On a problem of Bremermann concerning Runge domains (Q863418)

The author answers negatively to the question whether a Stein domain \(D\) in \(\mathbb C^{n}\) such that for each complex line \(l\) in \(\mathbb C^{n}\) \(l \backslash D\) is connected is itself Runge in \(\mathbb C^{n}\). He constructs first an example of a bounded strictly pseudoconvex domain \(W\) in \(\mathbb C^2\) with smooth, real analytic boundary which is Runge, but its closure is not holomorphically convex. By using a certain small perturbation of a strictly plurisubharmonic function used in the first step, he constructs a \(W\) as above but having moreover the following geometric property: for every complex line \(l\) the set of points where \(l\) is tangent to \( \partial W\) is at most finite. Finally, he finds a suitable neighbourhood of \(\overline {W }\) which is the counterexample desired.