On the Uniqueness of the G G -equivariant Spectral Flow

Marek Izydorek, Joanna Janczewska, Maciej Starostka, Nils Waterstraat · 2025

Read On the Uniqueness of the G G -equivariant Spectral Flow by Marek Izydorek, Joanna Janczewska, Maciej Starostka, Nils Waterstraat (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-2508-01061.txt.

Introduction

Marek Izydorek, Joanna Janczewska, Maciej Starostka and Nils Waterstraat wrote a 2025 math paper on a symmetry-aware version of spectral flow. Spectral flow counts how eigenvalues cross zero along a path of operators. It is known to be the only measure with a set of basic traits. The authors had built a version that respects a group symmetry.

Here they prove that the symmetric version is also unique. The same basic traits, restated for the symmetric case, pin it down fully. As a use case, they give a new definition of it through a symmetric Maslov index. The result firms up the base of this newer tool.

Catalog record

CREATORMarek Izydorek, Joanna Janczewska, Maciej Starostka, Nils Waterstraat
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE64 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-01061

Browse the full library · About the Tower of Babel

Full text

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

On the Uniqueness of the G G -equivariant Spectral Flow Marek Izydorek, Joanna Janczewska, Maciej Starostka, Nils Waterstraat Abstract The spectral flow is an integer-valued homotopy invariant for paths of selfadjoint Fredholm operators. Lesch as well as Pejsachowicz, Fitzpatrick and Ciriza independently showed that it is uniquely characterised by its elementary properties. The authors recently introduced a G G -equivariant spectral flow for paths of selfadjoint Fredholm operators that are equivariant under the action of a compact Lie group G G . The purpose of this paper is to show that the G G -equivariant spectral flow is uniquely characterised by the same elementary properties when appropriately restated. As an application we introduce an alternative definition of the G G -equivariant spectral flow via a G G -equivariant Maslov index. 1 1 footnotetext: 2010 Mathematics Subject Classification: Primary 58J30; Secondary 47A53 1 Introduction Let H be a real separable Hilbert space of infinite dimension and let us denote by \mathcal{L}(H) the bounded, by \GL(H) the invertible, by \mathcal{K}(H) the compact and by \mathcal{S}(H) the selfadjoint operators on H . An operator T\in\mathcal{L}(H) is called Fredholm if its kernel and cokernel are of finite dimension. For T\in\mathcal{S}(H)\subset\mathcal{L}(H) these conditions are equivalent to a finite dimensional kernel and a closed range. Atiyah and Singer considered in [ 1 ] the set \mathcal{FS}(H)\subset\mathcal{L}(H) of all selfadjoint Fredholm operators and showed that it consists of the three connected components \displaystyle\mathcal{FS}^{+}(H) \displaystyle:=\{T\in\mathcal{FS}(H):\,\sigma_{ess}(T)\subset(0,+\infty)\}, \displaystyle\mathcal{FS}^{-}(H) \displaystyle:=\{T\in\mathcal{FS}(H):\,\sigma_{ess}(T)\subset(-\infty,0)\}, and \displaystyle\mathcal{FS}^{i}(H):=\mathcal{FS}(H)\setminus\mathcal{FS}^{\pm}(H). Here \sigma_{ess}(T) denotes the essential spectrum of T , i.e., the set of all \lambda\in\mathbb{R} such that \lambda-T is not a Fredholm operator. Operators in \mathcal{FS}^{\pm}(H) have as only possible points in the spectrum below/above 0 eigenvalues of finite multiplicity. Instead operators in \mathcal{FS}^{i}(H) necessarily have essential spectrum above and below 0 . Atiyah and Singer also showed in [ 1 ] that the two components \mathcal{FS}^{\pm}(H) of \mathcal{FS}(H) are contractible as topological spaces, whereas \mathcal{FS}^{i}(H) has a rather rich topology and in particular an infinitely cyclic fundamental group. In a joint work with Patodi [ 2 ] they introduced an explicit isomorphismus between \pi_{1}(\mathcal{FS}^{i}(H)) and the integers. Actually, the idea of this map, which is called spectral flow, is sensible for any closed or open path and in any of the three components of \mathcal{FS}(H) . Following Phillips’ analytic approach [ 16 ] (cf. [ 7 ] ), its construction can be outlined as follows: Let I=[0,1] be the unit interval and L=\{L_{\lambda}\}_{\lambda\in I} a path in \mathcal{FS}(H) . As the spectra of the operators L_{\lambda} cannot accumulate at 0 (cf. [ 20 ] , [ 21 ] ), there is a partition 0=\lambda_{0}<\ldots<\lambda_{N}=1 of the unit interval and numbers a_{i}>0 , i=1,\ldots N , such that the maps \displaystyle[\lambda_{i-1},\lambda_{i}]\ni\lambda\mapsto\chi_{[-a_{i},a_{i}]}(L_{\lambda})\in\mathcal{L}(H) (1) are continuous families of finite rank projections. Here \chi_{[a,b]}(T) denotes the spectral projection of a selfadjoint operator T with respect to the interval [a,b]\subset\mathbb{R} . Consequently, the spaces E(L_{\lambda},[0,a_{i}]):=\im(\chi_{[0,a_{i}]}(L_{\lambda})) , \lambda_{i-1}\leq\lambda\leq\lambda_{i} , i=1,\ldots,N , are of finite dimension and the spectral flow of the path L can be defined by \displaystyle\sfl(L)=\sum^{N}_{i=1}{(\dim(E(L_{\lambda_{i}},[0,a_{i}]))-\dim(E(L_{\lambda_{i-1}},[0,a_{i}])))}\in\mathbb{Z}. (2) Note that \dim(E(L_{\lambda},[0,a])) is the number of eigenvalues of L_{\lambda} in the interval [0,a] including multiplicities. Roughly speaking, the spectral flow is the net number of eigenvalues of L_{0} crossing 0 from the negative to the positive half of the axis whilst the parameter is travelling along the interval. It can be shown that ( 2 ) is well defined, i.e., it only depends on the path L and neither on the choice of the partition \{\lambda_{i}\}_{i=0,\ldots,N} of I nor on the numbers a_{1},\ldots,a_{N} as long as the maps ( 1 ) are continuous and finite-rank valued. The spectral flow has the following properties: (\mathcal{Z}) If L=\{L_{\lambda}\}_{\lambda\in I} is a path in \mathcal{FS}(H)\cap\GL(H) , then \sfl(L)=0. (\mathcal{C}) If L^{1} , L^{2} are paths in \mathcal{FS}(H) , then \sfl(L^{1}\ast L^{2})=\sfl(L^{1})+\sfl(L^{2}), where L^{1}\ast L^{2} denotes the concatenation of L^{1} and L^{2} . (\mathcal{A}) If H=H_{1}\oplus H_{2} and L_{\lambda}(H_{i})\subset H_{i} , i=1,2 , for a path L=\{L_{\lambda}\}_{\lambda\in I} in \mathcal{FS}(H) , then \sfl(L)=\sfl(L\mid_{H_{1}})+\sfl(L\mid_{H_{2}}). (\mathcal{H}) If h:I\times I\rightarrow\mathcal{FS}(H) is a homotopy such that h(s,0),h(s,1)\in\GL(H) for all s\in I , then \sfl(h(0,\cdot))=\sfl(h(1,\cdot)). Let us note that (\mathcal{Z}) , (\mathcal{C}) , (\mathcal{A}) easily follow from the definition ( 2 ), and (\mathcal{H}) is a simple modification of the original argument in [ 16 ] (see, e.g., [ 19 ] ). It is a natural question to ask whether some of these properties uniquely characterise the spectral flow when a normalisation property is added that in particular excludes the trivial map which assigns 0\in\mathbb{Z} to any path in \mathcal{FS}(H) . Henceforth we denote for a pair of topological spaces (X,Y) by \Omega(X,Y) the set of all paths in X having endpoints in Y . Lesch considered in [ 15 ] maps \mu:\Omega(\mathcal{FS}^{i}(H),\mathcal{FS}^{i}(H)\cap\GL(H))\rightarrow\mathbb{Z} and proved that any such map which satisfies the same properties (\mathcal{C}) , (\mathcal{H}) as the spectral flow is already given by ( 2 ) if the following normalisation assumption holds: \mathcal{(N)} There is some T_{0}\in\mathcal{FS}^{i}(H)\cap\GL(H) and some rank-one orthogonal projection P in H commuting with T_{0} such that \mu(\{\lambda P+(I_{H}-P)T_{0}\}_{-\frac{1}{2}\leq\lambda\leq\frac{1}{2}}\})=1 Note that the path \{\lambda P+(I_{H}-P)T_{0}\}_{-\frac{1}{2}\leq\lambda\leq\frac{1}{2}} is in \GL(H) for any \lambda\neq 0 , whereas 0 is a simple eigenvalue for \lambda=0 . It is readily seen from ( 2 ) that indeed \sfl(\{\lambda P+(I_{H}-P)T_{0}\}_{-\frac{1}{2}\leq\lambda\leq\frac{1}{2}}\})=1 . Another approach to the uniqueness problem of the spectral flow was proposed by Fitzpatrick, Pejsachowicz and Recht in [ 6 ] . They considered maps \mu:\Omega(\mathcal{FS}(H),\mathcal{FS}(H)\cap\GL(H))\rightarrow\mathbb{Z} and showed that any such map which satisfies the same properties (\mathcal{Z}) , (\mathcal{A}) and (\mathcal{H}) as the spectral flow is already given by ( 2 ) if the following normalisation assumption holds: \mathcal{(M)} If H is of finite dimension and L=\{L_{\lambda}\}_{\lambda\in I} is a path in (\mathcal{FS}(H),\mathcal{FS}(H)\cap\GL(H)) , then \mu(L)=\mu_{Morse}(L_{0})-\mu_{Morse}(L_{1}), where \mu_{Morse}(L_{\lambda}) denotes the Morse index of L_{\lambda} , \lambda\in I , i.e., the number of negative eigenvalues including multiplicities. Inspired by [ 8 ] , the authors defined in [ 12 ] a G -equivariant spectral flow for paths L=\{L_{\lambda}\}_{\lambda\in I} in \mathcal{FS}(H) where each operator L_{\lambda} is equivariant under the orthogonal action of a compact Lie group G on H . The G -equivariant spectral flow is an element of the real representation ring RO(G) and it formally has the same properties (\mathcal{Z}) , (\mathcal{C}) , (\mathcal{A}) and (\mathcal{H}) from above. Moreover, also (\mathcal{M}) still holds, when appropriately restated in terms of virtual representations of G . The aim of this paper is to show that the G -equivariant spectral flow is uniquely characterised by these properties as well. If the group action is trivial, then our theorem is the uniqueness of the spectral flow ( 2 ) from [ 6 ] . We recall the definition of the G -equivariant spectral flow in the next section, where we also introduce the announced uniqueness of it as our main theorem. Section 3 is devoted to the rather elaborate proof of the latter theorem. In the final Section 4, we introduce a G -equivariant Maslov index, where we adopt an approach to the classical Maslov index from a celebrated paper by Cappell, Lee and Miller [ 5 ] . Finally, we show as first application of our uniqueness theorem that the G -equivariant spectral flow

Read the complete text