Multiplier ideals and klt singularities via (derived) splittings

This paper characterizes multiplier ideals on normal schemes over $\mathbb{Q}$ via maps from pushforwards of dualizing sheaves under regular alterations, thereby providing a derived splinter characterization of klt singularities and an analogous description of test ideals in characteristic $p>2$.

The Gist

Imagine you are an architect inspecting a building. Some buildings are perfectly smooth and straight (smooth manifolds), but many have cracks, corners, or uneven surfaces. In mathematics, these "cracks" are called singularities.

This paper is about a new way to measure how "bad" these cracks are in a specific type of mathematical structure called a scheme (which you can think of as a generalized geometric shape or space). The author, Peter McDonald, introduces a clever method to measure these imperfections using a concept called "splitting."

Here is the breakdown of the paper's ideas using simple analogies:

1. The Problem: Measuring the "Cracks"

In the world of algebraic geometry, mathematicians have long used a tool called a Multiplier Ideal to measure the severity of singularities.

• The Old Way: To measure a crack, you usually have to "smooth it out" first. You take a complex shape, replace it with a perfectly smooth version (a "resolution"), do your calculations there, and then try to translate the results back to the original cracked shape. It's like trying to measure the roughness of a bumpy road by first paving it with perfect asphalt, measuring the asphalt, and then guessing how bumpy the original road was.
• The Goal: The author wants a way to measure the cracks directly, without needing to smooth the whole road out first.

2. The New Tool: "Regular Alterations" and "Splitting"

Instead of smoothing the road, the author suggests looking at the road through a specific kind of lens called a Regular Alteration.

• The Analogy: Imagine you have a crumpled piece of paper (the bad shape). Instead of trying to flatten it perfectly, you project a shadow of it onto a wall using a very specific, high-quality projector. This projection (the "alteration") is smooth, but it comes from the crumpled paper.
• The Map: The author looks at maps (functions) that go from this smooth projection back to the original crumpled paper.
• The "Split": The core idea is splitting. Imagine you have a rope (the smooth projection) and you try to pull a knot (the original shape) out of it. If you can pull the knot out cleanly so that the rope snaps back into two distinct, independent pieces, we say the map "splits."
  • If the map splits, it means the original shape isn't too broken. It has a certain "goodness" or "tame" quality.
  • If the map doesn't split, the shape is too messy.

3. The Main Discovery: The "Sum of Images"

The paper proves a big theorem: The "Multiplier Ideal" (the measure of badness) is exactly the collection of all the things you can "pull out" (split) from these smooth projections.

Think of it like a sieve:

• You pour a mixture of "good" and "bad" mathematical elements through a sieve made of all possible smooth projections.
• The things that fall through (the "images" of the maps) are the "good" elements.
• The author proves that the set of all these "good" elements is exactly the Multiplier Ideal.
• Why this matters: It gives a new, direct definition. You don't need to find the "perfect" smooth version of the shape. You just need to check if you can "split" the shape through any smooth projection. If you can, the shape is "good" (specifically, it has klt type, which is a fancy way of saying "mildly singular").

4. The "Derived" Connection

The paper also touches on Derived Splinters.

• The Analogy: Imagine a complex machine with many gears (the "derived category"). A "splinter" is a machine where, no matter how you take it apart and put it back together, you can always find a way to reassemble the original core perfectly.
• The author shows that if a shape has "klt type" (mild cracks), it behaves like a "derived splinter." It's a robust structure that can be broken down and reassembled without losing its core identity. This connects the "crack measurement" directly to the "robustness" of the shape.

5. The "Positive Characteristic" Twist (The Hot Version)

Mathematics often has two versions: one for "cold" numbers (like 0, 1, 2...) and one for "hot" numbers (mathematics over fields with a prime number $p$, like 3, 5, 7...).

• In the "hot" world, the standard tools for smoothing shapes don't exist.
• However, there is a tool called the Frobenius map (which is like a magical "squaring" or "cubing" operation that only works in the hot world).
• The author shows that the same "splitting" logic works here too, but instead of using smooth projections, you use this magical Frobenius operation.
• The Result: They define a "Test Ideal" (the hot version of the Multiplier Ideal) using the same logic: "What can you pull out (split) using these Frobenius maps?"

Summary in One Sentence

This paper proves that you can measure how "broken" a geometric shape is by checking if you can "pull it apart" (split) through various smooth projections; if you can pull it apart cleanly, the shape is actually quite "healthy" (has mild singularities), and this rule works in both standard math and the specialized "hot" math of prime numbers.

The Takeaway:
Instead of trying to fix the broken shape to measure it, the author says: "Just try to pull it apart. If it splits cleanly, it's not that broken after all."

Technical Summary

Here is a detailed technical summary of the paper "Multiplier Ideals and KLT Singularities via (Derived) Splittings" by Peter M. McDonald.

1. Problem Statement

The paper addresses the characterization of multiplier ideals and Kawamata log terminal (klt) singularities in algebraic geometry, specifically for normal, excellent, Noetherian schemes over $\text{Spec}(\mathbb{Q})$ (characteristic 0) and certain rings in positive characteristic ($p > 2$).

• Context: Multiplier ideals, denoted $\mathcal{J}(X, \Delta)$, are fundamental tools for measuring the severity of singularities in pairs $(X, \Delta)$. The paper focuses on the multiplier ideal of the scheme itself, $\mathcal{J}(X)$, introduced by de Fernex and Hacon, which does not require the choice of a boundary divisor $\Delta$. A scheme $X$ has klt type if $\mathcal{J}(X) = \mathcal{O}_X$.
• Gap: While rational singularities have a well-known characterization via derived splinters (a result by Bhatt and Kovács stating that $X$ has rational singularities iff $\mathcal{O}_X \to R\pi_*\mathcal{O}_Y$ splits for all proper surjective $\pi$), a similar "splitting" characterization for klt singularities was lacking. The paper seeks to bridge this gap by relating multiplier ideals to maps involving canonical sheaves and derived splittings.

2. Methodology

The author employs a combination of birational geometry, Grothendieck duality, and trace maps to reframe the definition of multiplier ideals.

• Regular Alterations: Instead of restricting to log resolutions (which are birational), the paper utilizes regular alterations ($\pi: Y \to X$), which are surjective, proper, generically finite maps where $Y$ is non-singular. By Stein factorization, any such map factors into a resolution $\tau: Y \to Z$ and a finite map $\rho: Z \to X$.
• Trace Maps and Duality: The core technique involves analyzing maps of the form $\pi_* \omega_Y \to \mathcal{O}_X$. Using Grothendieck duality, the author relates the image of these maps to the trace of multiplier submodules on finite covers.
• Key Lemma Construction: The proof relies on a crucial lemma (Lemma 1.2/2.17) showing that the trace of a multiplier submodule on a finite cover $S \to R$ is contained within the multiplier ideal of the base ring $R$. This is achieved by constructing specific divisors and utilizing Galois invariance to ensure the trace lands in the correct ideal.
• Positive Characteristic Adaptation: For characteristic $p > 2$, the paper adapts these methods using Frobenius morphisms. It replaces multiplier ideals with test ideals ($\tau(R)$) and multiplier submodules with parameter test submodules ($\tau(\omega_R)$). A critical step involves constructing quasi-Gorenstein finite covers (Lemma 3.6) to simplify the trace calculations, a construction that requires $p \neq 2$ to ensure reducedness of general divisors.

3. Key Contributions and Results

A. Characterization of Multiplier Ideals (Characteristic 0)

The main result (Theorem 1.1 / Theorem 2.20) provides an alternate realization of the multiplier ideal $\mathcal{J}(X, I)$ (where $I$ is a formal combination of ideal sheaves) as a sum of images of evaluation maps:
$$\mathcal{J}(X, I) = \sum_{\pi: Y \to X} \text{Im}\left( \text{Hom}_X(\pi_* \omega_Y, \mathcal{O}_X) \otimes_{\mathcal{O}_X} \pi_* \omega_Y(-E_Y) \to \mathcal{O}_X \right)$$
Here, the sum ranges over all log regular alterations $\pi: Y \to X$, and $E_Y$ is the divisor associated with the ideal sheaves. This formulation interprets the multiplier ideal as the locus of functions that can be "lifted" and "traced" back from regular alterations.

B. Derived Splinter Characterization of KLT Singularities

As a direct corollary (Corollary 1.4 / Corollary 2.21), the paper establishes a derived splinter characterization for klt singularities, analogous to the rational singularity characterization:

• Theorem: A pair $(X, I)$ has klt type if and only if for all sufficiently large regular alterations $\pi: Y \to X$, the natural map $\mathcal{O}_X \to R\pi_*\mathcal{O}_Y$ splits in the derived category, and this splitting locally factors through $R\pi_*\omega_Y(-E_Y)$.
• Significance: This distinguishes klt singularities from general rational singularities by adding the condition that the splitting must factor through the twisted canonical sheaf of the alteration.

C. Characteristic $p > 2$ Analog

The paper extends these results to positive characteristic (Proposition 1.5 / Proposition 3.8), characterizing the test ideal $\tau(R)$:
$$\tau(R) = \sum_{R \subset S} \text{Im}\left( \text{Hom}_R(\omega_S, R) \otimes_R \tau(\omega_S) \to R \right)$$
where the sum is over finite extensions $R \subset S$ contained in specific algebraic closures.

• Result: This leads to a characterization of strongly F-regular rings (the positive characteristic analog of klt) as rings where $R$ is a summand of $\tau(\omega_S)$ for appropriate finite extensions.

4. Significance and Impact

• Unification of Singularities: The work unifies the study of multiplier ideals (char 0) and test ideals (char $p$) under a common framework of splittings and trace maps.
• New Characterization of KLT: It provides the first derived splinter characterization specifically for klt singularities, refining the existing characterization for rational singularities. This offers a new tool for detecting klt singularities without explicitly computing resolutions or boundary divisors.
• Methodological Innovation: The use of regular alterations (rather than just resolutions) and the explicit construction of quasi-Gorenstein covers in characteristic $p$ provide robust technical tools for future research in singularity theory.
• Local Algebra Applications: The results are formulated in a way that is directly applicable to local algebra questions, particularly regarding the structure of test ideals and multiplier ideals in local rings.

In summary, McDonald's paper successfully translates the geometric definition of multiplier ideals into a homological property involving derived splittings, thereby offering a powerful new perspective on the nature of klt and strongly F-regular singularities.