1 A.B.Sossinsky GEOMETRIES September2011. First13chaptersforRussianstudents. Notethat”chap- ter” means ”lecture”. 2 Chapter 1 TOY GEOMETRIES AND MAIN DEFINITIONS In this chapter, we study five toy examples of geometries (symmetries of the equilateral triangle, the square, the cube, and the circle) and a model of the geometry of the so-called elliptic plane. These examples are followed by the main definition of this course: a geometry in the sense of Klein is a set with a transformation group acting on it. We then present some useful gen- eral notions related to transformation groups. Finally, we study the relation- ships (called morphisms or equivariant maps) between different geometries, thus introducing the category of all geometries. The notions introduced in this chapter are illustrated by some problems (dealing with toy models of geometries) collected at the end of the chapter. But before we begin with these topics, we briefly recall some terminology from elementary Euclidean geometry. 1.1. Isometries of the Euclidean plane and space We assume that the reader is familiar with the basic notions and facts of Euclidean geometry in the plane and in space. One can think of Euclidean geometryasanaxiomatictheory(nottoorigorouslytaughtinhighschool)or as a small chapter of linear algebra (the plane R2 and the space R3 supplied with the standard metric). It is irrelevant for us which of these two points of view is adopted by the reader, and the aim of this subsection is merely to fix some terminology common to the two approaches. AnisometryoftheEuclideanplaneR2 (orspaceR3)isamapf : R2 R2 → (respectively f : R3 R3) which preserves the distance d between points, → i.e., d(f(P),f(Q)) = d(P,Q) for any pair of points P,Q of the plane (resp. of space). There are two types of isometries: those which preserve orientation (they are called motions) and those that reverse orientation (orientation- reversing isometries). In the plane, examples of motions are parallel translations (determined by a fixed translation vector) and rotations (determined by a pair (C,α), where C is the center of rotation and α is angle of rotation. In space, examples of motions are parallel translations and rotations (about an axis). Rotations in space are determined by pairs (l,α), where l is the axis of rotation, i.e., a straight line with a specified direction on it, and α is the angle of rotation; the rotation (l,α) maps any point M in space to the point M(cid:48) obtained by 3 rotating M in the plane Π perpendicular to l by the angle α counterclockwise if one looks at the plane from “above”, i.e., from some point of l obtained from the point l Π by moving in the direction specified on the axis. ∩ Examples of orientation-reversing isometries in the plane are reflections (i.e., symmetries with respect to a line). In space, examples of orientation- reversing isometries are given by mirror symmetries (i.e., reflections with respect to planes) and point symmetries (i.e., reflections with respect to a point). All other isometries of the Euclidean plane and space are compositions of those listed above. The reader who feels uncomfortable with the notions mentioned in this subsection is invited to study Appendix I at the end of the book. 1.2. Symmetries of some figures 1.2.1. Symmetries of the equilateral triangle. Consider all the isometries of the equilateral triangle = ABC, i.e., all the distance-preserving map- (cid:52) pings of this triangle onto itself. (To be definite, we assume that the letters A,B,C have been assigned to vertices in counterclockwise order.) Denote by s , s , and s the reflections in the bisectors of angles A, B, C of the A B C triangle. Denote by r , r , r the counterclockwise rotations about its center 0 1 2 of gravity by 0, 120, 240 degrees, respectively. Thus r takes the vertex A to 1 B, B to C, and C to A. These six transformations are all called symmetries of triangle ABC and the set that they constitute is denoted by Sym( ). (cid:52) Thus Sym( ) = r , r , r , s , s , s . 0 1 2 A B C (cid:52) { } Therearenootherisometriesof . Indeed, anyisometrytakesverticesto (cid:52) vertices, eachone-to-onecorrespondencebetweenverticesentirelydetermines the isometry. (For example, the correspondence A B, B A, C C → → → determines the reflection s .) But there are only six different ways to assign C the letters A,B,C to three points, so there cannot be more than 6 isometries of . (cid:52) In a certain sense, Sym( ) is the same thing as the family of all permu- (cid:52) tations of the three letters A,B,C; this remark will be made precise in the next chapter. We will use the symbol to denote the composition (or product) of isome- ∗ tries, in particular of elements of Sym( ), and understand expressions such (cid:52) 4 as r s to mean that r is performed first, and then followed by s . Ob- 1 A 1 A ∗ viously, when we compose two elements of Sym( ), we always obtain an (cid:52) element of Sym( ). (cid:52) What element is the composition of two given ones can be easily seen by drawing a picture of the triangle ABC and observing what happens to it when the given isometries are successively performed, but this can also be donewithoutanypictures: itsufficestofollowthe“trajectory”ofthevertices A,B,C. Thus, in the example r s , the rotation r takes the vertex A to 1 A 1 ∗ B, and then B is taken to C by the symmetry s ; similarly, B C B A → → and C A A, so that the vertices A,B,C are taken to C,B,A in that → → order, which means that r s = s . 1 A B ∗ The order in which symmetries are composed is important, because the resultingsymmetrymaychangeifweinversetheorder. Thus,inourexample, s r = s = s (as the reader will readily check), so that r s = s r . A 1 C B 1 A A 1 ∗ (cid:54) ∗ (cid:54) ∗ So for elements of Sym( ), composition is noncommutative. (cid:52) The compositions of all possible pairs of symmetries of can be conve- (cid:52) niently shown in the following multiplication table: r r r s s s 0 1 2 A B C ∗ r r r r s s s 0 0 1 2 A B C r r r r s s s 1 1 2 0 B C A r r r r s s s 2 2 0 1 C A B s s s s r r r A A C B 0 1 2 s s s s r r r B B A C 2 0 1 s s s s r r r C C A B 1 2 0 Here (for instance) the element s at the intersection of the fifth column V and the third row is s = r s , the composition of r and s in that order B 1 A 1 A ∗ (first the transformation r is performed, then s ). 1 A As we noted above, composition is noncommutative, and this is clearly seen from the table (it is not symmetric with respect to its main diagonal). The composition operation in Sym( ) is (obviously) associative, i.e., (i j) k = i (j k) for all i,j∗,k Sym((cid:52)). The set Sym((cid:3)) contains the ide∗ntit∗y transf∗orm∗ation r (also den∈oted id(cid:52)or 1). Any element i of Sym((cid:3)) 0 has an inverse i−1, i.e., an element such that i i−1 = i−1 i =1. ∗ ∗ The set Sym( ) supplied with the composition operation is called the (cid:52) ∗ symmetry group of the equilateral triangle. 5 1.2.2. Symmetries of the square. Consider all the isometries of the unit square (cid:3) = ABCD, i.e., all the distance-preserving mappings of the square to itself. A B r 2 r 1 S H S S bd ac S V D C Figure 1.1. Symmetries of the square Let us denote by s , s , and s , s the reflections in the horizontal and H V ac bd vertical mid-lines, and in the diagonals AC, BD, respectively. Denote by r , r , r , r the rotations about the center of the square by 0, 90, 180, 270 0 1 2 3 degrees, respectively. These eight transformations are all called symmetries of the square. We write Sym((cid:3)) = r , r , r , r , s , s , s , s . 0 1 2 3 H V ac bd { } Just as in the case of the equilateral triangle, the composition of any two symmetries of the square is a symmetry of the square, and a multiplication table, indicating the result of all pairwise compositions, can be drawn up: r r r r s s s s 0 1 2 3 H V ac bd ∗ r r r r r s s s s 0 0 1 2 3 H V ac bd r r r r r s s s s 1 1 2 3 0 ac bd V H r r r r r s s s s 2 2 3 0 1 V H bd ac r r r r r s s s s 3 3 0 1 2 bd ac H V s s s s s r r r r H H bd V ac 0 2 3 1 s s s s s r r r r V V ac H bd 2 0 1 3 s s s s s r r r r ac ac H bd V 1 3 0 2 s s s s s r r r r bd bd V ac H 3 1 2 0 Here (for instance) the element s at the intersection of the sixth column V and the fourth row is s = r s , the composition of r and s in that V 2 H 2 H ∗ 6 order (first the transformation r is performed, then s ). Composition is 2 V noncommutative. Obviously, composition is associative. The set Sym((cid:3)) contains the iden- tity transformation r (also denoted id or 1). Any element i of Sym((cid:3)) has 0 an inverse i−1, i.e., an element such that i i−1 = i−1 i =1.. The set Sym((cid:3)) supplied with the com∗position o∗peration is called the symmetry group of the square. 1.2.3. Symmetries of the cube. Let I3 = (x,y,z) R3 0 x 1,0 y 1,0 z 1 { ∈ | ≤ ≤ ≤ ≤ ≤ ≤ } betheunitcube. Asymmetryofthecubeisdefinedasanyisometricmapping of I3 onto itself. The composition of two symmetries (of I3) is a symmetry. How many are there? Let us first count the orientation-preserving isometries of the cube (other than the identity), i.e., all its rotations (about an axis) by nonzero angles that take the cube onto itself. 120 ◦ 90 ◦ 180 ◦ Figure 1.2. Rotations of the cube There are three axes of rotation joining the centers of opposite faces, and the rotation angles for each are π/2, π, 3π/2. There are four axes of rotation joining opposite vertices, the rotation angles for each being 2π/3 and 4π/3. There are six axes of rotation joining midpoints of opposite edges, with only one nonzero rotation for each (by π). This gives us a total of (3 3)+(4 2)+(6 1) = 23 orientation-preserving isometries, or 24 if we × × × include the identity. There are no other orientation-preserving isometries; at this point, we could prove this fact by a tedious elementary geometric counting argument, 7 but we postpone the proof to Chapter 3, where it will be the immediate result of more general and sophisticated algebraic method. There are also 24 orientation-reversing isometries of the cube. Listing them all is the task prescribed by Exercise 1.2 (see the end of the chapter), a task which requires little more than a bit of spacial intuition. Thus the cube has 48 isometries. All their pairwise compositions consti- tute a multiplication table, which is a 49 by 49 array of symbols, much too unwieldy to fit in a book page. The set Sym(I3) of all 48 symmetries of the cube supplied with the com- position operation is called the symmetry group of the cube; it is associative, noncommutative, has an identity, and all its elements have inverses, just as the symmetry groups in the two previous examples. 1.2.4. Symmetries of the circle. Let := (x,y) R2 x2 +y2 = 1 (cid:13) { ∈ | } be the unit circle. Denote by Sym( ) the set of all its isometries. The (cid:13) elements of Sym( ) are of two types: the rotations r about the origin by ϕ (cid:13) angles ϕ, ϕ [0,2π), and the reflections in lines passing through the origin, ∈ s , α [0,π), where α denotes the angle from the x-axis to the line (in the α ∈ counterclockwise direction). The composition of rotations is given by the (obvious) formula r r = r , φ ψ (φ+ψ)mod2π ∗ where mod 2π means that we subtract 2π from the sum φ+ψ if the latter is greater than or equal to 2π. The composition of two reflections s and s is a rotation by the angle α β α β , | − | s s = r . α β 2(α−β) ∗ Theinterestedreaderwillreadilyverifythisformulabydrawingapictureand comparingtheanglesthatwillappearwhenthetworeflectionsarecomposed. The set of all isometries of the circle supplied with the composition oper- ation is called the symmetry group of the circle and is denoted by Sym( ). (cid:13) The group Sym( ) has an infinite number of elements. As before, this (cid:13) group is associative, noncommutative, has an identity, and all its elements have inverses. 1.2.5. Symmetries of the sphere. Let S2 := (x,y) R3 x2 +y2 +z2 = 1 { ∈ | } 8 be the unit sphere. Denote by Sym(S3) the set of all its isometries and by Rot(S3) the set of all its rotations (by different angles about different axes passing through the center of the sphere). Besides rotations, the transforma- tion group Sym(S3) contains reflections in different planes passing through the center of the sphere, its symmetry with respect to its center, and the composition of these transformations with rotations. Reflections in planes, unlike rotations, reverse the orientation of the sphere. This means that a little circle oriented clockwise on the sphere (if we are looking at it from the outside) is transformed by any reflection into a counterclockwise oriented circle, and the picture of a left hand drawn on the sphere becomes that of a right hand. Now a reflection in a line passing through the sphere’s center does not reverse orientation (unlike reflections in the plane!) because a reflection of the sphere in a line is exactly the same transformation as a rotation about this line by 180◦. On the other hand, a reflection of the sphere with respect to its center reverses its orientation (again, thisisnotthecaseforreflectionsoftheplanewithrespecttoapoint). Note that the composition of two reflections in planes is a rotation (see Exercise 1.11), while the composition of two rotations is another rotation (by what angle and about what axis is the question discussed in Exercise 1.12). The set of all isometries of the sphere supplied with the composition oper- ation is called the symmetry group of the sphere and is denoted by Sym(S3). The group Sym(S3) has an infinite number of elements. As before, this group is associative, noncommutative, has an identity, and all its elements have in- verses. 1.2.6. A model of elliptic plane geometry. Consider the set Ant(S2) of all pairs of antipodal points (i.e., points symmetric with respect to the origin) on the unit sphere S2); thus elements of Ant(S2) are not ordinary points, but pairs of points. Now consider the family (that we denote O(3)) of all isometries of the space R3 that do not move the origin 1. Clearly, any such isometry takes pairs of antipodal points to pairs of antipodal points, thus it maps the set X = Ant(S2) to itself. The family O(3) of transformations of the set Ant(S2) is called the isome- try group of the Riemannian elliptic plane. This is a much more complicated object than the previous “toy geometries”. We will come back to its study in Chapter 6. 1InlinearalgebracoursessuchtransformationsarecalledorthogonalandO(3)iscalled the orthogonal group. 9 1.3. Transformation groups 1.3.1. Definitions and notation. Let X be a set (finite or infinite) of arbitrary elements called points. By definition, a transformation group G acting on X is a (nonempty) set G of bijections of X supplied with the composition operation and satisfying the following conditions: ∗ (i) G is closed under composition, i.e., for any transformations g,g(cid:48) G, ∈ the composition g g(cid:48) belongs to G; ∗ (ii) G is closed under taking inverses, i.e., for any transformation g G, ∈ its inverse g−1 belongs to G. These conditions immediately imply that G contains the identity trans- formation. Indeed, take any g G; by (ii), we have g−1 G; by (i), we have ∈ ∈ g−1 g G; but g−1 g =id (by definition of inverse element), and so id G. ∗ ∈ ∗ ∈ Note also that composition in G is associative (because the composition of mappings is always associative). If x X and g G, then by xg we denote the image of the point x under ∈ ∈ the transformation g. (The more usual notation g(x) is not convenient: we have x(g h) = (xg)h, but (g h)(x) = h(g(x)), with g and h appearing in ∗ ∗ reverse order in the right-hand side of this equality.) 1.3.2. Examples. The five toy geometries considered in the previous section all give examples of transformation groups. The five transformation groups Sym act (by isometries) on the equilateral triangle, the square, the cube, the circle, and the sphere, respectively. In the last example (1.2.5), the orthogonal group O(3) acts on pairs of antipodal points on the sphere, these pairs being regarded as “points” of the “elliptic plane”. More examples are given by the transformation group consisting of all the bijections Bij(X) of any set X. By definition of transformation groups, Bij(X) is the largest (by inclusion) transformation group acting on the given setX. Attheotherextreme, anysetX hasatransformationgroupconsisting of a single element, the identity transformation. When the set X is finite and consists of n objects, the group Bij(X) of all its bijections is called the permutation group on n objects and is denoted by Σ . This group is one of the most fundamental notions of mathematics, n and plays a key role in abstract algebra, linear algebra, and, as we shall see already in the next chapter, in geometry. 1.3.3. Orbits, stabilizers, class formula. Let (X : G) be some transfor- mation group acting on a set X and let x X. Then the orbit of x is defined ∈ 10 as Orb(x) := xg g G X, { | ∈ } ⊂ and the stabilizer of x is St(x) := g G xg = x G. { ∈ | } ⊂ Forexample, ifX = R2 andGistherotationgroupoftheplaneaboutthe origin, then the set of orbits consists of the origin and all concentric circles centered at the origin; the stabilizer of the origin is the whole group G, and the stabilizers of all the other points of R2 are trivial (i.e., they consist of one element – the identity id G). ∈ Suppose (X : G) is an action of a finite transformation group G on a finite set X. Then the number of points of G is (obviously) given by G = Orb(x) St(x) (1.1) | | | |×| | for any x X. Now let A X be a set that intersects each orbit at exactly ∈ ⊂ one point. Then the number of points of X is given by the formula G X = | | , (1.2) | | St(x) x∈A | | (cid:88) called the class formula. This formula, just as the previous one, follows immediately from definitions. 1.3.4. Fundamental domains. If X is a subset of Rn (e.g. Rn itself) and G is a transformation group acting on X, then a subset F X is called a ⊂ fundamental domain of the action of G on X if F is an open set in X ; • F Fg = ∅ for any g G (except g =id).; • ∈ X(cid:84)= Clos(Fg), where Clos(.) denotes the closure of a set. • g∈G (cid:83) For example, in the case of the square, a fundamental domain of the action of Sym((cid:3)) is the interior of the triangle AOM, where O is the center of the square and M is the midpoint of side AB; of course Sym((cid:3)) has many other fundamental domains. Thus fundamental domains are not necessarily