* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Algebraic Topology Introduction
Survey
Document related concepts
Michael Atiyah wikipedia , lookup
Sheaf cohomology wikipedia , lookup
Orientability wikipedia , lookup
Sheaf (mathematics) wikipedia , lookup
Brouwer fixed-point theorem wikipedia , lookup
Continuous function wikipedia , lookup
Surface (topology) wikipedia , lookup
Geometrization conjecture wikipedia , lookup
Grothendieck topology wikipedia , lookup
Homotopy groups of spheres wikipedia , lookup
General topology wikipedia , lookup
Homological algebra wikipedia , lookup
Transcript
1 Point Set Topology Partitions of unity, some common topologies, connectedness, compactness A good overflow thread discussing the definition of topology in terms of open sets http://mathoverflow. net/questions/19152/why-is-a-topology-made-up-of-open-sets/19156#19156 1.1 Basics Proposition 1.1. Let f : X → Y , with A ⊆ X and B ⊆ Y . • A ⊆ f −1 ◦ f (A) with equality if and only if f is injective. • f ◦ f −1 (B) ⊆ B with equality if and only if f is surjective. • A bijective continuous map is a homeomorphism if and only if it is open (or closed.) Include more properties about the inverse of a map e.g. inverse of union, intersection, etc. Proposition 1.2. A topological space X is Hausdorff if and only if the diagonal D = {(x, x) : x ∈ X} is a closed subset of X × X. Some results about compact spaces: See Terry Tao’s intuitive notes on compact spaces, http://www.math.ucla.edu/~tao/preprints/compactness. pdf Compact spaces can be roughly characterized as having tamed infinity. For example, points can’t quite run off to infinity like on a non-compact space e.g. R, For a metric space, compactness is equivalent to sequential compactness, that is, every sequence has a convergent sub-sequence. See also, Konig’s lemma. Theorem 1.3. 1. Every closed subset of a compact space is compact. 2. Every compact subset of a Hausdorff space is closed. 3. The image of a compact space under a continuous map is compact. 4. Let Y be Hausdorff with f : X → Y a continuous map. The inverse image of a compact set is closed. Further if X is compact then the inverse image of a compact set is compact. 5. Let f : X → Y be a bijective continuous function. If X is compact and Y is Hausdorff, then f is a homeomorphism. Proof. (2) Let X be a topological space with C ⊆ X a compact subspace. Let x ∈ X − C, we aim to show that there is an open set disjoint from C that contains X (i.e. C is closed.) Take any y ∈ C, and there exists an open neighborhood about y, Vy , and an open neighborhood about x, Uy , which are disjoint (definition of Hausdorff.) Now let the point y move around in C, so the set {Vy } makes an open cover of C. There exists a finite subset F ⊆ C, such that {Vy : y ∈ F } forms a finite open cover of C (definition of compact.) Now for each open set Vy with y ∈ F , take its partner Uy defined above. As y runs over F , the intersection . U = ∩y∈F Uy is a set of points containing x and disjoint from all of the Vy , and hence disjoint from C. Lastly, U must be open since the intersection of a finite number of open sets is open. We really did need the finite subcover from the compact property! (3) Let f : X → Y be continuous, with X compact. Let B be an open cover of f (X). Consider A = {f −1 (B) : B ∈ B}. First note that A is a collection of open sets in X, since f is continuous and each B is open in Y . Second, A is an open cover of X, since if x ∈ X there exists a B such that f (x) ∈ B i.e. x ∈ f −1 (B). Since X is compact, there exists a finite subcover of A, call this finite collection Ā. 1 Every open set in Ā has the form f −1 (B) for some open B in Y . Then, apply f to each open set in Ā, f ◦ f −1 (B) = B since f surjects onto f (X). This finite collection of B’s forms a finite open cover of f (X), and therefore f (X) is compact. (4) Since Y is Hausdorff, a compact subset C of Y must be a closed subset. Then, f −1 (C) is closed in X, by definition of continuity. If X happens to be compact, then f −1 (C) is compact, since closed subsets of compact spaces are compact. (5) Since f is bijective and continuous, to prove that it is a homeomorphism, all that remains to show that f −1 is continuous. We will show that f is a closed map, and hence f −1 is continuous. Let A be a closed subset of X. Since X is compact, A is compact (part 1). Then, f (A) is compact in Y (part 3). Finally by part 2, since Y is Hausdorff and f (A) ⊆ Y is compact, f (A) is also closed. Therefore f is a closed mapping, and f is a homeomorphism. 1.2 Quotient Space Let X and Y be topological spaces. A surjective, continuous map p : X → Y is called a quotient map if V ∈ TY if and only if p−1 (V ) ∈ TX . If X is a topological space and Y is a set, with p : X → Y surjective, then we can topologize Y with the quotient topology by defining V open in Y if and only if p−1 (V ) is open in X. If we suppose that X is a topological space with subspace Z, we’d like to construct the quotient space X/Z. However there is a problem in saying “the” quotient space, as topologizing X/Z depends crucially on the quotient map p : X → Z. I think I was initially confused because there is not a canonical “quotient” in the category of topological spaces, as there is for say, the category of groups or modules. Note that if Z ⊆ X, most(?) authors define X/Z to be the space obtained by identifying Z with a point i.e. p(x) = x if x ∈ X − Z and p(z) = ∗ if z ∈ Z. Example 1.4. Take X to be a rectangle and glue together the boundary in some particular way, for example we may form a torus or a Klein bottle. Both spaces are quotient spaces formed from a rectangle, taking the quotient by the boundary, but observe that the quotient map is the tool carrying all the information. In some circumstances there should be no confusion by what is meant by /. For example if G is a topological group with H a normal subgroup, then G/H denotes the usual quotient group, with the quotient topology. Frequently quotient spaces arise as equivalence relations, as any surjective map of sets p : X → Y defines an equivalence relation on X by x1 ∼ x2 if and only if p(x1 ) = p(x2 ). In general, if R is an equivalence relation on X, then the quotient space X/R is the set of all equivalence classes of X under R, with topology defined by the quotient map p : X → X/R, which assigns to each x its equivalence class. Example 1.5. Let 1.3 Homeomorphisms/ Homotopy Definition 1.6. Two spaces X and Y are homotopy equivlent, or of the same homotopy type, if there exist continuous maps f : X → Y and h : Y → X such that f ◦ h ∼ idY and h ◦ f ∼ idX . We may call f and h “homotopy inverses”. Proposition 1.7. Homotopy equivalence of spaces defines an equivalence relation. Example 1.8. A deformation retract from a space X to a subspace A is a homotopy between a retraction and the identity map on X. That is, there exists a continuous map F : X × [0, 1] → X such that for every x ∈ X and every a ∈ A, F (x, 0) = id(x) = x, F (x, 1) ∈ A (the retraction map on x) with F (a, 1) = a. If we add the condition that F (a, t) = a for all t ∈ [0, 1], so that the homotopy fixes A at all times t, we may call it a strong deformation retract. Hatcher takes this as the definition of deformation retract. n n For example, there is a “strong” deformation retract from R − {0} → S . Use the homotopy F (x, t) = (1 − t) + t kxk x. 2 Example 1.9. Every homeomorphism defines a homotopy equivilance, but the converse isn’t true in general. Let f : X → Y be a homeomorphism. By definition of homeomorphism, f −1 is a continuous map taking Y to X, so take h = f −1 as in the definition of homotopy equivalence. Then, not only are f ◦ f −1 and f −1 ◦ f homotopic to the respective identity maps, they are on the nose equal. The converse is certainly not true. That is, if X and Y are homotopy equivalent spaces, they are not necessarily homeomorphic. For example take the letter X and the letter Y. These spaces are homotopy equivalent but not homeomorphic. Certainly Y is a deformation retract of X so that X and Y are homotopy equivalent. To show this more rigorously, take the map f : X → Y taking three of the prongs of X to the three of Y and the fourth to the center of Y , and the map g : Y → X taking those three prongs back to X, define a homotopy equivalence between X and Y . On the other hand, removing the center point from X gives a space with 4 connected components, but removing any point from Y gives at most 3. Thus, the spaces are not homeomorphic. Another example of homotopy equivalent spaces which are not homeomorphic is the mobius band and the cylinder (some interesting answers on the math stack exchange question), or a point and the disk (another deformation retract.) 1.4 Connectedness A separation of a topological space X is a pair of disjoint, nonempty open subsets of X whose union is X. X is said to be connected if no separation exists (which is sort of a lame definition.) Proposition 1.10. Path connected implies connected, but the converse is not true. For example, the topologist’s sine curve is defined T = {(x, sin( x1 ) : x ∈ (0, 1)} ∪ {(0, 0)}, and is connected but not pathconnected. Take the origin and some other point on the curve, since the sine part of the curve is not continuous at the origin there is no path (continous image of the unit interval into the space) connecting the points. Example 1.11. For a manifold, path connected and connected are equivalent. A quick proof: Let X be a manifold with x ∈ X. If X is path-connected, then it’s connected as the above proposition states. Suppose on the other hand that X is connected and define the set U to be the set of all points in X that may be connected by a path to X. We will show that U is open. Take u ∈ U , by definition of manifold there is a neighborhood, Bu ⊆ X which contains u and is homeomorphic to an open ball in Rn . Since path-connectedness is an invariant under homeomorphism, Bu is path-connected. But x ∈ Bu , so the definition of U dictates that Bu ⊆ U . Therefore, U is open. . By the same reasoning, V = X − U is open in X. Clearly X = U ∪ V , and since x ∈ U , V must be empty, otherwise we would have a separation of the connected space X. Definition 1.12. Define a relation ∼ on a topological space X by x ∼ y if there exists a connected subspace C of X containing x and y. Prove that this is an equivalence relation. The equivalence classes are called connected components of X. We see that each connected component (equivalence class) is in fact connected, and is thus a maximal connected subset of X. In general if A is a connected set, with A ⊂ B ⊂ Ā, then B must be connected. So by maximality, every connected component C satisfies C = C̄. Therefore, connected components are closed. Proposition 1.13. A locally constant function on a connected component of a topological space X is constant. Example 1.14. Let p : E → X be a vector bundle (with finite dimensional fibers)[3]. Local triviality of vector bundles says that for any open U ⊆ X, p−1 (U ) is homeomorphic to U × V for some vector space V . Then, the function on X which maps x → vdim(Ex ) is locally constant. It is therefore constant on the connected components of X. We see that the fibers over a connected component are all vector spaces of the same dimension, and hence isomorphic. 3 Example 1.15. [2] pg. 3 Let G be a topological group with G0 the component of the identity e ∈ G. Then, G0 is a closed normal subgroup of G. Normality buys us that G/G0 is a group, and further it is also a topological group. As discussed above, any connected component is closed, so G0 is closed. Now we show that G0 is normal. Recall that left and right multiplication are homeomorphism in a topological group. Definition 1.16. A space X is called n-connected if πi (X) is trivial for 1 ≤ i ≤ n. A 1-connected space (trivial fundamental group) is also called simply connected. 1.5 Operations on Spaces e.g. suspension and loop 2 Homology Theory Topological spaces are often huge and complicated objects, one strategy to simplify their study is to convert topological questions into algebraic questions via the (co)homology functors from the category of topological spaces to an algebraic category. Different homology theories apply to different classes of topological spaces. Simplicial, singular, and cellular homology are three of the most common theories. Simplicial theory deals with polyhedron (or spaces homeomorphic/homotopic to polyhedron). Simplicial theory is combinatorial in nature, as we can count vertices, edges, faces, etc., and they are glued together linearly so everything is nice. The combinatorial aspect of simplicial theory is exploited more generally in abstract homotopy theory using simplicial sets. It is also exploited via computational topology software (e.g. Javaplex.) Singular theory is a generalization of simplicial theory which applies to all spaces, and was fundamental to rigorously establish the theory. Cellular theory 2.1 Simplicial Homology Basic definitions as in Munkres [1]. Given a set of vertices {a0 , a1 , . . . , an } ⊆ Rm , geometric independence is the condition that a1 − a0 , . . . , an − a0 forms a linearly independent set of vectors. An n-simplexP is built m from a geometrically independent set of vertices by taking the set of all points x ∈ R where x = i ti ai , P ti = 1, and ti ≥ 0. Geometrically, a 0-simplex is a point, a 1-simplex a line, a 2-simplex a triangle, a 3-simplex a solid tetrahedron, and so-on for the higher dimensional analogoues. A polyhedron is a space formed by joining together simplices in a way that satisfies a couple of axioms. This is captured in the definition of simplicial complex. The algebraic topology Definition 2.1. A simplicial complex K is a collection of simplices in Rm such that 1. Every face of a simplex of K is in K. 2. The intersection of any two simplexes of k is a face of each of them. The underlying space of a simplicial complex K is denoted |K| and signifies the union of all simplices in K. A space that is the underlying space of a simplicial complex is called a polyhedron. Lemma 2.2. A collection K of simplies is a simplicial complex if and only if the following hold: 1. Every face of a simplex of K is in K. 2. Every pair of distinct simplices of K have disjoint interiors. Specifying a map between two simplicial complexes K and L is fairly straight-forward. Let K and L be simplicial complexes with f : K 0 → L0 a map on the 0-simplices. Suppose f has the property that whenever v0 , . . . , vn are the vertices of a simplex in K, the points f (v0 ), . . . , f (vn ) are the vertices for a 4 simplex P in L. Then, we P can extend f to a continuous map on the underlying spaces g : |K| → |L| defined: x= ti vi ⇒ g(x) = ti f (vi ). This map is linear on the vertex set and is called a simplicial map. In the case that f is a bijection, and a set of vertices forms a simplex in K if and only if the image of the vertices form a simplex in L, then it can be shown that g : |K| → |L| is a homeomorphism, and is called a simplicial homeomorphism. As Munkres aptly observes, “In practice, specifying a polyhedron X by giving a collection of simplices whose union is X is not a very convenient way of dealing with specific polyhedra...it is messy to specify all of the simplices and to make sure they intersect only when they should.” An abstract simplicial complex simplifies this issue. It is defined as a collection S of finite nonempty sets, such that if A ∈ S, so is every nonempty subset of A. A convenient way of expressing an abstract simplicial complex Theorem 2.3. (Classification theorem of closed surfaces) Every connected closed surface is homeomorphic to a connected sum of real projective planes and g-tori. (The sphere is considered the connected sum of zero tori and zero projective planes.) Both the torus and the projective plane can be triangulated, and so any connected closed surface is triangulable. As Hatcher puts it (pg. 102), “A polygon with any number of sides can be cut along the diagonals into triangles, so in fact all closed surfaces can be constructed from triangles by identifying edges.” Proposition 2.4. To what extent do the homology groups classify the closed, connected surfaces? General formula for homology of connected sum of manifolds of higher dimension? Which manifolds can be triangulated? May be hard to answer in general: http://mathoverflow.net/ questions/44021/which-manifolds-are-homeomorphic-to-simplicial-complexes. 2.2 Singular Homology Proposition 2.5. ([4] pg. 109) Decompose a space X into its path-components Xα . There is an isomorphism of Hn (X) with the direct sum ⊕α Hn (Xα ). Proof. Let σ : ∆n → X be a singular n-simplex. The image of a path-connected space under a continuous map is path-connected, thus σ(∆n ) ⊆ Xα for some α. Then, the singular chain group Cn (X) (which is made of finite formal linear combinations of n-simplices) adds over the path-connected components of X i.e. Cn (X) = ⊕α Cn (Xα ). The kernel and image of the boundary map necessarily preserve this decomposition, and so the homology groups do too. Lemma 2.6. Let f : X → Y be a continuous map. Then, f induces a chain map f# on the singular chain groups. That is, for each n, f# : Cn (X) → Cn (Y ) is a homomorphism, and ∂f# = f# ∂ Moreover, f induces the map f∗ , with f∗ : Hn (X) → Hn (Y ), a homomorphism for all n. Proof. Lemma 2.7. Suppose that f and g are continuous maps taking X to Y with f ' g. Then, f∗ = g∗ as maps on homology. Proof. By definition there exists a continuous map F : X × [0, 1] → Y with the property that F (x, 0) = f (x) and F (x, 1) = g(x). Corollary 2.8. Homotopy equivalent spaces have isomorphic homology Proof. This is a simple application of functoriality of the homology functor. Let f : X → Y and g : Y → X be continuous maps such that f ◦ g ' IdY and g ◦ f ' IdX . Then as maps on homology,(f ◦ g)∗ ' (IdY )∗ and (g ◦ f )∗ ' (IdX )∗ . Since the homology functor takes identities to identities and splits over composition, we have that f∗ : H∗ (X) → H∗ (Y ) and g∗ : H∗ (Y ) → H∗ (X) are inverses as group homomorphisms, and therefore, H∗ (X) ∼ = H∗ (Y ). 5 Example 2.9. The converse to the above corollary is not true. Topological spaces with the same homology are not necessarily homotopic! For example, take X = S 1 ∨ S 1 ∨ S 2 and Y = S 1 × S 1 . Since homology is additive over the wedge, we see that these two spaces have the same homology groups. However, a computation of the fundamental groups shows that they cannot be homeomorphic. Poincare’s conjecture states that every every simply connected, closed 3-manifold is homeomorphic to the 3-sphere. The Poincare homology sphere is an example of a closed 3-manifold which has the same homology as the 3-sphere but has a non-trivial fundamental group (and thus is not homotopic to the 3-sphere.) 3 3.1 Cohomology Theory Differentiable manifolds and the de Rham cohomology It is typical that students of mathematics begin their education by passing through a sequence of calculus classes, so from a pedagogical perspective, it may be best to motivate the study of algebraic topology by thinking about topological spaces which are differentiable manifolds and the associated cohomology groups called de Rham cohomology groups. These objects generalize many of the concepts which students encounter in calculus. In particular the fundamental theorems of multivariable calculus: Stokes’ theorem, Green’s theorem, the Divergence theorem, and the fundamental theorem of line integrals all have a simple explanation in the language of differential topology. 3.2 Interpreting cup/cap product in differential manifolds Cup product should correspond to wedge product. Cap product is a pairing of a homology class and a cohomology class, should have to do with an integral. Poincare duality. 4 Homotopy Theory While (co)homology theory uses functors to address topological problems in an algebraic category, homotopy theory is more concerned with the intrinsic algebraic properties of the category T OP (topological spaces with continuous maps.) For example, while the algebraic notions of kernel or cokernel do not apply to a continuous map between topological spaces, there do exist the related ideas fibration and (co)fibration. I’ve often read that the point of attaching algebraic invariants to topological spaces is to Let X and Y be CW-complexes, with f : X → Y a continuous map. If f∗ : πn (X) → πn (Y ) Include Hurewicz and Whitehead theorems. Then define eilenberg maclane spaces Proposition 4.1. A CW-complex X is K(G, 1) if and only π1 (X) ∼ = G, and the universal cover of X is contractible. 4.1 Fibrations and Fiber Bundles pg. 19 Benson A Serre fibration is a map p : E → B that satisfies the homotopy lifting property with respect to all simplicial complexes. Serre fibrations give rise to a long exact sequence of homotopy groups. On homology and cohomology, a Serre fibration gives rise to a spectral sequence. Note that all fibers in a path component of B are homotopy equivalent. A covering space p : E → B is a fiber bundle such that the fibers are discrete sets. The fundamental group π1 (B, b0 ) acts on F = p−1 (b0 ). If E and B are both path-connected Thm: Every fiber bundle is a Serre Fibration. 6 5 Common spaces and their topological properties 5.1 The Sphere Topological Properties 1. Path connected for all n > 0 2. Simply Connected for n > 1 3. Orientable for all n. Proof: Hn (S n ; Z) = Z for all n. (See computation below) 4. S n is compact. It’s clearly closed and bounded, so apply the Heini-Borel theorem. 5. Smooth manifold for all n Computations Homology Groups Cohomology Homotopy Groups Special Cases/Misc. 1. S n is a Lie Group for n = 0, 1, 3 2. Poincare Conjecture - Every simply-connected closed 3-manifold is homeomorphic to the 3-sphere. 3. Exotic Spheres, Milnor’s paper on the 7-sphere. Constructing differentiable manifolds which are homeomorphic but not diffeomorphic to the 7-sphere. http://www.scientificamerican.com/article/ hypersphere-exotica/ 5.2 Projective Space Topological Properties 1. Projective space can be realized as an orbit space. The antipodal action of Z/2Z on S n has as orbit space RP n . Similarly, let S 1 act on S 2n+1 by multiplication in each factor. The orbit space S 2n+1 /S 1 can be identified with CP n . 2. Projective space is compact. Consider its realization as an orbit space. e.g. The quotient map p : S n → RP n /(Z/2Z) is a continuous map on a compact space, and since the continuous image of compact is compact, we see that RP n is compact. Computations Homology Groups Cohomology Homotopy Groups Special Cases/Misc. 1. The Riemann Sphere CP 1 . Mobius transformations... 7 5.3 Grassmann and Stiefel Manifolds Let W be a vector space. As a set, the Stiefel variety V n (W ) is the collection of all ordered, orthonormal n-frames, and the Grassmanian Gn (W ) is the collection of all n-dimensional subspaces of W . Topological Properties 1. How to topologize the sets... 2. The Grassmanian Gn (Rm ) is a homogeneous space. By basic linear algebra, any n-dimensional subspace of Rm can be transformed to any other ndimensional subspace. Thus O(n) acts on Gn (Rm ) transitively. If x ∈ Gn (Rm ), then since O(n) is a compact Lie group, we have the relationship between orbits and isotropy: O(n)/O(n)x ∼ = O(n)x. Now, transitivity buys us that O(n)x ∼ = Gn (Rm ). Similarly, if we replace R by C, and O(n) by U (n), the same argument holds. 3. The Grassmanian is a compact manifold. In the last item, we saw that the quotient O(n)/O(n)x is homeomorphic to Gn (Rm ). Then, the quotient map q : O(n) → O(n)/O(n)x is a continuous, surjective map on a compact space. Thus, its image is compact. 4. The Grassmanian Gn (V ) is a smooth manifold when V = R or C. Computations Homology Groups Cohomology Homotopy Groups Special Cases/Misc. 1. 5.4 The Unitary Group Topological Properties 1. Computations Homology Groups Cohomology Homotopy Groups Group Cohomology Special Cases/Misc. 1. 8 5.5 The Special Unitary Group Topological Properties 1. Computations Homology Groups Cohomology Homotopy Groups Group Cohomology Special Cases/Misc. 1. 5.6 The Orthogonal Group Topological Properties 1. Computations Homology Groups Cohomology Homotopy Groups Group Cohomology Special Cases/Misc. 1. 6 Abstract Homotopy Theory Eilenberg Maclane Spaces, Model Categories, K-Theory... 7 Things that split References [1] Munkres Intro to Alg Top [2] Bredon Transformation Groups [3] Atiyah K-Theory [4] Hatcher 9