Singular invariant trilinear forms and covariant (bi-)differential operators under the conformal group (Q425723): Difference between revisions

From the introduction: ``Let \(S=S^{n-1}\) be the unit sphere in a Euclidean space of dimension \(n\) (\(n\geqslant3\)). Let \(G=SO_0(1,n)\) be the connected component of the Lorentz group, acting on \(S\) by conformal transformations. Let \((\pi_\lambda)_{\lambda\in\mathbb C}\) be the non-unitary spherical principal series of \(G\), realized on \(\mathcal C^\infty(S)\). Let \(\lambda_1, \lambda_2, \lambda_3\) be three complex numbers. A con\-ti\-nu\-o\-us trilinear form \(\mathcal K\) on \(\mathcal C^\infty(S)\times\mathcal C^\infty(S)\times\mathcal C^\infty(S)\) is said to be conformally invariant with respect to \(\pi_{\lambda_1}\otimes\pi_{\lambda_2}\otimes\pi_{\lambda_3}\), if, for any three functions \(f_1,f_2,f_3\in\mathcal C^\infty(S)\), \[ \mathcal K\left(\pi_{\lambda_1}(g)f_1,\pi_{\lambda_2}(g)f_2,\pi_{\lambda_3}(g)f_3\right)=\mathcal K(f_1,f_2,f_3) \] for any \(g\) in \(G\). Generally, for \(\lambda=(\lambda_1,\lambda_2,\lambda_3)\) in \(\mathbb C^3\), there is such a unique (up to a multiple) invariant trilinear form. Viewing the trilinear form as a distribution on \(S\times S\times S\), it has a smooth density on the open set \[ \left\{(x_1,x_2,x_3)\in S\times S\times S;,\quad x_1\neq x_2, \quad x_2\neq x_3,\quad x_3\neq x_1\right\}, \] given by \[ k_{\pmb \alpha}(x_1,x_2,x_3)=|x_1-x_2|^{\alpha_3}|x_2-x_3|^{\alpha_1}|x_3-x_1|^{\alpha_2}, \] where \({\pmb \alpha}=(\alpha_1,\alpha_2,\alpha_3)\) is a triple of complex numbers, uniquely determined by \(\lambda\). The corresponding distribution \(\mathcal K_\lambda\) is defined by meromorphic continuation, and has simple poles along a certain plane in \(\mathbb C^3\). A pole \({\pmb \alpha}\) is said to be of type I if there exist \(j\in\{1,2,3\}\) and an integer \(k_j\) such that \(\alpha_j=-(n-1)-2k_j\), and of type II if there exists an integer \(k\) such that \(\alpha_1+\alpha_2+\alpha_3=-2(n-1)-2k\). The present paper deals with residues of this meromorphic family. The residues are distributions supported on proper submanifolds of \(S\times S\times S\). The computation of the residues of type I follows the classical approach to the meromorphic continuation of \(f^s\) as used by Gelfand and Shilov, which uses appropriate changes of variables (typically polar coordinates) leading to a meromorphic continuation problem for an elementary function of one variable. The expression of the residues involves conformally covariant differential operators on \(S\). The computation of the residues of type II is more difficult. First, the authors replace the compact re\-a\-li\-za\-tion of the representations \(\pi_\lambda\) by their noncompact realization, which transfers the problem to the flat setting (\(S\) being replaced by a Euclidean space of the same dimension). Next, a Bernstein-Sato identity is used in the proof of the general meromorphic continuation theorem for \(f^s\). After computing the residue for the ``first'' plane of poles by elementary techniques, the Bernstein-Sato identity allows, by induction, to compute the residues along the other planes of poles. The expression of the residues involves conformally covariant bidifferential operators on~ \(S\). The computation of the residues at the poles of type II leads to a new formula for conformally covariant bidifferential operators (generally, there is, up to a constant, only one such operator) which might be of independent interest.'' The Bernstein-Sato identities are known to exist, but are in general very difficult to find. In the case at hand, the authors explicitly construct such identities. The proof uses in a crucial way the properties of the classical Knapp-Stein intertwining operators and the generic uniqueness theorem for invariant trilinear forms.