On the tautological ring of \(\mathcal{M}_ g\) (Q1899741)

Fix an integer \(g\geq 2\) and denote by \({\mathcal C}^n_g\) the moduli space of tuples \((C, x_1, \dots, x_n)\), where \(C\) is a complex smooth connected projective curve of genus \(g\) and \(x_1, \dots, x_n\) are (not necessarily distinct) points of \(C\); we also write \({\mathcal M}_g\) for \({\mathcal C}^0_g\). Forgetting the \(i\)-th point defines a morphism \({\mathcal C}^n_g\to {\mathcal C}_g^{n-1}\) whose relatively dualizing sheaf is denoted by \(\omega_i\) \((i= 1, \dots, n)\). We write \(K_i\) for the first Chern class of \(\omega_i\), considered as an element of the Chow group \(A^1 ({\mathcal C}^n_g)\) (with rational coefficients); for \(n=1\) we also write \(K\). The direct image of \(K^{d+1}\) in \(A^d ({\mathcal M}_g)\) is the Mumford-Morita-Miller tautological class \(\kappa_d\). \textit{D. Mumford} showed in his fundamental paper [in: Arithmetic and geometry, Pap. dedic. Shafarevich, Vol. II, Prog. Math. 36, 271-328 (1983; Zbl 0554.14008)] that the \(\mathbb{Q}\)-subalgebra of \(A^\bullet ({\mathcal M}_g)\) generated by these classes (the tautological ring of \({\mathcal M}_g\)) is already generated by \(\kappa_1, \dots, \kappa_{g-2}\). On the basis of many calculations \textit{Carel Faber} has made the intriguing conjecture that this ring has the formal properties of the even-dimensional cohomology ring of the projective manifold of dimension \(g-2\), i.e., satisfies Poincaré duality and a Lefschetz decomposition. It is natural to define the tautological ring of \({\mathcal C}^n_g\) as the \(\mathbb{Q}\)-subalgebra of \(A^\bullet ({\mathcal C}^n_g)\) generated by \(K_1, \dots, K_n\), the (pullbacks of the) classes \(\kappa_i\) and the classes of the diagonal divisors on \({\mathcal C}^n_g\) defined by \(x_i= x_j\) \((1\leq i< j\leq n)\). -- The theorem below supports Faber's conjecture (take \(n=0\)) and answers two questions of \textit{S. Morita} [``Problems on the structure of the mapping class group of surfaces and the topology of the moduli space of curves'', in: Topology, geometry and field theory, 101-110 (1994)] affirmatively (take \(n=1\)). Theorem. Any element of degree \(d\) of the tautological ring of \({\mathcal C}^n_g\) is a linear combination of the classes of the irreducible components of the locus parametrizing tuples \((C, x_1, \dots, x_n)\) admitting a finite morphism \(C\to \mathbb{P}^1\) of degree \(\leq 2g-2 +n\) such that the fiber over 0 (resp. \(\infty\)) has at most \(g+ n- d- 1\) points (resp. is a singleton) and \(\{x_1, \dots, x_n\}\) is contained in the union of these two fibers. (Hence such a class is zero when \(d> g+ n-2\).) All such classes of degree \(g+ n-2\) are proportional to the class of the locus \(H^n_g\) parametrizing tuples \((C, x_1, \dots, x_n)\) with \(C\) hyperelliptic and \(x_1= \dots= x_n\) a Weierstraß point.