Lower Semicontinuity of Restricted Holonomy Groups

Linus Götzfried · 2025

Read Lower Semicontinuity of Restricted Holonomy Groups by Linus Götzfried (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-17340.txt.

Introduction

Holonomy measures how a vector twists as it is carried around a loop on a curved space, grouping spaces by that behaviour. Linus Götzfried, writing alone, proves a stability result for those groups. He asks what happens when a series of connections, the rules for carrying vectors, settles down toward a limiting connection.

Götzfried shows that if every member of the series keeps its holonomy inside one fixed closed group, then the limit must obey the same bound. A direct consequence covers curved spaces described by metrics: when metrics with special restricted holonomy converge, the limiting metric keeps that special type. In plain terms, the assignment of holonomy classes can only drop, never jump up, in a limit.

Catalog record

CREATORLinus Götzfried
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-17340

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Lower Semicontinuity of Restricted Holonomy Groups The author would like to thank Bernd Ammann for valuable supervision and comments. Linus Götzfried Abstract . A result on the monotony of holonomy groups of metric connections in limits is proven. The main outcome is that given a sequence of metric connections that are Lipschitz continuous and converging in C 0 C^{0} , if the connections in the sequence all have holonomy contained in a closed group H H , then also the limit connection has holonomy contained in H H . In particular, this finds an application for Riemannian holonomy groups, where it is proven that the map assigning to a Riemannian metric the conjugacy class of its restricted holonomy group is lower semicontinuous with respect to the C 1 C^{1} -topology on the space of metrics. 2020 Mathematics Subject Classification Primary 53C29; Secondary 53C05; 1. Introduction An important notion in differential geometry is the one of holonomy . Given any connection on a principal or vector bundle and a basepoint, one can define the holonomy group of the connection as the group of endomorphisms of the fibres over this point given by parallel translation along piecewise-differentiable loops. Of particular interest are the holonomy groups of the Levi-Civita connection on a Riemannian manifold, which are subgroups of the special orthogonal group. In many cases, a “special” holonomy of a Riemannian manifold implies interesting consequences for the resulting metric. For example, a smooth 4m -dimensional Riemannian manifold with holonomy contained in the group \mathrm{Sp}(m)\mathrm{Sp}(1) is automatically an Einstein manifold (see e.g. [ 3 ] , Theorem 14.39). In this work, we study the changes of the holonomy group under changes of the connection. We focus on the case of metric connections on a vector bundle and prove a result of monotony of the holonomy group in limits, Theorem 2.4 below. Roughly speaking, it is shown that in the case of C^{0} -convergence of metric connections \nabla^{k} to a metric connection \nabla , the connection \nabla can only have a more special holonomy than the connections \nabla^{k} , but not a more generic one. In particular, this finds an application for Riemannian holonomy groups. In this case, we also obtain a prettier (but slightly weaker) reformulation of the statement in terms of the lower semicontinuity of the map assigning to a C^{1} -Riemannian metric the conjugacy class of its restricted holonomy group. The motivation and some of the ideas of this paper are taken from the recent preprint [ 6 ] by Müller. However, the proof of the main theorem is a different one than in the cited preprint, and the resulting theorem is a stronger one (note that in [ 6 ] , C^{2} -convergence is required instead of only C^{1} -convergence). 2. Monotony of the holonomy group in limits We recall the definition of the holonomy group. Definition 2.1 . Let M be a connected manifold of class C^{1} , let x\in M and let E\to M be an orientable vector bundle over M with standard fibre \mathbb{R}^{l} , l\in\mathbb{N} . Let \nabla be a connection on E . (1) We denote by \Omega_{\mathrm{p}-C^{1}}(x) the set of all piecewise-differentiable loops c:[0,1]\to M based at x . (2) Let c\in\Omega_{\mathrm{p}-C^{1}}(x) . We define the parallel transport P_{c}^{\nabla} along c (with respect to the connection \nabla on E ) as follows: Given any e\in E_{x} , there exists a unique section s of E|_{c([0,1])} with \nabla_{\dot{c}(t)}s(t)=0 for all t\in[0,1] , and we let P_{c}^{\nabla}(e):=s(1) . Then \mathrm{Hol}_{x}(\nabla):=\left\{P_{c}^{\nabla}\,\middle|\,c\in\Omega_{\mathrm{p}-C^{1}}(x)\right\} forms a subgroup of \mathrm{GL}(E_{x}) , the holonomy group of \nabla (at x ). (3) We define the restricted holonomy group of \nabla at x , \displaystyle\mathrm{Hol}_{x}^{0}(\nabla):= \displaystyle\,\left\{P_{c}^{\nabla}\,\middle|\,c\in\Omega_{\mathrm{p}-C^{1}}(x)\textnormal{ such that $c$ is contractible}\right\} \displaystyle\subseteq \displaystyle\,\mathrm{Hol}_{x}(\nabla). Both \mathrm{Hol}_{x}(\nabla) and \mathrm{Hol}_{x}^{0}(\nabla) are Lie subgroups of \mathrm{GL}(E_{x}) . It can be shown (see e.g. [ 1 ] , Section 2, Lemma 4) that \mathrm{Hol}_{x}^{0}(\nabla) is the connected component of the identity in \mathrm{Hol}_{x}(\nabla) . (4) Choosing any basis of E_{x} , we may identify E_{x} with \mathbb{R}^{l} and consequently view \mathrm{Hol}_{x}(\nabla) and \mathrm{Hol}_{x}^{0}(\nabla) as subgroups of \mathrm{GL}(l) . However, this is only up to conjugation; the choice of different bases yields subgroups conjugate in \mathrm{GL}(l) . Note that since any C^{1} -map is in particular Lipschitz, a manifold of class C^{1} is a Lipschitz manifold , that is, the transition maps of its defining atlas are bi-Lipschitz ( [ 8 ] , Definition 1). It is well-defined to speak of locally Lipschitz continuous functions on such a manifold. In the present work, we will be concerned with the following setting. Situation 2.2 . • Let M be a connected manifold of class C^{1} , let x\in M and E\to M be an orientable vector bundle with standard fibre \mathbb{R}^{l} . We denote by E^{*} the dual bundle of E and by \mathrm{S}^{2}E^{*} the symmetrized tensor product of the dual bundle. • Let g\in C^{1}(M,\mathrm{S}^{2}E^{*}) be a bundle metric on E (that remains fixed throughout this section unless otherwise stated). Let \nabla^{g} be a connection on E whose coefficients are locally Lipschitz continuous and that is compatible with g , i. e. X(g(v,w))=g(\nabla^{g}_{X}v,w)+g(v,\nabla^{g}_{X}w) for all X\in TM , v,w\in C^{1}(M,E) . • Given any metric h on E , let \displaystyle\mathrm{SO}_{l}(E_{x})_{h}:=\{ \displaystyle A\in\mathrm{End}(E_{x})\,|\,\det(A)=1, \displaystyle\forall v,w\in E_{x}:h|_{x}(Av,Aw)=h|_{x}(v,w)\} be the special orthogonal group of E_{x} with respect to the metric h|_{x} . Then, since \nabla^{g} is metric, we have P_{c}^{\nabla^{g}}\in\mathrm{SO}_{l}(E_{x})_{g} for all c\in\Omega_{\mathrm{p}-C^{1}}(x) . • An orthonormal basis of E_{x} with respect to some metric h are vectors e_{1},\dots,e_{l} with \forall i,j=1,\dots,n:h(e_{i},e_{j})=\delta_{ij}, where \delta_{ij} is the Kronecker delta. We choose a positively oriented orthonormal basis (b_{1},\dots,b_{l}) of E_{x} with respect to the metric g and use it to identify E_{x} with \mathbb{R}^{l} . This then identifies \mathrm{SO}_{l}(E_{x})_{g} with \mathrm{SO}(l)\subset\mathrm{GL}(l) and consequently we may view \mathrm{Hol}_{x}(\nabla^{g}) as an explicit subgroup of \mathrm{SO}(l) . Now, given a (possibly different) metric h\in C^{1}(M,\mathrm{S}^{2}E^{*}) , we may still view \mathrm{SO}_{l}(E_{x})_{h} as a distinguished subgroup of \mathrm{GL}(l) , as follows: Definition 2.3 . (1) Using the basis (b_{1},\dots,b_{l}) of E_{x} , we may view all bilinear forms Q\in\mathrm{S}^{2}E^{*}_{x} and all endomorphisms Z\in\mathrm{End}(E_{x}) as l\times l -matrices over \mathbb{R} . In particular, we may associate to a bilinear form Q its representing matrix M_{Q} and via this an endomorphism Z_{Q} of E_{x}\cong\mathbb{R}^{l} . 1 1 1 To be very explicit, therefore Q(X,Y)=g(X,Z_{Q}Y) for all X,Y\in E_{x} , and if we view X,Y also as their respective coordinate vectors with respect to the basis (b_{1},\dots,b_{l}) , then this is nothing but X^{T}M_{Q}Y . This defines an embedding \mathrm{S}^{2}E^{*}_{x}\to\mathrm{End}(E_{x}) , Q\mapsto Z_{Q} . Now \mathrm{End}(E_{x}) can be turned into a normed vector space by using the operator norm induced by the metric g . By pullback along the embedding described above, also \mathrm{S}^{2}E^{*}_{x} becomes a normed vector space. Unless otherwise stated, in the following these norms will be used. (As all vector spaces involved are finite-dimensional, the resulting topologies do not depend on g or the basis.) (2) Let h\in C^{1}(M,\mathrm{S}^{2}E^{*}) with h|_{x}\in B_{1}(g|_{x}) . Let M_{h} be the representing matrix of h|_{x} and Z_{h} the corresponding endomorphism as in the previous point. We define \mathrm{SO}(l)_{h}:=\{A\in\mathrm{GL}(l)\,|\,A^{T}M_{h}A-M_{h}=0\} (the x -dependency is understood). Let W_{h} be a symmetric positive definite square root of

Read the complete text