Undergraduate Texts in Mathematics Richard S. Millman George D. Parker Geometry A Metric Approach with Models Second Edition 6 SpringerSpringer New York Berlin Heidelberg Barcelona Hong Kong London Milan Paris Singapore Tokyo Undergraduate Texts in Mathematics Editors S. Axler F, W. Gehring K. A. RibetUndergraduate Texts in Mathematics Abbott: Understanding Analysis. Anglin: Mathematics: A Concise History and Philosophy. Readings in Mathemauics. Anglin/Lambek: The Heritage of Thales. Readings in Mathematics. Apostol: Introduction to Analytic Number Theory. Second edition. Armstrong: Basic Topology. Armstrong: Groups and Symmetry. Axler: Linear Algebra Done Right. Second edition. Limits: A New Approach to Bak/Newman: Complex Analysis. Second edition. Banchoft/Wermer: Linear Algebra Through Geometry. Second edition. Berberian: A First Course in Real Analysis. Bix: Conics and Cubics: A Concrete Introduction to Algebraic Curves. Brémaud: An Introduction to Probabilistic Modeling. Bressoud: Factorization and Primality Testing. Bressoud: Second Year Calculus. Readings in Mathematics. Brickman: Mathematical Introduction to Linear Programming and Game Theory. Browder: Mathematical Analysis: An Introduction. Buchmann: Introduction to Cryptography. Buskes/van Roolj: Topological Spaces: From Distance to Neighborhood. Callahan: The Geometry of Spacetime: An Introduction to Special and General Relavitity. Carter/van Brunt: The Lebesgue- Stieltjes Integral: A Practical Introduction. Cederberg: A Course in Modern Geometries. Second edition. Childs: A Concrete Introduction to Higher Algebra. Second edition. Chung: Elementary Probability Theory ith Stochastic Processes. Third edition. Cox/Little/O'Shea: Ideals, Varieties, and Algorithms. Second edition. Croom: Basic Concepts of Algebraic Topology. Curtis: Linear Algebra: An Introductory Approach. Fourth edition. Devlin: The Joy of Sets: Fundamentals of Contemporary Set Theory. Second edition. Dixmler: General Topology. Driver: Why Math? Ebbinghaus/Flum/Thomas: Mathematical Logic. Second edition. Edgar: Measure, Topology, and Fractal Geometry. Elaydi: An Introduction to Difference Equations. Second edition. Exner: An Accompaniment to Higher Mathematics. Exner: Inside Calculus. Fine/Rosenberger: The Fundamental Theory of Algebra. Fischer: Intermediate Real Analysis. Flanlgan/Kazdan: Calculus Two: Linear and Nonlinear Functions. Second edition. Fleming: Functions of Several Variables. Second edition. Foulds: Combinatorial Optimization for Undergraduates. Foulds: Optimization Techniques: An Introduction. Franklin: Methods of Mathematical Economics. Frazier: An Introduction to Wavelets Through Linear Algebra. Gamelin: Complex Analysis. Gordon: Discrete Probability. Hairer/Wanner: Analysis by Its History. Readings in Mathematics. Halmos: Finite-Dimensional Vector Spaces. Second edition. (continued after index)Richard S. Millman George D. Parker Geometry A Metric Approach with Models Second Edition With 261 Figures & SpringerRichard S. Millman Vice President for Academic Affairs California State University San Marcos, CA 92096 USA Editorial Board S. Axler Mathematics Department San Francisco State University San Francisco, CA 94132 USA George D. Parker Department of Mathematics Souther Illinois University Carbondale, IL 62901 USA EW. Gehring Mathematics Department East Hall University of Michigan Ann Arbor, MI 48109 USA K.A. Ribet Mathematics Department University of California at Berkeley Berkeley, CA 94720-3840 USA Mathematics Subject Classification (2000): 51-01, 51Mxx Library of Congress Cataloging-in-Publication Data Available. Printed on acid-free paper. © 1981, 1991 Springer-Verlag New York, Inc. All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Springer-Verlag New York, Inc., 175 Fifth Avenue, New York, NY 10010, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or here- after developed is forbidden. The use of general descriptive names, trade names, trademarks, etc., in this publication, even if the former are not especially identified, is not to be taken as a sign that such names, as understood by the Trade Marks and Merchandise Marks Act, may accordingly be used freely by anyone. Printed and bound by R. R. Donnelley & Sons, Harrisonburg, Virginia. Printed in the United States of America. 987654 ISBN 0-387-97412-1 ISBN 3-540-97412-1 SPIN 10830910 Springer-Verlag New York Berlin Heidelberg ‘A member of BertelsmannSpringer Science+Business Media GmbHTo Sherry and Margie Sor their love and supportThis page intentionally left blankPreface This book is intended as a first rigorous course in geometry. As the title indicates, we have adopted Birkhoff’s metric approach (i.c., through use of real numbers) rather than Hilbert’s synthetic approach to the subject. Throughout the text we illustrate the various axioms, definitions, and theorems with models ranging from the familiar Cartesian Plane to the Poincaré Upper Half Plane, the Taxicab Plane, and the Moulton Plane. We hope that through an intimate acquaintance with examples (and a model is just an example), the reader will obtain a real feeling and intuition for non- Euclidean (and in particular, hyperbolic) geometry. From a pedagogical viewpoint this approach has the advantage of reducing the reader's tendency to reason from a picture. In addition, our students have found the strange new world of the non-Euclidean geometries both interesting and exciting. Our basic approach is to introduce and develop the various axioms slowly, and then, in a departure from other texts, illustrate major definitions and axioms with two or three models. This has the twin advantages of showing the richness of the concept being discussed and of enabling the reader to picture the idea more clearly. Furthermore, encountering models which do not satisfy the axiom being introduced or the hypothesis of the theorem being proved often sheds more light on the relevant concept than a myriad of cases which do. The fundamentals of neutral (i.¢., absolute) geometry are covered in the first six chapters. In addition to developing the general theory, these chapters include a rigorous demonstration of the existence of angle measures in our two major models, the Euclidean Plane and the Poincaré Plane. Chapter Seven begins the theory of parallels, which continues with an in- troduction to hyperbolic geometry in Chapter Eight and some classical Euclidean geometry in Chapter Nine. The existence of an area function inviii Preface any neutral geometry is proved in Chapter Ten along with the beautiful cut and reassemble theory of Bolyai. The last (and most sophisticated) chapter studies the classification of isometries of a neutral geometry and computes the isometry groups for our two primary models. The basic prerequisite for a course built on this book is mathematical maturity. Certain basic concepts from calculus are used in the development of some of the models. In particular, the intermediate value theorem as it is presented in calculus is needed at the end of Chapter Six. The latter part of the last chapter of the book requires an elementary course in group theory. Courses of various lengths can be based on this book. The first six chapters (with the omission of Sections 5.2 and 5.4) would be ideal for a one quarter course. A semester course could consist of the first seven chapters, culminating in the All or None Theorem and the Euclidean/hyperbolic dichotomy. Alternatively, a Euclidean oriented course could include Section 7.1 and parts of Chapter 9. (The dependence of Chapter 9 on Chapter 8 is discussed at the beginning of Section 9.1.) A third alternative would include the first six chapters and the first three sections of Chapter 11. This gives the student a thorough background in classical geometry and adds the flavor of transformation geometry. A two quarter course allows a wider variety of topics from the later chapters, including area theory and Bolyai’s Theorem in Chapter 10. The entire book can be covered in a year. Mathematics is learned by doing, not by reading. Therefore, we have included more than 750 problems in the exercise sets. These range from routine applications of the definitions to challenging proofs. They may in- volve filling in the details of a proof, supplying proofs for major parts of the theory, developing areas of secondary interest, or calculations in a model. The reader should be aware that an asterisk on a problem does not indicate that it is difficult, but rather that its result will be needed later in the book. Most sections include a second set of problems which ask the reader to supply a proof or, if the statement is false, a counterexample. Part of the challenge in these latter problems is determining whether the stated result is true or not. The most difficult problems have also been included in these Part B problems. In this second edition we have added a selection of expository exercises. There is renewed emphasis on writing in colleges and universities which extends throughout the four years of the undergraduate experience. Going under the name of “writing across the curriculum”, this effort involves writing in all disciplines, not just in the traditional areas of the humanities and social sciences. Writing in a geometry course is discussed in more detail in Millman [1990]. We feel that the expository exercises add another dimen- sion to the course and encourage the instructor to assign some of them both as writing exercises and as enrichment devices. We have found that a multi- ple draft format is very effective for writing assignments. In this approach,Preface ix there is no finished product for grading until the student has handed in a number of drafts. Each version is examined carefully by the instructor and returned with copious notes for a rewrite. The final product should show that the student has learned quite a bit about a geometric topic and has improved his or her writing skills. The students get a chance to investigate either a topic of interest to them at the present or one that will be used later in their careers. This approach is especially useful and effective when many of the students are pre-service teachers. A few words about the book’s format are in order. We have adopted the standard triple numbering system (Theorem 7.4.9) for our results. Within each section one consecutive numbering system has been used for all theo- tems, lemmas, propositions, and examples for ease in locating references. The term proposition has been reserved for results regarding particular models. Reference citations are made in the form Birkhoff [1932] where the year refers to the date of publication as given in the bibliography. We would like to thank our students at Southern Illinois University, Michigan Technological University, and Wright State University whose feed- back over the past ten years has led to the changes and (we hope) the improvements we have made in this new edition. Our sincere thanks go to Sharon Champion and Shelley Castellano, who typed the original manu- script, and to Linda Macak, who typed the changes for this edition. Finally, we would like to thank our wives for putting up without us while we closeted ourselves, preparing this new edition.Computers and Hyperbolic Geometry After teaching a course out of the first edition of this book for several years, it became clear to the second author that there were all sorts of interesting computational problems in the Poincaré Plane that were a bit beyond the range of the average student. In addition, graphical aids could be very useful in developing intuition in hyperbolic geometry. Out of this realization grew a computer program POINCARE. The program was written in Pascal over a three year period and runs on MS-DOS computers. It allows graphical explorations in the Poincaré Plane as well as various calculations such as finding the midpoint of a segment, finding an angle bisector, finding the common perpendicular of two lines (if it exists), carrying out geometric constructions, testing quadrilaterals for convexity and the Saccheri property, solving triangles, finding the cycle through three points, and finding the pencil determined by two lines. All of the code is based on the theory presented in this book, with the exception of the hyperbolic trigonometry. Except for transformation geome- try, all topics in the book are represented. Perhaps in the near future a module on isometries will be added to the program. POINCARE is currently in its third major version and is ready for distribution. Readers of this book who are interested in more information, who wish to obtain a personal copy, or who wish to obtain it for use in their school computer lab should contact the author: George D. Parker 1702 West Taylor Carbondale, IL 62901Contents Preface Computers and Hyperbolic Geometry CHAPTER 1 Preliminary Notions 1.1 Axioms and Models: 1.2 Sets and Equivalence Relations 1.3, Punctions CHAPTER 2 Incidence and Metric Geometry 2.1 Definition and Models of Incidence Geometry 2.2. Metric Geometry 2.3, Special Coordinate Systems CHAPTER 3 Betweenness and Elementary Figures 3.1 An Alternative Description of the Cartesian Plane 3.2 Betweenness 3.3. Line Segments and Rays 3.4 Angles and Triangles CHAPTER 4 Plane Separation 4.1. The Plane Separation Axiom 4.2 PSA for the Euclidean and Poincaré Planes 4.3 Pasch Geometries vii wun 7 7 2 37 42 42 47 52 59 63 18 xt4.4 Interiors and the Crossbar Theorem 4.5 Convex Quadrilaterals CHAPTER 5 Angle Measure 5.1 The Measure of an Angle 5.2 The Moulton Plane 5.3 Perpendicularity and Angle Congruence 5.4 Euclidean and Poincaré Angle Measure (optional) CHAPTER 6 Neutral Geometry 6.1 The Side-Angle-Side Axiom 62 Basic Triangle Congruence Theorems 6.3. The Exterior Angle Theorem and Its Consequences 6.4 Right Triangles 6.5 Circles and Their Tangent Lines 6.6 The Two Circle Theorem (optional) 6.7 The Synthetic Approach CHAPTER 7 The Theory of Parallels 7.1 The Existence of Parallel Lines 7.2 Saccheri Quadrilaterals 7.3 The Critical Punction CHAPTER 8 Hyperbolic Geometry 8.1 Asymptotic Rays and Triangles 8.2 Angle Sum and the Defect of a Triangle 8.3. The Distance Between Parallel Lines CHAPTER 9 Euclidean Geometry 9.1 Equivalent Forms of EPP 9.2. Similarity Theory 9.3 Some Classical Theorems of Euclidean Geometry CHAPTER 10 Area 10.1 The Area Punction 10.2 The Existence of Euclidean Area 10.3 The Existence of Hyperbolic Area 10.4 Bolyai's Theorem CHAPTER 11 The Theory of Isometries Contents 81 86 124 124 131 135 143 150 165 169 178 187 196 196 205 214 224 224 230 239 248 248 264 an 285Contents 11.1 Collineations and tsometries 11.2 The Klein and Poincaré Disk Models (optional) 11.3 Reflections and the Mirror Axiom 11.4 Pencils and Cycles 11.5 Double Reflections and Their Invariant Sets 11.6 The Classification of isometries 11,7 The tsometry Group 11.8 The SAS Axiom in 11,9 The lsometry Groups of 6 and 3 Bibliography Index xiii 285 297 305 313 320 328 336 341 351 359 361CHAPTER 1 Preliminary Notions 1.1 Axioms and Models Our study of geometry begins with two basic concepts. One is the notion of points, and the other is the notion of lines. These are then related to each other by a collection of axioms, or first principles. For example, when we discuss incidence geometry below, we shall assume as a first principle that if A and B are distinct points then there is a unique line that contains both Aand B. In the early development of geometry the point of view was that an axiom was a statement that described the true state of the universe. Axioms were thought of as “basic truths.” Such axioms should be “self-evident.” Of the basic axioms stated by Euclid in his Elements, all but one was accepted by the mathematical community as “true” and self-evident. However, his fifth axiom, which dealt with parallel lines, was not as well received. While everyone agreed it was true (whatever that meant) it was by no means obvious. For over two thousand years mathematicians tried to show that the fifth axiom was a theorem which could be proved on the basis of the remaining axioms. As we shall see, such efforts were doomed to fail. With great foresight Euclid chose an axiom whose value was justified not only by its intuitive content but also by the rich theory it implied. The modern view is that an axiom is a statement of a useful property. When we assume an axiom holds, usually in a definition, we are saying that we want to discuss only those objects which possess this special property. We are making no statement as to whether the axiom is a statement about the real world. Rather, we are saying “accept the following as a hypothesis.” (For a nice discussion of the modern axiomatic method see Kennedy [1972].)2 1 Preliminary Notions Although we may use any consistent axioms we wish, the choice of axioms is really guided by three underlying principles. First, the axioms must be “reasonable” or “appealing” because they correspond to some in- tuitive picture which we have of some geometric property. Second, the axioms should be useful and lead to a rich variety of theorems and hence a rich mathematical structure. Third, the axioms must be consistent—there must not be any internal inconsistency or contradiction. As we shall see, Euclid’s choice of axioms (or rather the modern version which we shall present), does satisfy these conditions and lead to a rich subject, Euclidean geometry. BUT, there are other different choices of axioms that also lead to tich theories which are significantly different from the Euclidean ones. In particular there is an alternative to Euclid’s fifth axiom which develops into a particularly beautiful and interesting subject, hyperbolic geometry. One of the goals of this book is to investigate this alternative structure. The moment we mentioned points and lines above you probably started to visualize a picture, namely the plane with straight lines from high school geometry. This is proper and also very useful in helping you understand all the definitions that will follow. However, it does have its drawbacks. It is very tempting to try to prove a theorem or proposition or to do a homework problem by looking at a picture. If this is done, the picture may be confused with the geometry itself. A picture may be misleading, either by not covering all possibilities, or, even worse, by reflecting our unconscious bias as to what is “correct.” This often leads to an incorrect “proof by picture.” It is crucially important in a proof to use only the axioms and the theorems which have been derived from them, and not depend on any preconceived idea or picture. (Of course, we may use a picture as an aid to our intuition. The point here is that when the “final” proof of the result is written it cannot depend on the picture.) The discussion of pictures leads us to the idea of a model for a geometry. A model is nothing more than an example. Each model of a geometry is determined by giving a set whose elements will be called “points” and a collection of subsets of this set which will be called “lines.” For instance, if we are given the definition of an incidence geometry, we may write down as an example the standard Euclidean geometry we met in high school. We then must check to see if this example satisfies all the axioms that are listed in the definition of an incidence geometry. When we are done we will have one example. But there are many other examples of an incidence geometry and hence many models, as we shall see. Throughout this book there will be several models, but we will concen- trate on two particular ones: the Euclidean Plane and the Poincaré Plane. This will mean that we will have at our fingertips two strikingly different examples. Two main purposes will be served. The first is to insure that we do not reason by pictures (pictures in the Poincaré Plane are very different from those in the familiar Euclidean Plane). The second purpose is to give valuable insight by allowing us to work with several examples while at-1.2 Sets and Equivalence Relations 3 tacking some of the problems in the text. We can also benefit from different examples by examining a newly defined concept in light of the definition. For example, a circle in the “taxicab metric” of Chapter 2 is an amusing phenomenon. This examination will add to our understanding as to what the definition is saying (and what is is not saying). At this point we cannot emphasize enough that the only proof that can be given of a theorem or proposition in a geometry is one based just on the axioms of that geometry. We must not go to a model and show that the theorem holds there. All we would have shown in that case is that the theorem is true in that particular model. It might be false in another model (and hence false in general). There are many statements that are true in the Euclidean Plane but false in the Poincaré Plane, and vice versa. Whenever we introduce a new axiom system or add axioms to an old system, we are changing the requirements for the system. For instance, consider the statement that “a slurb is a set which has only letters in it.” The set S = {X, Y,z} satisfies this statement and so is an example or model ofa slurb. If we further define that “a big slurb is a slurb that contains only capital letters” then the set S may or may not be a model for the new (en- larged) axiom system. (Of course, S is not a model for this system; that is, S is not a big slurb.) This means that if we continue to use certain models we must prove that they obey the new axioms that are cited. Do not confuse this with proving an axiom. Axioms are statements of desirable properties to be studied and cannot be proved. Verifying that something is a model of a certain axiom system is very much like making a general statement and then showing that it is applicable in the particular instance you want. What then is a geometry? Formally, a geometry consists of two sets—a set of points and a set of lines—together with a collection of relationships, called axioms, between those two sets. On the other hand, a model for the geometry is just an example. That is, a model for a geometry is a mathematical entity which satisfies all of the axioms for the geometry. It is important not to confuse a model with the geometry of which it is a model. 1.2 Sets and Equivalence Relations Intuitively a set, S, is a collection of objects which are called elements. It must be described by a very specific rule which lets us determine if any particular object belongs to the collection. (The collection of tall people is not a set because the terminology is not precise—it is subjective. The collec- tion of all living people at least 2 meters tall is a set-—the characteristic for tall is made precise.) We write a € S to mean that the object a belongs to S, and read this as “a is an element of S.” Similarly we write a¢ S to mean that a does not belong to the set S.4 1 Preliminary Notions Definition. The set T is a subset of the set S (written T < S) if every element of T is also an element of S. The set T equals the set S (written T = S) if every element of T is in S, and every element of S is in T. (Hence T=S if and only if Tc S and ScT) The empty set is the set with no members, and is denoted @. Note @ CS for every set S. As usual, the notation T = {x € S|---} means that the elements of T are precisely those elements of S which satisfy the property listed after the bar, |. Definition. The union of two sets A and B is the set A U B= {x|x Aor x € B}. The intersection of two sets A and B is the set A B= {x|xe A and xe B}. If An B= then A and B are disjoint. The difference of two sets A and B is the set A—B = {x|x € A and x ¢ B}. The following example illustrates several of the above ideas. Note in par- ticular the basic way we show two sets S and T are equal: we show that Sc Tand that TcS. Example 1.2.1. Show that A n (BU C) =(An B)U(AN C). SoLUTION. We first show that AN(BUC)C(ANB)U(ANC). Let xeAN (BUC). Then xe A and x e BUC. Since xe Bu C either xe BorxeC (or both!). If xe B then xe AT B. If xeC then xe ANC. Either way x€(4 0 B)U(AN C). Thus An (BUC) C(AN B)U(ANC). Next we show that(4 9 B)U(ANC)EAN(BU C).Letxe(An B)U (An C).Ifxe Ao B then xe A and xe B. Hence xe BUC and xe An (Bu C). Similarly, if xe A OC then xe A and xe C. Hence, xe BUC and x€AM(BUC). In either case, xe AN (BUC). Thus (49 B)U (ANC) cAN(BUC). Since An (BU C)c (AN B)U (An C)and (49 B)U(ANC)C AN (BU C), we have An (BU C) =(A0 B)U (ANC). o Definition. Let A and B be sets. An ordered pair is a symbol (a, b) where aé A and be B. Two ordered pairs (a, b) and (c,d) are equal if a=c and b =d. The Cartesian product of A and B is the set Ax B= {(a, b)lae A and be B}. Because of the use of the word “symbol”, the definition above is somewhat informal. The basic idea that the entries are “ordered” comes from the definition of equality: (a,b) and (b, a) are not equal unless a = b. Thus, changing the order of the entries leads to a different object. It is possible to give a purely set-theoretic definition of (a, 6). This is done in problem B16,1.2 Sets and Equivalence Relations 5 where the reader is asked to prove that (a, b) = (c, d) if and only if a = c and b = d using the formal definition given in that problem. Note that the notation R? to denote the set of ordered pairs of real numbers is an adaptation of exponential notation to represent R x R. As a first use of the concept of ordered pairs we present a way to say that two elements in a set are related in some particular way. A motivating example is given by the idea “less than”. The graph of the inequality x < y in R? consists of all ordered pairs (a, b) € R? such that a < b. (See Figure 1-1.) When we say that 2 <3, we are saying that (2, 3) is part of the graph. Conversely, since (— 3,2) is part of the graph, —3 <2. Thus the graph carries all the information of the “less than” relation. A binary relation is a generalization of “less than” that is described in terms of a graph. Figure 1-1 Definition. A binary relation, R, on a set S is a subset of S x S. If (s,t)eR then we say that s is related to ¢. Example 1.2.2. Each of the following is a binary relation on the set R of real numbers. A= {(s,)eR?|s = 1 + 2}. B= {(s, t) € R?|st is an integer}. C={(s,)eR*Is < th. D = {(s,t)¢ Rs? + =1} oO We frequently name relations using symbols such as < (for relation C above), ~, ll, or ~ instead of letters. We then indicate that two elements are related by placing the name of the relation between the ele- ments: (3, 5) C becomes (3, 5)€ <, which becomes 3 < 5. Thus we may make statements about “the relation ~” and write statements such as “a ~b". Note that if two elements a, b are not related by the relation ~ we write add.6 1 Preliminary Notions Because the idea of a relation depends on ordered pairs, the order that we write the symbols is important: 2 < 5 but 5 ¢ 2. For some special relations, like those below, the order is not important—the relation is symmetric. Note that if ~ is a relation on S and ae, then it is possible that there is no element b with a ~ b, For example, if S is the set of positive integers, and if the relation is > (greater than) then there is no b € S with 1 > b. In this case 1 is not related to anything. Definition. A binary relation, ~, on S is an equivalence relation if for every a,b,andceS (i) a ~ a (reflexive) (ii) a ~ b implies b ~ a (symmetric) (iii) a ~ band b ~ c implies a ~ c (transitive). Note that an equivalence relation is a binary relation that satisfies three axioms. Example 1.2.3. Let Z be the set of integers and define a ~ b if a—b is divisible by 2. Show that ~ is an equivalence relation. Sotution. To say that a — b is divisible by 2 means that there is an integer k such that a — b = 2k. Thus a~b_ ifand only ifthereiske Z witha —b = 2k. (i) Let ae Z. Then a — a= 0 = 2-050 that a ~ a and ~ is reflexive. (ii) Suppose that a, b € Z and a ~ b. Then there is a ke Z with a — b = 2k. This means that b — a = 2(—k). Since —k € Z, we have b ~ a. Thus ~ is symmetric. (iii) Ifa ~ b and b ~ c then there are numbers k, € Z and k, € Z with a-—b=2k, and b—c=2k). Adding these equations we obtain a — c = 2(k, + k,) and so a~c. Thus ~ is transitive. ~ is therefore an equivalence relation. a Definition. If a and b are integers then a is equivalent to b modulo n if a— b= kn for some integer k. This is written a = b(n) and means that a — b is divisible by n. The above example shows that =(2) is an equivalence relation. In Prob- lem A7 you will show that =(n) is an equivalence relation for any n. Example 1.2.4. Show that none of the binary relations in Example 1.2.2. is an equivalence relation. SoLuTion. A is not reflexive. It is certainly not true that a = a + 2 for alla. (Neither is it symmetric or transitive.)1,2 Sets and Equivalence Relations 7 Bis not reflexive or transitive. Cis not symmetric. Dis not reflexive or transitive. Oo In an equivalence relation we view several elements of S as being alike (or equivalent) if they have similar properties. In Example 1.2.3 all the odd numbers are related to each other and thus are equivalent. It is convenient to have a name for the set of all elements related (or equivalent) to a given element. Definition. If ~ is an equivalence relation on the set S and se S, then the equivalence class of s is the subset of S given by (s] = {x € S|x ~ s} = {xe S|s ~ x}. Example 1.2.5. In Example 1.2.3 the equivalence class of 3 is the set of odd integers, and the equivalence class of 2 is the set of even integers. Note in this case that if x, y € Z then either [x] = [y] or [x] n[y] = @. a Example 1.2.6. Let S = {1, 2,..., 100} and define x ~ y if x and y have the same number of digits in their base 10 representation. Then ~ is an equiv- alence relation and, for example, [5] = (1,2,3,4,5,6,7,8,9} = [7] = [9] =[8], etc. [11] = {s€S|10 < s < 99} = [63] = [43], etc. [100] = {100}. Note that, although different equivalence classes may have a different number of elements, we still have the result that two equivalence classes are either equal or disjoint. This is true in general as we now see. Oo Theorem 1.2.7. If ~ is an equivalence relation on S and if s,t € S then either [s]o[]=@ or [s]=[¢] Proor. We will show that if the first case is not true (ie. [s] > [¢] # @) then the second holds. This is the standard way we show either... or... results. Assume that [s] 0 [t] # &. Then there is an x € [s] > [ct]. Hence x € [s] and x e[¢]. Thus x ~ s and x ~ t. By symmetry s ~ x, and then by tran- sitivity, s~ x and x ~ t imply that s ~ ¢. We use this to show [s] < [¢]. Let y [s]. Then y ~ s and, since s ~ t, we also have y ~ t by transitivity. Thus y ¢ [1]. Hence [s] ¢ [#]. Similarly, since t ~ s, we can show [¢] < [s]. Hence [s] = [¢]. oO8 1 Preliminary Notions PROBLEM SET 1.2. Throughout this problem set, A, B, and C are sets. Part A. 1. ACB prove that A NCC BNC. 2. Prove that Cm (B—A) = (C 7 B)—A. 3. Prove that AU (BO C)=(AU B)A(AU C). 4. Suppose that 4 c C and Bc C. Show that Am B = @ implies that Bc C—A. If Bc C—A show that An B=. 6. a. If x ~ y means that x — y is divisible by 3, show that ~ is an equivalence rela- tion on the set of integers. b. What is [3]? [6]? [9]? (1)? [5]? 7. Show that =(n) is an equivalence relation on the integers for any n. What are the equivalence classes? 8 Let R? = {P =(x,y)]x and y are real numbers}. We say that P, =(x,,y,) and P, = (x3, y2) are equivalent if x} + y} = x3 + y}. Prove that this gives an equiva- lence relation on R?, What is [(1,0)]? [(0,1)]? [(2,2)]? [(0,0)]? What does an equivalence class “look like?” 9. The height, h, of a rectangle is by definition the length of the longer of the sides. The width, w, is the length of the shorter of the sides (thus h = w > 0). If the rec- tangle R, has height h, and width w, and the rectangle R, has height h, and width w,, we say that R, ~ R, if h,/w, = hz/w,. Prove that this defines an equivalence relation on the set of all rectangles. 10. Let the triangle T, have height h, and base b, and the triangle T, have height hz and base b,. We say that T, ~ T; if bh, = bh2. Show that this is an equivalence relation on the set of all triangles. Show that T, €[T)] if and only if T, has the same area as T. Part B. “Prove” may mean “find a counterexample”. 11. Prove that (A U B)—C = (A—C) u(B—C). 12. Prove that (4A—C) n(B—C) = (An B)—C. 13. Let S be the set of all real numbers. Let s, ~ s, if s} = s3. Prove that ~ is an equivalence relation on S. What are the equivalence classes? 14. Let X be the set of all people. We say that p, ~ p2 if p, and p, have the same father. Prove that this is an equivalence relation. What are the equivalence classes? 15. Let X be the set of all people. We say that p, ~ p> if p; lives within 100 kilo- meters of p2. Show that ~ is an equivalence relation. What are the equivalence classes? 16. A careful definition of ordered pair would be the following: If ae A and be B, then the ordered pair (a, b) is the set (a, b) = {{a, 1}, {b, 2}. Use this defini9 1,3 Functions prove that (a, b) = (c, d) if and only if a = c and b = d. (Be careful: {{a, 1}, {b, 2}} = {{c, 1}, {d, 2}} does not immediately imply that {a, 1} = {c, 1}.) 17. Give a careful definition of an ordered triple (a,b,c) and then prove that (a, b, &) = (d, e,f) ifand only ifa = d,b = e,andc =f. 18. Let R? = {3 = (x, y, z)lx, y, and ze R}. Let 3, # e R°—{G}. We say that 3 ~ iif there is a non-zero real number, A, with 7 = Aw. Show that ~ is an equivalence relation on R°—{0}. What are the equivalence classes? The set of all equivalence classes will be called the (real) Projective Plane. 19. Let X = {(x, y)|x eZ and ye Zand y 4 0}. We define a binary relation on X by (%1,¥1) ~ (2s Yo) ifand only if x,y2 = x2¥1- Prove that ~ is an equivalence relation. What are the equivalence classes? 20. Let ~ be a relation on S that is both symmetric and transitive. What is wrong with the following “proof” that ~ is also reflexive? “Suppose a ~ b. Then by symmetry b ~ a. Finally by transitivity a ~ b and b ~ a imply a ~ a. Thus ~ is reflexive,” Part C. Expository exercises. 21. How would you explain to a high school audience the notion of an equivalence relation? In that context, what would an equivalence mean? Where would it occur in their daily lives? 22. Think of a binary relation as a graph. What is the geometric (graphical) meaning of the three axioms for an equivalence relation? 1.3 Functions In this section we review the standard material about functions and bi- jections. The latter notion is an important part of the Ruler Postulate which will appear in the next chapter. We will continue to use R to denote the set of real numbers and Z to denote the set of integers. Definition. If S and T are sets, then a function f: S > T is a subset fc S x T such that for each se S there is exactly one ¢ € T with (s, t) € f. This unique element ¢ is usually denoted f(s). S is called the domain of f and T is called the range of f. In a very intuitive manner we may view a function as an archer who takes arrows (elements) from her quiver S and shoots them at a target T. In this analogy we say that the element s € S “hits” the element f(s) € T. Following standard conventions, we frequently describe a function f by giving a formula (or rule) for computing f(s) from s. Note that the function consists of this rule together with the two sets S and T.10 1 Preliminary Notions Example 1.3.1. Let f:R > R by the rule f(x) = x?. Let g:Z > R by the rule g(x) = x?. Note that f is not equal to g—they have different domains. Now let R* = {x € R|x > 0} and let h:R— R* by the rule h(x) = x”. Note that f and h are not equal—they have different ranges. Oo Definition. If f:S + T is a function then the image of f is Im(f) = {te T|t = f(s) for some se S}. Thus, Im(/) consists of the elements of T that are actually “hit” by f. Of course, Im(f) < Range(/), but these sets need not be equal. In Example 1.3.1, Im(f) = R* but Range(/) = R. (Some mathematicians use the word “range” to mean “image”. They then use “codomain” to mean what we call the “range”) Definition. f:S + T is surjective if for every te T there is an se S with Si =t. ‘In keeping with the target analogy above, this means that all elements of the target T are “hit.” Of course, an element may be hit more than once. That is, there may be several s € S such that f(s) = ¢. It is common usage to say that a function is “onto” instead of “surjective”. In this text we shall use the more correct terminology “surjective”. Example 1.3.2. Show that f:R—R by f(x) = x>—1 is surjective while g:R + R by g(x) = x? — 1 is not surjective. SoLuTION. To show that f is surjective we must show that for every te Range(f) = R there is an se Domain(f) with f(s) = t. That is, we must show that the equation -l=t (3-1) has a solution for every value of r. Since every real number has a cube root, we may set s = ./t + 1. Then f(s) =(ft+ IP -Laet1-lse Hence f is surjective. To show that g is not surjective we need only produce one value of ¢ such that the equation v—-l=t (3-2) does not have a solution. Let t = —2. Then a solution to Equation (3-2) must satisfy s-1=-2 or s=-1. This clearly cannot occur for any real number s. Hence g is not surjective. O1.3 Functions i Example 1.3.2 illustrates how we actually attempt to prove that a function is surjective. We set up the equation f(s) = t and try to solve for s (given ¢). At least one solution for every value of t must be found. Example 1.3.3. Let T = {t¢ R|t = e7'/*} and h:R > T be given by h(s) = e* 5. Prove that h is surjective. SoLution. Given te T, note that Int > —4. Thus 1+ 4Int>0. Let s be the “obvious choice”: 1+J/14+4int see. (3-3) It is easy enough to show that Equation (3-3) definesanse Rwithh(s)=t. O The above represents an absolutely correct (and totally unmotivated) solution to the problem. The way that we came up with the “obvious choice” (and the way to attack the problem) is to attempt to solve the equation h(s) = t. Thus on scrap paper you might write: For fixed 1, solve for s: erat. First take the natural logarithm of both sides to get st —s=iInt or s?—s—Int=0. An application of the quadratic formula gives saltvi+4ine = Because we only need to exhibit one solution, we take the “+” sign: s=4(1+,/1+4 4in¢). Of course, taking s = 4(1 — //1 + 41n zt) would also show that f is surjective. Once we have found what we believe is the solution, we must verify that it is correct as in Example 1.3.3. This is similar to solving an equation in algebra and then checking the solution to be sure that we have not found an extraneous solution. Note that it was crucial that we verified that 1 + 4 In ¢ > 0. If the range had included numbers t < e™!/* we would not always have 1 + 41Int > 0 and the function would not have been surjective. The concept ofa surjective function deals with whether or not an equation can be solved. Another important idea is the notion of an injective function, which deals with the number of solutions to an equation. Thus “surjective” deals with existence of a solution while “injective” deals with uniqueness. Definition. f:S + T is injective if f(s,) = f(s2) implies s; = s). In terms of the target analogy, f:S — T is injective if no two arrows hit12 1 Preliminary Notions the same place on the target. (Note that this says nothing about whether all of the target is hit.) It is common practice to use the term “one-to-one” to mean injective. In keeping with our practice regarding the word “surjective” we shall only use the term “injective” and not “one-to-one.” An alternative way of defining injective would be: f is injective if s, # s, implies f(s,) # f(s2). While this is not as convenient to use, there are times (e.g., Theorem 1.3.8) when it is helpful. Example 1.3.4. Let f:R > R be given by f(s) = e° *!. Show that f is injective. SoLution. We assume that there are real numbers s, and s, such that etl a psitt (3-4) We must show that the only way that this occurs is if s; = s,. If we take the natural logarithm of both sides of Equation (3-4) we obtain s}+1l=s3+1 or s}=s}. Since every real number has a unique cube root we must have s, = s,. Thus f is injective. Oo Example 1.3.5. Show that h:R* + R* by h(x) = x? is injective. Sotution. Assume that h(s,) = h(s) so that s} = s3. Then by taking square roots we have s; = +5). However, since the elements of R* are not negative, both s, and s, must be greater than or equal to zero. Hence s, # —sz (unless both are 0) and so s, = s,. Thus h is injective. oO The words “injective” and “surjective” are adjectives. If we wish to have a noun it is common to say “injection” for “injective function” and “sur- jection” for “surjective function.” Note that in Example 1.3.4 the function was an injection but not a surjection, whereas in Example 1.3.5 the function was both a surjection and an injection. The function h of Example 1.3.3 gives an example of a function which is a surjection but not an injection (since h(0) = h(1) = 1). There are many examples of functions which are neither injective or surjective. However, a function which is both has a special name. Definition. f:S > T is a bijection if fis both an injection and a surjection. Example 1.3.5 is a bijection. The term “one-to-one correspondence” is also common, but we shall not use it here because readers have a tendency to confuse the term with the idea of “one-to-one.” Recall the definition of the composition of functions. Definition. If f:S > T, g: U + V, and Im(f) ¢ U, then the composition of f and g is the function g o f:S + V given by (g © f)(s) = g(f(s)).1.3 Functions 13 Notice that the domain of g must contain the image of f in order for the composition of f and g to be defined. Example 1.3.6. If f: R--R— {0} is given by f(s)=2+ sin(s) and g:R—{0}>R is given by g(t) = 1/t find go f. SoLuTion. go f:R > R is given by 1 9° f(s) = A(S(s)) = Ts _ 1 ~ 2+ sin(s)" Theorem 1.3.7. If f:S—+ T and g:T — V are both surjections then go f is also a surjection. oO Proor. Let v c V. We must show that there is an s € S such that (g f)(s) = v. Since g is surjective, there is a tf T with g(t) = v. Then since f is surjective there is an se S with f(s) = t. Now (9° f(s) = a(f() = gl) = v so that go f is a surjection. a The next two results are left as Problems A6 and A7. Theorem 1.3.8. If {:S + T and g:T > V are both injections then go f:S + V is an injection. Theorem 1.3.9. If f:S + T and g:T +V are both bijections then g « f:S + V is also a bijection. If f:S + T is a bijection then for cach t € T there is a unique se S with S(s) = t. This allows us to assign to each t € T a corresponding element s € S. Thus we have manufactured a new function (called the inverse) which goes backwards. More formally we have Definition. If f:S— T is a bijection, then the inverse of f is the function g:T +S which is defined by g(t)=s, where sis the unique element of S with f(s)=t (3-5) The function g is frequently denoted f~. If f is the natural logarithm function given by f(s) = In(s), then the inverse of f is the exponential function g given by g(t) = e since e'™ = s. Definition. If S is a set, then the identity function ids: S — S is given by idg(s) = s.14 1 Preliminary Notions Theorem 1.3.10. If {:S— T, then f is a bijection if and only if go f =ids and f > g = id, for some function g:T — S. Furthermore, the inverse of f is g in this case. Proor. First we shall prove that if there is a function g: T+ S with fog = id; and g o f = ids then f is a bijection and g is its inverse. Assume there is a function g: T+ S with fog = id; and go f = ids. If te T then g(t)eS and f(g(t)) = id;(t) = t. Hence te Im(f) and f is sur- jective. If f(s;) = f(sz) for s,, s2 €S, then g(f(s,)) = g(f(s2)) or ids(s,) = ids(s.) or s; = 82. Thus f is injective and hence is a bijection. Finally ift = f(s) then g(t) = g(f(s)) = ids(s) = s, so that g satisfies Equation (3-5). Thus g is the inverse of f. Next we shall show that if f is a bijection then there is a function g:T + S with f » g = id; and go f = ids. Since f is a bijection it has an inverse. Call this inverse g: T + S. Then g(t) = s whenever f(s) = t. In particular, if t ¢ T then SG) = f(s) =t forallte T so that f og = idp. Also ifs € T let t = f(s). Then by Equation (3-5), g(t) = s so that g(f(s))=s forallseS. Thus go f = ids. fi Example 1.3.11. Let P* = {te Rt > 0} and set f:R>BP* by f(s) =e. What is f-':P* +R? SoLution. Equation (3-5) says that we must find a function g = f~*:P* > R with the property that g(t) = s whenever e* = t. This function is g(t) = In t. Since e'=7 and Ine=s, Theorem 1.3.10 gives a formal proof that our solution is correct. Qa Theorem 1.3.10 may be used to prove the next result. Theorem 1.3.12. If f:S—+T and h:T— V are bijections then (ho f)"* = fioh PRosLEM SET 1.3 Part A. 1. Prove that each of the following functions is surjective. a. f:R—{0} + R—{2}; fis) = 2+ I/s b. g:R-+R; g(s) =s° — 6s1.3 Functions 15 chi? R; h(x, y) = xy d. IR {x|-2 < x < 2}; () = 2cost Xv . Prove that each of the following functions is injective. a. f:R + R by fis) = fst +1 — {0} + R by g(x) = 3 - 1/(2x) c. AiR? + R? by h(x, y) = (y?, xy? +x + 1) d, [:P*—+ R? by I(t) = (t, Int), where P+ is the set of positive real numbers . Which of the functions in Problem 1 are bijections? In these cases, find the inverse. . Which of the functions in Problem 2 are bijections? In these cases, find the inverse. . Give an example of a function f:R -» R which is neither injective nor surjective. . Prove Theorem 1.3.8. . Prove Theorem 1.3.9. eI Awe wD . Prove Theorem 1.3.12. *9, If f:S + T is a bijection prove that f~!:T— S is also a bijection. Part B. “Prove” may mean “find a counterexample”. 10. If g: TV and f:S— T and if ge f is surjective prove that both g and f are surjective. Il. If g:T + V and f:S—+T and if ge f is injective, prove that both g and f are injective. 12. If f:S + T we define a binary relation on S by 5, ~ s; if f(s,) = f(s). Prove that ~ is an equivalence relation on S. 13. Ih: X + Y and g: Y + Z and go his a bijection, prove that g is surjective and h is injective. 14. If f:X + Yand A c Bc X, prove that f(B—A) < f(B)—f(A). 15. If f:S +R and g:S +R are both injective and h:S— R is defined by h(s) = S(8)+9(s), prove that jective. 16. If f:S-+R and g:S—R are both surjective and h:S—R is defined by h(s)= S(8) + 9(9), prove that h is surjective. 17. Prove that the set of all bijections from S$ to S forms a group. (Hint: Use composition for the multiplication.) 18. Let S be the set of all polynomials. Let f:S— S by f(p(x)) = d/dx(p(x)) and g:S + S by g(p(x)) = fi p(t) dt. Show that f og = ids but that f is not a bijection. Does this contradict Theorem 1.3.10?16 1 Preliminary Notions Part C. Expository exercises. 19. Discuss at an intuitive level what injective and surjective functions are. How would you explain them to an engineering student? to your parents or spouse? to high school students? Note how your answer changes depending on your audience!CHAPTER 2 Incidence and Metric Geometry 2.1 Definition and Models of Incidence Geometry In this section we shall define the notions of an abstract geometry and an incidence geometry. These are given by listing a set of axioms that must be satisfied. After the definitions are made, we will give a number of examples which will serve as models for these geometries. Two of these models, the Cartesian Plane and the Poincaré Plane, will be used throughout the rest of the book. As we discussed in Section 1.1, a geometry is a set Y of points and a set £ of lines together with relationships between the points and lines. What relationship shall we insist on first? We would certainly want every two points to be on some line, and we would want to avoid the pathology ofa line with only one point. We will add more relationships between points and lines later and so change the nature of the geometry. Definition. An abstract geometry »/ consists of a set *, whose elements are called points, together with a collection Y of non-empty subsets of YF, called lines, such that: (i) For every two points A, Be & there is a line le £ with Acland Bel. (ii) Every line has at least two points. If of ={/,L} is an abstract geometry with Pe Y, le Y and Pel, we say that P lies on the line |, or that | passes through P. In this language the first axiom of an abstract geometry reads: “every pair of points lies on some line.” A word of warning is necessary, however. Just because we use the word “line,” you should not think “straight line.” “Straight” is a biased 1718 2. Incidence and Metric Geometry term that comes from your previous exposure to geometry, and particu- larly Euclidean geometry. To us a “line” is just an element of Y. See Proposi- tion 2.1.2 below, where the “lines” do not “look straight.” Proposition 2.1.1. Let Y = R? = {(x, y)|x, ye R}. We define a set of “lines” as follows. A vertical line is any subset of RR? of the form L, = {(x, y) € R?|x = a} (1-1) where a is a fixed real number. A non-vertical line is any subset of R? of the form Lyn = {(x, 9) € R?| y = mx + B} (1-2) where m and b are fixed real numbers. (See Figure 2-1.) Let £, be the set of all vertical and non-vertical lines. Then € = {R?, Y,} is an abstract geometry. Proor. We must show that if P =(x,,y,) and Q =(x,y2) are any two distinct points of R? then there is an |e Y, containing both. This is done by considering two cases. Case 1. If x, = x, let a=x,=x,. Then both P and Q belong to /= Lie Le. Case 2.1fx, # x, we show how to find mand b with P, Qe L,,,,. Motivated by the idea of the “slope” of a line we define m and b by the equations: m= 22M and b=y,— mx. XX It is easy to show that y, = mx, + b and that y, = mx, + b, so that both P and Q belong to! = L,,, € Le. It is easy to see that each line has at least two points so that is an abstract geometry. Oo Definition. The model @ = {R?, %;} is called the Cartesian Plane (The nota- tion L, and L,,, will be reserved for the lines of the Cartesian Plane and certain other models that are developed later using the same set of points and lines.) Figure 2-12.1 Definition and Models of Incidence Geometry 19 We use the letter E in the name of the set of Cartesian lines (%_) to remind us of Euclid, the author (c. 300 B.C.E.) of the first axiomatic treat- ment of geometry. Later we shall add the additional structures of distance and angle measurement to the Cartesian model to obtain the familiar Euclidean model of geometry that is studied in high school. The name Cartesian is used to honor the French mathematician and philosopher René Descartes (1596-1650), who had the revolutionary idea of putting co- ordinates on the plane. Our verification that @ satisfied the axioms depended heavily on the use of coordinates. Descartes is also responsible for many of our conventions in algebra, such as using x, y, z for unknown quantities and a, b, c for known quantities, and for introducing the exponential notation x". Recall from your elementary courses that there are other ways to describe straight lines in R?. The way chosen above (that is, through L, and L,,,) is the best suited for this chapter. In Chapter 3, the results are proved most easily if the vector form of the equation of a line is used, and so we will start to use that approach there. Another approach is included in Problem Al4. Proposition 2.1.2. Let % =H = {(x, y) € R?|y > 0}. As in the case of the Cartesian plane, we shall describe two types of lines. A type I line is any subset of H of the form ol = {(x, ye H|x = a} (1-3) where a is a fixed real number. A type II line is any subset of H of the form cL, = {(x, ye H|(x — 0)? + y? = 7} (1-4) where c and r are fixed real numbers with r > 0. (See Figure 2-2.) Let Ly be the set of all type I and type II lines. Then # = {H, %y} is an abstract geometry. Proor. Let P =(x,,y,) and Q =(x2,y2) be distinct points in H so that y, > Oand y, >0. Case 1. If x, = xz then P and Q both belong to /= ,Le ¥y where a=x, =X). Case 2. If x; # xX, define c and r by pouciitd-¥ 2(x2 — x4) r=J/ (x1 — of + yi. (1-6) (In Proposition 2.1.5 below we will see what led to this choice of c and r.) In Problem A6 you will show that P and Q both belong to | = .L,€ Ly. It is easy to see that each line has at least two points so that 3 is an abstract geometry. im (1-5)20 2 Incidence and Metric Geometry Figure 2-2 Definition. The model # = {H, %,} will be called the Poincaré Plane. (The notation ,L and .L, will be used only to refer to lines in 9) 3 is called the Poincaré Plane in honor of the French mathematican Henri Poincaré (1854-1912) who first used it. Poincaré was a prolific re- searcher in many areas of pure and applied mathematics. He is particularly remembered for his work in mechanics, for his study of elliptic functions which tied analysis and group theory together, and for his work in geometry which led to the development of modern topology. The letters #, H, and H are used to remind us of the word “hyperbolic”. We shall see later in this chapter that the hyperbolic functions are important in this model, just as the trigonometric functions are important in Euclidean geometry. Once we have added more structure to # it will be a model of what we call a hyperbolic geometry. In the models given in Propositions 2.1.1 and 2.1.2 it seems clear that through any two points there is a unique line. This need not be the case in all abstract geometries as we see in the next example. This example will have a particular subset of R? = {(x, y, z)|x, y, z € R} as its set of points, % Definition. The unit sphere in R? is S? = {(x, y,2) € R3|x? + y? + 27 = 1}. A plane in R? is a set of the form {(x, y,2) € R°|ax + by + cz =d} where a, b, c, d are fixed real numbers, and not all of a, b, c are zero. Note that in the definition of a plane if the constant d = 0, then the plane goes through the origin (0,0,0). Definition. A great circle, Y, of the sphere S? is the intersection of S? with a plane through the origin. Thus is a great circle if there are a, b, ce R, not all zero, with GY = {(x, y, 2) € Slax + by + cz = 0}.2.1. Definition and Models of Incidence Geometry 21 Figure 2-3 Proposition 2.1.3. Let S = S? and let Ly be the set of great circles on S?. {S?, Lp} is an abstract geometry. Proor. We must show that if P = (x,, y,,2,)€ S? and Q = (x2, y2, 22) € S? then there is a great circle Y with Pe Y and Qe ¥Y. Thus we must find a, b, c real numbers (not all zero) such that ux, + by, +¢z;=0 and ax, + by, + cz, =0. (1-7) View Equations (1-7) as two equations in the three unknowns a, b, c. Since two homogeneous linear equations in three unknowns always have a non-zero solution (in fact, infinitely many solutions), we may always find a, b, and c solving Equations (1-7). Thus, there is a great circle with PeGand Q « G. Finally each great circle has at least two points. a Definition. The Riemann Sphere is the abstract geometry # = {S?, Lp }. The Riemann Sphere is named after G. B. F. Riemann (1826-1866) who wrote foundational papers in geometry, topology and analysis. His paper on geometry, Uber die Hypothesen, welche der Geometrie zu Grunde liegen (On the Hypotheses which lie at the Foundation of Geometry), which was written in 1854 (see Spivak [1970, vol. II] or Smith [1929]), provided geometry with a great unifying idea, that of a Riemannian metric. This concept, which is quite advanced, is the basis for modern differential geometry (see Millman and Parker [1977]) and the mathematics of Einstein's theory of general relativity. The name Riemann Sphere comes from Rie- mann’s work in functions of a complex variable and not from his work in geometry. Note that it is “geometrically obvious” and was proven above that any two points on S? lie on a great circle. However, unlike the first two examples, two points on S? may have more than one great circle joining them. Con- sider the north and south poles N and S as in Figure 2-4. There are infinitely many great circles joining N to S. The uniqueness of lines joining two points is such an important concept that it is singled out in the definition of in- cidence geometry.22 2 Incidence and Metric Geometry (7 Figure 2-4 Definition. An abstract geometry {/,#} is an incidence geometry if (i) Every two distinct points in ¥ lie on a unique line. (ii) There exist three points A, B, Ce ¥ which do not lie all on one line. Notation. If {%,#} is an incidence geometry and P, Qe ¥, then the unique line | on which both P and Q lie will be written as! = PQ. _It is useful to restate the second axiom of an incidence geometry in terms of the concept of collinearity. Definition. A set of points 9 is collinear if there is a line | such that P cl. is non-collinear if is not a collinear set. Sometimes we will say that “A, B, and C are collinear” instead of saying “{A, B, C} is a collinear set.” This abuse of notation and language makes it easier to state some results. Axiom (ii) of the definition above can be restated as (ii)’ There exists a set of three non-collinear points. Although the Riemann Sphere is not an incidence geometry both the Cartesian Plane and the Poincaré Plane are, as we shall now see. Proposition 2.1.4. The Cartesian Plane @ is an incidence geometry. Proor. We must show that two distinct points uniquely determine a Car- tesian line. Let P = (x,, y,) and Q = (x2, y2) with P # Q. We shall assume that P, Q belong to two distinct lines and reach a contradiction. Case 1. Suppose P, Q belong to both L, and L,. with a 4a’. Then a = xX, =X, and a’ =x, =x, so that a=a’, which is a contradiction. Case 2. If P, Q belong to both L, and L,,,, then P = (a, y:) and Q= (a, y2). Since both belong to L,,,, we also have yi =mx,+b=ma+b and y. =mx, + b=ma+b. Thus y, = y2, which contradicts (a, y,) = P # Q = (a, Va). Case 3. Suppose that P, Q belong to both L,,,, and L,,. and that L,,, # L,«. Then2.1 Definition and Models of Incidence Geometry 23 Yy=mx, +b, y= mx, +b. (1-8) By Case 2, P, Q cannot both belong to a vertical line so x, # x,. Hence we may solve Equation (1-8) for m: m= | (1-9) x2 —X From this value of m we obtain b: b=y,— mx, (1-10) A similar calculation for the line L,,. yields ary n . x2 Ms c=yy— nx. But this implies m = n and b = c, which contradicts L,,, # Ly,¢- Thus in all cases, the assumption that P, Q belong to two different lines leads to a contradiction so that P, Q belong to a unique line. In Problem AS you will show there exists a set of three non-collinear points. Hence @ is an incidence geometry. o Note that in the above proof we did not depend on any pictures or “facts” we already know about “straight lines.” Instead we were careful to use only the definition of the model and results from elementary algebra. Proposition 2.1.5. The Poincaré Plane 3 is an incidence geometry. Proor. Let P, Qe H with P # Q. If P and Q lie on two'type I lines ,L and aL then we can show that a = a’ just as in Proposition 2.1.4. Thus P and Q cannot lie on two different type I lines. In Problem A7 you will show that P and Q cannot lie on both a type I line and a type II line. We are left with proving that if P = (x,, y,) and Q = (x2, yz) are on both cL, and gL, then .L, = 4L,. We will show that c = d and r = s. This will be done by deriving Equations (1-5) and (1-6) and so will motivate the choice of c and r in Proposition 2.1.2. Since P and Q are on .L,, (xy - oP +yp=r? and (x, —- 0) +95 Subtracting, we obtain (x, — c)? — (x2 — 0)? = y3 — yf or x} — 2cx, — x3 + cx, = y} — yi. We then solve for c: 3 i adcsitxinxd 2(x2 — x1) which is Equation (1-5). An identical computation using the fact that P and Q are on gL, will yield yar yitxd— xt 4-20 =)24 2 Incidence and Metric Geometry so that c = d. Since r=JVei-P ty =Vei— dP + yas we see that r = s and so _L, = gL,. In Problem A8 you will show there is a set of three non-collinear points. 1 Theorem 2.1.6. Let 1, and |, be lines in an incidence geometry. If 1, 0 1, has two or more points then |, = 1. Proor. Assume that P#Q, Pel, V/,, and Qel, 1 1,. Then since both Pand Q are on |,, PO = |,. However, P and Q are also on I, so that PO = I. Hence |, = 1. Oo Definition. If |, and /, are lines in an abstract geometry then /, is parallel to (2 (written /, |[I,) if either I, = 1, or AL, = @. The study of parallel lines has a central place in the history of geometry. It and its history will be dealt with in detail later in this book. In Problems A10, All, and A12 the different “parallel properties” of our models are highlighted. Theorem 2.1.6 can be restated in terms of parallelism as the next result shows. Corollary 2.1.7. In an incidence geometry, two lines are either parallel or they intersect in exactly one point. PROBLEM SET 2.1 Part A. Find the Poincaré line through (1, 2) and (3, 4). Find the Poincaré line through (2, 1) and (4, 3). Find a spherical line (great circle) through (4, 4, //$) and (1, 0, 0). Find a spherical line (great circle) through (0, }, $./3) and (0, — 1, 0). yee Nn Show by example that there are (at least) three non-collinear points in the Cartesian Plane. 6. Verify that P = (x,, y,) and Q = (x3, y2) do lie on .L,, where c and r are given by Equations (1-5) and (1-6). Prove that if P and Q are distinct points in H then they cannot lie simultaneously on both gL and .L, 2 » Show by example that there are (at least) three non-collinear points in the Poincaré Plane. 9. Let Pand Q be in H and PG = .L,. Use your knowledge of Euclidean geometry to prove that c is the x-coordinate of the intersection of the Euclidean perpendicular bisector of the Euclidean line segment from P to Q with the x-axis. (Hint: Use Equation (1-5).)2.1 Definition and Models of Incidence Geometry 25 Figure 2-5 10. Find all lines through (0, 1) which are parallel to the vertical line L, in the Cartesian Plane. 11. Find all lines in the Poincaré Plane through (0, 1) which are parallel to the type I line gL. (There will be infinitely many!) 12, Find all lines of @ through N = (0,0, 1) which are parallel to the spherical line (great circle), @, defined by the plane z = 0. 13. Let ¥ = {P,Q,R} and ¥ = {{P,Q}, {P,R), {Q,R}}. Show that {¥,%} is an incidence geometry. Note that this example has only finitely many (in fact, three) points, It may be pictured as in Figure 2-5, It is called the 3-point geometry. The dotted lines indicate which points lie on the same line, 14, Let ¥ = R? and, for a given choice of a, b, and c, let J, = {(x, ye R2 Jax + by = ¢}. Let Y, be the set of all with at least one of a and b nonzero. Prove that {R?, 2} is an incidence geometry. (Note that this incidence geometry gives the same family ‘of lines as the Cartesian Plane. The point here is that there are different ways to describe the set of lines of this geometry.) 15. Let Y = R?—{(0,0)} and & be the set of all Cartesian lines which lie in Show that {Y, #} is not an incidence geometry. tea 1 1 ! ! tg s} Ponnnnn ne + @) R. tot VBA IS ‘/ Tbe we \I/ v26 2 Incidence and Metric Geometry 16. Prove that the Riemann Sphere is not an incidence geometry. 17, Show that the conclusion of Theorem 2.1.6 is false for the Riemann Sphere. Explain. 18. Prove Corollary 2.1.7. Part B. “Prove” may mean “find a counterexample”. 19. Some finite geometries are defined pictorially (as in the 3-point geometry of Problem A13) by Figure 2-6. i. In each example list the set of lines. |. Which of these geometries are abstract geometries? . Which of these geometries are incidence geometries? *20. Let {% #} be an abstract geometry and assume that 4, < % We define an F-line to be any subset of A of the form Im S, where | is a line of and where 10 ¥, has at least two points. Let ¥, be the collection of all 4-lines. Prove that {A, Y,} is an abstract geometry. {A, ¥,} is called the geometry induced from {A}. *21. If{%, G,} is the geometry induced from the incidence geometry {%, ¥}, prove that {Y,, ¥,} is an incidence geometry if % has a set of three non-collinear points. 22, Let {%, ¥,} and {%, Z} be abstract geometries. Let ¥ =.4 UA and 2 = L,UL,. Prove that {Y, #} is an abstract geometry. 23, Let {%,¥,} and {%, ¥,} be abstract geometries. If S=AAH, and ¥ = YL, 0 L, prove that {, #} is an abstract geometry. 24, Let {¥, } be an abstract geometry. If I, and J, are lines in & we write I, ~ I, if 1, is parallel to Iz. Prove that ~ is an equivalence relation. If {%, £} is the Cartesian Plane then cach equivalence class can be characterized by a real number or infinity. What is this number? 25. There is a finite geometry with 7 points such that each line has exactly 3 points on it. Find this geometry. How many lines are there? 26. Define a relation ~ on S? as follows. If A = (x;, y;, 2) and B =(x;, y,, 2) then A~ Bifeither A = Bor A = —B =(—x;, —y;, —2,). Prove ~ is an equivalence relation. 27. Let P = {[XJIX e S?} be the set of equivalence classes of ~ in Problem B26. If @ is a great circle (spherical line) let [6] = {[X]|X € @). Let ¥, = {[@]]¢ ¢ G}. Prove that = {P, %,} is an incidence geometry. ( is the Projective Plane. There is a natural bijection between P and the set described in Problem B18 of Section 1.2.) 28. Prove that there are no distinct parallel lines in (i., if 1, is parallel to !, then =) Part C. Expository exercises. 29. Discuss the mathematical career of René Descartes.2.2 Metric Geometry 27 30. Discuss the statement “parallel lines meet at infinity” in terms of the three models that are given in this section. Is there even a meaning to the phrase “at infinity” in the Poincaré Plane or the Riemann Sphere? 2.2 Metric Geometry At this level there are two fundamental approaches to the type of geometry we are studying. The first, called the synthetic approach, involves deciding what are the important properties of the concepts you wish to study and then defining these concepts axiomatically by their properties. This approach was used by Euclid in his Elements (around 300 B.C.E.) and was made complete and precise by the German mathematician David Hilbert (1862- 1943) in his book Grundlagen der Geometrie [1899; 8th Edition 1956; Second English Edition 1921]. Hilbert, as did Poincaré at the same time, worked in many areas of mathematics and profoundly affected the course of modern mathematics. He put several areas of mathematics on firm axiomatic footing. In an address to the International Congress of Mathematicians in 1900 he proposed a series of seventeen questions which he felt were the leading theoretical problems of his time. These questions (not all of which have been answered yet) directed mathematical research for years. The second approach, called the metric approach, is due to the American mathematician, George David Birkhoff (1884-1944) in his paper “A Set of Postulates for Plane Geometry Based on Scale and Protractor” [1932]. In this approach, the concept of distance (or a metric) and angle measurement is added to that of an incidence geometry to obtain basic ideas of betweenness, line segments, congruence, etc. Such an approach brings some analytic tools (for example, continuity) into the subject and allows us to use fewer axioms. Birkhoff is also remembered for his work in relativity, differential equations, and dynamics. A third approach, championed by Felix Klein (1849-1925), has a very different flavor—that of abstract algebra—and is more advanced because it uses group theory. Klein felt that geometry should be studied from the viewpoint of a group acting on a set. Concepts that are invariant under this action are the interesting geometric ideas. See Millman [1977] and Martin [1982]. In Chapter 11 we will study some of the ideas from this approach, which is called transformation geometry. In this book we will follow the metric approach because the concept of distance is such a natural one. (Modern treatments of the synthetic approach can be found in Borsuk and Szmielew [1960] or Greenberg [1980]. We will briefly outline the synthetic approach in Section 6.7.) Intuitively, “distance” is a function which assigns a number d(P, Q) to each pair of points P, Q. It should not matter whether we measure from P to Q or from Q to P (ie,28 2 Incidence and Metric Geometry d(P, Q) = d(Q, P)). Furthermore, the only time the distance between two points is zero should be when the points are actually the same. More formally we have the following definition. Definition. A distance function on a set ¥ is a function d: ¥ x Y +R such that for all P,Qe SF (i) d(P,Q) > 0; (ii) d(P, Q) = 0 if and only if P = Q; and (iii) a(P, Q) = d(Q, P). The following definition gives a distance function for the Cartesian Plane. See Problem Al. Definition. Let % = R?, P = (x;, y,) and Q = (x2, y2). The Euclidean distance de is given by (P,Q) = Je, — x2) + (i — ya). (2-1) To give an example of a reasonable distance function in the Poincaré Plane requires more thought. Suppose that P and Q belong to a type I line. A reasonable guess for the distance between P =(a,y,) and Q = (a, y2) might be |y, — y2|- However, this is somewhat displeasing because it means that as y, tends to zero (and thus Q goes toward the x-axis or “edge”) the distance from P to Q tends to y,, which is a finite number. It would be “nicer” if the “edge” were not a finite distance away. One way to avoid this is to use a logarithmic scale and say that the distance from (a, y,) to (a, y2) is |In(y,) — In(y2)| = |In(y,/y2)|. (Note that as y, 0, In(y1/y2)— 00.) This gives some justification for the following definition (which looks rather artificial.) The reasons for this definition will be clearer after we discuss the Ruler Postulate. Definition. If P = (x,, y,) and Q = (x2, y2) are points in the Poincaré Plane 3, the Poincaré distance d, is given by dy(P,Q) = in(22)| ifx, =x, (2-2) 1 xy c+ Yi . : =|inf —— . 2-3 dy(P,9) =n] ——**— }] iP and Q lie on L, (23) v2 The verification that dy, as defined by Equations (2-2) and (2-3), actually satisfies axioms (i) and (iii) of a distance function is left to Problem A2. Axiom (ii) is more difficult, especially for points on a type II line. Essentially,2.2 Metric Geometry 29 we need to show that the function f: .L, + R given by f(x, y)=In (=) is injective. We will do this in the proof of Proposition 2.2.6. We shall now present an example with a different twist. This example, called taxicab distance, comes from thinking ofa taxi driving on the rectangu- lar grid of a city’s streets. The taxicab distance measures the distance the taxi would travel from point P to point Q if there were no one way streets. See Figure 2-7. Definition. If P =(x,,y,) and Q =(x2,y) are points in R?, the taxicab distance between them is given by dy(P,Q) = |x, — x2] + [ys — yal: (2-4) Proposition 2.2.1. The taxicab distance is a distance function on R?. Proor.Note that d;(P,Q) = 0 since it is a sum of absolute values, each of which is always nonnegative. Thus axiom (i) for a distance holds. The second axiom states that d;(P,Q) = 0 if and only if P = Q. Clearly if P = Q then d;(P, Q) = 0 by Equation (2-4). On the other hand, ifd;(P,Q) = 0 then |x, — x2| + |y; — y2| = 0. Since each of these two terms is at east zero, they must both be zero: |x, — x2| = Oand|y, — y2| = 0. But this means xX, =x, and y, = y2. Therefore, if d;(P,Q) = 0 then P = Q. Finally axiom (iii), d7(P,Q) = d7(Q, P), holds because |a ~ b| = |b — a Note that dy and dy are both distance functions on the same underlying set R?. In general, a set may have many different distance functions on it (see, for example, Problem B16). Thus, when we want to talk about a property of distance on a set, we need to specify both the set / and the distance function d. The concept of a ruler is central to the remainder of this book. This was the idea introduced by Birkhoff to move geometry away from the very30 2 Incidence and Metric Geometry synthetic methods. Intuitively, a ruler is a line that has been marked so that it can be used to measure distances. We shall “mark” our lines by assuming that for every line there is a bijection between that line and R in such a way that the “markings” measure distance. Definition. Let | be a line in an incidence geometry {¥,%}. Assume that there is a distance function don Y. A function f:| > Risa ruler (or coordinate system) for / if (i) f is a bijection; (ii) for each pair of points P and Q on | [f(P) - {(Q)| = a(P,Q). (2-5) Equation (2-5) is called the Ruler Equation and f(P) is called the coordinate of P with respect to f. Example 2.2.2. Let | be the nonvertical line L,,, in the Cartesian Plane © with the Euclidean distance. Show that if Q = (x, y) then f(Q) = J5x gives aruler f for | and find the coordinate of R = (1,5) with respect to f. SoLution. f is certainly a bijection so all we need verify is the Ruler Equation. Note that (x, y) € L2,3 if and only if y = 2x + 3 so that if P =(x,,y,) then (P,Q) = x, — x + (v1 — vy)? = Oy — x) + 40) — x? = J5ix1 — xl =Lf(P) - SQ). Thus the Ruler Equation holds. The coordinate of R = (1,5) is f(R) = J. oO Some comments are in order. The terms ruler and coordinate system are typically used interchangeably in the literature, and we will use both. Note also that since a point may lie on more than one line it may have different “coordinates” with respect to the various lines or rulers used. In particular, if P lies on the line I', and if!’ has a ruler f’, then there need not be any relation between the coordinate of P with respect to | and the coordinate of P with respect to I’. (See Problem A4.) In addition, we shall see that if a line has one ruler f, it has many rulers and thus many possible coordinates for P. (The analogous situations for coordinates in analytic geometry are that the rectangular coordinates of a point may be quite different from its polar co- ordinates and that by translating the origin we also get different coordinates.) Definition. An incidence geometry {7,£} together with a distance function d satisfies the Ruler Postulate if every line |< Y has a ruler. In this case we say M ={/,L,d} is a metric geometry. Why do we study metric geometries? It is because many of the concepts in the synthetic approach which must be added are already present in the2.2 Metric Geometry 31 metric geometry approach. This happens because we can transfer questions about a line | in ¥ to the real numbers R by using a ruler f. In R we under- stand concepts like “between” and so can transfer them back (via f~ ') to I. This is the advantage of the metric approach alluded to in the beginning of the section. After we have more background (ie., in Chapter 6), we will return to the question of a synthetic versus metric approach to geometry. The definition states that in order to prove {, ¥,d} isa metric geometry, we need to find for each |e ¥ a function f:!— R which is a bijection and which satisfies Equation (2-5). However, because of the next lemma, we do not really have to prove that f is an injection. This lemma will then prove useful in the problems at the end of the section as well as in Propositions 2.2.4 and 2.2.7. Lemma 2.2.3. Let le Y and f:I— R be surjective and satisfy Equation (2-5). Then f is a bijection and hence a ruler for I. Proor. Since we assume that f is surjective we need only show that it is injective. Suppose that f(P) = (Q). Then by Equation (2-5) we have (P,Q) =| f(P) — f(Q)| =0 so that P = Q by the second axiom of distance. Oo Proposition 2.2.4. The Cartesian Plane with the Euclidean distance, dg, is a metric geometry. Proor. Let | be a line. We need to find a ruler for I. This will be done in two cases. Case 1. If | = L, is a vertical line then P ¢ L, means P = (a, y) for some y. We define f:1—+ R by S(P) = f((a, y)) = y. (2-6) Ff is clearly surjective. If P = (a, y,) and Q = (a, y.), then IS(P) — f(Q)| = In — Y2l = a(P, Q)- Therefore f is a ruler by Lemma 2.2.3. Case 2. If | = L,,., then P € L,,,, means that P = (x, y) where y = mx + b. Define f:L,,,— R by SUP) = f(x, y)) = xT + mi? (2-7) IfteR let x=t/,/1+m?, y=(mt/,/1 +m?) +b. Certainly, P=(x, ye Ln,s- Furthermore, SP) = 4: JT +m =t Ji+m so that f is surjective.32 2 Incidence and Metric Geometry Now suppose that P = (x,, y,) and Q = (x2, y2). Then LCP) = £(Q)| = [xa VT + om? — x, JT + m?| = J1 + m\x, — x,]. On the other hand de(P,Q) = Vx1 — x2)? + (1 = 2)? HV - Xa)" + mx, - x22 = J/1+mJ(x, — x, =J1+ m|x, — x). Combining these two sets of equations we have |f(P) — f(Q)| = (P,Q). Hence by Lemma 2.2.3, f is a ruler. Oo Definition. The Euclidean Plane is the model & = {R’, Z,, dg}. Our next step is to show that the Poincaré Plane with the Poincaré dis- tance is a metric geometry. To do this it will be useful to use hyperbolic functions. Recall that the hyperbolic sine, hyperbolic cosine, hyperbolic tan- gent and hyperbolic secant are defined by sinh(t) = cosh(t) = G +e : inh(e) _—e™! 1 2° tanh(t) = sin a 7 sech(t) = cosh(t) e+e! cosh(t) e& +e" From the above it is easy to prove Lemma 2.2.5. For every value of t: (i) [cosh(t)]? — [sinh(Q}? = 1; (ii) [tanh(t)}? + [sech()}? = 1. The first equation of Lemma 2.2.5 is particularly suggestive. Whereas the trigonometric (or circular) functions sine and cosine satisfy sin? t+cos?t=1 and remind us of a circle: x? + y? = 1, the hyperbolic sine and cosine lead to an equation of a hyperbola: x? — y? = 1. We should also note that if x = tanh(t) and y = sech(t) then (x, y) lies on the circle x? + y? = 1.2.2 Metric Geometry 33 Proposition 2.2.6. dy, is a distance function for the Poincaré Plane and {H, Ly, dy} is a metric geometry. Proor. By Problem A2, dy satisfies axioms (i) and (iii) of a distance function. We must verify axiom (ii) and find appropriate rulers. Clearly, if P = Q then dy(P, Q) = 0. We need to show that if dy(P, Q) = 0 then P = Q. To do this we consider two cases depending on the type of line that P and Q belong to. Suppose that P, Q belong to a type I line ,L with P =(a,y,) and Q=(a, y2). If dy(P, Q) = 0 then |In(y,/y2)| = 0 so that y,/y, = 1 and y, = y. Thus if P, Q belong to a type I line and dy(P,Q) =0 then P=Q. In Problem A8 you will show that the function g:,L + R given by g(a, y) = In(y) is a bijection and satisfies the Ruler Equation. Thus g is a ruler for ,L. Now suppose that P, Q belong to a type II line .L, and that d,(P, Q) = 0. Let f:,L, + R be given by fix, ) = n(5=£ = that f is a ruler. First we must show it is a bijection. (Lemma 2.2.3 cannot be used because we do not yet know that dy is a distance function.) To show that f is bijective we must show that for every te R there is one and only one pair (x, y) which satisfies . We will eventually show (k-c?ty=r, y>0, and fix,yyae (2-9) We try to solve f(x, y) = ¢ for x and y. If fix, y) = In (==) =tthen*—£*+" ~ ot Thus y y eta en) _¥e-e = _ yWx-c-" x-ctr (x-ctr(x-c-n (x-c 9 -y __xceor y since (x, y) € .L,. Hence 1 x¥-¢tr x-c-r_2r eb+et= y y y or y =rsech(t). Also x-ctr x-c-r ee y y = 4(x-o e+e" 2r 2r y or x —c=rtanh(t).34 2 Incidence and Metric Geometry Hence the only possible solution to Equation (2-9) is x=c+rtanhe, y =rsech ¢. (2-10) A simple computation using Lemma 2.2.5 shows that x and y as given in Equations (2-10) satisfy (x — c)? + y? =r? and that y > 0. Thus Equations (2-10) define a point in .L,. Finally, a straightforward substitution verifies that for this x, y we have f(x, y) = t. Thus Equations (2-9) have one and only one solution for each t € R and therefore f:.L, — R is a bijection. Next, if P =(x,, y,) and Q@ = (x, y2) belong to .L,, then by Equation (2-3) and the properties of logarithms dylP, Q) = |f(x1, 91) — £2, Yadl- Hence, f satisfies the Ruler Equation. Finally, if dy(P, Q)=0, then f(x,,y,)= S(X2, 2). Since f is bijective, this means (x,, y,) = (x2, y2) and dy satisfies axiom (ii) of a distance function. Since we have proved that dy is a distance and each line in # has a ruler (g and f above) {H, Ly, dy} is a metric geometry. oO Convention. From now on, the terminology Poincaré Plane and the symbol # will include the hyperbolic distance dy: 3 ={H, Ly, dy}. Note that with the given rulers in 9, if P, Q € | and if we let Q tend to the “edge” (ie., the x-axis) along I, then f(Q) tends to + co so that d(P,Q) = |f(P) - f(Q)| > ~- That is, the “edge” of the Poincaré Plane is not a finite distance away from any point P. To a creature living in the geometry the edge is not reachable, hence not observable. The x-axis that we sketch in our pictures of the Poincaré Plane is the “horizon”. Proposition 2.2.7. The Cartesian Plane with the taxicab distance is a metric geometry. Proor. If | is a vertical line L, we define f:!— R by f((a, y)) = y. If lisa nonvertical line L,,, we define f:1—> R by f((x, y)) = (1 + |m|)(x). We leave the proof that these really are coordinate systems to Problem A12. o Definition. The model J = {R?, £47} will be called the Taxicab Plane. Note that we started with a single incidence geometry (the Cartesian Plane), put two different distances on it, and obtained two different metric geometries. Thus we have two metric geometries with the same underlying incidence geometry. In general, there are many metric geometries which have the same underlying incidence geometry.2.2 Metric Geometry 35 How do we actually construct models of a metric geometry? In our three examples we started with an incidence geometry, defined a distance and hunted for rulers so that Equation (2-5) was satisfied. We can reverse this process in a certain sense. That is, we can start with a collection of bijections from the lines to R and use them to define a distance function which has these bijections as rulers. In fact, this method (which is described in Theorem 2.2.8 below) is really how we decided what the “right” definition was for a distance in H. Theorem 2.2.8, Let {/, £} be an incidence geometry. Assume that for each line le & there exists a bijection f,:|-+R. Then there is a distance d such that {f,£,d} is a metric geometry and each fy:1— R is a ruler. Proor. If P, Qe ¥ we must define d(P,Q). If P=@Q let d(P,Q) =0. If P # Q let | be the unique line through P and Q, and f,:! > R be the bijection described in the hypothesis. Define d(P, Q) = | f,(P) — f,(Q)|- In Problem A13 you will verify that d satisfies the three properties of a distance. Finally each f, is clearly a ruler for the line [. Oo The opposite problem in which we start with a distance and ask if there is metric geometry with that distance is more subtle. In Problem B15 we give an example of a distance on the incidence geometry, R?, which does not have rulers and hence does not give a metric geometry. We close this section with a table which summarizes the rulers which we have discussed for the three major models of a metric geometry. Standard Ruler or Model Type of line coordinate system for line Euclidean L, = {(a yy R} fla,y)=y Plane, Lino = {(x, y) € R?| y = mx + B} Slxy) = x1 +m? Poincaré aL = {(a, ye Hy > 0} Slay) =Iny Plane, cL = (x eH|xe-—o ty =P} fey) =In (=) Taxicab {a y)|ye R} fla y)=y Plane, 7 Law = {(%y) € R2| y = mx + B} Sx y) = (1 + [rmx Convention. In discussions about one of the three models above, the coordinate of a point with respect to a line / will always mean the coordinate with respect to the standard ruler for that line as given in the above table. In the next section we will discuss some special rulers for a line. These should not be confused with the standard rulers defined above.36 2 Incidence and Metric Geometry PROBLEM Set 2.2 Part A. . Prove that the Euclidean distance function as defined by Equation (2-1) is a distance function. v . Verify that the function dy defined by Equations (2-2) and (2-3) satisfies axioms (i) and (iii) of the definition of a distance function, . Prove Lemma 2.2.5. In the Euclidean Plane, (i) find the coordinate of (2,3) with respect to the line x = 2; (ii) find the coordinate of (2, 3) with respect to the line y = —4x + 11. (Note that your answers are different.) aw . Find the coordinate of (2, 3) with respect to the line y= —4x + 11 for the Taxicab Plane. (Compare with Problem 4.) » . Find the coordinates in H of (2,3) (i) with respect to the line (x — 1)? + y? = 10; (ii) with respect to the line x = 2. a . Find the Poincaré distance between i. (1, 2) and (3, 4) (See Problem A1 of Section 2.1.) ii, (2, 1) and (4, 3) (See Problem Al of Section 2.1.) ~ . Show that the function g:,L — R given by g(a, y) = In(y) is a bijection and that it satisfies the Ruler Equation. Show that the inverse of g is given by g~'(t) = (a, e'). - 9. Find a point P on the line L», 5 in the Euclidean Plane whose coordinate is — 2. 10. Find a point P on the line L»,-; in the Taxicab Plane whose coordinate is — 2. 11. Find a point P on the line -,L /7 in the Poincaré Plane whose coordinate is In 2. 12. Complete the proof of Proposition 2.2.7. 13. Complete the proof of Theorem 2.2.8. Part B. “Prove” may mean “find a counterexample”. 14. We shall define a new distance d* on R? by using dg. Specifically: _ fae(P,Q) ifd.(P,Q) <1 a(P, Q)= {i if d,(P, Q) > 1. (i) Prove that d* is a distance function. (ii) Find and sketch all points P ¢ R? such that d*((0, 0), P) < 2. (iii) Find all points P € R? such that d*(0, 0), P) = 2. 15, Let d* be the distance function of Problem B14. Prove that there is no incidence geometry on R? such that {R?, % d*} is a metric geometry. (Thus not every distance gives a metric geometry.) Hint: Suppose by way of contradiction that there is a ruler f: ! + R and that Pp € [has coordinate zero. Consider the set of all points on / with coordinate +2. 16. If dy and d, are distance functions on Y, prove that if s > 0 and ¢ > 0, then sdo + td, is also a distance function on .2.3 Special Coordinate Systems 37 17. If {¥, ¥d} is a metric geometry and Pe Y, prove that for any r > 0 there is a point in # at distance r from P. 18. Define the max distance (or supremum distance), ds, on R? by dg(P, Q) = max{|x, — x2h ly, — yal} where P = (x,, y,) and Q = (x3, y2). i. Show that dg is a distance function. ii. Show that {R?, &., ds} is a metric geometry. 19. In a metric geometry {¥, ¥, d} if Pe ¥ and r > 0, then the circle with center P and radius r is € = {Q € ¥|d(P, Q) = r}. Draw a picture of the circle of radius 1 and center (0, 0) in the R? for cach of the distances dg, d;, and ds. 20. Let {%, ¥, d} be a metric geometry, let Pe SY let le L with Pel, and let ¢ bea circle with center P. Prove that 1 contains exactly two points. 21. Find the circle of radius 1 with center (0, e) in the Poincaré Plane. Hint: As a set this circle “looks” like an ordinary circle. Carefully show this. 2. N We may define a distance function for the Riemann Sphere as follows. On a great citcle @ we measure the distance dp(A, B) between two points A and B as the shorter of the lengths of the two arcs of @ joining A to B. (Note dg(A, — A) = =.) Prove that dg is a distance function. Is {S?, Za, dp} a metric geometry? On the Projective Plane (see Problem B26 of Section 2.1) define d,([A], [B]) = minimum of the two numbers da(A, B) and dx(A, — B). Prove that dp is a distance function. Is {P, Zp, dp} a metric geometry? 2. Part C. Expository exercises. 24. Compare and contrast the definition of the taxicab metric as given in this section with that of Byrkit [1971]. 2.3 Special Coordinate Systems In this section we shall prove the existence of a special kind of coordinate system. This coordinate system will play an important role in our study of betweenness in Chapter 3. We shall also see that, as a consequence of the Ruler Postulate, every line in a metric geometry must have infinitely many points. Theorem 2.3.1. Let f be a coordinate system for the line | in a metric geometry. If ae Rand ¢ is +1 and if we define h,,,.:1+ R by h,.e(P) = e(f(P) — a) then h,,, is a coordinate system for |. Proor. By Lemma 2.2.3 we need only show that h,,, is surjective and satisfies the Ruler Equation. If te R is given we know that there is an Re! with38 2 Incidence and Metric Geometry S(R) = t/e + asince f is surjective. But then hyd) = ef f(R) ~ 0) = ((G + 2) - a) = so that h,,, is surjective. As for the Ruler Equation, Vhra,e(P) — ha,e(Q) = le( f(P) — a) — e(£(Q) - @)| = lel LSP) - £(Q)| = |f)- f)| = d(P,Q) since f is a coordinate system for I. Qo Geometrically, when a = 0 and ¢ = — 1 the coordinate system of Theorem 2.3.1 interchanges the positive and negative points of | with respect to f. More precisely, if Po is that point of | with f(Po) = 0 then ho,-, is the result of reflecting the ruler f about Po. See Figure 2-8. We may also translate a coordinate system by an element a¢R. This amounts to changing the origin (i¢., the point which corresponds to 0). In Figure 2-9, we assume that f(P,) = aand f(Po) = 0so that P, corresponds to a and Po corresponds to the origin in the coordinate system f. If we apply Theorem 2.3.1 with & = 1 then P, corresponds to the origin and Py to —a in the new coordinate system h, 1. P P P P —— + ———_2 | Iss * R — R - o + + 0 - Figure 2-8 P P P P 0 1 1 0 1 7 | Is o —> R ae — R 0 a =a o Figure 2-92.3 Special Coordinate Systems 39 Theorem 2.3.2 (Ruler Placement Theorem). Let | be aline in a metric geometry and let A and B be points on the line. There is a coordinate system g on | with g(A) = 0 and g(B) > 0. Proor. Let f:/-+ R be a coordinate system for ! and let a = f(A). If f(B) > a let e = +1. If f(B) < a let e = —1. By Theorem 2.3.1 g = h,,, is a coordinate system for I, and (A) = h,, (A) = &( f(A) — a) = €-0 = 0; 9(B) = h,,.(B) = e(f(B) — a) = |f(B) — al > 0. Thus g is a coordinate system with the desired properties. a The special coordinate system of Theorem 2.3.2 is so useful that it merits a special name. Definition. Let | = AB. If g:! + R is a coordinate system for / with g(A) =0 and g(B) > 0, then g is called a coordinate system with A as origin and B positive. It is reasonable to ask if there are any other operations (besides reflection and translation) that can be done to a coordinate system to get another coordinate system; that is, is every coordinate system of the form h,,? The next theorem says the answer is yes. This result will not be used in the rest of the book. It is included for the sake of completeness and is optional. Theorem 2.3.3. If | is a line in a metric geometry and if f:1+ Rand g:|>R are both coordinate systems for |, then there is anaeé Rand ane = +1 ae g(P) = e(f(P) — a) for all Pel. Proor. Let Po €! be the point with g(P)) = 0. Let a = f(Po). Since both f and g are rulers for |, we have for each P € / that la(P)| = lg(P) — g(Po)| = 4(P, Po) = |f(P) — f(Po)l =|f(P) - al. Thus for each Pel, g(P) = +(f(P) - a). (3-1) We claim we can use the same sign for each value of P. Suppose to the contrary that there is a point P, # Py with g(P,) = +(f(P,) — a) and another point P, # Po with g(P,) = —(f(P,) — a). Then d(P,, Pz) = l9(Pr) — 9(Pa)l = If(P1) — a + f(P2) — al = [f(Pi) + f(P2) — 2al.40 2 Incidence and Metric Geometry But d(P,, P2) = Lf(Pi) — SP) Thus IfPr) — FOP) = 1P(P,) + f(P2) — 2a and either S(Pi) — SPs) = f(P1) + f(P2) — 2a or S(P1) — f(Pa) = —f(P,) — f(P2) + 2a. In the first case f(P,) = a = f(Pp) and in the second case f(P,) = a = f(Pp). Either way we contradict the fact that f is injective. Thus by Equation (3-1), either g(P)=f(P)-a forall Pel or g(P)= -(f(P)—a@) forall Pel. Thus for an appropriate choice of e (either + 1 or — 1), g(P) = e(f(P) — a) for all Pel. oO A metric geometry always has an infinite number of points (Problem A5). In particular, a finite geometry (Problems A13, B19, and B25 of Section 2.1) cannot be a metric geometry. On the other hand, Problem A6 shows that not every distance on an incidence geometry gives rise to a metric geometry even if it has infinitely many points. The points must “spread out.” (Problems B14 and B15 of Section 2.2.) The Ruler Postulate is therefore a very strong restriction to place on an incidence geometry. PROBLEM SET 2.3 Part A. 1. In the Euclidean Plane find a ruler f with f(P) = 0 and f(Q) > 0 for the given pair PandQ: i P=(2,3),Q= ti, P= (2,3),.0 2. In the Poincaré Plane find a ruler f with f(P) = 0 and f(Q) > 0 for the given pair PandQ: i P=(2,3),0=(2,1) it, P = (2,3), Q =(—1,6). 3. In the Taxicab Plane find a ruler f with f(P) =0 and f(Q) > 0 for the given pair Pand Q:2.3 Special Coordinate Systems 41 = (2,3), Q = (2, -5) = (2,3), Q = (4,0). 4, Let P and Q be points in a metric geometry. Show that there is a point M such that Me PG and d(P, M) = d(M,Q). ii 5. Prove that a line in a metric geometry has infinitely many points. 6. Let {¥,%,d} be a metric geometry and Qe ¥. If lis a line through Q show that for each real number r > 0 there is a point P € / with d(P,Q) = r. (This says that the line really extends indefinitely.) Part B. 7. Let g:R + R by g(s) = s/(|s| + 1). Show that g is injective. 8, Let {¥, Ld} be a metric geometry. For each le Y choose a ruler f,. Define the function d by a(P, Q) = la LCP) — 9(F(Q))| where | = PQ and g is as in Problem B7. Show that d is a distance function. 9. In Problem B8 show that {, &, d} is not a metric geometry.CHAPTER 3 Betweenness and Elementary Figures 3.1 An Alternative Description of the Cartesian Plane In Chapter 2 we introduced the Cartesian Plane model using ideas from analytic geometry as our motivation. This was useful at that time because it was the most intuitive method and led to simple verification of the incidence axioms. However, treating vertical and non-vertical lines separately does have its drawbacks. By making it necessary to break proofs into two cases, it leads to an artificial distinction between lines that really are not different in any geometric sense. Furthermore, as we develop additional axioms to verify we will need a more tractable notation. For these reasons we introduce an alternative description of the Cartesian Plane, one that is motivated by ideas from linear algebra, especially the notion of a vector. Definition. If A = (x,, y,), B = (x2, y2) € R? and re R then (i) A+ B=(x, + x2, y1 + y2) ER? (ii) rA = (rxy,ry,) € R? (iii) A- B= A + (-1)B = (x, — x2, 1 — Ya) (iv) (A, BD = x1x2 + iy ER (y) ||Al| = V&A, 4d € R. For those of you familiar with the ideas, all we are doing is viewing R? asa vector space with its standard addition, scalar multiplication, and inner product. (Note that you probably wrote = ¢B, A> (vi) (7A, B> = r¢A, BD (vil) (A + B,C) =¢A,C> + (viii) [|r] = [1A] (ix) ||Al] > Oif A # 0,0). Using this notation we make R? into an incidence geometry by defining the line through the distinct points A and B to be L4, where Lan = {X € R?|X = A + ¢(B — A) for some t € R}. (1-1) Proposition 3.1.2. If £' is the collection of all subsets of R? of the form Lan, then {R?, £'} is the Cartesian Plane and hence is an incidence geometry. Proor. Let %, be the set of Cartesian lines as given in Chapter 2. We will show that Yc #' and Lc &. Step 1. Let le Y, be a Cartesian line. If | is the vertical line L, choose Ato be (a,0) and Bio be(a, 1). A, Bel. T= {(a,0)|t © R} = {(a,0) + 10, 1)|te R} = Lage #. Thus le £’. If | is the non-vertical line L,,, choose A to be (0,b) and choose B to be (1,b +m). A, Bel. 1 = {(x, yy = mx + b} = (x, y) = (t, me + b)|te RB} = {(x, y) = (0,b) + (1,m)|t€ R} = Lape L” Thus |e ¥' and hence Y, c L. Step 2. Let Lape L' with A = (x, y,), B= (x2, y2), and A ¥ B.If x, = x2, then (since A # B) y, — y; # Oand Lap = {(%1, 91) + 0, y2 — y,)|t€ R} = {0 1 + (2 — yn) te R} = {(x, y)e R2|x = x1} = Ly, € Le. Thus Lape Le. Ifx, # x2 then x, — x, £0, and we let ma 22a and b=y,—mx,. x2 —%144 3 Betweenness and Elementary Figures Then Lan = {(X1, 1) + (x2 — x1, y2 — yy) te RB} = {(x1, mx, +b) + they — x4, m(xz — x1))|te R} (x1 + (2 — x4), Moxy + t(x2 — x4) + b)|r |e R} = {(x, mx + b)|x € R} = {(x, ye Rly = mx + D} = Le Le. Hence Luge LY, and £'c L,. Thus we have shown that Y, = Y’ so that {R?, ¥’} is the Cartesian Plane. o In Problem B6 you are asked to prove directly that {R?, Y'} is an incidence geometry without any reference to the initial model {R?, Z,}. In terms of our new notation, the distance function dz is described slightly differently as you will show in Problem A2. We also have a nice description of some rulers. Proposition 3.1.3. [f A, B ¢ R? then d,(A, B) = ||A — Bll. Proposition 3.1.4. If L4, is a Cartesian line then f:L4 — R defined by S(A + (B — A)) = t||B — All isa ruler for {R?, Zp, dg}. Proor. The function f makes sense only if for each point P € Ly, there is a unique value of ¢ with P = A + ¢(B — A). This can be seen to be true as follows. Suppose P = A + r(B — A) and P = A + s(B — A). Then (0,0) = P— P=(A+r(B - A)) -(A + 5(B- A)) =(r — s(B- A) so that either r — s = 0 or B— A = (0,0). Since A # B, B— A # (0,0) and so r —s=0, That is, r = sand there is a unique value of ¢ with P = A + ¢(B— A). Hence the function f makes sense. The proof that f actually is a ruler is Problem A3. Oo In a college algebra or linear algebra course you probably learned that the dot product of two vectors is given by the product of the lengths of the vectors and the cosine of the angle in between: a+b = |lal|||b|| cos 8.3.1 An Alternative Description of the Cartesian Plane 45 Since |cos 6| < 1, we have |a - b| < |{all |[b||. The Cauchy-Schwarz Inequality (Proposition 3.1.5) is a careful statement of this result without any reference to angles or the measurement of angles. It will be used in Chapter 5 when we develop angle measurement. We will apply it in this section to prove a special property of the Euclidean distance function (Proposition 3.1.6). Proposition 3.1.5 (Cauchy-Schwarz Inequality). If X, Y « R? then Kx, Y>] s [XI I¥- (1-2) Furthermore, equality holds in Inequality (1-2) if and only if either Y = (0,0) or X =tY for somete R. Proor. If ¥ = (0,0) we clearly have |{X, Y >| = 0 = ||X||- |||] and Inequality (1-2) is true. Hence we assume Y # (0,0). Consider the function g:R + R by g(t) = ||X — tY||?. Then g(t) = 4 Oand g(r) is a quadratic function. Now g(t) = 0 for all ¢ so that g cannot have two distinct real zeros. Since a quadratic function at? + 2bt + c has distinct real zeros if and only if b? — ac > 0, it must be that X,Y)? — CY, Y + [YIP s ||AIP + X,Y] + [YIP (|X|? + 20XTI¥I + [YIP = (x| + 11D? Hence IX + YIP s (IX + eb? oe I + YI [XI + IIE To complete the proof let X = A— Band Y=B-C. Oo We shall see later that the triangle inequality is a consequence of certain other axioms that we will want our geometries to satisfy. In particular, it will hold in the Poincaré Plane. (A direct proof of this fact is for the mas- ochistic.) However, it does not hold in every metric geometry as Problem B9 shows. PROBLEM SET 3.1 Part A. 1. Prove Proposition 3.1.1. . Prove Proposition 3.1.3. . Complete the proof of Proposition 3.1.4. . Complete the proof of Proposition 3.1.6. ween . Show that the ruler in Proposition 3.1.4 is a coordinate system with A as origin and B positive. Part B. “Prove” may mean “find a counterexample”. 6. Let Y' be the collection of subsets of R? of the form given by Equation (1-1). Prove directly that {R?, £’} is an incidence geometry without using our previous model of the Cartesian Plane. 7. Prove that the Taxicab distance dy satisfies the triangle inequality. 8. Prove that the max distance ds on R? satisfies the triangle inequality. (See Problem BI8 of Section 2.2.) 9. Define a function d, for points P and @ in R? by3.2 Betweenness 47 0 ifP=@Q 4,(P,Q)=4de(P,Q) if Lpg is not vertical 3de(P,Q) if Lpg is vertical. a. Prove that d, is a distance function on R? and that {R?, %;,dp} is a metric geometry. b. Prove that the triangle inequality is not satisfied for this distance, dp. Part C. Expository exercises. 10. What other descriptions of the Cartesian Plane can you find in various mathe- matic books? Why is it useful to have more than one description of an object such as the Cartesian Plane? (The answer could deal with either technical reasons or the level of the intended audience.) 11. Another example of different descriptions of the same mathematical concept is given by the notion of a “vector”. Discuss different definitions of a vector, why they are the same, and what their possible uses are. (Note: a use need not be an application of vectors to another subject—it might be to use vectors to do mathematics.) 3.2 Betweenness The concept of one point being between two others is an extremely important, yet at the same time, an extremely intuitive idea. It does not appear formally in Euclid, which leads to some logical flaws. (Euclid made certain tacit assumptions about betweenness. These often occurred as he reasoned from a figure—a shaky practice at best!) These flaws were first rectified by Pasch [1882] who axiomatized betweenness. Without a precise definition of between it is possible to produce erroneous “proofs.” (What would Euclid have thought of the fallacious “proof” which will appear in Problem Set 6.4B that every triangle is isosceles?) In this section we shall use the distance function to define betweenness. In turn, betweenness will allow us to define elemen- tary figures such as segments, rays, angles, and triangles. Definition. B is between A and C if A, B, and C are distinct collinear points in the metric geometry {¥, ¥,d} and if d(A, B) + d(B, C) = d(A, C). (2-1) Note that the definition of between requires that the three points all lie on the same line. (See Problem A10.) Because we will be using betweenness and distance constantly throughout the rest of the book we adopt the fol- lowing simplified notation.48 3 Betweenness and Elementary Figures Notation. In a metric geometry {7, #,d} (i) A—B—C means B is between A and C (ii) AB denotes the distance d(A, B). Thus in this notation, Equation (2-1) becomes, for distinct collinear points, A—B—C ifandonlyif AB+BC= AC. (2-2) The axioms of the distance function are written in this notation as (i) PQ 20; (ii) PQ =O ifand only if P = Q; (iii) PQ = QP; and 7 (iv) PQ =| f(P) — f(Q)| for a ruler f on PQ. (2-3) Note that by using PQ for the distance we have dropped all reference to which distance function we are using. Since an incidence geometry may have more than one distance function, whenever we use the notation PQ for d(P,Q) it must be clear which distance is involved. In our basic models we will continue to use the notation ds, dy, and dy. Example 3.2.1. Let A = (—4, \/3/2), B = (0, 1), and C = (4, /3/2) be points in the Poincaré Plane. Show that A—B—C. Sotution. A, B, and C are on the type Il line o£, = {(x, y)¢ H|x? + y? = 1}. From Equation (2-3) of Chapter 2 -t+1 AB = dy(A, B) = ine =InJ3 t BC = dy(B,C) = in = InJ3 and AC = d,(A,C) = In 3. Thus dy(A, B) + dy(B, C) = dy(A, C) and A—B—C, Note that in Figure 3-1 the point B “looks” like it is between A and C. Oo Figure 3-13.2 Betweenness 49 Theorem 3.2.2. If A—B—C then C—B—A. Proor. If A, B, and C are distinct and collinear, then so are C, B, and A. Since A—B—C, Equation (2-2) shows that AB + BC = AC. Since PQ = QP for all P and Q, we have BA + CB = CA or CB+BA=CA which is what we needed to show. Oo If/is a line with a ruler, the next theorem will allow us to interpret between- ness on | in terms ofa corresponding notion of betweenness for real numbers. This will be a useful method of proving certain results involving betweenness. Thus we will be using the notion of betweenness on the real line to help us with the betweenness in a metric geometry. Definition. If x, y, and z are real numbers, then y is between x and z (written x* y» 2) if either x 0}. Theorem 3.3.4, In a metric geometry (i) if Ce AB and C # A, then AC = AB, (i) if AB = CD then A=C. Theorem 3.3.4 (i) tells us that_a given ray can be named in many ways. Part (ii) says that one point of AB is special. (A can be shown to be the only extreme point of AB.) For this reason we can give it a special name. Definition. The vertex (or initial point) of the ray AB is the point A. Theorem 3.3.5. If A and B are distinct points in a metric geometry then there isa ruler {: AB — R such that AB = {X € AB] f(X) = 0}. Proor. Let f be the special coordinate system with origin A and B positive. We claim that this ruler f is the one we desire. We first show that {X € AB| f(X) = 0} < AB. (3-5) Suppose X ¢ AB with f(X) > 0. Let x = f(X) and let f(B) = y, which is positive by assumption. If x = 0 then X = id X € AB. If x = y then X = Band X ¢ AB. There are only two possi ies left. Either 0 0, we have x <0 < y. This means that D—A—B which is impossible if D ¢ AB. Hence all elements of AB have a nonnegativeco- ordinate with respect to f. Oo Note that Theorem 3.3.5 says there is a ruler with a certain property. However, there was only one possible choice for f: it must be the ruler which is zero at the (unique) vertex and positive elsewhere on the ray. One of the most familiar (and most basic) topics in geometry is the study of congruences, especially the congruence of triangles. In order to reach the point where we can develop this rigorously, we must consider the congruence of line segments and the congruence of angles. We consider the former here and the latter in Chapter 5. Definition. Two line segments AB and CD in a metric geometry are con- gruent (written AB ~ CD) if their lengths are equal; that is AB~CD if AB=CD. The next result will be used continually when dealing with congruence in triangles. It allows us to mark off (or construct) on a ray a unique segment which is congruent to a given segment. Theorem 3.3.6 (Segment Construction). If AB is a ray and PO is a line segment in a metric geometry, then there is a unique point C € AB with PQ ~ AC. Proor. Let f be a special coordinate system for the line AB with A as origin and B positive. Then f(A)=0 and AB = {X € AB| f(X) > 0}. Let r= PQ and set C = f(r). Since r = PQ > 0, we have C € AB. Furthermore, AC = | f(A) — f(©)| = [0 - 7 r=PQ so that AC = PQ. Thus we have at least one point C on AB with AC ~ PQ. Now suppose C’e AB with AC’ ~ PQ. Then since C’e AB, f(C’)>0 and SIC) = fC) = f(A) = |(C) - fA) = AC = PQ = f(C).3.3 Line Segments and Rays 57 Since f is injective, C’ = C and so there is exactly one point Ce AB with AC ~ PQ. Qa Example 3.3.7. In the Poincaré Plane let A = (0,2), B = (0, 1), P = (0, 4), Q = (1, 3). Find C € AB so that AC = PQ. SOLUTION. First we must determine PQ. Both P and Q lie on ,L, so that —34+5 a 445 3 Since C = (0, y) is on the type I line AB, dy(A,C) = |In y/2|. In order that AC = PO we need In y/2 = +1n 6. Hence = [In =| = In6. PQ = dy(P, Q) = {In é y y_l ¥ 36 Yas 2 " 376 Thus 1 y=12 or y= z Since we want C ¢ AB we must take C = (0,4). See Figure 3-6. Oo Figure 3-6 Note that in this example a segment from a type I line is congruent to a segment from a type II line. Of course, two such segments could never be equal but they can be congruent. Do not confuse “congruent” and “equal”! Congruence of segments means “equal in length” whereas equality of segments means “equal as sets.” Theorem 3.3.8 (Segment Addition). In a metric geometry, if A—B—C, _—Q—R, AB = PO, and BC =~ OR, then AC ~ PR. Theorem 3.3.8 says that we create congruent segments by “adding” congruent segments. The following theorem, whose proof is also left as an exercise, says we may also “subtract” congruent segments.58 3 Betweenness and Elementary Figures Theorem 3.3.9 (Segment Subtraction). In a metric geometry, if A—B—C, P—Q—R, AB ~ PQ, and AC ~ PR, then BC = QR. PROBLEM Set 3.3 Part A. . Complete the solution of Example 3.3,1. v Prove Proposition 3.3.3. Prove Theorem 3.3.4. » *4, Prove that “congruence” is an equivalence relation on the set of all line segments in a metric geometry. 5. Prove Theorem 3.3.8. 6. Prove Theorem 3.3.9. 7. In the Taxicab Plane show that if A =(—4, 2), B=(4, 2), C = (2.2, P= (0,0), Q=(2,1) and R=(3,4)then A—B—C and P—Q—R. Show that AB ~ PQ, BC ~ QR and AC ~ PR. Sketch an appropriate picture. 8. Let A = (0,0), B = (4s, 1), and C = (1, 1) be points in R? with the max distance ds(P,Q) = max{|x, — x2[, ly; — y2l}. Prove that AB ~ AC. Sketch the two seg- ments. Do they look congruent? (ds was defined in Problem B18 of Section 2.2.) 9. In the Poincaré Plane let P = (1,2) and Q = (1,4). If A = (0, 2) and B= (1, /3), find C € AB with AC ~ PO. 10. In the Taxicab Plane let P = (1, —2), @ = (2, 5), A = (4, 1) and B = (3, 2). Find Ce AB with AC ~ PO. “1 . Suppose that A and B are distinct points in a metric geometry. M € AB is called a midpoint of AB if AM = MB. a. If M is a midpoint of AB prove that A—M—B. b. If A = (0,4) and B = (0, 1) are points in the Poincaré Plane find a midpoint, M, of AB. Sketch A, B and M on a graph. Does M look like a midpoint? *12. If A and B are distinct points of a metric geometry, prove that a. the segment AB has a midpoint M. (See Problem A11.) b, the midpoint M of AB is unique. 13. Prove that AB = BA for any distinct points A and B in a metric geometry. 14, If D © AB—AB in a metric geometry, prove that AB = AB U AD. 15. If A # Bin a metric geometry, prove that AB = AB u BA and AB = AB BA. Part B. “Prove” may mean “find a counterexample”. 16. Prove that in a metric geometry, AB is the set of all points C € 4B such that A is not between C and B.3.4 Angles and Triangles 59 17. Prove that in a metric geometry any segment can be divided into n congruent parts for any n > 0. More formally: Let A and B be distinct points in a metric geometry. a. Prove there are points Xo, X,,...,X, on AB such that Xo Xiai—Xio2 for i= 0, 1,...,0 — 23 XiXie1 = ABn, and AB b. Prove that the points X; given by the above are unique. 18. If AB = CD in a metric geometry, prove that A = C and B = D. 19. If D ¢ ABin a metric geometry, prove that AD o AB = {A}. 20. If D € AB in a metric geometry, prove that either AD = AB or ADU AB = AB. 21. In a metric geometry suppose that A—B—C, AB ~ PQ, AC ~ PR, BC ~ Prove that P—Q—R. S| Part C. Expository exercises. 22. Rays, segments, and points can be quite beautiful. Go to an art book such as Feldman [1981] or McCall [1970] and identify pictures with significant geometric content. The work of artists such as Seurat, Mondrian, and Kandinski show manifold geometric ideas, See Millman-Speranza [1990] for a presentation of these ideas at the elementary or middle school level. 23. Discuss the statement “Congruent triangles are the same” on both a mathematical and a philosophical level. 3.4 Angles and Triangles In this section we will define angles and triangles in an arbitrary metric geometry. Just as in the case of segments and rays, they will be defined as sets using the concept of betweenness. We will also show that the idea of a vertex of an angle or a triangle is well defined. It is important to note that an angle is a set, not a number like 45°. We will view numbers as properties of angles when we define angle measure in Chapter 5. For us an angle will consist of two rays which are not collinear but have the same initial point. Definition. If A, B and C are noncollinear points in a metric geometry then the angle £ ABC is the set LABC = BAU BC. Note that a line is not permitted to be an angle nor is a ray since {A, B, C} in the definition must be noncollinear. This is for convenience: If “straight angles” or “zero angles” were allowed, we would have to make assumptions to deal with those special cases when stating theorems. Some0 3 Betweenness and Elementary Figures Figure 3-7 Figure 3-8 angles in the Euclidean and Poincaré Planes are sketched in Figures 3-7 and 3-8. It is customary to talk about the vertex of / ABC as the point B. However, a priori, it is not clear that the point B is well defined. After all, it might be possible for 1 ABC = 1 DEF without B = E. (Of course, we will prove that B does in fact equal E.) If this seems unnecessarily pedantic, ask someone if £ ABC = L DEF implies that A = D and C = F. Theorem 3.4.2 below contains the “well defined” result referred to above and is similar in spirit to Theorem 3.3.2. Its proof needs a preliminary lemma. Lemma 3.4.1. In a metric geometry, B is the only extreme point of L ABC. Proor. We first show that if Ze ABC and Z # B then Z is a passing point of 2 ABC. If Ze £ ABC and Z # B then either Z€ BA or Ze BC. Since the two cases are similar we may assume Z € BA. Since Z 4 B, Theorem 3.3.4 implies that J BA = BZ. There exists a Dé BZ such that B—Z—D. (Why?) Thus D ¢ BA and Z is between two points of / ABC, namely B and D. Next we show that B is not a passing point of 2 ABC. We do this with a proof by contradiction. Suppose that X—B—Y with X, Ye ABC. X belongs to either BA or BC. Both cases are similar so that we may assume that X € BA. Since X # B, BA = BX by Theorem 3.3.4. Since Y—B—X, Y ¢ BX = BA. Since Y € / ABC, this means that Y e BC and BC = BY. But then Ae BA = BX < XY, Be XY c XY, and Ce BC = BY < XY. Thus A, B, and C are collinear, which is impossible since we are given 2 ABC. Thus B is not between two points of {ABC and is the only extreme point of 2 ABC. ao Theorem 3.4.2. In a metric geometry, if L ABC = L DEF then B= E. PROOF. {B} = {Z © L ABC|Z is an extreme point of 2 ABC} = {Ze LDEF|Z isan extreme point of 2 DEF} = {E}. a3.4 Angles and Triangles 61 After Theorem 3.4.2 we may make the following definition without any ambiguity. Definition. The vertex of the angle 2 ABC ina metric geometry is the point B. Definition. If {A, B, C} are noncollinear points in a metric geometry then the triangle ABC is the set AABC = ABU BCUCA. Triangles in the Euclidean Plane and the Poincaré Plane are given in Figures 3-9 and 3-10. The Poincaré triangles certainly do not look standard! Figure 3-9 Figure 3-10 We now know that the vertex of a ray, the pair of endpoints of a segment, and the vertex of an angle are all uniquely determined by the ray, segment or angle. We shall show that the points A, B and C of A ABC are also uniquely determined in Theorem 3.4.4. This also needs a preliminary result. Lemma 3.4.3. In a metric geometry, if A, B, and C are not collinear then A is an extreme point of AABC. Proor. Our proof is by contradiction. Suppose that D—A—E with D, E AABC. We show that this implies that both D and E are in BC, which leads to a contradiction. If De AB then either D = B so that E—A—B or D # B so that E—A— D—B and E—A—B (D # A because D—A—E). Either way E ¢ AB. If E belongs to AC or BC then either C—E—A—B or C = Eso that C—A—B. But A, B, C are not collinear. Hence we cannot have D € AB because E must belong to one of AB, AC, or BC. _ In a similar fashion D ¢ AC. Since De A ABC it must be that De BC. A similar proof shows that E ¢ BC also. Thus D, E € BC. But D—A—E implies A € BC also, which is contrary to the hypothesis that A, B, C are noncollinear. Hence it cannot be that A is between two points of AABC. oO62 3 Betweenness and Elementary Figures Theorem 3.4.4. In a metric geometry, if AABC = ADEF then {A, B,C} = {D, E, F}. Proor. If X € AABC and X ¢ {A, B, C} then X is in one the segments AB, BC, or AC but is not an end point. Then X is a passing point of that segment and hence a passing point of AABC. By Lemma 3.4.3 we have {A, B, C} = {X € AABC|X is an extreme point of AABC} = {¥ € ADEF|X is an extreme point of ADEF} = {D, E, F}. oO Definition. In a metric geometry the vertices of A ABC are the points A, B, C. The sides (or edges) of A ABC are AB, AC, and BC. The exercises suggest several alternative proofs of Theorem 3.4.4. PROBLEM SET 3.44. Part A. 1, Prove that £ ABC = CBA in a metric geometry. In problems 2 through 8 do not use Lemma 3.4.3 or Theorem 3.4.4. 2. Let D, E, and F be three noncollinear points of a metric geometry and let | be a line that contains at most one of D, E, and F. Prove that each of DE, DF and EF inter- sects / in at most one point. . Prove that if AABC = ADEF in a metric geometry then AB contains exactly two of the points D, E and F. 4. Use Problem A3 to give an alternative proof of Theorem 3.4.4. 5. In_a metric geometry, prove that if A, B and C are not collinear then AB = AB ABC. 6. Use Problem AS to prove Theorem 3.4.4. 7. Prove that, in a metric geometry, if AABC = ADEF then AB contains two of the three points D, E, and F. 8. Use Problem A7 to prove Theorem 3.4.4, Part C. Expository exercises. 9. Prior to Lemma 3.4.1 there is a discussion of the idea of what it means for a concept to be “well defined”. What examples do you know about from your previous mathematics courses where a concept needed to be well defined? What concepts in this course need to be well defined? Explain what the notion of “well defined” is in your own words.CHAPTER 4 Plane Separation 4.1 The Plane Separation Axiom The Plane Separation Axiom is a careful statement of the very intuitive idea that every line has “two sides.” Such an idea seems so natural that we might expect it to be a consequence of our present axiom system. However, as we shall see in Section 4.3, there are models of a metric geometry that do not satisfy this new axiom. Thus the Plane Separation Axiom does not follow from the axioms of a metric geometry, and it is therefore necessary to add it to our list of axioms if we wish to use it. In this section we will introduce the concept of convexity, use it to state the Plane Separation Axiom, and develop some of the very basic results coming from the new axiom. In the second section we will show that our two basic models, the Euclidean Plane and the Poincaré Plane, do satisfy this new axiom. In the third section we will introduce an alternative formulation of plane separation in terms of triangles. This substitute for the Plane Separation Axiom is called Pasch’s Postulate. We shall see that it is equivalent to the Plane Separation Axiom: any metric geometry that satisfies one of these axioms satisfies the other. Definition. Let {%, ¥,d} be a metric geometry and let Y, cS. S, is said to be convex if for every two points P, Qe Y,, the segment PQ is a subset of SF. In Figure 4-1 each of the individual subsets of R? is convex while each of those in Figure 4-2 is not convex. In Figure 4-2 the segment PQ is contained in S, but the segment PQ’ is not. This means that ¥, is not convex: convexity requires that the segment between any two points of % be in 4, 6364 4. Plane Separation o| @ Figure 4-1 Figure 4-2 not just some. To show that a set is convex we must show that for every pair of points in the set, the segment joining them is contained in the set. To show a set is not convex, we need only find one pair of points such that the line segment joining them is not entirely contained in the set. We should also note that the concept of convexity depends on the metric geometry. This is because convexity involves line segments, which in turn involve betweenness, which is defined in terms of distance. Thus a change in the distance function affects which sets are convex. For example, consider the set of ordered pairs (x, y) with (x — 1)? + y?=9,0 =0 implies that Z =tX for some te R. Proor. We leave part (a) to Problem A1. For part (b) we proceed as follows. Let X = (x, y) and Z = (z,w) so that X+ = (—y, x). Then ¢Z, X+) = 0 means ~zy + wx =0. (2-2) Since X ¥ (0,0) one of x and y is not zero. If x # 0 then we may solve Equa- tion (2-2) for w = zy/x so that Z =X with ¢ = 2/x. If y #0 then z = xw/y so that Z = tX with t = w/y. Either way Z = tX for some te R. Oo Using X+ we can give an alternative description of a line in R?. Our mo- tivation here is the idea from linear algebra that a line can be described by giving one point on the line and a vector normal to it. See Figure 4-9. Proposition 4.2.2. If P and Q are distinct points in R? then PO = {Ae R*| =0. Now Q — P # (0,0) since Q # P. Thus by Lemma 4.2.1 there is a real number ¢ with A — P = t(Q — P). Hence A=P+t(Q— Pye PO. Thus {A € R?|(A — P,(Q — P)') = 0} c PG. We now have containment in both directions so the sets are equal. o Definition. Let |! = PQ be a Euclidean line. The Euclidean half planes deter- mined by / are H* = {Ae R?|(A — P), (Q — P)*) > 0} H” = {Ae R*|(A — P), (Q — P)*) <0} (See Figure 4-10.) 2-4) Figure 4-102 4 Plane Separation Proposition 4.2.3. The Euclidean half planes determined by | = PQ are convex. Proor. We will handle only the case of H* and leave H~ as Problem A2. Let A, Be H* so that (A-P)(Q—-P)*)>0 and (B-P)(Q—P)*>>0 (2-5) We must show that if C ¢ AB then Ce H*. Since A, Be H* we need only consider the case A—C—B. Thus by Proposition 3.3.3 we may assume that there is a number t with 0 = (1 = (A — P) + (B — P)),(Q — P)*> = (1 — t)<(A — P),(Q — P)*> + t(B — P), (Q — P)*> Since 0 > O and Ce H*. oO Proposition 4.2.4. The Euclidean Plane satisfies PSA. Proor. Let 1 = PQ be a line. If A eR? then <(A — P), (Q — P)*» is either positive (so that A € H*), zero (so that A € | by Proposition 4.2.2), or negative (so that Ae H~). Thus R? —!=H* UH. Since H* and H™ are clearly disjoint and Proposition 4.2.3 says they are convex, we need only show that condition (iii) of PSA holds. Let Ae H* and Be H~. To show that AB 1! 4 @ we must find ¢ with 0 (2-7) Since Ae H*, the left hand side of (2-7) is positive. We now show that <(A — B), (Q — P)") is also positive. Since A — B =(A — P) — (B — P) we have «(A = B), (Q — P)') = (A — P),(Q — P)'>) — (B— P),(Q— P)*> (2-8) The first term on the right is positive because A ¢ H*, whereas the second term is negative because Be H~. Thus the difference is positive. Hence4.2. PSA for the Euclidean and Poincaré Planes 3 we may divide Equation (2-7) by <(A — B),(Q — P)*> to obtain «(4 = P),(Q — Py) t=-———_-~__~ >0 2-9) (4B, (Q—P)> el To finish the proof we must show that the value of t in Equation (2-9) is less than one. Note that Equation (2-8) implies that the numerator of t is less than the denominator. Hence t < 1. With the value of ¢ given by Equation (2-9) we have a point X = 4+ ¢(B— A)e ABO l. Qo An alternate proof that condition (iii) is satisfied is given in Problem A3. Now we turn our attention to the Poincaré Plane. In this proof we shall use calculus. A reader who has not had calculus should skip the proof and go on to Section 4.3. The results we need from calculus are (i) if f'(0) > 0 for all ¢ then f(t) is an increasing function; (ii) the Intermediate Value Theorem, which says that if f(t) is a continuous function and f(a) a} H_ = {(x,y)e H|x 17} H- = {(x,y)e H|(x — 0)? + y? 0, g; is always increasing, . If AB = JL, is a type II line we parametrize AB by (x, y) ¢ AB if and only if (x, y) = (d + s tanh(t),s sech(t)) = fr,(t). Sir is the inverse of the standard ruler for ,L,. Again we let guilt) =(x- oP +? — 7? =(d—c +s tanh(t))? + (s sech(t))? — r? and find that Git) = 2(d — c + s tanh(t))s sech?(t) + 2(s sech(t))(—s sech(t)tanh(t)) = 2(d — c)s sech?(t), so that gj; is either increasing (d > c) or decreasing (d <‘c) or constant (d=c). We let f =f, and g =g, if AB is a type I line and we let f = f,, and 9=491 if AB is a type II line. Thus we have a function f from the real numbers to AB and real numbers t, < tz with f(t,) = A and f(t,) = B. We also have a continuous real valued function g such that g(r) > 0 if f(t) € H,, a(t) < Of f(t) € H., and g(t) = Oif fit)el. We now prove that H, is convex. Suppose A, Be H, and let f, g be as above so that A = f(t,) and B = f(t2) with t, < tz. If A—C—B then C= S(ts) with t, R by Sil, y) ifx <0 gil, y) = ie y—VJi+m ifx>1. The next result is Problem AS. Proposition 4.3.4. If {,£} is the Missing Strip Plane and |= L,,, then 9:1 S > R is a bijection. The coordinates of several points on the lines Lo; VY and Ly, 20 Ff are shown in Figure 4-15. For each vertical line | in # let g, be any Euclidean ruler. By Theorem 2.2.8 this collection of rulers g, determines a distance d’ on ¥ that makes {Y, L,d'} a metric geometry.80 4 Plane Separation 2-1 012 3 2 { 1 i % t Ae ' Figure 4-15 Proposition 4.3.5. The Missing Strip Plane is not a Pasch geometry. Proor. Consider AABC where A = (2,0), B=(2,3) and C=(—2,0). The line 1 Y, where |= Lpo,, intersects AB at D = (2,2). However, (ln A) AAC # Band (ln Y) BC = @, contradicting PP. See Figure 4-16. o t A Lo f Figure 4-16 Note that if PSA were to follow from the axioms of a metric geometry then every model of a metric geometry would satisfy PSA. Proposition 4.3.5 gives a model of a metric geometry which doesn’t satisfy PSA. Thus, Proposition 4.3.5 shows us that PSA really is an addition to our list of axioms and cannot be deduced from the previous ones. In Problems B6 and B8 there are two more examples where PSA is not satisfied. PROBLEM SET 4.3 Part A. 1. (Peano’s Axiom) Given a triangle A. ABC in a metric geometry which satisfies PSA and points D, E with B—C—D and A—E—C, prove there is a point F e DE with A—F—B, and D—E—F.4.4 Interiors and the Crossbar Theorem 81 v . Given AABC in a metric geometry which satisfies PSA and points D, F with B—C—D, A—F—B, prove there exists E € DF with A—E—C and D—E—F. y . Given A ABC and a point P in a metric geometry which satisfies PSA prove there is a line through P that contains exactly two points of A ABC. S }. Prove that the Missing Strip Plane is an incidence geometry. oy Prove Proposition 4.3.4. Part B. “Prove” may mean “find a counterexample”. 6. Let {R?, Z-, dy} be the metric geometry of Problem B20 of Section 4.1. Prove that PP is not satisfied. 7. Given AABC in a metric geometry and points D, E with A—~D—B and C—E—B, prove AENCD#¥@. 8. Let R? = {(x, y, 2)|x, y, ze R}. If A, Be R? define L4g = {A + t(B — A)|t € R}. Let L ={LyglA, Be R, A B}. If A, Be R? let d(A, B)= 1A — BI. Prove that {R°, Y, d} is a metric geometry but that it does not satisfy PSA. Part C. Expository exercises. 9. We have just shown (Theorems 4.3.1 and 4.3.3) that two axioms are equivalent. Write a short essay on the equivalence of axiom systems using as a reference an appropriate book on mathematical logic. 10. Find some middle or high school students, ask them to draw a triangle and to pick a point on the triangle. Then ask them to draw a line through the point. They will probably construct it so that it crosses one of the other sides. Ask them how they know it would cross the side and write up their reactions. (The answer that Euclidean geometry satisfies Pasch’s Postulate will be too subtle for them.) After the experiment be sure to tell them not to feel bad about not knowing the answer—neither did Euclid! 4.4 Interiors and the Crossbar Theorem In this section we will be interested in interiors—the interior of a ray, of a segment, of an angle, and of a triangle. These concepts will aid us in proving the main theorem of this section which says that a ray starting at a vertex of a triangle and which passes through a point in the interior of the angle at that vertex must intersect the opposite side; that is, it must “cross the bar.” Theorem 4.4.1. In a Pasch geometry if sf is a nonempty convex set that does not intersect the line I, then all points of of lie on the same side of |.82 4 Plane Separation Proor. Let A € of and let B be any other point of . Since of is convex, AB < of. Since 01= @, AB l= @. Thus A and B are on the same side of |. Thus every point of is on the same side of / as A is. Oo Definition. The interior of the ray 4B in a metric geometry is the set int(AB) = AB—{A}. The interior of the segment AB in a metric geometry is the set int(AB) = AB—{A, B}. In Problem A1 you will show that the interior of a ray or a segment is convex. Theorem 4.4.1 can then be applied in a number of interesting cases. The proof of the next result is left to Problem A2. Theorem 4.4.2. Let of be a line, ray, segment, the interior of a ray, or the interior of a segment in a Pasch geometry. If | is a line with of 1 = @ then all of of lies on one side of |. If there is a point B with A—B—C and AC 01 = {B} then int(BA) and int(BA) both lie on the same side of | while int(BA) and int(BC) lie on opposite sides of |. Theorem 4.4.3 (Z Theorem). In a Pasch geometry, if P and Q are on opposite sides of the line AB then BP \ AQ = &. In particular, BP AQ = S. Proor. See Figure 4-17. By Theorem 4.4.2, int(BP) lies on one side of AB and | int(AQ) lies. on the other (since J P and Q are on opposite sides). Thus int(BP) 0 int(AQ) = @. Since B ¢ AG ( (because A, B, Q are not collinear) we have BP nint(AQ) @. Since A ¢BP, we have BP n AQ = &. The rest follows from BP c BP and AQ c AQ. Oo Figure 4.17 Theorem 4.4.3 is surprisingly useful. The key to using it is to recognize a “Z configuration” in the picture you have sketched. With a little imagination Figure 4-17 looks like a Z. The name of the theorem comes from this observa- tion. The Z Theorem will be used repeatedly in the proof of the Crossbar Theorem.