Lipschitz estimates for the \(\overline\partial\)-equation on the minimal ball (Q5954600)

The authors give an example of convex circular and non-piecewise smooth domain, with defining function that is not differentiable for which the \(\overline\partial\)-equation has the Lipschitz \(\frac 12\)-estimate (it is well known that if the domain is smooth convex of type \(m\), the sharp gain of smoothness is \(\frac 1m\); the particular case of smooth strictly pseudoconvex corresponds to \(\frac 12)\). Namely, if we note \(z\circ w=\sum^n_{j=1} z_jw_j\) for \(z,w\in \mathbb{C}^n\), let \(B_*=\{z\in \mathbb{C}^n\), \(p(z)=|z|^2+|z \circ z_{k_1}\}\), the minimal ball. This domain has \(B\)-regular boundary in the sense of \textit{N. Sibony} [Duke Math. J. 55, 299-319 (1987; Zbl 0622.32016)], and the singular part of its boundary is \(\partial B_*\cap V\), where \(V=\{z\in \mathbb{C}^n\setminus \{0\}\), \(z\bullet z=0\}\); the regular part of the boundary consists of all strictly pseudoconvex points. Let \({\mathcal A}^\infty_{0,1} (\mathbb{B}_*)\) be the space of \(\overline\partial\)-closed \((0,1)\) form on \(B_*\), with \(L^\infty\)-coefficients, endowed with the norm \(\|\cdot \|_\infty\). Let \[ \Lambda_{\widetilde 1\over 2}(\mathbb{B}_*)= \left\{f:\|f \|_\infty +\sup_{\substack{ z,z+h \in\mathbb{B}_*\\ 0<|h|<\frac 12}}{\bigl|f(z+h)- f(z)\bigr |\over|h|^{1\over 2}|\log|h||} \equiv \|f\|_{\Lambda_{\widetilde 1\over 2}}< \infty\right\}. \] The main result of this work is: There exists finite constant \(C\) and an explicitly defined linear integral operator \[ T:{\mathcal A}^\infty_{(0,1)} (\mathbb{B}_*)\to \Lambda_{\widetilde 1\over 2}(\mathbb{B}_*) \] such that \(\overline\partial Tf=f\), and \(\|Tf\|_{\Lambda_{\widetilde 1\over 2}}\leq C\|f\|_\infty\) for any \(f \in{\mathcal A}^\infty_{(0,1)} (\mathbb{B}_*)\). The \(\overline\partial\)-solving operator has the form \[ (Tf)(z)= \int_{\mathbb{B}_*} K[f](z,\xi){d\xi\Lambda d \overline \xi\over|\xi\cdot \xi|} +\int_{\partial \mathbb{B}_*}S[f](z,\xi) {d\theta (\xi)\over |\xi \cdot\xi|} \] where \(K\) and \(S\) are explicit. Here \(\theta\) is the Lebesgue measure on \(\partial\mathbb{B}_*\setminus V\). The idea of proof is to solve the \(\overline\partial\)-equation on an auxiliary manifold \(M\), that is a ramified covering of degree 2 of \(\mathbb{B}_* \setminus \{0\}\). If \(\pi:\overline \mathbb{M}\to \overline B_*\setminus\{0\}\) is the proper map defined by \(\pi(z_1,\dots, z_n,z_{n+1})= (z_1,\dots,z_n)\), and \(T\) the solution-operator for the \(\overline\partial\)-equation in \(\mathbb{M}\), then \(T(\pi^*f) (z, z_{n+1}) =T(\pi^* f|(z,\dots,z_{n+1}))\) and we can define \((Tf)(z)= T(\pi^* f)(z,z_{n+1})\). The manifold \(\mathbb{M}\) is defined as \(\{z\in \mathbb{C}^{n+1} \setminus \{0\}\), \(z\bullet z=0\), \(|z|<1\}\). In order to solve the \(\overline\partial\)-equation in \(\mathbb{M}\), one approximates \(\mathbb{M}\) by some regular varieties, that are complete intersections, and one can apply to these the result of \textit{B. Berndtsson} [Math. Ann. 263, 399-418 (1983; Zbl 0499.32013)].