$h$-function, Hilbert-Kunz density function and Frobenius-Poincaré function

Abstract

Given ideals $I,J$ of a noetherian local ring $(R,\mathfrak{m})$ such that $I+J$ is $\mathfrak{m}$-primary and a finitely generated $R$-module $M$, we associate an invariant of $(M,R,I,J)$ called the $h$-function. Our results on $h$-functions allow extensions of the theories of Frobenius-Poincaré functions and Hilbert-Kunz density functions from the known graded case to the local case, answering a question of V.Trivedi. When $J$ is $\mathfrak{m}$-primary, we describe the support of the corresponding density function in terms of other invariants of $(R,I,J)$. We show that the support captures the $F$-threshold: $c^J(I)$, under mild assumptions, extending results of V. Trivedi and Watanabe. The $h$-function encodes Hilbert-Samuel, Hilbert-Kunz multiplicity and $F$-threshold of the ideal pair involved. Using this feature of $h$-functions, we provide an equivalent formulation of a conjecture of Huneke, Mustaţă, Takagi, Watanabe; recover a result of Smirnov and Betancourt; give a new proof of a result answering Watanabe-Yoshida's question comparing Hilbert-Kunz and Hilbert-Samuel multiplicity and establish lower bounds on $F$-thresholds. We also point out that a conjecture of Smirnov-Betancourt as stated is false and suggest a correction which we relate to the conjecture of Huneke et al. We develop the theory of $h$-functions in a more general setting which yields a density function for $F$-signature. A key to many results on $h$-functions is a `convexity technique' that we introduce, which in particular proves differentiability of Hilbert-Kunz density functions almost everywhere on $(0,\infty )$, thus contributing to another question of Trivedi.

v3: substantial changes: applications, results added, sec 7 of v2 subsumed into other sections, rewritten for better exposition