Invariant hypersurfaces in holomorphic dynamics (Q2275738)

The author proves that if the dominant endomorphism \(g\) of a compact complex manifold \(M\) has a large enough number of totally invariant hypersurfaces \(W_i\subset M\), then there exists a non constant meromorphic function \(\Phi\) such that \(\Phi\circ g=\alpha\Phi\), \(\alpha=\text{constant}\). The number \(k\) of the required totally invariant hypersurfaces satisfies the inequality \(k>\dim M+\dim H^1(M,\Omega_M^1)\), where \(\Omega_M^1\) is the sheaf of holomorphic \(1\)-forms and \(H^1(M,\Omega_M^1)\) is the 1st Čech cohomology group. Modifying appropriately the lower bound for \(k\), the author also extends the above assertion to the case of meromorphic transformations \(g\) of \(M\). The idea of the proof is the following. Let \(l\geq k\) be the number of irreducible components of the totally invariant hypersurface \(W=\bigcup_iW_i\). Let \(Z\) be the \(l\)-dimensional vector space generated by the logarithmic \(1\)-forms \(\sigma_W=d\log(\phi)\), where \(\phi\in{\mathcal O}({\mathcal U})\) is a local equation of \(W\). Let \(Z_0=Z\cap K_M^1\) be the intersection of \(Z\) with the kernel \(K_M^1\) of the morphism mapping the vector space of global sections of the quotient sheaf \({\mathcal M}_M^1/\Omega_M^1\) (\({\mathcal M}_M^1\) is the sheaf of meromorphic \(1\)-forms) into \(H^1(M,\Omega_M^1)\), and let \(h^{1,1}\) be the codimension of \(Z_0\) in \(Z\). The author shows that there exist \(m\geq l-h^{1,1}(M)\) linearly independent logarithmic \(1\)-forms \(\nu_j\), \(j=1,\dots,m\), with poles along \(W\) and satisfying the eigenvector condition \(g^*\nu_j=\lambda_j\nu_j\). However, if \(m>\dim(M)\), then the \(\nu_j\) are linearly dependent as vectors of the cotangent space \(T_x^*M\) at all points \(x\in M\). Thus, permuting indices if necessary, there is the number \(n\) such that \(2\leq n\leq\dim M <m\), and \(\nu_1,\ldots,\nu_n\) are linearly independent at the generic point \(x\) of \(M\) while \(\nu_{n+1}=\sum_{j=1}^{n}b_j\nu_j\). Finally, the meromorphic function \(\Phi\) of the author's interest appears as a nontrivial coefficient \(b_j\) in the above expansion of a \(1\)-form \(\nu_{n+1}\).