\(F\)-signature of pairs and the asymptotic behavior of Frobenius splittings (Q1759376): Difference between revisions

Throughout \((R,\mathfrak{m},k)\) will denote a commutative Noetherian local ring with unit, and having characteristic \(p > 0\). Let \(F^e:R \to R\) denote the \(e\)-th iteration of the Frobenius endomorphism. In this article, the authors investigate the natural question: ``How many splittings does the \(e\)-th iterated Frobenius map have for \(e \gg 0\)?'' Let \(M\) be an \(R\)-module, viewing \(M\) as an \(R\)-module via restriction of scalars for \(F^e\) yields an \(R\)-module denoted by \(F^e_* M\). Also, let \(J \subseteq R\) be an \textit{\(F\)-compatible} ideal, i.e. an ideal such that for any \(e \in \mathbb{Z}_{\geq 0}\) and any \(\varphi \in \) Hom\(_R(F^e_*R,R)\) there is an induced map \(\varphi_J \in \Hom_{R/J}(F^e_*(R/J),R/J)\) (equivalently \(\varphi(F^e_*J) \subseteq J\) for all \(e\) and all \(\varphi\) as above). Then any \(F\)-splitting of \(R\) induces an \(F\)-splitting of \(R/J\). Although, not every \(F\)-splitting of \(R/J\) arises in this way in general, hence one has to enumerate only certain kinds of \(F\)-splittings, and this leads the authors to consider Cartier subalgebras of \(R\). Suppose \(R\) is \(F\)-finite (i.e. \(F_*R\) is a finite \(R\)-module), and for all \(e \in \mathbb{Z}_{\geq 0}\) let \(\mathcal{C}^R_e = \Hom_R(F^e_*R,R)\) denote the set of all \(p^{-e}\)-linear maps, and set \(\mathcal{C}^R = \bigoplus_{e \geq 0} \mathcal{C}^R_e\). A \textit{Cartier subalgebra on \(R\)} is a graded \(\mathbb{F}_p\)-subalgebra \(\mathcal{D}\) of \(\mathcal{C}^R\) such that \(\mathcal{C}_0 = \mathcal{D}_0 = R\) and \(\mathcal{D}_e \neq 0\) for some \(e \in \mathbb{Z}_{>0}\). Let \(a_e^\mathcal{D}\) denote the maximal number of \(F\)-splittings contained in \(\mathcal{D}_e\). The main result of the article is the following: Set \(\Gamma_\mathcal{D} = \{e \;| \;a_e^\mathcal{D} \neq 0\}\) and \(\alpha(R) = \text{log}_p[k:k^p]\). Then the limit \[ \displaystyle s(R,\mathcal{D}) = \lim_{e \in \Gamma_\mathcal{D} \to \infty} \frac{a_e^{\mathcal{D}}}{p^{e(d+\alpha(R))}} \] exists and is called the \(F\)-signature of \((R,\mathcal{D})\). Furthermore \(s(R,\mathcal{D})\) is positive if and only if \((R,\mathcal{D})\) is \(F\)-regular, i.e. for all \(c \in R\) not contained in any minimal prime, there exists \(e \in \mathbb{Z}_{\geq 0}\) and \(\varphi \in \mathcal{D}_e\) such that \(\varphi(F^e_*c) = 1\). \(F\)-signature of a ring was first introduced by \textit{C. Huneke} and \textit{G. J. Leuschke} in [Math.Ann. 324, No. 2, 391-404 (2002; Zbl 1007.13005)]. The existence part in this theorem recovers and generalizes the existence results of the third author [Invent. Math. 190, No. 3, 743--765 (2012; Zbl 1271.13012)]. The positivity part vastly generalizes the work of \textit{I. M. Aberbach} and \textit{G. J. Leuschke} [Math. Res. Lett. 10, No. 1, 51--56 (2003; Zbl 1070.13006)]. The authors also focus their attention on the \(F\)-splitting ratio (defined here in the more general setting of Cartier subalgebras): \[ \displaystyle r_F(R,\mathcal{D}) = \lim_{e \in \Gamma_\mathcal{D} \to \infty} \frac{a_e^{\mathcal{D}}}{p^{e(\dim(R/P_{\mathcal{D}})+\alpha(R))}} \] exists and it is positive when \(R\) is \(F\)-pure (i.e. when there is a surjective homomorphism \(\varphi \in \mathcal{D}_e\) for some \(e \in \mathbb{Z}_{\geq 0}\)). Here \(P_\mathcal{D}\) is the \(F\)-splitting prime of \(R\), which was first defined, together with the \(F\)-splitting ratio, in [Math. Z. 250, No. 4, 791--806 (2005; Zbl 1102.13001)] by \textit{I. M. Aberbach} and \textit{F. Enescu}. Finally, exploiting the crucial point that Cartier subalgebras facilitate changing settings from one ring to another, the authors derive some methods to compute explicitly the \(F\)-signature and the \(F\)-splitting ratio, in the spirit of the work of \textit{R. Fedder} [Trans. Ame. Math. Soc. 278, 461--480 (1983; Zbl 0519.13017)], and they provide explicit computations in some interesting examples.