The Abel, Fourier and Radon transforms on symmetric spaces (Q855600)

The author shows some new results on the Radon, Abel, and Fourier transfoms on symmetric spaces \(X=G/K\) of noncompact type, and explains their relationship to known results. 1) \(X\)-ray inversion: For \(p\geq0\) let \(\Gamma_p\) be the set of all geodesics \(\gamma\) in \(X\) lying in some flat \(\ell\)-dimensional \((\ell=\text{rank}\;X>1)\) totally geodesic submanifold of \(X\) through the origin \(o=eK\) and \(d(o,\gamma)=p\). Then the \(X\)-ray transform \(\widehat f(\gamma)=\int_\gamma f(x)\,dm(x)\) on \(X\) can be inverted as \[ f(o)=-\frac{1}{\pi} \int_0^\infty\biggl(\frac{d}{dp}\int_{\Gamma_p}\widehat f(\gamma) \,d\omega_p(\gamma)\biggr)\frac{dp}{p}. \] 2) Support theorem on the horocycle transform: Let \(f\to\widehat f\) be the horocycle transform from \({\mathcal D}(X)\) to \({\mathcal D}(\Xi)\), \(\Xi=G/MN\). The support theorem is refined as follows. Suppose that \(f\in{\mathcal D}(X)\) satisfies \(\hat f(ka\cdot\xi_0)\) for \(k\in K\), \(a\in A^+\), \(| \log a| >R\), then it holds for \(a\in A\) with \(| \log a| >R\), and \(f(x)=0\) for \(d(x,0)>R\). 3) Identity for the Abel transform: Let \({\mathcal A}f=e^\rho\widehat f\) be the Abel transform for \(K\)-invariant functions on \(X\) and \(L\) the Fourier multiplier on \({\mathfrak a}^*\) corresponding to \(| c(\lambda)| ^{-2}\). Then for \(\phi\in{\mathcal D}_W(A)\), \(\psi\in{\mathcal E}_W(A)\), \[ {\mathcal A}^*(L\phi)=w{\mathcal A}^{-1}(\phi),\;{\mathcal A}^*(\phi*\psi)=w^{-1} {\mathcal A}^*(L\phi)\times{\mathcal A}^*, \] where \(F^*\) denotes the Euclidean Fourier transform of \(F\), \(w=| W| \), and \(*\), \(\times\) the convolutions on \(A\) and \(X\) respectively. 4) Analog of Riemann-Lebesgue's Lemma: Let \(\widetilde f(\lambda,b)\), \((\lambda,b)\in{\mathfrak a}_{C}^*\times B\), \(B=K/M\), be the Fourier transform on \(X\). Let \(C(\rho)\) be the convex hull of \(\{s\phi;s\in W\}\) in \({\mathfrak a}^*\). Suppose \(f\in L^1(B)\). Then there exists \(B'\subset B\) such that \(B-B'\) is of measure \(0\) and for each \(b\in B'\), \({\widetilde f}(\lambda,b)\) is defined for \(\lambda\in{\mathfrak a}^*+iC(\rho)\) and \(\lim_{\xi\to\infty}{\widetilde f}(\xi+i\eta,b)=0\) uniformly for \(\eta\in C(\rho)\). These subjects have been investigated by several authors, and there are many related results. From this paper we can understand their historical relation, and find a future direction in our research.