Logo image
Spectral theory for borel sets in Rn of finite measure
Journal article   Open access   Peer reviewed

Spectral theory for borel sets in Rn of finite measure

Palle E.T Jorgensen and Steen Pedersen
Journal of functional analysis, Vol.107(1), pp.72-104
1992
DOI: 10.1016/0022-1236(92)90101-N
url
https://doi.org/10.1016/0022-1236(92)90101-NView
Published (Version of record) Open Access

Abstract

Let (Ω, Λ) be a pair of subsets in R n such that Ω has finite, positive Lebesgue measure. The pair is called a spectral pair if the functions e iλ · x , for λ ϵ Λ, form an orthogonal, total family in L 2 (Ω) , i.e., when restricted to Ω. Our present work continues earlier papers on simultaneous diagonalization of vector fields, for the special case when Ω is also assumed open . However, the setting here is more general, and the results are stated in the measure theoretic category . (In considering fundamental domains for lattices, for example, we allow distinct lattice point translates of the domain to overlap on sets of measure zero , rather than the traditional condition of empty overlap.) Two classes of examples of spectral pairs are known, fundamental domains and simple factors . The latter may be identified in terms of a certain finite group action. We show that every spectral pair can be factored (in the sense of Cartesian products) as a product of two pairs of lower dimension where the first factor is a fundamental domain, and the second a simple factor. (The dimensions add.) A simple factor yields in a natural way a finite covering of an associated compact torus; but not every such covering corresponds to a spectral pair. We give a complete spectral characterization of the finite coverings and identify those which correspond to simple factors . The latter identification is phrased in terms of generators and relations for certain C ∗ -algebras which are analogous to (but different from) algebras studied earlier by Cuntz and Arveson.

Details

Metrics

Logo image