Deformations of fibered Calabi--Yau varieties

This paper extends Kollár's result on the preservation of elliptic fibrations under small deformations to general fibered smooth K-torsion varieties with vanishing second cohomology of the structure sheaf, utilizing Hodge theory and the Kawamata–Ran $T^1$-lifting criterion to further establish that semiample line bundles remain semiample up to homological equivalence even without this cohomological assumption.

The Gist

Imagine you have a very special, intricate piece of origami. In the world of mathematics, this piece of paper is called a Calabi–Yau variety. It's a complex shape that physicists and mathematicians love because it helps explain the hidden dimensions of the universe (like in String Theory).

Now, imagine this origami isn't just a random crumple; it's folded in a very specific way. It looks like a bundle of strings or a stack of pancakes. In math terms, we say it is "fibered." It's made of many smaller shapes (the fibers) stacked on top of each other to form the big shape.

The big question this paper asks is: What happens if you gently wiggle or deform this origami?

If you push the paper slightly, does it stay in its neat, stacked structure? Or does it crumple into a messy ball where the "stacking" disappears?

The Main Characters

1. The Shape (X): A smooth, perfect Calabi–Yau variety. Think of it as a perfectly balanced, magical sculpture.
2. The Stack (The Fibration): The way the sculpture is built from layers.
3. The Deformation: A gentle nudge or a slight change in the environment (like changing the temperature or humidity) that tries to reshape the sculpture.
4. The "Magic" Condition ($H^2(X, O_X) = 0$): This is a fancy math way of saying the sculpture has a specific kind of "rigidity" or "emptiness" in its internal structure. It's like saying the sculpture has no hidden pockets or secret compartments that could absorb the wiggle.

The Story of the Paper

Part 1: The Rigid Case (The "Easy" Answer)

The authors start by looking at a specific type of sculpture where that "Magic Condition" is true.

• The Old Idea: A mathematician named Kollár already proved that if your sculpture is a stack of elliptic curves (think of them as donuts), and you wiggle it, it stays a stack of donuts.
• The New Discovery: This paper says, "Great! But what if the layers aren't donuts? What if they are any shape at all?"
• The Result: They prove that as long as the Magic Condition holds, the stack never breaks. No matter how you wiggle the sculpture, it will always remain a neat stack of layers. The "stacking" is permanent.

Analogy: Imagine a deck of cards. If the cards are glued together in a specific, rigid way (the Magic Condition), you can shake the deck, and it will always stay a deck. It won't turn into a pile of loose cards.

Part 2: The Tricky Case (When the Magic Condition is Missing)

But what if the sculpture doesn't have that Magic Condition? What if it's a bit more flexible?

• The Problem: In this case, the stack can break. If you wiggle a generic sculpture, the layers might dissolve, and the neat structure disappears.
• The Twist: However, the authors found a clever workaround. They realized that even if the exact layers change, the essence of the stack remains.
• The Result: They proved that if you start with a "semiample" bundle (a fancy way of saying a bundle that wants to create a stack), then in any small wiggle, there is always a new bundle that creates a stack. It might not be the exact same stack you started with, but a very similar one (mathematically equivalent) will always exist.

Analogy: Imagine you have a tower of Jenga blocks. If you shake it too hard, it might fall. But this paper says: "Even if the tower falls, if you look closely, the blocks are still arranged in a way that could be a tower again. You just might need to rearrange them slightly to see the new tower."

The Secret Weapon: The "T1-Lifting" Trick

How did they prove this? They used a tool called the T1-lifting criterion (invented by Kawamata and Ran).

• The Metaphor: Imagine you are trying to build a tower of blocks on a moving train. You want to know if you can keep stacking blocks as the train moves.
• The "T1-lifting" is like a magical blueprint. It checks if the tiny, microscopic movements of the blocks (the "infinitesimal deformations") can be extended to the whole train ride.
• The authors used this blueprint, combined with Hodge Theory (which is like analyzing the "vibrations" or "colors" of the shape), to show that the blueprint always works for these specific shapes.

The "Gotcha" Examples (Section 3)

The paper ends with a warning. They show that while the stacking usually survives, the individual pieces inside the stack might not.

• The Scenario: Imagine a sub-structure inside the sculpture (like a specific curve drawn on the paper).
• The Surprise: Sometimes, even if the big sculpture stays a stack, that specific little curve might get "stuck" and refuse to move smoothly when the sculpture wiggles.
• The Lesson: The whole system is stable, but the tiny details can be stubborn and get "obstructed."

Why Does This Matter?

1. For Physicists: These shapes are used to model the universe. If the shapes change their structure when you wiggle them, the physics changes too. Knowing that the "stacking" stays stable helps physicists trust their models.
2. For Mathematicians: It solves a long-standing puzzle about how these complex shapes behave. It connects different areas of math (geometry, topology, and algebra) to show that nature prefers order and stability, even when things are wiggly.

Summary in One Sentence

"Even if you gently wiggle these complex mathematical shapes, their fundamental 'stacked' structure is so strong that it refuses to break, ensuring that the beautiful order of the universe remains intact."

Technical Summary

1. Problem Statement and Context

The paper addresses a fundamental question in the deformation theory of algebraic varieties, specifically within the class of $K$-torsion varieties (smooth proper varieties with numerically trivial canonical bundle, $K_X \sim_{\mathbb{Q}} 0$).

• The Core Question: Given a $K$-torsion variety $X$ equipped with a fibration $f: X \to Y$ (induced by a semiample line bundle $L$), does this fibration structure persist under small deformations of $X$?
• The Obstruction: In general, fibrations do not deform. For example, a product of an abelian variety and an elliptic curve admits an elliptic fibration, but a generic deformation of this product is a simple abelian variety with no lower-dimensional fibrations. Similarly, a very general deformation of an elliptically fibered K3 surface often loses its elliptic fibration because the Picard rank drops to 1.
• Previous Work: Kollár [Kol15] proved that for elliptically fibered $K$-torsion varieties, if the cohomological vanishing condition $H^2(X, \mathcal{O}_X) = 0$ holds, then all small deformations remain elliptically fibered.
• The Gap: It was unknown whether this result generalizes to arbitrary fibrations (not just elliptic) or to $K$-torsion varieties without the $H^2$ vanishing assumption.

2. Methodology

The authors employ a combination of Hodge theory, deformation theory, and the structural decomposition of $K$-torsion varieties.

A. Hodge-Theoretic Techniques and $T^1$-Lifting

The primary tool is the Kawamata–Ran $T^1$-lifting criterion. The authors analyze the deformation functor of the pair $(X, D)$, where $D$ is a general divisor in the linear system of a sufficiently high multiple of the semiample line bundle inducing the fibration.

• They study the forgetful map $\text{Def}(X, D) \to \text{Def}(X)$.
• Using the Global Invariant Cycle Theorem and the vanishing of $H^2(X, \mathcal{O}_X)$, they prove this map is dominant and smooth onto its image. This ensures that deformations of $X$ can be lifted to deformations of the pair $(X, D)$, preserving the fibration structure.

B. Beauville–Bogomolov Decomposition

To handle the general case (without assuming $H^2(X, \mathcal{O}_X) = 0$), the authors utilize the Beauville–Bogomolov decomposition theorem.

• They pass to a finite étale cover $\tilde{X} \to X$ that splits the variety into a product of:
  1. Strict Calabi–Yau varieties ($K_X \sim 0$, $H^1(\mathcal{O}_X)=0$).
  2. Irreducible holomorphic symplectic (IHS) manifolds.
  3. Complex tori (Abelian varieties).
• The problem is reduced to proving the deformation persistence for each factor separately, and then showing the results descend to the original variety.

C. Specific Factor Analysis

• Strict Calabi–Yau: Handled via the $T^1$-lifting argument (Proposition 2.5).
• IHS Manifolds: Relies on Matsushita's [Mat16] results regarding the persistence of Lagrangian fibrations.
• Complex Tori: Requires a specific construction involving the quotient of the torus by the sub-torus defined by the degenerate locus of the Chern class of the line bundle (Lemma 2.10).

3. Key Contributions and Results

The paper establishes two main theorems:

Theorem 1.1 (Generalization of Kollár's Result)

Statement: Let $\pi: \mathcal{X} \to T$ be a smooth projective family of $K$-torsion varieties over an analytic germ. If the central fiber $X_0$ satisfies $H^2(X_0, \mathcal{O}_{X_0}) = 0$ and admits a fibration $\psi_0: X_0 \to Y_0$, then there exists a commutative diagram where the fibration deforms to a flat family $\psi: \mathcal{X} \to \mathcal{Y}$ over $T$.

• Significance: This extends Kollár's result from elliptic fibrations to arbitrary fibrations, provided the cohomological vanishing condition holds. It implies that if the Generalized Abundance Conjecture holds for a nef line bundle $L_0$, it holds for all its deformations in the miniversal deformation space.

Theorem 1.2 (Deformation of Semiample Line Bundles)

Statement: Let $\pi: \mathcal{X} \to T$ be a smooth proper family of $K$-torsion Kähler manifolds. If $L$ is a line bundle on $\mathcal{X}$ such that $L|_{X_0}$ is semiample, then there exists a line bundle $M$ on $\mathcal{X}$ such that:

1. $M|_{X_0} \cong L|_{X_0}$.
2. $M$ is $\pi$-semiample (relatively semiample).
3. $M$ is homologically equivalent to $L$ (specifically, $M \equiv_T L$).

• Significance: This removes the $H^2(X, \mathcal{O}_X)=0$ assumption. While the original line bundle $L$ might not deform to a relatively semiample bundle (due to obstructions in the Picard group), there always exists a homologically equivalent bundle $M$ that does deform relatively semiample. This confirms that the "fibration structure" persists up to homological equivalence.

Counterexamples and Obstructions (Section 3)

The authors investigate subvarieties with trivial normal bundles.

• Question: Do deformations of a subvariety $Y \subset X$ with trivial normal bundle remain unobstructed and form fibers of a fibration?
• Result: The answer is negative in general.
  • Example 3.2: A subvariety in a hyperkähler manifold of $K3^{[2]}$-type may have trivial normal bundle but cannot be a fiber of any fibration of the ambient space.
  • Example 3.3: A strict Calabi–Yau 3-fold containing an elliptic curve with trivial normal bundle. While the curve moves in a pencil on a specific fiber, its embedded deformations in the total space are obstructed (it cannot deform past the first order for general fibers).

4. Significance and Impact

1. Resolution of Deformation Problems: The paper provides a definitive answer to the deformation of fibrations on $K$-torsion varieties, distinguishing between cases where the structure is rigid (due to cohomology vanishing) and cases where it persists only up to homological equivalence.
2. Connection to Generalized Abundance: The results strongly support the Generalized Abundance Conjecture. They show that if a nef line bundle is semiample on a central fiber, it remains semiample (up to homological equivalence) in the deformation family.
3. Methodological Advancement: The synthesis of the $T^1$-lifting criterion with the Beauville–Bogomolov decomposition offers a robust framework for studying deformations of complex geometric structures on $K$-trivial varieties.
4. Clarification of Obstructions: By providing explicit counterexamples in Section 3, the authors delineate the precise limits of these deformation theorems, showing that trivial normal bundles do not guarantee unobstructed deformations or fibration structures in higher dimensions.

In summary, the paper establishes that while specific fibration structures may not deform rigidly without cohomological assumptions, the underlying "semiample" nature of the defining line bundles is robust under deformation, ensuring the persistence of the fibration geometry up to homological equivalence.