🔢 mathematics

A note on b-divisors and filtrations on a local ring

This paper establishes a correspondence between filtrations and b-divisors over a general class of Noetherian local domains and applies this result to prove a recent conjecture by Roé-Urbinati in the global setting.

Original authors: Lu Qi

Published 2026-04-09

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Lu Qi

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 describe a very complex, multi-layered mountain range. You have two different ways to map this terrain:

The "Top-Down" Map (b-divisors): This is like looking at the mountain from a helicopter that can zoom in and out infinitely. You see the peaks, valleys, and ridges. But because the mountain is so complex, you can't just draw it on a single flat piece of paper. Instead, you have a whole stack of maps, each showing a different level of detail. A b-divisor is like a "super-map" that keeps track of all these layers simultaneously, ensuring that if you zoom in on a specific valley, the details match up perfectly with the view from above.
The "Bottom-Up" Filter (Filtrations): This is like a sieve or a colander used in cooking. Imagine you have a pile of mixed ingredients (numbers or functions). You pour them through a series of sieves with increasingly fine holes.
- The first sieve catches the big chunks.
- The next one catches medium chunks.
- The finest one catches the dust.
  A filtration is the rulebook that tells you exactly what gets caught in each sieve. It organizes the chaos of the ingredients into neat, ordered layers.

The Big Problem

For a long time, mathematicians knew these two ways of looking at the mountain (the super-map and the sieve) were related, but they didn't have a perfect dictionary to translate between them.

If you start with a Sieve (filtration), you can build a Super-Map (b-divisor).
But if you start with a Super-Map, can you always reconstruct the exact Sieve that created it? Sometimes, the map is too vague, or the sieve is too messy, and the translation fails.

The Paper's Discovery

The author, Lu Qi, proves that if you restrict your attention to a specific, well-behaved type of mountain (a "normal, excellent local ring"—think of a smooth, well-defined spot on the ground), there is a perfect, one-to-one correspondence between these two methods.

Here is the simple breakdown of the magic:

The Translation Rule:
- From Sieve to Map: If you have a perfectly organized sieve (called a "saturated filtration"), you can draw a unique Super-Map.
- From Map to Sieve: If you have a Super-Map that isn't too wild (it's "bounded"), you can build a unique, perfectly organized sieve from it.
The "Aha!" Moment: The paper shows that if you take a sieve, turn it into a map, and then turn that map back into a sieve, you get the exact same sieve you started with. No information is lost.

Why Does This Matter? (The Real-World Analogy)

Imagine you are a detective trying to solve a mystery about a specific location (a "valuation").

The Old Way: You had to check every single clue (every possible angle of the mountain) to see if the location was "special" (b-divisorial). It was tedious and prone to error.
The New Way (This Paper): The author proves a shortcut. You don't need to check every angle. You just need to look at the "Super-Map" generated by the clues.
- If the map is empty (zero), the location is ordinary.
- If the map has content (is not zero), the location is "special" (b-divisorial).

This solves a recent conjecture (a mathematical guess) by Roe and Urbinati. It's like proving that if a shadow exists, the object casting it must be real, without needing to touch the object itself.

The "Secret Sauce"

The paper relies on a clever trick called Saturation.
Think of a sieve that has a few tiny holes that are slightly too big. It lets some dust through that it shouldn't. "Saturation" is the process of fixing those holes so the sieve is perfect. The author proves that in this specific mathematical world, any well-behaved sieve is already "fixed" (saturated), which makes the translation between the sieve and the map possible.

Summary

In short, this paper builds a perfect bridge between two different languages used to describe geometric shapes and algebraic rules. It tells us that if we organize our data correctly (using "saturated" rules), the "shape" of the data (the b-divisor) and the "rules" of the data (the filtration) are actually the same thing, just viewed from different angles. This allows mathematicians to solve difficult problems about the shape of space by using simpler algebraic tools, and vice versa.

1. Problem Statement

The paper addresses the relationship between two fundamental objects in birational geometry and commutative algebra:

Shokurov's b-divisors: Families of Weil divisors on birational models of a scheme, compatible with push-forwards.
Filtrations (and graded sequences of ideals): Decreasing sequences of ideal sheaves (or ideals in a local ring) satisfying multiplicative properties.

While the correspondence between ideals and b-divisors (via the "Zariski divisor" construction) is well-known, the paper seeks to establish a rigorous, bijective correspondence between linearly bounded saturated filtrations and bounded b-divisors in the local setting of a normal Noetherian local domain.

Furthermore, the paper aims to resolve a specific conjecture by Roé and Urbinati regarding b-divisorial valuations. The conjecture posits that a valuation $v$ on a projective variety is "b-divisorial" (meaning it is determined entirely by its associated b-divisor) if and only if the b-divisor associated with its filtration is non-zero.

2. Methodology

The author employs a combination of commutative algebra, valuation theory, and birational geometry. The methodology relies on the following key steps:

Local Reduction: The problem is first reduced to the affine local setting, where $X = \text{Spec}(R)$ for a normal, excellent Noetherian local domain $(R, \mathfrak{m})$ .
Definitions of Spaces:
- Filtrations: The author focuses on $\mathfrak{m}$ -filtrations $\mathfrak{a}_\bullet = \{\mathfrak{a}_\lambda\}_{\lambda \in \mathbb{R}_{>0}}$ that are linearly bounded (contained in powers of $\mathfrak{m}$ ) and saturated (defined via divisorial valuations).
- B-divisors: The author considers the space of b-divisors over the closed point $x \in X$ , specifically the subset of bounded b-divisors ( $\text{Div}_b(X, x)$ ), which are anti-effective and bounded below by a multiple of the canonical b-divisor $Z(\mathfrak{m})$ .
Construction of Maps:
- From Filtration to B-divisor: Using the construction $Z_X(\mathfrak{a}_\bullet) := -\sum_{P/X} \text{ord}_P(\mathfrak{a}_\bullet) \cdot P$ , where the sum runs over prime divisors over $X$ .
- From B-divisor to Filtration: Defining ideals via global sections: $\mathfrak{a}_\lambda(W) := \{f \in R \mid Z(f) \leq \lambda W\}$ .
Key Technical Tools:
- Rees' Theorem for Filtrations: Cited from recent work (BLQ24) to establish that linearly bounded valuations yield saturated filtrations.
- Izumi's Inequality: Used to establish bounds on valuations in normal excellent domains.
- Coefficientwise Topology: The convergence of b-divisors is defined by the convergence of coefficients on prime divisors.

3. Key Contributions

The paper makes three primary technical contributions:

Establishment of a Correspondence (Theorem 1):
The author proves that there is a natural correspondence between the space of linearly bounded saturated $\mathfrak{m}$ -filtrations ( $\text{Fils}^s_{R,\mathfrak{m}}$ ) and the space of bounded b-divisors ( $\text{Div}_b(X, x)$ ).
- The map $\mathfrak{a}_\bullet \mapsto Z_X(\mathfrak{a}_\bullet)$ is injective and continuous.
- The map $W \mapsto \mathfrak{a}_\bullet(W)$ is surjective.
- Crucially, for any saturated filtration $\mathfrak{a}_\bullet$ , the reconstruction holds: $\mathfrak{a}_\bullet(Z_X(\mathfrak{a}_\bullet)) = \mathfrak{a}_\bullet$ .
Resolution of the Roé-Urbinati Conjecture (Corollary 3):
The paper proves that for a normal projective variety $X$ and a valuation $v \in \text{Val}_X$ , $v$ is b-divisorial if and only if $Z_X(\mathfrak{a}_\bullet(v)) \neq 0$ .
- This generalizes previous results known only for surfaces ( $\dim X = 2$ ) to arbitrary dimensions.
- The proof avoids complex approximation techniques used in lower dimensions, relying instead on the local correspondence and Izumi's inequality.
Clarification of Saturated Filtrations:
The paper highlights the necessity of the saturation condition (defined via divisorial valuations) to ensure the bijection. It demonstrates that linearly bounded valuations naturally produce saturated filtrations, bridging the gap between valuation theory and the geometry of b-divisors.

4. Main Results

Theorem 1 (Local Correspondence):
Let $(R, \mathfrak{m})$ be a normal, excellent Noetherian local domain.

The map $Z_X: \text{Fils}^s_{R,\mathfrak{m}} \to \text{Div}_b(X, x)$ is an injective, continuous map.
The map $\mathfrak{a}_\bullet: \text{Div}_b(X, x) \to \text{Fils}^s_{R,\mathfrak{m}}$ is surjective.
For any $\mathfrak{a}_\bullet \in \text{Fils}^s_{R,\mathfrak{m}}$ , $\mathfrak{a}_\bullet(Z_X(\mathfrak{a}_\bullet)) = \mathfrak{a}_\bullet$ .
For any $W \in \text{Div}_b(X, x)$ , $Z_X(\mathfrak{a}_\bullet(W)) \leq W$ .

Corollary 3 (Global Application):
Let $X$ be a normal projective variety over a field $k$ , and $v \in \text{Val}_X$ . Then $v$ is b-divisorial if and only if $Z_X(\mathfrak{a}_\bullet(v)) \neq 0$ .

Lemma 13 (Saturation Property):
If a valuation $v \in \text{Val}_{R,\mathfrak{m}}$ is linearly bounded, then the associated filtration $\mathfrak{a}_\bullet(v)$ is saturated. This relies on Rees' theorem for filtrations.

5. Significance

Unification of Theories: The paper provides a clean geometric framework linking the algebraic theory of filtrations (central to K-stability and singularity theory) with the geometric theory of b-divisors (central to the Minimal Model Program).
Advancement in K-Stability: Filtrations are central to the study of K-stability of Fano varieties. By establishing a correspondence with b-divisors, this work allows techniques from b-divisor theory (such as intersection theory on the Zariski-Riemann space) to be applied to filtration problems.
Generalization of Dimension 2 Results: The proof of the Roé-Urbinati conjecture removes the dimension restriction, providing a robust tool for studying valuations on higher-dimensional varieties without relying on surface-specific geometry.
Foundational for Future Work: The paper sets up a "geometric object" structure for the space of filtrations, suggesting a metric space structure analogous to the space of Kähler potentials, which is a growing area of interest in non-Archimedean geometry and complex geometry.

In summary, Lu Qi's note successfully formalizes the duality between bounded b-divisors and saturated filtrations in the local setting, using this duality to solve a significant conjecture in the global theory of valuations.