arXiv reader

Homotopy invariance of the space of metrics with positive scalar curvature on manifolds with singularities

Boris Botvinnik and Mark Walsh

arXiv
2005.03073v3 · PDF · 6 April 2021
Categories
math.AT, math.DG
Journal
SIGMA 17 (2021), 034, 27 pages
DOI
10.3842/SIGMA.2021.034
Licence
CC BY SA 4.0 The site shows the text.

Abstract

In this paper we study manifolds, X_{\Sigma}, with fibred singularities, more specifically, a relevant space {\mathcal R}^{\rm psc}(X_{\Sigma}) of Riemannian metrics with positive scalar curvature. Our main goal is to prove that the space {\mathcal R}^{\rm psc}(X_{\Sigma}) is homotopy invariant under certain surgeries on X_{\Sigma}.

Items 44

Theorem

Theorem 1.1 (Gromov–Lawson , Stolz )

93CT

Let M be a smooth closed simply connected manifold of dimension n\geq 5.

    =0pt

  1. (i) If M is spin, then M admits a psc-metric if and only if the index \alpha(M)\in KO^{-n} of the Dirac operator on M vanishes.

  2. (ii) If M is not spin, then M always admits a psc-metric.

Question

Question 1

KWYP

Does there exist a psc-metric g_Y on Y, such that the product metric g_Y+ g_L on \partial X = Y\times L can be extended (being a product near \partial X) to a psc-metric g_X on X?

Remark

Remark 1.2

USOZ

We emphasize that, according to our construction, the metrics g_{\mathrm{att}(L)} and {g_{C(L)}} are both invariant under the action of the isometry group \mathrm{Iso}(g_L) of the metric g_L. In our work, the role of L will be played by either the circle S^1 (with U(1) as the isometry group), some other appropriate Lie group or a homogeneous space, such as S^n or \mathbb{HP}^n, equipped with the natural metric g_L of non-negative constant scalar curvature.

Proposition

Proposition 1.3

0WDO

The construction of g_{X_{\Sigma}} above yields a smooth (singular) metric of non-negative scalar curvature on X_{\Sigma}.

Theorem

Theorem 1.4 (Botvinnik )

QP2C

Let X be a simply connected spin manifold with nonempty \eta-singularity of dimension n\geq 7. Then X admits a metric of positive scalar curvature if and only if \alpha^{\eta}([X])=0 in the group KO_n^{\eta} \cong KU_n.

Theorem

Theorem 1.5 (Botvinnik–Rosenberg )

BJVW

Let X_{\Sigma}= X\cup N(\beta X) be a simply connected spin manifold with fibred \eta-singularity (i.e., X and \beta X are simply-connected and spin) of dimension n\geq 7. Then {\mathcal R}^{\mathrm{psc}}(X_{\Sigma})\neq\varnothing if and only if \alpha^{\eta\mbox{-}\mathrm{fb}}([X_{\Sigma}])=0 in the group KO^{\eta\mbox{-}\mathrm{fb}}_n.

Theorem

Theorem 1.6 (Botvinnik–Piazza–Rosenberg )

Y0Q9

Let (X, \partial X, f) define a closed (L,G)-singular spin manifold X_\Sigma. Assume that X, \beta X, and G are all simply connected, that n-\ell\ge 6, and suppose that L is a spin boundary, say L=\partial\bar L, with the standard metric g_L on L extending to a psc-metric on \bar L, and with the G-action on L extending to a G-action on \bar L. Assume that the two obstructions \alpha(\beta X)\in KO_{n-\ell-1} and \alpha_{{\rm cyl}}(X)\in KO_n both vanish. Then X_\Sigma admits a well-adapted psc-metric.

Theorem

Theorem 1.7 (Botvinnik–Piazza–Rosenberg )

CHC4

Let (X, \beta X, f) define a closed (L,G)-singular spin manifold X_\Sigma, with L=\mathbb{HP}^{2k} and G={\rm Sp}(2k+1), n\ge 1. Assume that \partial X = \beta X \times L, i.e., the L-bundle over \beta X is trivial, or in other words that the singularities are of Baas–Sullivan type. Then if X and \beta X are both simply connected and n-8k\ge 6, X_\Sigma has an adapted psc-metric if and only if the \alpha-invariants \alpha(\beta X) \in KO_{n-8k-1} and \alpha_{{\rm cyl}}(X) \in KO_n both vanish.

Theorem

Theorem 1.8

WSC6

Let X_{\Sigma}=X\cup_{\partial X} -N(\beta X) be a pseudo-manifold with (L,G)-singularities, with \operatorname{{\text{\rm dim}}} X = n, \operatorname{{\text{\rm dim}}} L=\ell.

    =0pt

  1. (i) Let i\colon S^p\subset X be a sphere with trivial normal bundle, and X_{\Sigma}' be the result of surgery on X_{\Sigma} along S^p. Then if 2\leq p\leq n-3, the spaces {\mathcal R}^{\mathrm{psc}}(X_{\Sigma}) and {\mathcal R}^{\mathrm{psc}}(X_{\Sigma}') are weakly homotopy equivalent.

  2. (ii) Let i\colon S^p\subset \beta X be a sphere with trivial normal bundle, and with f\circ \iota\colon S^p \to BG homotopic to zero. Let \bar B be a trace of the surgery along S^p\subset \beta X with \partial \bar B = \beta X\sqcup -\beta X_1 and a structure map \bar f\colon \bar B\to BG. Then if 2\leq p \leq n-\ell-3, the spaces {\mathcal R}^{\mathrm{psc}}(X_{\Sigma}) and {\mathcal R}^{\mathrm{psc}}(X_{\Sigma,1}) are homotopy equivalent, where X_{\Sigma,1} is given by (?).

Remark

Remark 1.9

F4XM

As it deals with surgery on the interior, part (i) of Theorem 1.8 above follows directly from Chernysh’s theorem [Ch]. Our contribution in this paper is in proving part (ii).

Corollary

Corollary 1.10

QY44

Let X_{\Sigma} be a spin (\left<k\right>\!\mbox{-}\mathrm{fb})-manifold. Assume \operatorname{{\text{\rm dim}}} X\geq 7 and that X and \beta X are simply-connected. Then the homotopy type of the space {\mathcal R}^{\mathrm{psc}}(X_{\Sigma}) is a bordism invariant and depends only on the bordism class [X_{\Sigma}]\in \Omega^{\mathrm{spin},\left<k\right>\mbox{-}\mathrm{fb}}_n.

Corollary

Corollary 1.11

JLZA

Let X_{\Sigma} be a spin (\eta\mbox{-}\mathrm{fb})-manifold. Assume \operatorname{{\text{\rm dim}}} X\geq 9 and that X and \beta X are simply-connected. Then the homotopy type of the space {\mathcal R}^{\mathrm{psc}}(X_{\Sigma}) is a bordism invariant and depends only on the bordism class [X_{\Sigma}]\in \Omega^{\mathrm{spin},\eta\mbox{-}\mathrm{fb}}_n.

Remark

Remark 1.12

AO22

To simplify the presentation, we consider only the spin case here. However, there are also non-spin cases where similar results hold. We leave this intentionally outside of this paper.

Corollary

Corollary 1.13

DPWG

In both of the cases described in (3) and (4), the homotopy type of the space {\mathcal R}^{\mathrm{psc}}(X_{\Sigma}) is a bordism invariant and depends on the bordism class [X_{\Sigma}]\in \Omega^{\mathrm{spin},(L,G)\mbox{-}\mathrm{fb}}_n, provided n-\ell \geq 6 (where \ell =2k in the case (4)).

Remark

Remark 1.14

6PJE

To see how Theorem 1.8 implies the corollaries, it is enough to notice that in all those cases, the dimensional assumptions imply that a cobordism W_{\Sigma}\colon X_{\Sigma}\rightsquigarrow X_{\Sigma,1} could be modified so that the embedding X_{\Sigma,1} \subset W_{\Sigma} is two-connected in an appropriate sense, i.e., the induced embeddings \beta X_1\subset \beta W and (X_1,\partial X_1) \hookrightarrow (W,\partial W) are both 2-connected. Then the argument goes exactly as in the case of closed manifolds.

Lemma

Lemma 2.1 (surgery lemma; see )

PYW6

Let g_0\in {\mathcal R}^+(Y_0) be any metric. Assume q\geq 2. Then there exist metrics g_1\in {\mathcal R}^+(Y_1) and \bar g\in {\mathcal R}^+(Z,\partial Z)_{g_0,g_1} such that (Z,\bar g)\colon (Y_0,g_0)\rightsquigarrow (Y_1,g_1) is a psc-bordism.

Theorem

Theorem 2.2

8EY2

Let Z\colon Y_0\rightsquigarrow Y_1 be an elementary bordism as above with p,q\geq 2. Then for any metric g_0\in {\mathcal R}^+(Y_0) there exist metrics g_1\in {\mathcal R}^+(Y_1) and \bar g\in {\mathcal R}^+(Z,\partial Z)_{g_0,g_1} such that the map

\begin{gathered} \mu_{Z,\bar{g}}\colon\ {\mathcal R}^{\mathrm{psc}}(X, \partial X)_{g_0} \xrightarrow{\simeq} {\mathcal R}^{\mathrm{psc}}(X\cup Z, Y_1)_{g_1} \end{gathered}

defined by (?), is a weak homotopy equivalence.

Proposition

Proposition 2.4

GSRI

For any horizontal distributions, \mathcal{H}, \mathcal{H}'\in \mathcal{HD}(p), the maps \mathcal{S}_{\mathcal{H}}^{\mathrm{psc}} and \mathcal{S}_{\mathcal{H}'}^{\mathrm{psc}} are homotopy equivalent.

Lemma

Lemma 2.5

7WFY

The maps \mathcal{B} and \mathcal{S}_{\mathcal{H}}^{\mathrm{psc}} are homotopy inverse, thus determining a homotopy equivalence:

\begin{gathered} {\mathcal R}^{\mathrm{psc}}(B)\simeq{\mathcal R}_{\mathcal{S}}^{\mathrm{psc}} (M). \end{gathered}
T

T

MMC2

he composition \mathcal{B}\circ\mathcal{S}_{\mathcal{H}}^{\mathrm{psc}} is precisely the identity map on {\mathcal R}^{\mathrm{psc}}(B). Consider now an arbitrary element \bar{g}\in{\mathcal R}_{\mathcal{S}}^{\mathrm{psc}} (M). This element is uniquely represented as a triple \bar{g}=(\bar{g}_{B}, \mathcal{H}_{\bar{g}}, \tau_{\bar{g}}), consisting of a base metric \bar{g}_{B}=\mathcal{B}(\bar{g}), a horizontal distribution and a scaling constant. Now,

\begin{gathered} \mathcal{S}_{\mathcal{H}}^{\mathrm{psc}}\circ \mathcal{B}(\bar{g})=\mathcal{S}(\bar{g}_{B}, \mathcal{H}, \bar{\tau}(\bar{g}_{B}, \mathcal{H})). \end{gathered}

Thus, the desired homotopy is given by

\begin{gathered} H_{t}(\bar{g}):=\mathcal{S}(\bar{g}_{B}, (1-t)\mathcal{H}+t\mathcal{H}_{\bar{g}}, \bar{\tau}(\bar{g}_{B}, (1-t)\mathcal{H}+t\mathcal{H}_{\bar{g}})), \end{gathered}

where t\in [0,1]. That the resulting submersion metric H_{t}(\bar{g}) has positive scalar curvature for all t is guaranteed by the construction of the function \bar{\tau}.

Definition

Definition 2.6

8XB2

A horizontal distribution, \bar{\mathcal{H}} on the bundle \bar{p}\colon \bar{E}\to \bar{B} is said to be admissible provided it takes the form of a product \mathcal{H}_{i}\times\mathbb{R} on the restrictions \bar{E}_{c_i}\to \bar{B}_{c_i} for each i\in\{0,1\}.

Used by Lemma 2.7, W
Lemma

Lemma 2.7

Q87Y

Let \bar{p}\colon \bar{E}\to \bar{B} denote the bundle described above with structure map \bar{f}\colon\bar{B}\to BG and fibre L, equipped with a fixed metric g_L of non-negative scalar curvature. Let

\begin{gathered} p_0=\bar{p}|_{E_0}\colon\ (E_0, g_0)\to (B_0, h_0) \end{gathered}

be a psc-Riemannian submersion with respect to the fibre (L, \tau_0 g_L) for some constant \tau_0>0, and horizontal distribution \mathcal{H}_0\in\mathcal{HD}(p_0). For compact spaces K_1 and K_2, let \bar{h}(x)\in {\mathcal R}^{\mathrm{psc}}\big(\bar{B}, \partial \bar{B}\big), be a continuous family of psc-bordism metrics on \bar{B}, indexed by x\in K_1, and let \bar{\mathcal{H}}(y)\in \mathcal{HD}_{\mathrm{adm}}(\bar{p}) be a continuous family of admissible horizontal distributions indexed by y\in K_2. We further assume that, for all x\in K_1 and y\in K_2, \bar{h}(x)|_{B_0}=h_0 and \bar{\mathcal{H}}(y)|_{E_0}=\mathcal{H}_0.

Then there is a constant \bar{\tau}_{\operatorname*{{\text{\rm min}}}}>0 and a corresponding family of psc-metrics \bar{g}(x,y), (x,y)\in K_1\times K_2 on \bar{E} which satisfies the following conditions.

    =0pt

  1. (i) \bar{p}\colon \big(\bar{E}, \bar{g}(x,y)\big)\to \big(\bar{B},\bar{h}(x)\big) is a Riemannian submersion with respect to the horizontal distribution \bar{\mathcal{H}}(y).

  2. (ii) Each psc-metric \bar{g}(x,y) takes the form of a product {g}_{i}(x,y)+{\rm d}t^{2} near the boundary components E_i, and satisfies {g}_{0}(x,y)=g_0 on E_0.

  3. (iii) The Riemannian submersion obtained by restriction to E_1,

    \begin{gathered} p_1=\bar{p}|_{E_1}\colon\ \big(E_1, g_1(x,y):=\bar{g}(x,y)|_{E_1}\big)\to\big(B_1, h_1(x):=\bar{h}(x)|_{B_1}\big), \end{gathered}

    has fibre (L, \bar{\tau}_{\operatorname*{{\text{\rm min}}}} g_L).

W

W

7XSE

e begin by specifying \bar{g}=\bar{g}(x,y) as, for all (x,y), the product g_0+{\rm d}t^2 near the boundary component E_0. More precisely, \bar g:= g_0+{\rm d}t^2 on \bar{E}_{c_0}\cong E_{0}\times [0,1]. Thus, near E_0, the metric \bar g forms part of a submersion p_0\times \mathrm{Id}\colon \big(E_0\times I, g_0+{\rm d}t^2\big)\rightarrow \big(B_0\times I, h_0+{\rm d}t^{2}\big) with fibre (L,\tau_0 g_L) and horizontal distribution \bar{\mathcal{H}}|_{\bar{E}_{c_0}}=\mathcal{H}_0\times \mathbb{R}.

Our next task is to extend this metric as a psc-submersion over the rest of \bar{E}. To maintain positive scalar curvature it may be necessary to continuously rescale the fibre metric, but in such a way as to preserve the original fibre metric near E_0. To simplify things, we make adjustments to the fibre metric only in \bar{E}_{c_0} (and away from E_{0}), where the metric, \bar g, takes the form of a product \bar g=g_0+{\rm d}t^{2}. Consider, for constants \tau>0 and b_{\tau}\geq 2, smooth monotonic functions \gamma_{\tau}\colon [0, b_{\tau}]\rightarrow [0, \tau] satisfying

    =0pt

  1. (i) \gamma_{\tau}(t)=\tau_0 when t\in[0,1],

  2. (ii) \gamma_{\tau}(t)=\tau when t\in[b_{\tau}-1,b_{\tau}].

We define g_0(t) to be the submersion metric obtained on E_0\times\{t\}\to B_0\times\{t\} with base metric h_0 and horizontal distribution \mathcal{H}_0 as before but replacing the fibre metric \tau_0 g_{L} with \gamma_{\tau}(t)g_{L}. Now replace the submersion p_0\times \mathrm{Id}\colon \big(\bar{E}_{c_0}=E_0\times I, g_0+{\rm d}t^{2}\big)\rightarrow \big(B_0\times I, h_0+{\rm d}t^{2}\big), with p_0\times \mathrm{Id}\colon \big(E_{0}\times[0,b_\tau], g({\gamma_{\tau}(t)})+{\rm d}t^{2}\big)\rightarrow \big(B_0\times [0,b_\tau], h_0+{\rm d}t^{2} \big). The following proposition is an easy consequence of a well-known rescaling technique; see for example [GL, Lemma 3] or [W0, Lemma 1.3].

Proposition.

For any constant \tau>0, the smooth function \gamma_{\tau} above can be chosen so that the metric g({\gamma_{\tau}(t)})+{\rm d}t^{2} has positive scalar curvature.

We will specify the choice of \tau>0 shortly. In the mean time, let us assume that \gamma_{\tau} has been chosen to guarantee positive scalar curvature. Note that near t=0, the metric remains unchanged.

Consider now the other end of the cylinder, where t=\tau_b. Here we have a product of psc-submersion metrics differing only from that at the t=0 end in that the fibre metric is now \tau g_{L}. No change has been made to the base metric and it is assumed to extend over the base bordism \bar{B} as a psc-metric \bar{h} which takes the form of a product metric {h}_{1}+{\rm d}t^{2} near the boundary component B_{1}. For each y, the admissible horizontal distribution \bar{\mathcal{H}}(y) extends \mathcal{H}_0\times \mathbb{R} over the rest of \bar{E}, inducing, outside of the collar \bar{E}_{c_0}, Riemannian submersion metrics with base metric \bar{h}|_{\bar{B}_{c_0}'}(x) and fibre metric \tau g_{L}. We need to ensure positive scalar curvature by choosing appropriately small \tau.

Suppose for any (x,y), we choose \tau=\bar{\tau}\big(\bar{h}(x), \bar{\mathcal{H}}(y)\big)>0, as defined in (?). This ensures, via the canonical variation formula [Besse, Proposition 9.70], that the submersion metric

\begin{gathered} \mathcal{S}\big(\bar{h}(x), \bar{\mathcal{H}}(y), \bar{\tau}\big(\bar{h}(x), \bar{\mathcal{H}}(y)\big)\big) \end{gathered}

on the bundle \bar{E}_{c_0}'\to \bar{B}_{c_{0}}' with base metric \bar{h}(x)|_{\bar{B}_{c_{0}'}}, fibre metric \bar{\tau}\big(\bar{h}(x), \bar{\mathcal{H}}(y)\big) g_{L} and horizontal distribution \bar{\mathcal{H}}(y)|_{\bar{E}_{c_0}'} has positive scalar curvature. Thus, we set

\begin{gathered} \tau=\bar{\tau}_{\operatorname*{{\text{\rm min}}}}:=\operatorname*{{\text{\rm min}}}\big\{\bar{\tau}\big(\bar{h}(x), \bar{\mathcal{H}}(y)\big)\colon (x,y)\in K_1\times K_2\big\}, \end{gathered}

to obtain a single scaling factor which works for all (x,y). Finally, the metric \bar{g}(x,y) is defined to be the submersion metric determined by the base \big(\bar{B}, \bar{h}(x)\big), fibre L with varying metric \gamma_{\bar{\tau}_{\operatorname*{{\text{\rm min}}}}}(t) g_L on E_{c_0} and \bar{\tau}_{\operatorname*{{\text{\rm min}}}} g_{L} on the rest of \bar{E}, as well as horizontal distribution \bar{\mathcal{H}}(y).

Proposition

Proposition 2.8

ANHS

For any constant \tau>0, the smooth function \gamma_{\tau} above can be chosen so that the metric g({\gamma_{\tau}(t)})+{\rm d}t^{2} has positive scalar curvature.

Definition

Definition 2.9

DB56

A link (L,g_L) is simple if it satisfies one of the following conditions:

    =0pt

  1. (a) (L,g_L) is homogeneous with scalar curvature, s_{g_L}, a positive constant,

  2. (b) (L,g_L)=\big(S^1,{\rm d}\theta^2\big),

  3. (c) L =\mathbb{Z}_k.

Used by Remark 2.14
Remark

Remark 2.10

ESWV

We emphasise that the fibre metric g_{L} and structure map f\colon \beta X\to BG are assumed to be fixed throughout. Thus, while the map \mathcal{S} (along with \mathcal{S}^{\mathrm{psc}} and \mathcal{S}_{\mathcal{H}}^{\mathrm{psc}}) depend of course on these objects, and could be denoted by something like \mathcal{S}_{f, g_{L}}, we will not feature them in the notation.

Theorem

Theorem 2.11

XOJH

The restriction map \mathrm{res}\colon {\mathcal R}^{\mathrm{psc}}(X, \partial X) \to {\mathcal R}^{\mathrm{psc}}(\partial X) is a Serre fibre bundle.

Definition

Definition 2.12

60MN

The space

\begin{gathered} {\mathcal R}_{\mathcal{S}}^{\mathrm{psc}}(X, \partial X):=\mathrm{res}^{-1}\big({\mathcal R}_{\mathcal{S}}^{\mathrm{psc}} (\partial X)\big)\subset {\mathcal R}^{\mathrm{psc}}(X, \partial X) \end{gathered}

is called the space of well-adapted psc-metrics on X.

Definition

Definition 2.13

1I24

The image of the map i_{\Sigma}, denoted {\mathcal R}^{\mathrm{psc}}(X_{\Sigma}), is called the space of well adapted psc-metrics on X_{\Sigma}.

Remark

Remark 2.14

H6MD

It is worth emphasising that the space {\mathcal R}_{\mathcal{S}}^{\mathrm{psc}}(\partial X) (as well the space {\mathcal R}_{\mathcal{S}}^{\mathrm{psc}}(X, \partial X)) could be empty even though the boundary \partial X has a psc-metric. A simple example is the following. Let \beta X= K3. Then for any prime generator c\in H^{2}(K3;\mathbb{Z})=\mathbb{Z}^{22} a corresponding map c\colon K3\to \mathbb{CP}^{\infty} gives a bundle Z \to \beta X, where Z is a simply-connected spin manifold which, of course, admits a psc-metric. Moreover, Z is a spin boundary: Z=\partial X. However, no psc-metric on Z could be invariant under S^1-action, otherwise, according to Bérard-Bergery [B-B, Theorem C], it would imply the existence of a psc-metric on K3.

Proposition

Proposition 2.15

0NRL

The spaces \mathcal{F}(\mathcal{HD}(p_0), \mathcal{HD}_{\mathrm{adm}}(\bar{p})) and \mathcal{F}_0(\mathcal{HD}(p_0), \mathcal{HD}_{\mathrm{adm}}(\bar{p})) are convex.

Corollary

Corollary 2.16

NX8E

The homotopy type of the map \mu_{(\bar B,\bar f, \bar h^{\beta})} is invariant of the psc-isotopy type of the base metric \bar h^{\beta}, as well as the choice of distribution map \xi\in \mathcal{F}_0(\mathcal{HD}(p_0), \mathcal{HD}_{\mathrm{adm}}(\bar{p})).

Theorem

Theorem 2.17

UUL9

Let (X,\beta X, f) be a manifold with fibred singularities, i.e., the boundary \partial X is a total space of an (L,G)-fibration \partial X\to \beta X given by the structure map f\colon \beta X \to BG, where \operatorname{{\text{\rm dim}}} X = n, \operatorname{{\text{\rm dim}}} L= \ell. Furthermore, we assume \big(\bar B,\bar f\big)\colon (B_0,f_0)\rightsquigarrow (B_1,f_1) is an elementary bordism with p,q\geq 2, where B_0=\beta X. Then for any psc-metric h^{\beta}_0 on B_0 there exist psc-metrics h^{\beta}_1 on B_1 and \bar h^{\beta}\in {\mathcal R}^{\mathrm{psc}}\big(\bar B\big)_{h^{\beta}_0,h^{\beta}_1} such that the map

\begin{aligned} \mu_{(\bar B,\bar f, \bar h^{\beta})}\colon\ & {\mathcal R}_{\mathcal{S}}^{\mathrm{psc}}(X,\partial X)_{h^{\beta}_0} \longrightarrow {\mathcal R}_{\mathcal{S}_1}^{\mathrm{psc}}(X_1,\partial X_1)_{h^{\beta}_1}, \qquad g \longmapsto g\cup \bar g^{\partial},\\ & X_1=X\cup_{\partial X=E_0} \bar E, \qquad \beta X_1= E_1, \end{aligned}

defined as in (?) above, is a weak homotopy equivalence.

Remark

Remark 2.18

SFOW

The construction of the metrics h^{\beta}_1 on B_1 and \bar h^{\beta} on \bar{B} used in this theorem, follows from surgery techniques pioneered by Gromov and Lawson [GL], and later generalised by others. It should be pointed out however that, given such a metric, h^{\beta}_1, the theorem works equally well if h^{\beta}_1 is replaced by any psc-metric which is psc-isotopic to h^{\beta}_1, i.e., lies in the same path component of the space {\mathcal R}^{\mathrm{psc}}(B_1).

Proposition

Proposition 3.1

56HM

Let n\geq3, \delta>0 and \lambda, \lambda_1, \lambda_2\geq 0.

    =0pt

  1. (i) The metrics g_{\mathrm{torp}}^{n}(\delta)_{\lambda}, g_{\mathrm{torp}+}^{n}(\delta)_{\lambda} and \hat{g}_{\mathrm{torp}}^{n}(\delta)_{\lambda_1, \lambda_2} have positive scalar curvature.

  2. (ii) For any constant b\geq 0 and any \lambda, \lambda_1, \lambda_2\geq 0, there exists \delta>0 so that the scalar curvature of the metrics g_{\mathrm{torp}}^{n}(\delta)_{\lambda}, g_{\mathrm{torp}+}^{n}(\delta)_{\lambda} and \hat{g}_{\mathrm{torp}}^{{n}}(\delta)_{\lambda_1, \lambda_2} is bounded below by b.

Theorem

Theorem 5.1 (Botvinnik–Ebert–Randal-Williams , and Perlmutter )

HDKW

Assume M is a spin manifold with \operatorname{{\text{\rm dim}}} M \geq 5 and {\mathcal R}^{\mathrm{psc}}(M)\neq \varnothing with a base point g_0\in {\mathcal R}^{\mathrm{psc}}(M). Then the index-diffence map (?) induces a non-trivial homomorphism in the homotopy groups

\begin{gathered} (\mathsf{inddiff}_{g_0})_*\colon\ \pi_q({\mathcal R}^{\mathrm{psc}}(M))\to KO_{q+n+1}, \end{gathered}

when the target group KO_{q+n+1} is non-trivial.

Conjecture

Conjecture 5.2

0JKQ

Let X_{\Sigma} be a spin (\left<k\right>\!\mbox{-}\mathrm{fb})-manifold. Assume \operatorname{{\text{\rm dim}}} X\geq 7 and X and \beta X\neq \varnothing are simply-connected and {\mathcal R}^{\mathrm{psc}}(X_{\Sigma})\neq \varnothing with a base point g_0\in {\mathcal R}^{\mathrm{psc}}(X_{\Sigma}). Then there is an index-difference map

\begin{gathered} \mathsf{inddiff}_{g_0}^{\left<k\right>}\colon\ {\mathcal R}^{\mathrm{psc}}(X_{\Sigma})\to \Omega^{\infty+n+1}\mathbf{KO}^{\left<k\right>}, \end{gathered}

which induces a non-trivial homomorphism in the homotopy groups

\begin{gathered} \big(\mathsf{inddiff}_{g_0}^{\left<k\right>}\big)_*\colon\ \pi_q({\mathcal R}^{\mathrm{psc}}(X_{\Sigma}))\to KO^{\left<k\right>}_{n+q+1} \end{gathered}

when the target group KO^{\left<k\right>}_{n+q+1} (KO with \mathbb{Z}_k-coefficients) is non-trivial.

Conjecture

Conjecture 5.3

HOAC

Let X_{\Sigma} be a spin manifold with (\eta\mbox{-}\mathrm{fb})-singularity of dimension n\geq 9. Assume \beta X\neq \varnothing, and {\mathcal R}^{\mathrm{psc}}(X_{\Sigma})\neq \varnothing with a base point g_0\in {\mathcal R}^{\mathrm{psc}}(X_{\Sigma}). Then there is an index-difference map

\begin{gathered} \mathsf{inddiff}_{g_0}^{\eta\mbox{-}\mathrm{fb}}\colon\ {\mathcal R}^{\mathrm{psc}}(X_{\Sigma})\to \Omega^{\infty+n+1}\mathbf{KO}^{\eta\mbox{-}\mathrm{fb}}, \end{gathered}

which induces a non-trivial homomorphism in the homotopy groups

\begin{gathered} \big(\mathsf{inddiff}_{g_0}^{\eta}\big)_*\colon\ \pi_q({\mathcal R}^{\mathrm{psc}}(X_{\Sigma}))\to KO^{\eta\mbox{-}\mathrm{fb}}_{q+n+1} \end{gathered}

when the target group KO^{\eta\mbox{-}\mathrm{fb}}_{q+n+1}= KO_{q+n+1}(\mathbb{CP}^{\infty}) is non-trivial.

Theorem

Theorem 5.4 (see )

WQA9

Let X_\Sigma be an (L,G)-fibred compact pseudo-manifold with L a simply connected homogeneous space of a compact semisimple Lie group. Assume {\mathcal R}^{\mathrm{psc}}(X_{\Sigma})\neq \varnothing. Then there exists a section s\colon \mathcal{R}^{\mathrm{psc}}(\beta X)\to \mathcal{R}^{\mathrm{psc}} (X_\Sigma) to \mathrm{res}_{\Sigma}. In particular, there is a split short exact sequence:

\begin{gathered} 0 \to \pi_q(\mathcal{R}_w^+ (X_\Sigma)_{g_{\beta X}})\xrightarrow{i_*} \pi_q(\mathcal{R}_w^+ (X_\Sigma))\xrightarrow{(\mathrm{res}_{\Sigma})_*}\pi_q(\mathcal{R}^+(\beta X))\to 0 , \qquad q=0,1,\ldots. \end{gathered}
Corollary

Corollary 5.5 (see )

Z304

Let X_\Sigma be an (L,G)-fibred compact pseudo-manifold with L a simply connected homogeneous space of a compact semisimple Lie group, and n-\ell-1\geq 5, where \operatorname{{\text{\rm dim}}} X=n, \operatorname{{\text{\rm dim}}} L =\ell. Let g_0\in {\mathcal R}^{\mathrm{psc}}(X_{\Sigma})\neq \varnothing be a base point giving corresponding base points, the metrics g_{\beta X,0}\in {\mathcal R}^{\mathrm{psc}}(\beta X), g_{\partial X,0}\in {\mathcal R}^{\mathrm{psc}}(\partial X) and g_{X,0}\in {\mathcal R}^{\mathrm{psc}}(X)_{g_{\partial X,0}}.

If M_\Sigma is spin and simply connected, then we have the following commutative diagram:

\begin{gathered} \begin{diagram} \setlength{\dgARROWLENGTH}{1.95em} \node{0\to \pi_q{\mathcal R}^{\mathrm{psc}}(X_\Sigma)_{g_{\beta X,0}}} \arrow{e,t}{j_*} \arrow{s,l}{\mathrm{inddiff}_{g_{X,0}}} \node{\pi_q\mathcal{R}^{\mathrm{psc}} (X_\Sigma)} \arrow{e,t}{(\mathrm{res}_{\Sigma})_*} \arrow{s,l}{\mathrm{inddiff}_{g_0}} \node{\pi_q\mathcal{R}^{\mathrm{psc}}(\beta X)\to 0} \arrow{s,l}{\mathrm{inddiff}_{g_{\beta X,0}}} \\ \node{\!\!\! 0\longrightarrow KO_{q+n+1}} \arrow{e} \node{KO_{q+n+1}\oplus KO_{q+n-\ell} } \arrow{e} \node{KO_{q+n-\ell}\to 0,} \end{diagram} \end{gathered}

where the homomorphisms \mathrm{inddiff}_{g_{X,0}} and \mathrm{inddiff}_{g_{\beta X,0}} are both nontrivial whenever the target groups are. In particular, the homomorphism

\begin{gathered} \mathrm{inddiff}_{g_0}\colon\ \pi_q\mathcal{R}^{\mathrm{psc}} (X_\Sigma) \to KO_{q+n+1}\oplus KO_{q+n-\ell} \end{gathered}

is surjective rationally and surjective onto the torsion of KO_{q+n+1}\oplus KO_{q+n-\ell}.

Remark

Remark 6.1

AI1C

Regarding the earlier setting when g_L is scalar flat, we will typically only be interested in the case when L=S^{1} and so l=1. Thus, metric formulae in both cases coincide.

Proposition

Proposition 6.2

2T4J

Attaching \big(\big[0, \frac{1}{2}\big]\times L, g_{C(L)}\big) to \big([0, 1]\times L, g_{\mathrm{att}(L)}\big) by identifying \big\{\frac{1}{2}\big\}\times L in the former with \{0\}\times L in the latter determines a smooth metric of non-negative scalar curvature g_{\mathrm{att}(L)}\cup g_{C(L)}.

S

S

RZ7D

moothness is immediate from the properties of the transition function a. The component, g_{C(L)}, has zero scalar curvature and that g_{\mathrm{att}(L)} has non-negative scalar curvature is easily seen from the formula (?).

Source and licence

The text on this page is from arXiv:2005.03073v3 by Boris Botvinnik and Mark Walsh, under the licence CC BY SA 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY SA 4.0.

The record of each item is in analysis/objects/2005/2005.03073.jsonl.