Existence of the minimal model program for log canonical generalized pairs

This paper establishes the existence of the minimal model program for arbitrary log canonical generalized pairs by introducing linearly decomposable pairs as a substitute for rational decompositions and proving the existence of flips without relying on the klt, NQC, or $\mathbb{Q}$-factoriality conditions.

The Gist

Imagine you are an architect trying to renovate a massive, ancient, and slightly crumbling cathedral. Your goal is to find the "perfect" version of this building—a version that is stable, beautiful, and follows the laws of physics (mathematics) without any unnecessary clutter. This process of renovation is called the Minimal Model Program (MMP).

For decades, mathematicians have been successful at renovating "nice" buildings (called klt pairs). However, they hit a wall when trying to renovate "rougher" buildings (called log canonical pairs) that had some cracks, weird angles, or extra structural beams that didn't quite fit the standard rules. Specifically, they couldn't figure out how to perform a specific type of surgery called a "flip" on these rough buildings without breaking them apart.

This paper by Zhengyu Hu and Jihao Liu is the breakthrough that finally solves this problem. Here is how they did it, explained simply:

1. The Problem: The "Rough" Building

In the world of algebraic geometry, a "Generalized Pair" is like a building with two parts:

• The Walls ($B$): The visible structure.
• The Invisible Blueprint ($M$): A hidden set of rules or forces (called the "nef part") that live on a higher, abstract level and influence the building's shape.

For a long time, mathematicians could only safely renovate buildings where the Blueprint ($M$) was very simple and predictable (called the NQC condition). But in the real world, blueprints are often messy, complex, and don't follow simple rules. When the blueprint is messy, the standard renovation tools fail. The architects couldn't perform the "flip" (a surgery that swaps a bad angle for a good one) because they couldn't predict how the messy blueprint would react.

2. The New Tool: "Linearly Decomposable" (LD) Pairs

The authors realized that even if the whole blueprint is a mess, you can still find order if you look at it in a specific way. They invented a new concept called Linearly Decomposable (LD) pairs.

The Analogy:
Imagine you have a complex, multi-colored smoothie (the messy building). You can't separate the ingredients perfectly because they are blended. However, you can prove that this smoothie is just a mix of a few specific, simple juices (like orange, apple, and grape) in certain proportions.

• Old Method: Tried to separate the smoothie into pure, distinct ingredients (Rational Decomposition). This failed for messy blueprints.
• New Method (LD): Instead of separating the ingredients, they realized they could describe the smoothie as moving along a straight line between a few simple, known recipes. Even if the exact recipe is messy, it stays within a "safe zone" defined by these simple recipes.

This "Linear Decomposition" acts as a safety net. It allows the architects to say, "Even though we don't know the exact shape of the building right now, we know it's just a mix of these safe, simple shapes, so we can proceed with the renovation."

3. The Renovation Process (The Flip)

With this new "LD safety net," the authors could finally perform the Flip.

• The Flip: Imagine a room in the cathedral has a weird, unstable corner. To fix it, you tear down that corner and rebuild it in a new, stable shape. In math, this is a "flip."
• The Gluing: When you tear down a corner, you have to make sure the rest of the building stays connected. The authors used a technique called Kollár-type gluing. Think of this as a super-strong, flexible tape that holds the building together while the surgery happens. They proved that even with the messy blueprint, this tape works perfectly if you use their new LD method.

4. The Result: The Complete Renovation Plan

By proving that flips exist for these "rough" buildings, the authors completed the entire Minimal Model Program for this class of structures.

• Before: We had a renovation plan that worked for 90% of buildings, but got stuck on the last 10% (the rough, non-standard ones).
• Now: We have a complete, universal rulebook. No matter how rough, cracked, or complex the building is (as long as it follows the basic laws of log canonical geometry), we can now guarantee that we can renovate it into its perfect, minimal form.

Why Does This Matter?

This isn't just about abstract math. These "buildings" represent the fundamental shapes of our universe in higher dimensions.

• Predictability: It tells us that even in chaotic, complex systems, there is an underlying order we can find and stabilize.
• New Horizons: This opens the door to solving other huge problems in geometry, like understanding how shapes evolve over time or how they behave in complex analytic spaces (like those found in string theory or complex analysis).

In a nutshell: Hu and Liu found a new way to "decompose" a messy, complex mathematical object into simple, manageable pieces. This allowed them to finally perform the final, difficult step of the renovation process, proving that we can always find the "perfect" version of these complex geometric shapes.

Technical Summary

Here is a detailed technical summary of the paper "Existence of the Minimal Model Program for Log Canonical Generalized Pairs" by Zhengyu Hu and Jihao Liu.

1. Problem Statement

The paper addresses a fundamental gap in the Minimal Model Program (MMP) for generalized pairs.

• Context: Generalized pairs $(X, B, M)$, introduced by Birkar and Zhang, extend standard pairs by including a nef part $M$ (a b-divisor) living on a higher model. They are crucial for the effective Iitaka fibration conjecture and the canonical bundle formula.
• The Gap: While the MMP was established for klt (Kawamata log terminal) generalized pairs and for log canonical (lc) generalized pairs under restrictive conditions (specifically the NQC condition, where the nef part is a positive linear combination of nef b-Cartier b-divisors, and/or $\mathbb{Q}$-factoriality), the general case remained open.
• The Challenge: The authors aim to prove the existence of flips and the full MMP for arbitrary log canonical generalized pairs without assuming:
  1. The klt condition (allowing lc singularities).
  2. The NQC condition (allowing the nef part to be non-rational or non-semi-ample).
  3. $\mathbb{Q}$-factoriality of the ambient variety.

2. Methodology and Key Innovations

The authors overcome three major obstacles that prevented the extension of previous NQC-based strategies to the general non-NQC setting:

A. Linearly Decomposable (LD) Generalized Pairs

• The Obstacle: In the NQC setting, generalized pairs admit rational decompositions (expressing the pair as a convex combination of $\mathbb{Q}$-generalized pairs). This allows the application of gluing theory and finite generation arguments. In the non-NQC setting, such decompositions generally fail.
• The Solution: The authors introduce Linearly Decomposable (LD) generalized pairs.
  • Definition: A pair is LD if its log canonical class $K_X + B + M_X$ varies within a rational affine subspace such that the pair remains lc in a neighborhood.
  • Significance: LD pairs serve as a workable substitute for rational decompositions. They allow for "controlled convex decompositions" into $\mathbb{Q}$-classes even when the nef part is not NQC.
  • Key Lemma: Any lc generalized pair polarized by an ample $\mathbb{R}$-divisor admits an LD structure up to $\mathbb{R}$-linear equivalence. This allows the reduction of the general problem to the LD case.

B. Special Termination via Codimension 2 Control

• The Obstacle: Standard "special termination" relies on a "difficulty function" defined by the coefficients of the nef part. In the non-NQC setting, these coefficients are not well-defined.
• The Solution: The authors develop a new termination argument based on codimension 2 control.
  • They observe that the invariants relevant to special termination arise from surface plt (purely log terminal) points.
  • By utilizing the boundedness of local Cartier indices at codimension 2 points (derived from the $\epsilon$-plt condition and the preservation of Weil indices along the MMP), they define a new difficulty function dependent on the specific generalized pair and an induction hypothesis.
  • This allows them to prove special termination for LD generalized pairs without relying on the undefined "coefficients of the nef part."

C. Avoiding ACC for lc Thresholds

• The Obstacle: Previous proofs for abundant generalized pairs relied heavily on the Ascending Chain Condition (ACC) for lc thresholds and global ACC, which are not available or not applicable in the general non-NQC setting.
• The Solution: The authors adopt a strategy inspired by Hashizume's work on pairs, proving the existence of good minimal models for relatively potentially log Calabi–Yau pairs and in families without invoking ACC inputs. This is combined with the LD structure and the new special termination to establish the necessary finiteness properties.

D. Kollár-Type Gluing Theory for LD Pairs

• The authors establish a Kollár-type gluing theory specifically for LD generalized pairs. This theory allows them to glue semi-ample models from lower-dimensional strata (lc centers) to the whole space, a crucial step in proving the existence of good minimal models when the canonical class is not big.

3. Key Results

Main Theorems

1. Existence of Flips (Theorem 1.1):
   For any log canonical generalized pair $(X, B, M)/U$ and a flipping contraction $f: X \to Z$, the flip $f^+: X^+ \to Z$ exists. This holds without NQC, klt, or $\mathbb{Q}$-factorial assumptions.

2. Existence of the MMP (Theorem 1.2):
   Combining the existence of flips with the Cone and Contraction theorems (previously established for the general case by CHLX23), the authors prove that one can run a $(K_X + B + M_X)$-MMP for arbitrary log canonical generalized pairs.

3. Existence of Good Minimal Models (Theorems 1.3 & 1.4):

   • Theorem 1.3: For a generalized pair where $K_X + B + A + M_X \sim_{\mathbb{R}, U} 0$ (relatively potentially log Calabi–Yau), a good log minimal model or a log Mori fiber space exists.
   • Theorem 1.4: If an LD generalized pair has a good minimal model on a non-empty open subset $U_0$ and every lc center intersects $U_0$, then it has a good log minimal model globally.

Structural Results

• Theorem 1.5: For a $\mathbb{Q}$-factorial lc generalized pair $(X, B, M)$ and an ample divisor $A$, there exists an lc pair $(X, \Delta)$ such that $K_X + \Delta \sim_{\mathbb{R}, U} K_X + B + M_X + A$. This effectively reduces the study of the log canonical class of generalized pairs to standard lc pairs after adding an ample divisor.
• Corollary 1.6: The base of a Fano contraction of a generalized pair is "potentially lc" (i.e., it admits an lc structure).

4. Significance and Impact

• Completeness of the MMP: This paper closes the last major open case in the Minimal Model Program for generalized pairs. It unifies the theory for klt and lc singularities, and for both NQC and non-NQC nef parts.
• Foundational Tool: The introduction of LD generalized pairs provides a robust framework for handling non-NQC settings. This technique is likely to be applicable to other problems in birational geometry where rational decompositions fail.
• Applications: The results are essential for:
  • The Effective Iitaka Fibration Conjecture.
  • The Moduli theory of varieties with anti-nef canonical classes.
  • The Kähler and analytic MMP (where the NQC condition often fails).
  • The study of foliations and their minimal models.
• Analytic Extensions: The authors note that their methods, combined with recent analytic developments, suggest the possibility of extending the MMP to compact Kähler varieties and complex analytic spaces, a major frontier in higher-dimensional geometry.

In summary, Hu and Liu have successfully generalized the Minimal Model Program to the most inclusive setting currently known for generalized pairs, overcoming deep technical hurdles related to the lack of rationality and factorization in the nef part through the novel concept of Linearly Decomposable pairs and refined termination arguments.