On octonionic regular functions and the Szegő projection on the octonionic Heisenberg group (Q2513298)

The title gives explicitly the aim of the paper. Let \(\mathcal{H}\) denote the octonionic Heisenberg group, an octonionic regular operator on \(\mathcal{H}\) is defined as \(Zf=\sum^{7}_{j=0}e_j Y_j f\) for \(f\in C^{1}(\mathcal{H},\mathbb{O})\), where \(\mathbb{O}\) stands for the set of octonionic numbers and \(Y_j\) are left invariant vector fields on \(\mathcal{H}\). A distribution on \(\mathcal{H}\) is called octonionic regular (OR) if \(Zf=0\) in the distributional sense. An analog of the Hardy space is \(\mathscr{H}(\mathcal{H},\mathbb{O}):=\left\{f\in L^{2}(\mathcal{H},\mathbb{O});\; Zf=0\right\}\), which is proved to be a closed subspace of \(L^{2}(\mathcal{H},\mathbb{O})\) and also a real Hilbert space. Identifying \(\mathbb{O}\) with \(\mathbb{R}^{8}\) allows to consider \(Z\) as a matrix differential operator. It generates the matrix Szegő projection operator \(P:L^{2}(\mathcal{H},\mathbb{R}^{8})\rightarrow \mathscr{H}(\mathcal{H},\mathbb{R}^{8}).\) As a matter of fact, such a transition to matrices changes essentially the initial setting, since the matrix multiplication is associative whereas the octonionic one is not. But in return much more information can be obtained. For instance, the authors introduce the associated Laplace operator \(\square:=Z^{*} Z\), where \(Z^{*}\) is the formal adjoint of \(Z\) and both operators are considered as matrices; this is impossible if the operators are taken in the octonionic form. Section 2 gives a wide description of the properties of \(\mathbb{O}\) and \(\mathcal{H}\) with a special emphasis on the matrix form. Section 3 deals with the OR operator together with the associated Laplace operator, again representing them explicitly as matrices. Section 4 is entitled ``The Fourier transform on the octonionic Heisenberg group \(\mathcal{H}\) and representations of \(\square\)'' and the last Section 5 is entitled ``The \(L^{2}\) projection on the null space of \(\square\)''. In particular, the Szegő kernel is given by the following theorem. Theorem 1.1. The orthonormal projection \(P\) of \(L^{2}(\mathcal{H},\mathbb{R}^{8})\) onto \(\mathscr{H}(\mathcal{H},\mathbb{R}^{8})\) maps a \(\mathbb{R}^{8}\)-valued functions \(f\) to \(f\ast Re\, (S)\), where \(S\) is an \(\text{End}(\mathbb{C}^{8})\)-valued distribution on \(\mathcal{H}\) and coincides with the function \[ \frac{1}{(2\pi)^{11}}\int_{S^{6}}\frac{P^{\lambda+i\varepsilon \hat{s}}_{+}}{\omega^{11}(y,s;\lambda+i\varepsilon\hat{s})}d\lambda, \] for \((y,s)\neq (0,0),\) where \(\omega(y,s;\lambda)\) is given by \[ \omega(y,s;\lambda):=\frac{1}{4}\left|y\right|^{2}-i\lambda(s), \] and where \(\lambda(s)=\sum^{7}_{\beta=1}\lambda_\beta s_\beta\), \(\lambda, s\in\mathbb{R}^{7}\), \(y\in\mathbb{R}^{8}\), and \(\hat{s}\) denotes the unit vector \(s/\left|s\right|\) if \(s\neq 0\), and \(0\) if \(s=0\). Here \(0<\varepsilon<\varepsilon_0\) is sufficiently small. The proof of the theorem is rather involved and contains a number of fine details.