UC Santa Barbara UC Santa Barbara Previously Published Works Title Describing the universal cover of a compact limit Permalink https://escholarship.org/uc/item/1t60830g Journal DIFFERENTIAL GEOMETRY AND ITS APPLICATIONS, 24(5) ISSN 0926-2245 Authors Ennis, J Wei, Guofang F Publication Date 2006-09-01 Peer reviewed escholarship.org Powered by the California Digital Library University of California
Describing the Universal Cover of a Compact Limit John Ennis 1504 Psychology Building Psychology Department UC - Santa Barbara Santa Barbara, CA 93106 ennis@psych.ucsb.edu Guofang Wei Department of Mathematics UC - Santa Barbara Santa Barbara, CA 93106 wei@math.ucsb.edu Abstract If X is the Gromov-Hausdorff limit of a sequence of Riemannian manifolds Mi n with a uniform lower bound on Ricci curvature, Sormani and Wei have shown that the universal cover X of X exists [13, 15]. For the case where X is compact, we provide a description of X in terms of the universal covers Mi of the manifolds. More specifically we show that if X is the pointed Gromov- Hausdorff limit of the universal covers M i then there is a subgroup H of Iso( X) such that X = X/H. We call H the small action limit group and prove a similar result for compact length spaces with uniformly bounded dimension. 1 Introduction In the early 1980 s Gromov proved that any finitely generated group has polynomial growth if and only if it is almost nilpotent [7]. In his proof, Gromov introduced the Gromov-Hausdorff distance between metric spaces [7, 8, 9]. This distance has proven to be especially useful in the study of n-dimensional manifolds with Ricci curvature uniformly bounded below since any sequence of such manifolds has a convergent subsequence [10]. Hence we can follow an approach familiar to analysts, and consider the closure of the class of all such manifolds. The limit spaces of this class have path metrics, and one can study these limit spaces from a geometric or topological perspective. 2000 Mathematics Subject Classification. Primary 53C20. Research partially supported by NSF Grant # DMS-0204187. Keywords: universal cover, equivariant Gromov-Hausdorff convergence, fundamental group Corresponding author 1
Much is known about the limit spaces of n-dimensional Riemannian manifolds with a uniform lower bound on sectional curvature. These limit spaces are Alexandrov spaces with the same curvature bound [1], and all points have tangent cones that are metric cones. Since Perelman has shown that Alexandrov spaces are locally homeomorphic to their tangent cones [12], these limit spaces are locally contractible. In this case, an argument of Tuschmann s shows that there is eventually a surjective map from the fundamental groups of the manifolds in the sequence onto the fundamental group of the limit space [17]. We seek similar results when Ricci curvature is uniformly bounded from below. Cheeger and Colding have made considerable progress studying the geometric and regularity properties of the limit spaces of this class [2, 3, 4], but the local topology of the limit spaces could be very complicated. For instance, Menguy has shown that the limit spaces can have infinite topology on arbitrarily small balls [11], even when the sequence has nonnegative Ricci curvature. In addition, it is not known whether the limit space is locally or even semilocally simply connected. Sormani and Wei have shown that the limit space X of a sequence of manifolds M n i with uniform Ricci curvature lower bound has a universal cover X (see Definition 2.4). This cover is not assumed to be simply connected, see e.g. the double suspension of Hawaiian earring [16]. In this note we use the notation M i GH X to mean that the manifolds M i converge to the space X in the Gromov-Hausdorff sense. For a compact limit space we describe the universal cover X in terms of the universal covers of the manifolds. Theorem 1.1. Let X i and X be a sequence of compact length spaces with universal covers X i and X and dimension n. Assume and that X i GH X ( X i, p i ) GH ( X, x). Then there is a closed subgroup H Iso( X), called the small action limit group, such that X/H is the universal cover of X. Corollory 1.2. Suppose M n i have Ric Mi (n 1)H and diam Mi D. Assume M n i GH X and that ( Mi, p i ) GH ( X, x). Then there is a closed subgroup H Iso( X) such that X = X/H. As above, we call H the small action limit group. 2
We start by reviewing some results, then give an example to show that the universal cover of the limit space may not be the limit of the universal covers of the manifolds. This example leads us to consider equivariant Hausdorff convergence, due to Fukaya [5, 6], which extends Gromov-Hausdorff convergence to include group actions. We then combine results of Fukaya and Yamaguchi with results of Sormani and Wei [13, 14] to prove Theorem 1.1. The authors thank Daryl Cooper for numerous helpful conversations. 2 Background In this paper, a manifold is a complete Riemannian manifold without boundary. An essential result in the study of Gromov-Hausdorff limits of manifolds with uniform lower bound on curvature is the Gromov Precompactness Theorem. Theorem 2.1 (Gromov Precompactness Theorem). Let R be the set of all closed, connected, Riemannian n-manifolds with diam D and Ric (n 1)H, and let M be the set of all isometry classes of compact metric spaces. Let d GH denote the Gromov-Hausdorff distance, which is a metric on M. Then R (M, d GH ) is precompact. The Gromov-Hausdorff limit of length spaces is a length space. An effective way to study the coverings of these spaces is using δ-covers, which were introduced by Sormani and Wei in [13]. Suppose X is a complete length space. For x X and δ > 0, let π 1 (X, x, δ) be the subgroup of π 1 (X, x) generated by elements of the form = [α β α 1 ], where α is a path from x to some y X and β is a loop contained in some open δ-ball in X. x Open ball of radius δ > 0 α β Figure 1: A typical generator for π 1 (X, x, δ) Definition 2.2 (δ-cover). The δ-cover of a metric space X is the covering space π δ : Xδ X with (π δ ) (π 1 ( X δ, x)) = π 1 (X, x, δ). 3
Note that (π δ ) : π 1 ( X δ, x) π 1 (X, x) is the map [γ] [π δ (γ)]. Intuitively, a δ-cover is the result of unwrapping all but the loops generated by small loops in X. Remarks 2.3. 1. Xδ covers X δ for δ δ. 2. δ-covers exist for connected, locally path connected spaces. See [16] for more details. Definition 2.4. [Universal Cover, [16, pp. 62,83]] We say X is a universal cover of X if X is a cover of X such that for any other cover X of X, there is a commutative triangle formed by a continuous map f : X X and the two covering projections. Note that the universal covers of a sequence compact length spaces may not have any converging subsequence even if the sequence itself converges. For example, a sequence of finite sets of circles joined at a common point which converges to the Hawaii ring (see [14, Example 7.5] for more examples). On the other hand Sormani- Wei [14, Prop. 7.3] showed that this doesn t happen for δ-covers. Proposition 2.5 (Sormani-Wei). If a sequence of compact length spaces X i converges to a compact length space X in the Gromov-Hausdorff topology, then for any δ > 0 there is a subsequence of X i such that their δ-covers also converges in the pointed Gromov-Hausdorff topology. Sormani and Wei have studied the properties of δ-cover. In particular, if a compact length space has a universal cover, then the universal cover is a δ-cover for some δ > 0 [13, Prop. 3.2]. Using δ-covers they show that when manifolds with a uniform Ricci curvature lower bound and a uniform diameter upper bound converge in the Gromov-Hausdorff sense to a limit space X, the universal cover of X exists. Theorem 2.6 (Sormani-Wei). Suppose M n i have Ric Mi (n 1)H, diam Mi D, and M i GH X. Then the universal cover X exists, and there is δ = δ(x) > 0 such that ( X, x) = GH lim i ( M δ i, p i). Note that X may not be the limit of the universal covers of the manifolds in the sequence. The following example shows this need not be the case. 4
Example 2.7. Consider S 3 /Z p GH S 2. Then S 3 GH S 3 S 3 /Z p GH S 2 Here the loops in the lens spaces S 3 /Z p shrink to points as p goes to infinity, so S 3 /Z p collapses to S 2. In this case the fundamental group Z p of S 3 /Z p fills up S 1 as p grows and we have S 3 /Z p GH S 3 /S 1 = S 2. Example 2.7 indicates the need for considering group actions as well as convergence of spaces. For this reason we use equivariant Hausdorff convergence, introduced by Fukaya [5, 6]. Consider pointed group metric spaces (X, G, x), where X is a complete metric space, G is a group of isometries of X and x X. Set for each R > 0. G(R) = {g G d(g(x), x) < R} Definition 2.8 (Equivariant ǫ-hausdorff Approximation). Suppose (X 1, G 1, x 1 ) and (X 2, G 2, x 2 ) are pointed group metric spaces. Let d be a metric on B 1/ǫ (x 1, X 1 ) B 1/ǫ (x 2, X 2 ), and let φ : G 1 (1/ǫ) G 2 (1/ǫ) and ψ : G 2 (1/ǫ) G 1 (1/ǫ) be maps. The triple (d, φ, ψ) is said to be an equivariant ǫ-hausdorff approximation if 1. d extends the original metrics on B 1/ǫ (x i, X i ) for i = 1, 2. 2. For each y 1 B 1/ǫ (x 1, X 1 ) there is y 2 B 1/ǫ (x 2, X 2 ) such that d(y 1, y 2 ) < ǫ, and for each y 2 B 1/ǫ(x 2, X 2 ) there is y 1 B 1/ǫ(x 1, X 1 ) with 3. d(x 1, x 2 ) < ǫ. d(y 1, y 2 ) < ǫ. 4. For each y i B 1/3ǫ (x i, X i ) with d(y 1, y 2 ) ǫ and g i G i (1/3ǫ) we have d(y 1, g 1 y 1 ) d(y 2, φ(g 1 )(y 2 )) < ǫ, d(y 2, g 2 y 2 ) d(y 1, ψ(g 2 )(y 1 )) < ǫ. 5
Definition 2.9 (Equivariant Hausdorff Convergence). The sequence (X i, G i, x i ) of pointed group metric spaces converges to the pointed group metric space (X, G, x) in the equivariant Hausdorff sense if there are equivariant ǫ i -Hausdorff approximations between (X i, G i, x i ) and (X, G, x), where ǫ i 0 as i. We write (X i, G i, x i ) eh (X, G, x). Note that equivariant Hausdorff convergence implies Gromov-Hausdorff convergence of (X i, x i ) to (X, x). Fukaya and Yamaguchi have given a constructive proof of the following important theorem [6]. Theorem 2.10 (Fukaya-Yamaguchi). If (X i, p i ) GH (Y, q) in the Gromov-Hausdorff sense, and G i Iso(X i ) are closed subgroups then there is G Iso(Y ) such that, after passing to a subsequence, (X i, G i, p i ) eh (Y, G, q). In addition, Fukaya has shown a natural relationship between equivariant convergence and Gromov-Hausdorff convergence [5]. Theorem 2.11 (Fukaya). If (X i, G i, p i ) eh (Y, G, q) then Remark 2.12. If (X i /G i, [p i ]) GH (Y/G, [q]). G i = π 1 (M i ), (M i, p i ) GH (X, x) and ( M i, π 1 (M i ), p i ) eh ( X, G, x), then Theorem 2.11 implies X = X/G. 6
3 Description of Universal Cover Suppose M n i is a sequence of manifolds with Ric Mi (n 1)H and diam Mi D. By Theorem 2.1, there is a length space X such that, after passing to a subsequence we have M i GH X. If we pick a sequence of points p i M i, where M i is the universal cover of M i, a further subsequence of ( M i, p i ) converges to a length space ( X, x) in the pointed Gromov-Hausdorff sense. Since π 1 (M i ) is a discrete subgroup of Iso( M i ), π 1 (M i ) is closed. Thus Theorem 2.10 implies that there is G Iso( X) such that, after passing to a subsequence, ( M i, π 1 (M i ), p i ) eh ( X, G, x). Set for each i. Then set G i = π 1 (M i, p i ) G ǫ i =< g G i d(g q, q) ǫ for some q M i > for each ǫ > 0. Lemma 3.1. G ǫ i is closed, normal subgroup of G i. Proof. Since G i is a discrete group, G ǫ i is closed. For normality, suppose g is a generator of G ǫ i which has d(g q, q) ǫ for some q M i. For each h G i, d(hgh 1 (h q), h q) = d(hg q, h q) so hgh 1 G ǫ i. It follows that G ǫ i is normal in G i. = d(g q, q) ǫ, Thus we may consider the quotient G i /G ǫ i, and its isometric action on M i /G ǫ i by [g][ q] = [g q]. Lemma 3.2. G i /G ǫ i is a discrete group that acts freely on M i /G ǫ i. Proof. If [g] G i /G ǫ i is not trivial, then d([g][ q], [ q]) > ǫ for all [ q] particular, G i /G ǫ i acts freely on M i /G ǫ i. M i /G ǫ i. In Remark 3.3. Lemma 3.2 implies that M i /G ǫ i covers ( M i /G ǫ i )/(G i/g ǫ i ) = M i. 7
Next we prove two lemmas relating the covering spaces M i /G ǫ i to the δ-covers M δ i. Lemma 3.4. For 0 < ǫ/2 < δ, Mi /G ǫ i covers M δ i. Proof. We show that G ǫ i π 1 (M i, δ, p i ). Suppose g is a generator for G ǫ i. There is q i M i with d( q i, g q i ) ǫ. Connect q i to g q i by a distance minimizing path β, and connect p i to q i by a path α. Note that the length of β, l( β), is at most ǫ. Set α = π i ( α) and β = π i ( β). By uniqueness of path lifting, the lift of α β α 1 beginning at p i is α β (g α) 1. M i α q i β g α g q i p i g p i M i π i α δ q i β p i Figure 2: α β α 1 lifts to α β (g α) 1 Thus [α β α 1 ] p i = g p i, so g = [α β α 1 ]. Moreover, l( β) ǫ implies that β is contained in B(β(0), ǫ/2), which lies in the open δ-ball centered at β(0). Thus g π 1 (M i, δ, p i ), whence G ǫ i π 1 (M i, δ, p 1 ). Lemma 3.5. For each 0 < δ < ǫ/5, Mδ i covers M i /G ǫ i. 8
Proof. Here we show that π 1 (M i, δ, p i ) G ǫ i. Suppose g is a generator for π 1 (M i, δ, p i ). Then g = [α β α 1 ], where α is a path in M i from p i to some q i and β δ is a loop in B(q i, 2δ). Let α be the lift of α to M i beginning at p i, set q i = α(1) and let β be the lift of β to M i beginning at q i. M i α q i β g q i p i π i M i β q i α p i Figure 3: Lemma 3.5 Observe that if l(β) < ǫ, d( q i, g q i ) l( β) = l(β) < ǫ. In this case, g = [α β α 1 ] G ǫ. In general, if β B(q i, 2δ) is a loop based at q i, Figure 4 shows we can find loops β 1,..., β k based at q i with l(β j ) < 5δ and [β] = [β 1 ][β 2 ] [β k ]. For each j, [α β j α 1 ] G 5δ G ǫ. Then g = [α β α 1 ] = [α β 1 α 1 ] [α β k α 1 ] G ǫ. Now we show a key relationship between δ-covers and the group actions coming from the sequence of manifolds. 9
β 1 β 2 β β k q i Figure 4: Dividing β Lemma 3.6. Suppose and that for δ > 0, ( M i, G i, p i ) eh ( X, G, x) (X δ, x δ ) = GH lim i ( M δ i, p δ i). Then there is ǫ > 0 such that X/H ǫ is a covering map of X that covers X δ. Proof. By Lemma 3.4 we may pick ǫ > 0 so that φ i : Mi /G ǫ i M δ i are covering maps. In particular, each φ i is distance nonincreasing. By Lemma 2.10 we may pass to a subsequence and obtain a closed subgroup H ǫ of G Iso( X) such that ( M i, G ǫ i, p i) eh ( X, H ǫ, x). Note that by Theorem 2.11, M i /G ǫ i GH X/H ǫ. Thus the Arzela-Ascoli lemma implies that some subsequence of {φ i } converges to a distance nonincreasing map φ : X/H ǫ X δ. Set δ 1 = ǫ/5. By Lemma 3.5, 1 Mδ i covers M i /G ǫ i. As above, if φ i : Mδ 1 i M i is a covering map, we may pass to a subsequence and obtain a distance nonincreasing map φ : X δ 1 X. We have 10
M δ 1 i GH X δ 1 M i /G ǫ i GH X/H ǫ φ i φ i M i δ GH φ X δ ψ φ i M i GH X where we have chosen basepoints so the downward pointing arrows commute. Now each φ i is an isometry on balls of radius less than δ 1, so φ = lim i φ i is an isometry on balls of radius less than δ 1. In particular, φ is a covering map. Thus X/H ǫ covers X δ. To see that X/H ǫ covers X, observe that a similar argument as above shows that φ is an isometry on balls of radius less than δ 1. Since this map factors through ψ : X/H ǫ X, ψ is also an isometry on balls of radius less than δ 1. Thus ψ is a covering and the proof is complete. Now we are ready to prove Theorem 1.1. Proof of Theorem 1.1. By Prop. 3.2 in [13], the universal cover of X is some δ- cover. By Proposition 2.5 and [13, Theorem 3.6], for any δ > 0, a subsequence of the δ-covers of X i converges to a cover of X which is an almost δ-cover of X. Therefore there is δ > 0 such that the universal cover of X is X = X δ = GH lim i ( X i δ, pδ i ). By Lemma 3.6 we may choose H = H ǫ so that X/H covers both X and X δ. Thus X/H = X. Corollary 1.2 follows using Theorem 2.6. References [1] Dmitri Burago, Yuri Burago, and Sergei Ivanov. A course in metric geometry, volume 33 of Graduate Studies in Mathematics. American Mathematical Society, Providence, RI, 2001. [2] Jeff Cheeger and Tobias H. Colding. On the structure of spaces with Ricci curvature bounded below. I. J. Differential Geom., 46(3):406 480, 1997. [3] Jeff Cheeger and Tobias H. Colding. On the structure of spaces with Ricci curvature bounded below. II. J. Differential Geom., 54(1):13 35, 2000. 11
[4] Jeff Cheeger and Tobias H. Colding. On the structure of spaces with Ricci curvature bounded below. III. J. Differential Geom., 54(1):37 74, 2000. [5] Kenji Fukaya. Theory of convergence for Riemannian orbifolds. Japan. J. Math. (N.S.), 12(1):121 160, 1986. [6] Kenji Fukaya and Takao Yamaguchi. The fundamental groups of almost nonnegatively curved manifolds. Ann. of Math. (2), 136(2):253 333, 1992. [7] Mikhael Gromov. Groups of polynomial growth and expanding maps. Inst. Hautes tudes Sci. Publ. Math., (53):53 73, 1981. [8] Mikhael Gromov. Structures métriques pour les variétés riemanniennes, volume 1 of Textes Mathématiques [Mathematical Texts]. CEDIC, Paris, 1981. Edited by J. Lafontaine and P. Pansu. [9] Mikhael Gromov. Infinite groups as geometric objects. In Proceedings of the International Congress of Mathematicians, Vol. 1, 2 (Warsaw, 1983), pages 385 392, Warsaw, 1984. PWN. [10] Misha Gromov. Metric structures for Riemannian and non-riemannian spaces, volume 152 of Progress in Mathematics. Birkhäuser Boston Inc., Boston, MA, 1999. Based on the 1981 French original [MR 85e:53051], With appendices by M. Katz, P. Pansu and S. Semmes, Translated from the French by Sean Michael Bates. [11] X. Menguy. Examples with bounded diameter growth and infinite topological type. Duke Math. J., 102(3):403 412, 2000. [12] G. Ya. Perel man. Elements of Morse theory on Aleksandrov spaces. Algebra i Analiz, 5(1):232 241, 1993. [13] Christina Sormani and Guofang Wei. Hausdorff convergence and universal covers. Trans. Amer. Math. Soc., 353(9):3585 3602 (electronic), 2001. [14] Christina Sormani and Guofang Wei. The covering spectrum of a compact length space. J. Differential Geom., 67(1):35 77, 2004. [15] Christina Sormani and Guofang Wei. Universal covers for Hausdorff limits of noncompact spaces. Trans. Amer. Math. Soc., 356(3):1233 1270 (electronic), 2004. [16] Edwin H. Spanier. Algebraic topology. McGraw-Hill Book Co., New York, 1966. [17] Wilderich Tuschmann. Hausdorff convergence and the fundamental group. Math. Z., 218(2):207 211, 1995. 12