Embeddings from the Point of View of Immersion Theory : Part II Thomas G Goodwillie Michael Weiss

Embeddings from the Point of View of Immersion Theory : Part II Thomas G Goodwillie Michael Weiss

ISSN 1364-0380 (on line) 1465-3060 (printed) 103 Geometry & Topology G T GG T T T Volume 3 (1999) 103–118 G T G T G T G T Published: 28 May 1999 T G T G T G T G T G G G G T T Embeddings from the point of view of immersion theory : Part II Thomas G Goodwillie Michael Weiss Department of Mathematics, Brown University Box 1917, Providence RI 02912–0001, USA and Department of Mathematics, University of Aberdeen Aberdeen AB24 3UE, UK Email: [email protected] and [email protected] Abstract Let M and N be smooth manifolds. For an open V ⊂ M let emb(V, N) be the space of embeddings from V to N . By the results of Goodwillie [4], [5], [6] and Goodwillie–Klein [7], the cofunctor V 7→ emb(V, N) is analytic if dim(N) − dim(M) ≥ 3. We deduce that its Taylor series converges to it. For details about the Taylor series, see Part I (reference [15]). AMS Classification numbers Primary: 57R40 Secondary: 57R42 Keywords: Embedding, immersion, calculus of functors Proposed: Ralph Cohen Received: 10 May 1998 Seconded: Haynes Miller, Gunnar Carlsson Revised: 5 May 1999 c Geometry & Topology Publications 104 Thomas G Goodwillie and Michael Weiss 0 Introduction This is a continuation of [15]. The ideas in this second part are mostly due to Goodwillie; notes and illustrations mostly by Weiss. For introductions and notation generally speaking, see [14] and [15]. We fix smooth manifolds M m and N n , without boundary unless otherwise stated, and we assume m ≤ n. As in [15] we write O for the poset of open subsets of M . Then V 7→ emb(V, N) is a cofunctor from O to Spaces (= fibrant simplicial sets), and from [14] and [15] we have a fairly thorough understanding of its Taylor approximations, the cofunctors V 7→ Tk emb(V, N). Here we show that the Taylor approximations are good approximations. The main result is Corollary 2.5. Sections 4 and 5 contain illustrations. Following are some conventions. Let R be a set with r elements. A functor X from the poset PR of subsets of R to Spaces is an r–dimensional cube of Spaces [3, section 1]. The cube is k–Cartesian if the canonical map X (∅) −→ holim X (S) {S|∅6=S⊂R} is k–connected. A Cartesian cube is one which is k–Cartesian for all k. Since PR as a category is isomorphic to its own opposite, we can use similar termi- nology for cofunctors from PR to Spaces. Such a cofunctor will also be called a cube. It is k–Cartesian if the canonical map X (R) −→ holim X (S) {S|S⊂R,S6=R} is k–connected, and Cartesian if it is k–Cartesian for all k. n+2 n Let R x := [0, ∞)×[0, ∞)×R . An (n+2)–dimensional smooth manifold triad is a paracompact Hausdorff space Q together with a maximal atlas consisting n+2 of open subsets Vi ⊂ Q and open embeddings φi: Vi → R x satisfying the following conditions: • The union of the Vi is Q. −1 • The changes of charts φi(φj ) are smooth where defined, and take points n+2 in R x with vanishing first coordinate (vanishing second coordinate) to points with vanishing first coordinate (vanishing second coordinate). Let ∂0Q be the set of those x ∈ Q which under some chart φi map to a point in n+2 R x with vanishing first coordinate. Also, let ∂1Q be the set of those x ∈ Q n+2 which under some chart φi map to a point in R x with vanishing second coordinate. Then Q is an (n + 2)–dimensional smooth manifold with corners; its boundary is ∂0Q ∪ ∂1Q and its corner set is ∂0Q ∩ ∂1Q. Geometry & Topology, Volume 3 (1999) Embeddings from immersion theory : II 105 The handle index of the manifold triad Q is the smallest integer a such that Q can be built from a collar on ∂0Q by attaching handles of index ≤ a. (It may happen that Q is a collar on ∂0Q. When that is the case we say that the handle index is −∞.) Example Suppose that P is smooth, with boundary, and let f: P → R be smooth. If 0 is a regular value both for f and for f|∂P , then the inverse image of [0, ∞) under f is a manifold triad Q, with ∂0Q = ∂P ∩ Q. (If ∂P = ∅, then of course ∂0Q = ∅.) Any Q ⊂ P which can be obtained in this way will be called a codimension zero subobject of P . 1 Excision Estimates Let Y be a smooth n–manifold with boundary. Let Q0,Q1,...,Qr be smooth compact n–manifold triads with handle index qi where n − qi ≥ 3. Suppose that smooth embeddings ei: ∂0Qi → ∂Y have been selected for 0 ≤ i ≤ r, and that their images are pairwise disjoint. For S ⊂ [r] = {0,...,r} let QS be the disjoint union of the Qi with i ∈ S . By emb(QS,Y ) we mean the Space of smooth embeddings f: QS → Y which satisfy f | ∂0Qi = ei and −1 f (∂Y )= ∂0QS . 1.1 Theorem ([4], [5], [6], [7]) The (r + 1)–cube taking a subset S of [r] to r emb(QS,Y ) is (3 − n +Σi=0(n−qi −2))–Cartesian, provided r ≥ 1. Comment Assuming that emb(Q[r],Y ) is nonempty, fix a base point f in emb(Q[r],Y ). Let the image of f in emb(QS,Y ) serve as base point for emb(QS,Y ). Goodwillie shows in [6] that the cube S 7→ Ω emb(QS,Y ) is r (2 − n +Σi=0(n−qi−2))–Cartesian, provided r ≥ 1. The delooped statement, Theorem 1.1 just above, has been proved by Goodwillie and Klein and will appear in [7]. We will need a slight generalization of 1.1 where the Qi are allowed to have dimension m ≤ n. For this and other purposes we need a lemma. 1.2 Lemma Let u: X →Y be a map of (r + 1)–cubes. That is, X and Y are functors from the poset of subsets of [r] to Spaces, and u is a natural transformation. Suppose that Y is k–Cartesian and, for every y ∈ Y(∅), the (r + 1)–cube defined by S 7→ hofiber[X (S) −→Yu (S)] is k–Cartesian. (The homotopy fiber is taken over the image of y in Y(S).) Then X is k–Cartesian. Geometry & Topology, Volume 3 (1999) 106 Thomas G Goodwillie and Michael Weiss Proof of 1.2 Combine [3, 1.18] with [3, 1.6]. Now let Y and Qi be as in 1.1, assuming however dim(Qi) = m ≤ n. As before, embeddings ei: ∂0Qi → ∂Y are specified and n − qi ≥ 3, where qi is the handle index of Qi . We want to show that the conclusion of 1.1 still holds. Without loss of generality, Y is a smooth submanifold of some Rt . Then we can define maps emb(QS,Y ) −→ map(QS, Gn−m) ; f 7→ νf where Gn−m is the direct limit of the Grassmannians of (n − m)–dimensional u linear subspaces of R , for u ≥ 0, and where νf takes x ∈ QS to the intersection of the tangent space of Y at f(x) with the normal space of f(QS) at f(x). In other words, νf classifies the normal bundle of the embedding f . We have therefore a map of (r + 1)–cubes: (•) {S 7→ emb(QS,Y ) } −→ {S 7→ map(QS, Gn−m) } . Since the codomain cube in (•) is Cartesian, the contravariant analog of 1.2 tells us that the domain cube is k–Cartesian provided that, for every ξ in map(Q[r], Gn−m), the cube of homotopy fibers (••) S 7→ hofiberξ [ emb(QS,Y )} −→ map(QS, Gn−m) ] is k–Cartesian. Now it is easy to construct a homotopy equivalence from the ′ ′ homotopy fiber of emb(QS,Y ) → map(QS, Gn−m) to emb(QS,Y ) where QS is the total space of the disk bundle on QS determined by ξ. This construction ′ is natural in S , so (••) is k–Cartesian if the cube S 7→ emb(QS,Y ) is k– r Cartesian, which by 1.1 is the case if k = (3 − n +Σi=0(n−qi−2)). Therefore: 1.3 Observation Theorem 1.1 generalizes to the situation where the Qi have dimension m ≤ n. The corollary below is a technical statement about the cofunctor on O given by V 7→ emb(V, N). Suppose that P is a smooth compact codimension zero subobject of M , and that Q0,...,Qr are pairwise disjoint compact codimension zero subobjects of M r int(P ). For S ⊂ {0,...,r} = [r] let VS be the interior of P ∪ QS where QS is the union of the Qi for i ∈ S . We write V instead of V∅ . Suppose that Qi has handle index qi ≤ n − 3. 1.4 Corollary Assume r ≥ 1. The (r + 1)–cube taking a subset S of [r] to r emb(VS, N) is (3 − n +Σi=0(n−qi−2))–Cartesian. Geometry & Topology, Volume 3 (1999) Embeddings from immersion theory : II 107 Proof Let V¯S be the closure of VS in M . Let emb(V¯S, N) be the inverse limit of the Spaces (simplicial sets) emb(U, N) where U ranges over the neigh- borhoods of VS in M . The restriction from emb(V¯S, N) to emb(VS, N) is a homotopy equivalence. Using this fact and 1.2, we see that it is enough to show that for every embedding f: V¯ → N , the (r + 1)–cube res S 7→ hofiber [ emb(V¯S, N) −−→ emb(V¯ , N) ] r is (3 − n +Σi=0(n−qi −2))–Cartesian.

View Full Text

Details

  • File Type
    pdf
  • Upload Time
    -
  • Content Languages
    English
  • Upload User
    Anonymous/Not logged-in
  • File Pages
    16 Page
  • File Size
    -

Download

Channel Download Status
Express Download Enable

Copyright

We respect the copyrights and intellectual property rights of all users. All uploaded documents are either original works of the uploader or authorized works of the rightful owners.

  • Not to be reproduced or distributed without explicit permission.
  • Not used for commercial purposes outside of approved use cases.
  • Not used to infringe on the rights of the original creators.
  • If you believe any content infringes your copyright, please contact us immediately.

Support

For help with questions, suggestions, or problems, please contact us