Propagation des singularités des courants positifs fermés (Q1059183)

Let T be a closed positive current in a bounded Runge open subset \(\Omega \subset {\mathbb{C}}^ n\). One studies sufficient conditions to be verified by the mass density of T in order that there exists a global extension of T to \({\mathbb{C}}^ n\). Assume that T is of bidegree (1,1) (resp. (q,q), \(q>1)\), that the cohomology class of T in \(\Omega\) is 0 and that the trace measure \(\sigma_ T(z,r)\) of the ball B(z,r) satisfies \((*)\quad \sup_{z\in K}\int^{\epsilon}_{0}(\sigma_ T(z,r)/r^{2n- 1})dr<+\infty\) (resp. \(\int^{\epsilon}_{0}(\sup_{z\in K} \sigma_ T(z,r)^{1/2}/r^ n)dr<+\infty)\) for every \(K\subset \subset \Omega\) and \(\epsilon >0\) small enough. Then for every \(\omega_ 1\subset \subset \omega_ 2\subset \subset \Omega\), one proves the existence of a closed positive current \(\Theta\) in \({\mathbb{C}}^ n\) which is equal to T in \(\omega_ 1\) and \(C^{\infty}\)-smooth in \({\mathbb{C}}^ n\setminus {\bar \omega}_ 2\). Conversely, using Skoda - El Mir structure theorems for closed \(\geq 0 currents\), one constructs various counterexamples to the extension problem. When T has bidegree (1,1), one shows that (*) is essentially the best possible condition allowing extension, whereas in bidegree (q,q) there exists a current T such that \(\sup_{z\in K} \sigma_ T(z,r)\leq C_ Kr^{2n-2q-1}\), whose singularities propagate up to \(\partial \Omega\) along (2n-2q-1)-CR submanifolds. This last example rests upon the existence of totally real complete pluripolar (n- 1)-submanifolds in \({\mathbb{C}}^ n\), a result due to Diederich-Fornaess which one proves again in a new and simpler way.