Pseudoconcavity of flag domains: the method of supporting cycles (Q2324864)

Let \(G\) be a simple complex Lie group, \(P\subset G\) a parabolic subgroup and \(Z=G/P\) the corresponding flag manifold. If \(G_0\) is a non-compact real form of \(G\), then \(G_0\) acts on \(Z\) with finitely many orbits. Some of these are open orbits and are called flag domains. In this paper the complex function theoretic properties of flags domains \(D\) are investigated, under the assumption that \(D\not= Z\). If \(\mathcal{O}(D)\not =\mathbf{C}\), then \(G_0\) is of Hermitian type, \(D\) is holomorphically convex, and admits a holomorphic fibration over a bounded symmetric domain with fiber a compact submanifold \(C\) of \(D\), canonically associated to the group action at hand. Here the authors are interested in the case when \(\mathcal{O}(D) =\mathbf{C}\) and therefore \(D\) cannot be holomorphically convex. Their main result is the following theorem. Theorem 1. A flag domain is pseudoconcave if and only if it is not holomorphically convex. Recall that a complex manifold \(X\) is pseudoconcave of degree \(d\) if it contains a relatively compact open set \(Y\) so that for every \(p\in\partial Y\) there exists a holomorphic map \(F\colon \Delta^d(r)\to X\) with boundary \(F( \Delta^d(r))\) contained in \(Y\) (here \(\Delta^d(r)\) denotes a complex \(d\)-dimensional polydisk of radius \(r\)). The relatively compact open set \(Y\) in \(D\), which displays the pseudoconcavity of \(D\), is (roughly) obtained as follows. Let \(K=K_0^{\mathbf C}\subset G\) be the complexification of a maximal compact subgroup of \(G_0\). Then a flag domain contains a unique \(K_0\)-orbit \(C_0\) which is also a \(K\)-orbit. In particular, \(C_0\) is a compact complex submanifold of \(D\), referred to as the base cycle in \(D\). Denote \(\mathcal{M}_Z:=G\cdot [C_0]\) and \(\mathcal{M}_D\) the open subset of \(\mathcal{M}_Z\) defined as the connected component containing \(C_0\). Define \(\mathcal{H}_D=\{ (p,[C])\in Z\times\mathcal{M}_D~:~p\in C\}\). The projections \(\mu\colon\mathcal{H}_D\to D\) and \(\nu\colon\mathcal{H}_D\to\mathcal{M}_D\) are holomorphic \(G_0\)-equivariant and have open image. Set \(Y:=\mu(\nu^{-1}(B))\), where \(B\) denotes any relatively compact domain with smooth boundary in \(\mathcal{M}_D\) containing \([C_0]\). Then set \(Y\subset D\) has the property that every boundary point \(p_0\in\partial Y\) lies in at least one cycle \(C\), which is contained in the closure \(cl(Y)\) and is referred to as a supporting cycle at \(p_0\). A crucial fact in the proof of Theorem 1 is that is that for every supporting cycle, the intersection \(C\cap Y\) is open and dense in \(C\). This is the content of Theorem 2 in the paper (and of its reformulations in some specific cases). For an estimate of the degree of pseudoconcavity of \(D\), observe that every supporting cycle \(C\) at \(p_0\) displays a certain degree \(c(C,p_0)\) of pseudoconcavity of \(Y\) at \(p_0\). Then then supporting cycle pseudoconcavity of \(Y\) at \(p_0\) is the maximum \(c(p_0)\) of the \(c(C,p_0)\) as \(C\) runs through all supporting cycles containing \(p_0\) and the pseudoconcavity of \(Y\) is defined to be the minimum \(c(Y)\) of the \(c(p_0)\) as \(p_0\) runs through \(\partial Y\). In the equal-rank case, the authors are able to estimate the degree of pseudoconcavity of \(D\) by below in terms of a root theoretic invariant \(d_{ma}\) which only depends on the closed \(K\)-orbit in \(Z\). In the final part of the paper, lower bounds for the pseudoconcavity of flag domains in Grassmannians in the equal-rank cases of the mixed signature unitary and real symplectic groups are explicitly computed.