Range characterization of the cosine transform on higher Grassmannians (Q1826890): Difference between revisions

Let \(Gr_{k,n}\) denote the Grassmann space of \(k\)-dimensional subspaces of~\(\mathbb R^n\). Using a fixed Euclidean structure, on defines the cosine \(\cos(E,F)\) of the angle between subspaces \(E\in Gr_{i,n}\) and \(F\in Gr_{j,n}\) as the volume distortion in \(E\) under orthogonal projection along~\(F\). The cosine transform \(T_{j,i}:C(Gr_{i,n})\to C(Gr_{j,n})\) is defined by \((T_{j,i}f)(E):=\int_{G_{i,n}}| \cos(E,F)| \,f(F)\,dF\), where \(dF\) denotes the Haar measure on the homogeneous (in fact, symmetric) space~\(Gr_{i,n}\). (The authors use a nonstandard order of the indices in their notation for \(T_{j,i}\).) The cosine transformation is invariant under the orthogonal group~\(O(n)\). Therefore, the image under \(T_{j,i}\) is a module for \(SO(n)\), and splits as a (multiplicity-free) sum of irreducible modules. Theorem~1.2 characterizes the highest weights that determine these irreducible components. The proof uses Radon transforms to reduce to the case \(i=j\). In this case, the cosine transform is interpreted as an intertwining operator between certain representations of \(GL(n,\mathbb R)\), and the image of \(T_{i,i}\) is recognized as an irreducible \(GL(n,\mathbb R)\)-module (Theorem~1.3). This assertion also has a non-Archimedean analogue (over a local field~\(F\), Theorem~2.1), stating also that one obtains an unramified \(GL(n,F)\)-module.