🔢 mathematics

A characterization of virtually free actions via arc spaces and its application to the lower semi-continuity conjecture

This paper introduces a characterization of virtually free actions via arc spaces to establish a necessary and sufficient condition for the precise inversion of adjunction conjecture in arbitrary hyperquotient singularities and to unconditionally prove the lower semi-continuity conjecture for the same setting.

Original authors: Yusuke Nakamura, Kohsuke Shibata

Published 2026-06-24

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Yusuke Nakamura, Kohsuke Shibata

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to measure the "roughness" or "sharpness" of a point on a geometric shape. In the world of algebraic geometry, mathematicians have a specific tool for this called the Minimal Log Discrepancy (MLD). Think of the MLD as a "smoothness score." A higher score means the point is very smooth; a lower score means it's jagged or singular.

For a long time, mathematicians believed two big things about these scores:

The LSC Conjecture: If you walk around a shape, the smoothness score shouldn't suddenly jump up and then drop down unexpectedly. It should generally stay steady or get smoother as you move.
The PIA Conjecture: If you have a shape and you slice off a piece of it (a divisor), the smoothness of the slice should be directly related to the smoothness of the original shape right next to the cut.

The Problem:
Previously, these rules were known to work perfectly if the shape was "nice" (mathematically called klt, which is like saying the shape is well-behaved and doesn't have too many weird twists). However, the authors found a specific type of shape called a hyperquotient singularity where these rules broke down. It was like finding a road that suddenly dips into a pothole when you thought it was flat.

The big mystery was: Why did the rules break? And could we fix the rules to work for all these shapes, even the messy ones?

The Solution: The "Virtually Free" Detective Work
The authors, Yusuke Nakamura and Kohsuke Shibata, introduced a new concept called "Virtually Free Actions."

To understand this, imagine a group of dancers (the Group) performing on a stage (the Shape).

Free Action: Every dancer moves around the stage without ever bumping into anyone or standing on the same spot as another dancer. Everything is clear.
Virtually Free Action: This is a slightly more relaxed rule. It says that even if the dancers bump into each other occasionally, if you zoom in on the specific spot where they are standing, you can rearrange the stage (resolve the singularity) so that the dancers never actually stand on the same spot again. They are "virtually" free.

The authors realized that the "broken rules" happened exactly when the dancers were not virtually free at a specific point.

The Magic Tool: Arc Spaces
To prove this, they used a tool called Arc Spaces. Imagine taking a movie camera and filming every possible path a tiny particle could take as it moves through the shape.

If the dancers are "virtually free," the paths (arcs) in the movie behave nicely and give you clear information.
If they are not virtually free, the paths get tangled and "thin" (like a ghostly, empty set), hiding the true geometry.

The authors proved a brilliant connection: A point is "virtually free" if and only if the "movie" of paths through that point is full and rich, not thin and empty.

The Results
Using this new "movie camera" insight, they achieved two major things:

Solved the Mystery of the Broken Rules (PIA Conjecture): They found a precise "Yes/No" test. The rule about slicing the shape (PIA) works if and only if the "bad" dancers (those who aren't virtually free) are the same ones causing trouble on the slice as they are on the whole shape. This explained exactly why the previous counterexamples failed: the "bad" dancers were behaving differently on the slice than on the whole shape.
Fixed the Smoothness Rule (LSC Conjecture): Even though the "slicing" rule sometimes fails for messy shapes, the authors proved that the smoothness score (MLD) never jumps up and down unexpectedly. They showed that you can calculate the smoothness of a messy shape by looking at the smoothness of simpler pieces (quotients) and taking the lowest score among the "bad" dancers. Since these simpler scores behave well, the whole shape behaves well too.

In a Nutshell
The paper is like a detective story where the authors:

Found a crime scene (counterexamples to old math rules).
Introduced a new magnifying glass (Arc Spaces and "Virtually Free" actions).
Discovered that the crime was committed by specific "bad dancers" (non-virtually free actions).
Proved that even with these bad dancers, the overall "smoothness" of the shape is still predictable and stable.

They didn't just patch the hole; they rewrote the rulebook so it works for all these complex shapes, removing the need for the "nice shape" (klt) requirement that was previously thought to be necessary.

Technical Summary: A Characterization of Virtually Free Actions via Arc Spaces and Its Application to the Lower Semi-Continuity Conjecture

Problem Statement
The paper addresses two central conjectures in birational geometry concerning minimal log discrepancies (mld): the Lower Semi-Continuity (LSC) conjecture and the Precise Inversion of Adjunction (PIA) conjecture. While these conjectures have been established for various classes of singularities, including quotient singularities and hyperquotient singularities under the assumption that the varieties are Kawamata log terminal (klt), a significant gap remains for non-klt settings.

Previous work by the authors (in [NS22, NS25]) proved the LSC conjecture for hyperquotient singularities provided the variety is klt. However, a counterexample to the PIA conjecture was previously identified for hyperquotient singularities that are log canonical but not klt. This failure of the PIA conjecture in the non-klt setting obstructed the standard reduction strategy used to prove the LSC conjecture, leading to the suspicion that the LSC conjecture might also fail for non-klt hyperquotient singularities. The primary objective is to uncover the mechanism behind the PIA counterexample, clarify the role of the klt condition, and determine if the LSC conjecture holds unconditionally for hyperquotient singularities.

Methodology
The authors employ the theory of arc spaces, originally developed by Denef and Loeser, to analyze the geometry of quotient varieties. The core methodological innovation involves the introduction and characterization of virtually free actions and their localized versions via arc spaces.

Virtually Free Actions: The paper utilizes the concept of a "virtually free" action, where a group action is free in codimension one after any equivariant proper birational modification. This notion serves as an intermediate between free actions and actions free in codimension one.
Arc Space Characterization: The authors introduce a new $k[t]$ $k [t]$ -scheme, denoted $X[\gamma]$ $X [γ]$ , constructed as the quotient of $X \times \text{Spec } k[t^{1/d'}]$ $X \times Spec k [t^{1/ d^{'}}]$ by the cyclic group $\langle \gamma \rangle$ $⟨ γ ⟩$ . They establish a correspondence between the "virtually free" property of an action at a point and the "thinness" of specific subsets in the arc space of the quotient variety.
- Specifically, they prove that the action of an element $\gamma$ is not virtually free at a point $x$ if and only if the local arc space associated with $\gamma$ is not a thin set in the total arc space of the quotient.
Comparison of Schemes: The paper bridges the gap between the classical construction $X(\gamma)$ (used by Denef and Loeser, which is algebraically tractable for computing canonical divisors) and the new functorial construction $X[\gamma]$ (which is compatible with resolutions). By showing a natural bijection between their arc spaces, the authors combine the algebraic advantages of $X(\gamma)$ with the geometric functoriality of $X[\gamma]$ .

Key Contributions and Results

Characterization of PIA Failure: The paper provides a necessary and sufficient condition for the PIA conjecture to hold for arbitrary hyperquotient singularities (Theorem 1.6). The conjecture holds if and only if the minimum value of the minimal log discrepancy, taken over the set of elements $\gamma$ whose action is not virtually free, is attained by an element $\gamma$ that is also not virtually free on the divisor $D$ . This characterization precisely explains the mechanism of the known counterexample: the minimum is attained by an element that is virtually free on the ambient space but not on the divisor, or vice versa, breaking the equality.
Refinement of mld Formulas: The authors refine the formula for the minimal log discrepancy of a quotient variety. They show that the minimum over all group elements $G$ can be restricted to the subset $NVF(X, x)$ (elements whose action is not virtually free at $x$ ). This is a stronger statement than previous results which considered the minimum over all of $G$ .
Unconditional Proof of the LSC Conjecture: Contrary to the pessimistic expectation that non-klt hyperquotient singularities might serve as counterexamples to the LSC conjecture, the authors prove that the LSC conjecture holds unconditionally for arbitrary hyperquotient singularities (Theorem 1.3).
- The proof bypasses the need for the PIA conjecture as a reduction tool. Instead, it utilizes the localized characterization of virtually free actions and the semi-continuity of the set of non-virtually-free elements to express the mld as a finite minimum of lower semi-continuous functions.

Significance
The paper claims to resolve a critical obstacle in the study of singularities in birational geometry. By characterizing virtually free actions via arc spaces, the authors clarify the fundamental role of the klt condition in the PIA conjecture, demonstrating that the conjecture fails specifically when the "virtually free" property breaks down in a way that affects the minimal log discrepancy calculation.

Most significantly, the work removes the restrictive klt assumption from the LSC conjecture for hyperquotient singularities. This establishes the lower semi-continuity of minimal log discrepancies for a broad class of singularities (finite group quotients of complete intersections) without requiring the variety to be klt. This result is achieved not by extending the PIA conjecture, but by developing a new approach that directly analyzes the arc spaces of the quotient varieties, thereby circumventing the limitations exposed by the PIA counterexamples.

Technical Summary: A Characterization of Virtually Free Actions via Arc Spaces and Its Application to the Lower Semi-Continuity Conjecture

More like this