. An algorithm's running time is some function f(N) of the size of the problem given. Our example above would take 3 steps if the list of numbers had length 3, 17 steps if the list were length 17, and so on, so in this case f(N) = N. This is simple, buta complex algorithm might havea very complicated function f(N), like 18N? log N— 12N(log NY? + 7. In practice, though, the complicated details of the function f aren’t all that interesting, What really determines the running time is the term of f which is largest as N > 00. To classify an algorithm, then, computer scientists use only this first term. The algorithm with the complicated function above would thus be called “an N? log N algorithm.” To denote this type of approximation, we use a capital O, as in O(N? log N). The most important types of algorithms are, from fastest to slowest, O(1), O(log N), O(N), O(N?), and O(2%). The last is of particular importance, because an O(2N) algorithm is very slow for large N. This can be seen in the fact that if the problem size goes from N to N + 1, the running time goes from 2" to 2+1—it doubles! For even moderately-sized problems, an 0%) algorithm is impractical. On the other end of the spectrum, the running time of an O(logN) algorithm grows very slowly with the problem size N, and the running time of an O(1) algorithm is a constant regardless of problem size. (For example, the problem of finding the 12th element of a list is an O(1) algorithm.) 87Chapter 9 Complex Numbers 9.1 Drawing the Complex Numbers ‘The standard picture of the real numbers is a line: -6 -5 -4 -3 -2 -1 0 2345 6 Each real number is associated with a point on the line, and each point with a real number. Can we devise a similar representation for all complex numbers? Suppose we place the pure imaginaries on a line perpendicular to the reals. If we then put the two lines together, so that they overlap at their zero points (since 0i = 0), then we can associate every point in the plane with a complex number, where the x-coordinate is the real part and the y-coordinate is the com- plex part, To see how this works, examine the figure at right. It should be fairly clear that we can represent any complex number by a point in this plane, and that every point in the plane represents a different complex number. EXERCISE 9-1. Draw a complex plane with the points } + fi and -3 + V2i. EXERCISE 9-2. Draw the set of points z in the complex plane such that: i. Re(z)=1. ii, Im(2) + Rez) = 1 iii, Im(@) <1. « 88 >the ART of PROBLEM SOLVING: Volume 2 < 89 EXERCISE 9-3 For z = 3 + 2i, draw z,Z,2",andz-1. Once we understand the complex plane, we can use it to expand our understanding of complex numbers. For example, we have used rectangular coordinates to locate points in the complex plane. But these are not the only coordinates we could use. Let's apply polar coordinates instead. In our discussion of polar coordinates (page 47) we saw that the polar coordinates (r, 8) of a point (x, y) are such that x=roos0 Jy [r=Vee¥ y=rsind @ = arctan(y/x) For a complex number z, we have chosen the rectangular coordinates to be (Re(z), Im(z)). Thus the associated polar coordinates (r, 8) of z will be such that Re(z) = reos@ Im@) = rsin@. (9.1) But since z = Re(z) + iIm(z), we then have 2 =rcos6 + irsind. (9.2) This is called the trigonometric representation of a complex number, and is sometimes written simply z = rcis@. Of course, just giving the polar coordinates (r, 0) is enough to determine z completely; this is called the polar representation. EXAMPLE 9-1 Find the polar representation for i. Solution: We use equations (9.1) to get rcos@ = 0 and rsin@ = 1, Clearly @ must be n/2 and r must be 1, so the polar representation is (1, 72/2). EXERCISE 9-4 Draw the points r = 2,0 = 7/3 and r = 3,6 EXERCISE 9-5 Draw the curves r = 1 and @ = n/3 in the complex plane. As with regular polar coordinates, the numbers r and @ are the distance from the origin to z and the angle between z and the positive real axis, respectively. It is important to remember the ambiguity in the value of 6 as well; for a given @ we could just as well use 6 + 2n, 0 + 4m, etc. See page 47 for a discussion of this ambiguity. 9.2 The Complex Absolute Value The value of r is particularly important. Like the absolute value for real numbers, it is simply the distance to the origin of coordinates. We therefore call r the absolute value, or sometimes the magnitude or the modulus, of a complex number z and denote it by |z|. Fot a generic complex number x + iy, we have [x + iy| = r= YP +9. The complex absolute value has some interesting properties. Perhaps the most important is that90 > CHAPTER 9. COMPLEX NUMBERS {zw| = lellw| for any 2 and w. To see this, we just write lzwl = (2, +22i)(wi + w2i)) = [iwi - 2202) + @rwa + za01)il = Vea ~ Zot)? + rto2 + z2001)* (cancelling terms) Aut + hw} + ziw} + zwp VEE +B) wi + 9) B+ hut + wh lelleol, and we're done. Similarly we can show that z/1| = |2|/l|. EXERCISE 9-6 Find the absolute value of the complex number (12 + 5i)(7 ~ 24i), EXERCISE 9-7 Find a counterexample to the false claim |w + z| = |wl + [2|- Let’s examine the relationship between |w + 2| and {to| + |2|. Above you should have found that these two quantities are not always equal, but you should also have seen, or at least suspected, that \w +2] is never larger than the sum |wo| + |2|. To prove our suspicion, we write jo| + |2| > ko + z| in terms of w = wy + wei and z = 21 + 22! /R+B+ fur+uh> Vea tm)? +e +m). Squaring both sides and cancelling common terms leaves 2y/zh +23 lw? + wh > 221) + Azar. Cancelling the 2's and squaring again, we get huh + dh + Dot + uh > Bu} + 2eyzawywe + Bro, or Bu} - 2eyzuywr + Buh 20 Recognizing the left as a perfect square, we write (12 ~ 2211)? 2 0, which is clearly always true. Our steps are reversible, so our original inequality must hold. We can find when || + |2| = ko + 2| by noting that these can only be equal when (2,202 ~ z2t1)* = 0. (Why?) This occurs when z1w2 = z2tv1, or 21/1, = z2/w2. Letting this common ratio be c, our equality condition is 21 = cw; and zp = cwz, soz = cw and one number is just a real multiple of the other! The inequality |u| + [z| > lw + 2| is called the Triangle Inequality for complex numbers. Given the name, you might think the inequality has something to do with geometry. You're right; using a geometric representation of complex numbers and complex addition, we can prove the Triangle Inequality quite easily. (This is done on page 103.)the ART of PROBLEM SOLVING: Volume 2 s 91 EXERCISE 9-8 What is the equality condition of the Triangle Inequality when w and z are written in polar form? EXERCISE 9-9 Prove that |2| + lu] > z ~ wl 9.3 Complex Multiplication and Coordinates We now ask the question: what is the effect of complex multiplication in our new coordinate perspective? We can answer this immediately. Consider two complex numbers z; = r1(cos 61 + isin 01) and zp = r2(cos 62 + isin 02). Their product is 2122 = nyra(cos 6; cos 2 + icos Oj sin Oy + isin 0, cos 62 — sin 6} sin 8) = rinr((cos61 cos 2 ~ sin 6; sin 6) + i(c0s 6; sin 82 + sin 6; cos 82) = rira(cos(@, + 02) + isin(@y + 6). The coordinates of the product of (r1, 01) and (r2, 62) are thus (r1r2, 01 + 62). When we multiply two complex numbers, the magnitudes multiply and the angles add. EXAMPLE 9-2 Take the complex number } + i%8 = (cos § + isin 3). When it is multiplied by any complex number r(cos 6 + isin 8), the product is r(cos(@ + 3) + isin(@ + 3). Thus multiplication by 3+ 19 is a rotation by the angle 3. EXERCISE 9-10 To what geometrical motion does multiplication by i correspond, in the sense of Example 9-2? EXERCISE 9-11 On a complex plane, draw the points 2 + 3i, 1 + 2i, and (2 + 3i)(1 + 2/) to convince yourself that the magnitudes multiply and the angles add to form the product. While the polar method is a more satisfying way to look at complex multiplication, for routine calculation it is usually easier to fall back on the distributive law as used in Volume 1. However, for more complicated calculations and clever applications, the polar approach is often much more fruitful. 9.4 Complex Powers and Geometry Once we have a representation for multiplication, we can tackle powers. Consider the generic com- plex number z = r(cos @+-isin 8). The power z" (where n is a positive integer) is just a product zzz---z with 1» z's multiplied together. Thus the magnitude is the product of the individual magnitudes, or , and the angle is the sum of the individual angles, or 10. For positive integers n, we immediately have [r(cos6 + isin 6)]" = r*(cos(n8) + isin(n6)), which can be written in the more memorable form (1, 0)" = (r",n6). (9.3)92 > CHAPTER 9. COMPLEX NUMBERS EXERCISE 9-12 How could we express the powers of i in polar form? EXAMPLE 9-3 To see what exponentiation looks like, let's consider the powers of ¥2+iV2 = 2(cos } + isin$). The second power is ¢ 4(cos ¥ +isin §) = 4i, the third power 8(cos *f +isin 38) = ~4 v2+4 V2i, B the fourth power 16(cos 7 + isin) = —16, etc. The first three powers are at points A, B, and C at right, where one tick mark equals 2 units. 3 To go from one power to the next, we rotate 45° counterclockwise and double the distance from the origin. EXAMPLE 9-4 We can easily extend the result of equation (9.3) to negative integers as well. Take (r, 6)", where 1 is a positive integer. We can set (7, 6)~" = (s,) and write 1 = (r,6)"(r, 6)" = (6, 6)(r", n8), where the last equality comes from equation (9.3). Using the fact that magnitudes multiply and angles add, we then have (sr",@ +6) = (1,0), $08 =r and @ = —n6. Thus (r,6)"" = (r"", ~n8), as desired. EXERCISE 9-13 What is the geometrical relationship of z = (1,6) and its reciprocal for any angle 6? EXAMPLE 9-5 Let's extend Example 9-3 to some negative powers of 2+ iV2 = 2(cos ¥ + isin ¥). The —1st power is 3 (cos(—§) + isin(—})), the -2nd power } (cos(-¥) + isin(~¥)), the -3rd power 3 (cos(~"}) + isin(—%)), etc. The 0th power is 1(cos 0 + isin 0) = 1, as we expect. The first three negative powers are at A, B, and C, where we get from each or one to the next by rotating by 45° clockwise and halving the distance to By YA the origin. EXERCISE 9-14 Duplicate Example 9-3 and Example 9-5 for the numbers > + Pi and } +} 9.5 DeMoivre’s Theorem In equation (9.3) we found a very nice property of integral powers of a complex number. However, we can just as easily look at fractional powers. Consider a complex number which we have written in the polar form (1, 0). How can we find (r, 6)?/4, where p and q are integers? To begin with, we get rid of the p. By equation (9.3), we immediately have (0/9 = (, poy", Thus our problem boils down to finding the qth roots of (r”,p6). This isn’t so hard. We can by guesswork write down one qth root: (r°/4, p8/q). This is a qth root because its qth power is (4, pO /q)" = ((?!4)!, q(p/q)) = (r?,P@),the ART of PROBLEM SOLVING: Volume 2 2 93 using equation (9.3). So we have one qth root immediately. How do we find others? Remember that the complex number (?, p6) is equal to (r”, pO + 27k) for any integer k; adding a full circle to the angle doesn’t change our location. From these equivalent numbers we can easily write down the qth roots (v0 + =) . A) a4 In fact, the numbers generated by equation (9.4), as k ranges over the integers, are all the p/q powers of (7,6). This assertion, the simple proof of which is below, is called DeMoivre’s Theorem (de MAUVE’s theorem). EXAMPLE 9-6 Find all cube roots of 4 + 4-V3i. Solution: We can readily convert this number into the trigonometric form 8(cos 60° + isin 60°), which gives the polar form (8,60°). Then by DeMoivre, the cube roots (1/3 powers) are given by (8°8,20° + 360°/3) = (2,20° + 120°k). For k = 0, we have (2,20°). For k = 1, we have (2,20° + 120°) = (2,140°). For k = 2, we have (2,260°). For k = 3, we have (2,380°), which is equal to (2,20°) again. Similarly, as k takes on other integral values, we will just get our three basic values over and over again. Thus our roots are (2, 20°), (2,140°), and (2, 260°). EXERCISE 9-15 Find all values of (-4-V2 + 4 ¥2i)*/4 Let’s now prove DeMoivre's Theorem. Clearly every number of the form (9.4) with k an integer is a p/q power of (, 6). We verify this by taking the expression to the qth power and observing that the result is a pth power of (r, 6) in agreement with equation (9.3). Since the expression (9.4) is a qth root of a pth power of (r, 6), it is a p/q power. Wenow need only to prove that every p/q power of (r, 8) can be written in the form (9.4). Consider some general complex number (s, $) such that (r,6)"/7 = (s,$). We can write (r, 8° = (5, @)!, so (9.3) gives (7°, p6) = (s!,4). The only way these two sets of polar coordinates can be equal is if Psst and pO +2nk=4¢, which clearly forces (s, $) to be of the form (9.4), so we're done. EXAMPLE 9-7 In Example 9-6 above, we found that there are three cube roots of 4 + 4-V3i. This is no accident. In fact, DeMoivre guarantees that there will be exactly q qth roots of any nonzero complex number. Why? Because the allowed angles of (r, 6)!/7 are @ Ink _ O86. IMO. 4n Sy ie Seem 909099 49 G All of these values will be different until we get to k = q, when we will have 9/q + 2nq/q = 0/q + 2n. This is the same as just 0/4, so we are back where we started: adding 2n/q just gives 0/q + 2n/q again, and so on. Similarly, negative values of k give the same roots as well.94 > CHAPTER 9. COMPLEX NUMBERS 9.6 Exponential Form We have in De Moivre’s Theorem a very powerful tool for examining the powers of a complex number. We can make this tool still sharper, however. Consider the powers of the complex number cos @ + isin @. De Moivre’s Theorem tells us that (cos @ + isin 6)" = cos(n6) + isin(n6). ‘This is a very special property of the function f(8) = cos @ + isin 8, namely that (OM = f(r). Which known functions behave like this? Only the exponentials, as in (a*)" = a”. This possible connection was closed by Euler, who showed that the usual exponential e*, x real, could be extended to the imaginaries by taking, for any real number y, é¥ = cosy + isiny, (9.5) We should emphasize that this is neither pure coincidence nor arbitrary choosing. In fact, students familiar with calculus should be able to prove the necessity of this extension using Taylor series. EXAMPLE 9-8 Evaluate the complex number e/?+#"/4, Solution: We split the given expression into the product ¢*/e'"/*. The first part cannot be simplified, but the second part becomes + + 92 using (9.5). Thus our number is (47 + Ba. The relation (9.5) is very useful for the understanding of complex numbers. We define the exponential form of a complex number to be re", where r and @ are the same as for polar form. In. fact, the exponential form turns out to be just a more powerful way to express polar form. For one thing, De Moivre’s Theorem is obvious in exponential form, saying that (rey = ren, Just by writing exponential form, we get De Moivre’s Theorem for free! However, always remember that if the exponent is not an integer, we must write (retov20)® = pence where all the gth powers are achieved by ranging k over the integers. To write the exponential form of an arbitrary complex number we just find r and @ as for polar form.the ART of PROBLEM SOLVING: Volume 2 1 95 EXAMPLE 9-9 Find i. Solution: It’s easy with exponential form. We just write i = el(V/2+20H), go jf = (ei(r/2+2nhyi = e-("2+2"8), There are thus infinitely many such powers ('), all of which are real: = fe 8 et eF eF,.. This should not be too disturbing, given that even fractional powers give many answers. The reason this is so weird is that we are using the notion of “powers” very abstractly; we can no longer expect intuitive results. We can use this method to take complex powers of any complex number which we can put into exponential form. Note that in Example 9-9, we used De Moivre’s Theorem for the complex power i, though the theorem was proven only for real powers. Fortunately, the theorem still holds for complex powers. (Can you prove it?) EXAMPLE 9-10 Equation (9.5) tells us that e* = cosx + isin x; substituting —x for x, we similarly gete™ = cos(—x) + isin(—x) = cosx —isinx. Adding these two equations and dividing by 2, we get cosx while subtracting them and dividing by 2i yields This is not some kind of trick; these expressions are perfectly valid ways to express sine and cosine, and are useful in many ways, EXERCISE 9-16 Use the expressions of Example 9-10 to prove that sin?x + cos?x = 1 and sin2x = 2sin xcosx. EXAMPLE 9-11 Using the expressions of the previous example as a basis, we can define two new functions, the hyperbolic sine and cosine sinh x and cosh x. The functions are defined as fe z and coshx = z sinhx = © The similarity to the expressions given for sine and cosine in Example 9-10 is obvious. The tie between hyperbolic and ordinary sine and cosine is made still clearer by the fact that the hyperbolics satisfy similar identities; for example, cosh? x — sinh? x = 1. EXERCISE 9-17 Prove the validity of cosh?x — sinh” x = 1 and investigate formulas for sinh 2x, cosh 2x, etc.96 > CHAPTER 9. COMPLEX NUMBERS ‘The hyperbolic tangent, secant, etc. are defined in the familiar way; for example, tanhx = sinhx/coshz = (& -e~*)/(e +e). 9.7 Two for One ‘When we are working with rational variables and functions, the equation f(x) = g(x) isjust that, one equation. Let’s consider the complex variable equation a + bi = c+ di. The points in the complex plane represented by a + bi and c + di can only be the same if a = cand b = d,soa+bi=c+di gives us two equations, not just one. Now let's look at how this can be useful. EXAMPLE 9-12 Use DeMoivre’s Theorem to get an expression for cos 3@ in terms of cos 0. Solution: If we write DeMoivre's Theorem for (cos 6 + isin @)°, we have cos36+isin3@ = (cos +isin6) = cos? @ + 3icos’ @ sin @ — 3.cos @sin® 0 — isin’ 6. To get cos 30, we equate the real parts of the above equation, yielding cos30 = cos*@-3cos sin’ O cos* @ ~ 3.cos 6(1 ~ cos? 8) 4cos* @ -— 30s 8. EXERCISE 9-18 Find an expression for sin 30 in terms of sin 6. EXERCISE 9-19 Show that i i Im (& we) = )oksinnd. ea ma When is this helpful? Clearly our new technique is very useful for problems involving cos @ and sinn6. 9.8 The Roots of Unity DeMoivre’s Theorem can easily be used to find the n nth roots of 1. Since 1 = e**™ for integers k, the nth roots are given by 4n . 2n Qn 4n 14" = Artin = 1, cos — + isin —, cos — +isin—,. 1” ” n 7 EXERCISE 9-20 Write the three 3rd roots and the four 4th roots of 1 in Cartesian form.the ART of PROBLEM SOLVING: Volume 2 3 97 These nth roots are interesting in a number of ways. First, we consider plotting i them all in the complex plane. Since all their amplitudes are 1, all the roots lie on the unit circle. Moreover, their angles are equally spaced from 0 to 2r, so they form a regular polygon! This is shown at right for n = 7. (In fact, the roots of 1 any number form a regular polygon, but this fact is generally most often used for roots of unity.) EXERCISE 9-21 For what n is ~1 an nth root of 1? For what n does —1 lie on the polygon drawn above? EXERCISE 9-22 How many 17th roots of 1 are there in the second quadrant? Another property of the nth roots of 1 is that they all satisfy the equation x"-1=0, Since we know 1 is a root, (x ~ 1) divides the polynomial on the left, which we can then factor as (eA 4x2 4-41) =0. For any root # 1, we thus have a relation which is useful for many problems. EXAMPLE 9-13 If q is one of the imaginary roots of the equation x° = 1, then find the product (1-@ +0%)(1 + w ~ w), (AHSME 1971) Solution: Writing the given equation as x° - 1 = 0, we factor and find (x — 1)(x2 + x +1) = 0. Since w is imaginary, it is a root of x7 + x + 1, so w* + w +1 = 0. Hence we have 1+ @ = -w? and 1+? =-w and -«@ — @)(-w? ~ @*) = 4@ = 4, (-o+a%\(1+o-0)= since w = 1. Problems to Solve for Chapter 9 141. Find | 3#/|. (mao 1991) 142. Find (1 + i)4(2 — 219°. (mae 1987) 143. Find the product of the n nth roots of 1 in terms of 7. 144, If f(z) = 244, then find f1991(2 + i). (MA@ 1991) 145. For how many positive real values of K will (2 + Ki)® be a real number? (MA@ 1992) 146. Show the line through the complex points w and z plotted in the complex plane has slope Im(z - w)/ Re(z - w).a 98 > CHAPTER 9. COMPLEX NUMBERS 147. The diagram at right shows several numbers in the complex plane. The Fe circle is the unit circle centered at the origin. Which of the shown points could possibly be the reciprocal of F? (AHSME 1983) 148. If the six solutions of x° = -64 are written in the form a + bi, where a and b are real, then find the product of those solutions with a > 0. (AHSME 1990) Ee 149, What is the graph of |z| = Imz + 1? (MA@ 1991) 150. Write cos 50 as a function of cos 8. (MA@ 1990) 151. Find Im((cos 12° + isin 12° + cos 48° + isin 48°)°). 152. If 0 is a constant such that 0 < 6 < mand x + 1 = 2cos@, then for each positive integer n, find x" + J. in terms of n and 6. (AHSME 1981) 153. Evaluate = cos(n6) 2 = where cos @ = 1/5. (MA@ 1991) 154, Suppose that the coefficients of the equation x" +a,-1x"* ++++ +a, +9 = Oare real and satisfy 0 1. Show that 2”? = 1. (This problem originally appeared on a contest used to determine the Chinese national team.) (MOP)the ART of PROBLEM SOLVING: Volume 2 * >the BIG PICTURE Carl Friedrich Gauss was, among his many other achievements, one of the primary popu- larizers of complex numbers. One of the discoveries of which Gauss himself was proudest was the constructibility of the regular 17-gon. The ancient Greeks were able to construct the equilateral triangle, and the regular pentagon, but no other regular polygons with a prime number of sides, but Gauss was at last able to extend this repertory. Given a segment of length 1, Gauss knew that any integer-length segment could be constructed. Moreover, the sum, difference and quotient of two segments can be constructed, and the square root of a segment can be constructed. Thus any segment whose length is an expression made up of sums, differences, quotients, and square roots of integers can be constructed. Asa simple example of such an expression, Gauss used the 17 seventeenth roots of unity to show that yn7 = 144 2 sa c0s860"/7 = — Fe + Fe V7 + 7 V84—2V17 +i Vi7+3 V7 V34—2-V17 -2 V3442VI7. Once he could construct a segment of length cos 360°/17, Gauss could construct the point (cos 360°/17, sin 360°/17), by laying off cos 360°/17 along the x axis and drawing a perpendicu- lar to the x axis at that point. The intersection, P, of this line and the unit circle has polar angle 360°/17; copying the angle between the positive x axis and OP (O js the origin) seventeen times around the unit circle provides the seventeen vertices of a 17-gon. Gauss's construction of the 17-gon is one of the most compelling examples of the geometry of complex numbers, and Gauss asked that his tombstone be made in the shape of this wonderful, constructible polygon.Chapter 10 Vectors and Matrices 10.1 What is a Vector? or, abstractly, even more dimensions. A vector is typically given a variable-type name, like v, and is denoted by @. Also, the base point of a vector is called the tail and the end of the arrow the head. ‘A vector is simply an arrow from one point to another. For example, at right XN, we have drawn some vectors in two dimensions (2D). Vectors can also be in three —m— ‘The length of the vector @ (distance from tail to head) is denoted by ld. A vector is typically regarded as depending only on its length and direction; the location of the starting point of the arrow is immaterial. Since the starting point doesn’t matter, we can “add” two vectors by moving the tail of one vector to the head of the other, as at left, where the boldface vector is the sum of the other two. ‘We can easily verify that vector addition defined in this way is commutative, a so that # + @ = @ +7. Just draw the two additions and note that the two copies each of and @ are parallel, as at right. Using our definition of vector addition we can quickly expand to multipli- cation by a positive real number: the vector cd for c a positive real is the vector < in the same direction as # but with length c|jdl]. (Convince yourself that this e makes sense.) The vector @ is defined to be the vector with length zero. se Similarly, the vector ~@ is just a vector with the same length as d, but in the opposite direction. This way we get 3 + (-#) 6, as we would normally expect. We can then define i - # = 0 + (-d), and we see from the diagram that i — dis the vector that runs from the head of d to the head of @. This | should not be a surprise since # + (@- 8) = B+ @-d) =a. a8 < 100 >the ART of PROBLEM SOLVING: Volume 2 < 101 10.2. The Dot Product The length of the vector i — 7 can be found using the law of cosines. Since d, , ba and @ — 8 form a triangle whose sides have lengths ||, and {|i — a], we have Uleb ~ AIP = aA? + [BI - 2}:AI:3I| cos 8, (10.1) a where 0 is the angle between # and The expression |[@|l||| cos @ is called the dot product of # and a; it is denoted by &-@. We can then write (10.1) as (lab — AIP = [AP + I? - 20°-w, so that the dot product is given explicitly as. oB= tan — 055 We can establish certain nice properties of the dot product. 1. -@=B-d. (The dot product is commutative.) 2. d-w = 0 for nonzero dand w if and only if Sand are perpendicular. 3. (c8)-@ = e(@-@) for any real number c. 4. i. (8+@) =i. 8+ it-w. (The dot product is distributive.) EXAMPLE 10-1 Prove property 1 above. Proof: Let O,3_azbe the angle from «to ¥, so that 8,3 = —O 93. Now we just write D-F= [BIA] co5 6,5, = lll cos(—O g) = lll cos(8g 3) = 3, where we have used the fact that cos(—8) = cos 0. EXERCISE 10-1 Prove properties 2 and 3 above. Property 4 is proved using coordinates in the next section (see Example 10-2). Properties 3 and 4 mean that the dot product is linear. The use of vectors as abstract “arrows” is most useful in vector geometry, which we do not treat antil Chapter 12. In the next section we will begin to examine vectors in a particular coordinate system, which is more pertinent to elementary problems.102 > CHAPTER 10. VECTORS AND MATRICES 10.3 Coordinate Representation of Vectors The standard way to represent vectors in a coordinate system is to define an origin and place the tails of our vectors there. We can then use regular rectangular coordinates with the given origin at the center, as at right. We associate a vector with the coordinates of its head; if the head coordinates are (x, y), the vector is represented as (x yor (5): The former representation is called a row vector and the latter a column vector. We'll generally use row vectors because they take up less space. ‘The power of the coordinate representation comes from the fact that we can use regular coordinate techniques. For example, the vectorsum (x; y1)+(%2 yn) isjust (Gi +x) (ity). EXERCISE 10-2 What is the length of the vector (2_3) ? In the rectangular coordinate form, the dot product has a nice form. Consider two vectors a = (x1 y1) and & = (x2 yz) which form angles 6; and 62 with the positive x axis. Their dot product is found using polar coordinates: ah [laili@all cos(@x - 82) = |\@illlll(cos @; cos 2 + sin 6; sin 2) ((ldi}lcos 6,)(|Idll cos 62) + (till sin @1)(i@al sin 62) = met yryr EXAMPLE 10-2 Prove that the dot product is distributive. Proof: Leti?= (xo yo),d=(x1 y1),and@= (x2 ya). We have = (F4 w) = oles + x2) + yolys + ¥2) = ox: + yorn) + ore + Yoyr) = 8) + -W), as desired. We can easily extend the coordinate representation into three (or more) dimensions. The dot product in three dimensions (3D) is 1X2 + Yiy2 + 2122 and a similar expression holds in higher dimensions, even though we run out of letters. EXERCISE 10-3 Show that (1 17 ~3 2) isperpendicularto(-6 1 5 2).the ART of PROBLEM SOLVING: Volume 2 < 103 The notion of vector addition gives us a swift, nice proof of the Triangle o Inequality for complex numbers (page 90). If we view d and w as complex numbers v and w, so that = (0 v2) = v1 + opi and [fal] = |o|, the graph at right represents the addition v + w in the complex plane. Since the vectors @, @, and @ + form either a triangle or a straight line (when @ and @ are in the same direction), we have |a}| + || = || + wl] (since |[a,|k||, and a+ @|| are the sides o ofa triangle, which may be degenerate). In complex number notation, this becomes [o| + tol > |v +t, and the equality condition of v = cw follows from the observation that [ll] + lal = lid + al only if ¢ and @ are in the same direction, so 01 = cw, and v2 = cw, 10.4 What is a Matrix? ‘To understand what a matrix is, we will solve the following standard problem: Given a point (x, y), what is the new point obtained after rotating (x, y) by an angle 0 in the plane? We can do this using polar coordinates as we did on page 49. Here, we will rotate the point, not the axes, through an angle of 6 counterclockwise. If the polar coordinates of the point are initially (r, a), then after the rotation the coordinates are (r,a+ 6). Then we can go back to rectangular coordinates. Let's call the new coordinates (x’, y’); thus, we find x =rcos(a+6) = r(cosacos 6 -sinasin 6) y'=rsin(a+0) = r(sinacos6 +cosasin6). Since rcosa = x and rsina = y we can write these in terms of the original coordinates: x’ = xcos@—ysind ¥ = xsind+ycos@. (10.3) (Notice that these equations are different from those on page 49. This is because here we are rotating the point counterclockwise rather than the axes. We can see that the above equations are the same of those by noting that @ on page 49 equals -@ here.) We have boiled the rotation down to a function. from the old coordinates to the new coordinates. The functions for x’ and y' are linear in x and y, so they are completely specified by four coefficients; we can use this to develop an efficient notation for the transformation, If we write the old and new points as the vectors ) and (3). then we can encapsulate the information in the transformation equations (10,3) with the form x ‘cos@ -sin@\ (x (7) - (Ss 8 cosé ) () on) Compare this closely to (10.3). The object by which we “multiply” the vector (3) isa matrix; each ‘tem in itis called an entry. Each entry corresponds to a particular coefficient in equations (10.3), as104 CHAPTER 10. VECTORS AND MATRICES a comparison shows. The upper left entry is the coefficient from x to x’, the upper right is from y to x, the lower left is from x to y’, and the lower right is from y to y’. We have gone from regarding the transformation (10.3) s transforming each coordinate according, to an equation, to seeing it as the application of a transformation matrix to a vector. Let us examine the method by which, given only the right side of (10.4), we can get the left. In other words, given only the initial vector and the transformation matrix, how do we compute the transformed vector? "The answer is quite simple. To compute the first element of the new vector, we go across the first row of the matrix, multiplying the elements of the row by the corresponding elements of the vector. ‘The resulting sum is the first element of the new vector. To compute the second element, we do the same with the second row of the matrix. EXAMPLE 10-3 Let's carty out the process described above to multiply the vector () by the matrix ( 3): ‘The first element of the product will be found by going across the top row of the matrix and down the vector to get (2)(a) + (3)(b). The second element is formed with the second row of the matrix, going down the vector as before to get (4)(a) + (5)(6). Thus we have (3) @)- (Gass): EXERCISE 10-4 Find the product Gs %) () EXERCISE 10-5 i, What is the effect of the matrix ( ‘) ona general vector? Geometrically, what kind of a transformation is this? ii, Find the matrix which takes any vector to itself. This is called the identity matrix. iii, Can you guess what matrix multiplication looks like for 3D vectors? Matrices are often named with letters, like vectors. Usually these are underlined, so the matrix named A is written A. It is easy to verify that matrix multiplication of a vector is linear, meaning that A(é+ @) = Ad+ Ad and A(cd) = cAd. EXERCISE 10-6 Verify that matrix multiplication of a vector is linear. 10.5 Matrix Multiplication Once we understand the role of matrices as transformations of vectors, we can ask: what happens when we apply two transformations in a row? Say we have the vector (i). and we apply first thethe ART of PROBLEM SOLVING: Volume 2 < 105 matrix G ~)otnen the matrix G 2) «The result will be -1 2\[/3 -4) (1 3 -4)|\5 -6) \a)] EXERCISE 10-7 If you still feel uncomfortable with multiplying a vector by a matrix, evaluate the above product explicitly. Applying the matrices one at a time is fine; but suppose we wish to consider the two transfor- mations as one, composite transformation? That is, we want to find a matrix C such that (3 216 )@]-e@): ‘We shall define “matrix multiplication” so that the above holds with C product of the two matrices. How shall we define this product matrix C? In general we have CDC MO] - € DG) = (Mex+ fy) + (gx + hy) (a + fy) + d(gx+ 1) = [ae + bg)x + (af + bh)y ~ \(cotdg)x + (cf + dhdy, _ (aerbg af ron (x ~ \ce+dg of +dh) \y)* Weshall thus define the product (: ’) ( f) tobe the matrix (er a a oh ; then everything will work nicely. If this seems like a rigged game, it is! We have chosen matrix multiplication so that it means something we want. Observe that the way to do matrix multiplication is the same as when we multiplied a vector by a matrix. In fact, if we consider only the first column of ( f) and the first column of the product, we have matrix multiplication of a vector; the same is true for the second columns. “1 2) (3 -4 EXAMPLE 10-4 Find ( 3 2) G =). Solution: To get the top left entry in the product, we go across the top row of the first ma- trix and down the left column of the second, to get (~1)(3) + (2)(5) = 7. To get the top right entry, We go across the top row of the first matrix and down the right column of the second, getting106 > CHAPTER 10. VECTORS AND MATRICES (-1)(-4) + 2)(-6) = -8. To get the bottom left, we go across the bottom row of the first matrix and down the left column, getting (3)(3) + (~4)(6) = -11. To get the bottom right, we use the bottom row of the first matrix and the right column of the second, to get (3)(~4) + (~4)(-6) = 12. Thus the product is ( i B): EXERCISE 10-8 Verify that column by column, matrix multiplication of matrices looks just like matrix multiplication of vectors. EXAMPLE 10-5 By equation (10.4), a 90° counterclockwise rotation is given by ( ° ). anda 180° rotation by CC ‘) Doing one and then the other gives the transformation (GG @): To find the upper left entry in the product, go across the first row of ( 0 4 ) and down the first column of ( ) to get (-1)(0) + (O(1) = 0. For the upper right, go across the same row of the first matrix, but down the second column of the second, to get (—1)(~1) + (0)(0) = 1. For the bottom entries we do the same thing, but going across the second row of the first matrix. The result is ot -1 0)" EXERCISE 10-9 To what rotation does the product above correspond? Is this what you would expect? 2 -3\(1 1 EXERCISE 10-10 Evaluate (3 3) G 2): EXERCISE 10-11 Geometrically, what do you get when you reflect through the x axis and then through the y axis? Show you get the right result in matrices by multiplying the matrix for x -1 0 o1 reflection, ( by the matrix for y reflection. Since matrix multiplication is associative, A(BC) = (AB)C. We usually just write ABC and do the multiplication in whichever order we want. WARNING It seems sensible that matrix multiplication would be commutative, so that AB = BA. However, this is NOT true! EXERCISE 10-12 Show that matrix multiplication is not commutative by finding a simple coun- terexample. Since we can multiply matrices, we can also take them to positive integral powers, just writing AAAA as A‘, for example.the ART of PROBLEM SOLVING: Volume 2 < 107 EXERCISE 10-13 Write down a matrix A for a rotation by 60°. Find A® without any computation. 10.6 Matrices in Higher Dimensions A.2x2 matrix has been used to represent a transformation from one 2D vector to another. If you thought about Exercise 10-5, you may have figured out how to extend this to 3D vectors. In three dimensional space each point has an x and y coordinate as in 2D, but also has a z coordinate to denote its distance above or below the xy plane. The positive x, y, and z axes are situated as shown. Here a transformation froma vector (x y 2) toavector(x’ y/ 2/)hasthe % form x’ = ax + by + cz, ete, There are nine coefficients to the transformation, that from x to x’, from y to x’, from z to x’, from x to y’, etc. (compare this to the discussion of 2 x2 matrix entries on page 104.) We represent the coefficients in exactly the same way as for 2D matrices. y EXERCISE 10-14 Write down the 3 x 3 identity matrix. EXAMPLE 10-6 We can easily write down the 3 x3 matrix for a rotation by angle 0 about the z axis. Clearly, the new z coordinate is the same as the old z coordinate, so the coefficient from z to 2’ is 1, while the coefficients from x and y to 2’ are 0 (no contribution). Also, neither x’ nor y’ is affected by z, so these two coefficients are 0. Finally, the other coefficients come from the standard rotation matrix of equation (10.4). The matrix is thus ‘cos sind 0 sin@ cos@ 0 0 oo. EXERCISE 10-15 Find the 3 x 3 matrix for each of the following: i. Rotation about the x axis. . Squashing all vectors to , the origin. iii. Reflection in the xy plane. Itis pretty simple to show that 3 x 3 matrices work the same as 2 x 2's in terms of associativity, noncommutativity, etc. Since we can imagine vectors of more than 3 dimensions, we can similarly write down matrices which transform those vectors, such as 4X 4,5 x5, etc. We can even write down matrices which are not square! For example, a 2 x 3 matrix (number of rows goes first) takes 3D vectors to 2D ones: @ 123 a+ 2b+3¢ (i 5 ) , = (0). (105) All you need to remember is that to find an entry in the product, we go across the corresponding row and down the corresponding column. Study the above example if this is still unclear.108 © CHAPTER 10. VECTORS AND MATRICES EXERCISE 10-16 Write down and multiply: i, Two 3 x 3 matrices. ii, A2x4 and a4x3 matrix. iii, A1X3 and a3 x1 matrix. (To what does this correspond?) Compare the dimensions of the products to the dimensions of the original matrices in these three cases. Is there a pattern? It seems strange, but only under certain circumstances can we multiply a kx] matrix by anm x1 matrix. To see why, let the first matrix be A and the second B. EXAMPLE 10-7 Consider the multiplication Az. Find the dimensions of # and Az, where Ais akx! matrix. Solution: In the multiplication we will be going across rows of A and down %. Since A has I columns, each of its rows is / entries long. Thus Zhas I entries also, so is dimension I. On the other hand, there will be one entry in the product A¥ for each row of A, as each entry in the product is formed by going across one row of A. Thus the product will have k entries, so will be dimension k. EXERCISE 10-17 Compare the preceding discussion to equation (10.5). Do the two agree? For vectors we have (AB)z = A(B3), by associativity. Clearly # must be an n-dimensional vector if B, which takes n-dimensional vectors to m-dimensional ones, can transform it. The product BY will be n-dimensional. But this product must be I-dimensional to be transformed by A! Thus we must have m = 1. Furthermore, the dimension of ABZ will be k, since A was used last and its outputs are k- dimensional. Thus the product AB takes n-dimensional vectors to k-dimensional ones, and must be akxn matrix. It is this pattern that we were looking for in Exercise 10-16—go back and verify that itis true in those cases if you have not already. 10.7 Better Matrix Notation Up to now we have written a general 2x 2 matrix as A = (¢ ‘): A better way to write this is by labelling each entry by its row and column numbers, We would then write A= @ a) . 40.6) 221 ax, e WARNING: As always with matrices, ROWS GO FIRST—for example, a2 is an entry in the fourth row, second column of A, not the other way around. We'll often use the a;; notation because it is very efficient.the ART of PROBLEM SOLVING: Volume 2 < 109 EXAMPLE 10-8 The 3 x 3 identity matrix is completely specified in this notation by aft isi i= {o ial Do you see why? EXAMPLE 10-9 The product rule takes a nice form in this notation. Given AB = C, with Aan xm matrix, B an m x n, and C therefore a1 x n, we have ij = ainda; + dinbaj +... + ainbnj, or to compress even more, Gy = Saaby. If these expressions are unclear, write A and B in the form (10.6) and multiply out to see. Try starting with small values, like | = 2,m=3,n= 4. o 12 EXERCISE 10-18 Express the matrix { -1 0 1 J in the manner of Example 10-8. -2.-1 0 Problems to Solve for Chapter 10 155. Write down the 2D matrix for a rotation by 45°. 156. Show how the multiplication of rotation matrices can be used to remember the trig identities sin(x + y) = sinxcosy + sin ycosx and cos(x + y) = cosxcos y ~ sinxsin y. 219 -2 1 -1 60 -3)(4 4 -3 157, Find the product 13 2/\3 21 158. Three vertices of parallelogram PQRS are P(-3,-2), Q(1,-8), R(9,1) with P and R diagonally opposite. What is coordinate 5? (AHSME 1963) 159. Find the cosine of the angle between the vectors (3 4 5) and (-1 4 3) 160. Matrix A has two rows, three columns. Matrix B has four rows, two columns. The existing product of these two matrices consists of how many elements? (MA@ 1992) (° 3) 161. Let A be the matrix% —~ 110 > CHAPTER 10. VECTORS AND MATRICES and let x be the sum of the entries of a matrix B such that AB = BA. Find the smallest value of x over all matrices B whose entries are positive integers. (Mandelbrot #3) 3 141 162. What is the image of | 1 ] under the mapping { -2 0 0 }? (mae 1991) 2 32-3 4163. Find min f and max f where x and y are real numbers and f(x,y) = 2sinxcos y + 3sinxsin y + 6cosx. (MaiQ 1991)the ART of PROBLEM SOLVING: Volume 2 < 111 ;——the BIG PICTURE Imagine an atom (particle of matter) sitting in space. There are many possible “states” the atom could be in: vibrating fast or slow, spinning around in different ways, and so on. But as we pointed out in a BIG PICTURE in Volume 1, quantum mechanics can’t tell us exactly which state the atom is in, just the probabilities of its being in each state. These probabilities are often thought of as a vector, B = (P, Pz P3 -+-), where P; is the probability of being in the ith state. So far we're on solid ground. But what if the atom has infinitely many possible states? Suddenly our vector has become infinite-dimensional! Now imagine a particle of light—a photon—flies onto the scene and our atom absorbs it. This will cause our atom to switch from its initial state to some other state. How do we know what other state? We don’t! Again, we only know the probabilities. If the atom started in state i, there is some probability A;-,; that it ends up in state 1, some probability A;2 that it ends up in state 2, and so on. (What does 4j..; represent?) Our original probability vector P thus transforms into a new probability vector @, which describes the probabilities after the absorption. How do we transform one vector to another? With a matrix, of course! We define a transition matrix A, so that AP = G: Aa Ara \ /Pi Qi Ats2 Are | { Po} = [ @ If you are familiar with the rules of probability, you might be able to understand this in more depth. Why, for example, is Q;, the probability we end up in state 1, equal to P)A1 41+ P2Az1+ PsAso1 +--+? Can you see that we must have P, + P2 + P3+--+=Qi+Qr+Q34---=1? To calculate the probability vectors and transition matrices takes some doing, but the math- ematical apparatus of the infinite dimensional vectors involved is more or less the same as that for the humble 2D vectors we define in the text.Chapter 11 Cross Products and Determinants 11.1 The Cross Product ‘We have seen in Chapter 10 that there is a connection between the dot product and vector lengths. We now define a special product between vectors which allows us to discuss areas. So, given two 3D vectors 3 and @, we shall define the cross product x tobe the vector # such that » Zis perpendicular to both and w (so that 2d = #-B = 0); » the length of 7'is the area of the parallelogram spanned by d and , as in the figure at right. EXAMPLE 11-1 For any vectors d and i, we have (8x @) -# = 0, since we have defined the cross product to be such that (8x ) 1 EXERCISE 11-1 What is the area of the parallelogram spanned by @ and w in terms of ||@, |||, and 6, the angle between the two vectors? Given vectors and @, there are two vectors which satisfy our criteria for the cross product, one which points ‘up’ from the plane containing # and @ and another which points ‘down’ from the plane. How do we know which vector is x a? There's no sound mathematical reason to choose ‘one or the other, so we must adopt a convention which we can apply to any pair of vectors. This brings us to the dreaded right hand rule. This is nothing more than a way to remember in which of the two possible directions we choose our cross product to be. > Consider the equation # x @ = #. If you extend the index finger of your right hand along and the middle finger along @, then # will be along your thumb pointing perpendicular to the other two fingers. EXAMPLE 11-2 Ifwe take #=(0 1 0)and@=(1 0 0), then find ox 3. Solution: Take the desired vector to be # = (@ b c). Then the dot products of # with J and Bare b and a respectively; by the first condition on the cross product these must both equal zero, < 112 >the ART of PROBLEM SOLVING: Volume 2 <_ 113 so¥=(0 0 c). By the second condition, the length [|i] = Ic| must be 1 (why?), so that c is +1. The right hand rule tells us that Zis pointing down, so#= (0 0 -1). EXERCISE 11-2 As @ stays fixed and # rotates in a full circle, describe the path followed by the tip of dxd. EXERCISE 11-3 Using the right hand rule, find the relationship between # x w and dx @. WARNING: Exercise 11-3 shows an important difference between the cross product and other products with which you are familiar. The cross product is NOT commutative; that is, 3x @ # Bx@. 11.2 The Cross Product in Coordinates Like the dot product, the cross product takes ona fairly simple form in the coordinate representation. Ifwetaked= (x 1 z)and@=(x2 yo 2), then it can be shown that the vector BxB=(VYi2-yem) (am—2™m) (ry2-x2y1)) (11.1) has the desired properties. Since we have seen that there is only one vector satisfying all three defining conditions, this must be the desired cross product. EXERCISE 11-4 Show that the vector 3 x @ defined in (11.1) is perpendicular to both # and w. Although the cross product is not commutative, it is still linear, so #X (w; +B.) = (Bx) + (Px) and dx (ci) = c( x @). These properties can be easily verified using (11.1). When we take the cross product of two vectors which are only two dimensional, we extend to three dimensions: we pretend that our vectors are actually three dimensional, with a z-component Of 0. So if the vectors are (11 yi) and (x1 ya), we write them as (1) y: 0) and (x y2 0), tacking on a component of zero in the third dimension. We can then use (11.1). EXERCISE 11-5 Use (11-1) to verify that the cross product of two 2D vectors points either straight up or straight down (in three dimensions) 11.3 The Determinant We are now able to ask (and answer) a question which has been hanging about since the introduction of matrices, namely, is there a “size” for matrices? The size of a vector is simply its length. For matrices, however, the size depends on area, which we are only now ready to tackle. A very common way to represent a vector is to break it down in terms of the fundamental unit vectors i= (1 0) and j'=(0 1): (a 6) =ai+0j- (Similarly, in 3D we have (a b c) =ai+bj'+ck, @.