The behaviour of solutions with singularities on a characteristic surface to linear partial differential equations in the complex domains (Q2367111)

The author extends his previous results on the existence of solutions to a linear partial differential equation with holomorphic coefficients, solutions which are singular on a characteristic (smooth) hypersurface [\textit{S. Ōuchi}, J. Fac. Sci., Univ. Tokyo, Sect. IA 32, 457-498 (1985; Zbl 0603.35014); idem. Proc. Japan Acad., Ser. A 65, No. 4, 102-105 (1989; Zbl 0706.35002); idem. Publ. Res. Inst. Math. Sci. 26, No. 5, 735- 783 (1990; Zbl 0751.35006)] to the following case: Consider the equation \((*)\) \(L(z, \partial_t) u(z) = f(z)\) defined in a neighborhood \(\Omega\) of \(z = 0\) in \(\mathbb{C}^{n + 1}\), let \(K = \{z \in \Omega, z_0 = 0\}\) (here \(z = (z_0, z_1, \ldots, z_n))\) be a connected smooth characteristic hypersurface, and suppose that \(f(z)\) and \(u(z)\) are holomorphic in a sector \(\Omega (\theta_0)\), whose edge is \(K\). Suppose moreover that \(u(z)\) has at most exponential growth on \(\Omega (\theta_0)\) and \(f(z)\) has some asymptotic expansion with bounds on \(\Omega (\theta_0)\). Under some conditions on the characteristic indices of \(L(z, \partial_z)\), \(u(z)\) has an asymptotic expansion of the same type as \(f\). This type of results gives, as particular cases, the existence of holomorphic solutions. Just to state the author's main results we need some definitions. If \(\Omega = \Omega_0 \times \Omega'\) an open polydisk in \(\mathbb{C}^{n + 1}\), \(\Omega_0 (a,b) = \{z \in \Omega_0 \widetilde \backslash \{0\}\), \(a < \arg z_0 < b\}\) (here \(\Omega_0 \widetilde \backslash \{0\}\) is the universal covering of \(\Omega_0 \backslash \{0\} )\), \(\Omega (a,b) = \Omega_0 (a,b) \times \Omega'\) and \(\Omega (- a,a)\) is simply denoted by \(\Omega (a)\). By \(\text{Asy}_K \Omega (a,b)\), where \(0 < K \leq + \infty\), we denote the class of all \(f \in {\mathcal O} (\Omega (a,b))\) having an asymptotic expansion \[ \left |f(z) - \sum^N_0 a_K(z') z^K_0 \right |\leq AB^N \Gamma \left( {N \over K + 1} \right) |z_0 |^N \] for any \(N\) as \(z_0 \to 0\) in any \(\Omega_0 (a',b') \times D\), with \(D \Subset \Omega'\), and the \(a_k(z') \in {\mathcal O} (\Omega')\) (if \(k = \infty\), \(f \in \text{Asy}_{+ \infty} \Omega (a,b)\) means that \(f \in {\mathcal O} (\Omega))\). By \(\widetilde {\mathcal M} - \text{Asy}_{(K)} (\Omega (a,b))\) the author means the class of all \(f \in {\mathcal O} (\Omega (a,b))\) having an asymptotic expansion with polar and logarithmic terms as \(z_0 \to 0\). The main results are that (under some conditions on the characteristic indices), if \(f \in \text{Asy} (\Omega (\theta_0))\) and \(u \in {\mathcal O}_{(\gamma) } (\Omega (\theta_0)\) then \(u \in \text{Asy}_{(k)} (\Omega (\theta_0))\) (respectively if \(f \in \widetilde u - \text{Asy}_{(K)} (\Omega (\theta_0))\) then \(u \in \widetilde {\mathcal M} - \text{Asy}_{ (k)} (\Omega (\theta_0))\), where \(\Omega (\theta_0)\) is the sector \(\{z \in \Omega \backslash \{z_0 = 0\}\), \(|\arg z_0 |< \theta_0\}\), \({\mathcal O}_{(\gamma)} (\Omega (a,b))\) is the set of holomorphic functions such that \[ \bigl |f(z) \bigr |\leq A_{\varepsilon, a',b',D} \exp \bigl( \varepsilon |z_0 |^{-K} \bigr) \text{ in } \Omega \bigl( (a',b'), D \bigr) \] for any \(a', b'\) with \(a < a' < b' < b\), any \(\varepsilon > 0\) and \(D \Subset \Omega\). The indices \(\gamma \) and \(K\) are related to the characteristic indices of \(L\). These very precise results are obtained via an integral representation of the solutions; for this the author investigates an integro-differential operator \({\mathcal L}_\alpha (z, \lambda, \xi, \partial_z, \partial_\xi\), \(\lambda t_0 - \lambda \partial_\lambda)\), \(0 < \alpha < 1\), derived from \(L(z, \partial_z)\). As an interesting technical tool, he introduces the \(k\)-Laplace transform and studies the relations between functions with \(k\)-asymptotic expansion and their \(k\)-Laplace transforms. It is impossible to give here more details; let me say that the proof of these results depends on a very careful analysis of the integral representation of the solutions, and that a lot of hard computations is involved.