🔢 mathematics

Obstructions for codimension one multiple fibers of Lagrangian and Calabi--Yau fibrations

This paper proves that compact hyper-Kähler manifolds with Lagrangian fibrations and simply-connected K-trivial varieties with Calabi-Yau fibrations generally lack codimension-one multiple fibers, with a single exceptional case realized by Borisov–Nuer examples, thereby advancing the understanding of singular fiber structures and related conjectures.

Original authors: Yoon-Joo Kim, Keiji Oguiso, Evgeny Shinder

Published 2026-08-12

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Yoon-Joo Kim, Keiji Oguiso, Evgeny Shinder

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 an architect designing a massive, multi-story building where every floor is a perfect, self-contained world. In the realm of mathematics, specifically a field called algebraic geometry, scientists study shapes that exist in many dimensions, far beyond the three we can see. These shapes, called "manifolds," can be incredibly complex, but some are special because they are perfectly balanced and smooth, like a crystal that never breaks. Among these, there are two famous types: "hyper-Kähler manifolds" (which are like super-symmetrical, multi-dimensional dance floors) and "Calabi–Yau varieties" (which are the hidden geometric shapes that string theory suggests make up the fabric of our universe).

To understand these giant shapes, mathematicians often try to "slice" them. They imagine projecting the shape down onto a simpler base, like casting a shadow or peeling an orange to reveal its layers. This process is called a "fibration." Usually, when you peel an orange, every slice is a perfect, single piece of fruit. But sometimes, a slice might be "double-layered" or "triple-layered"—mathematicians call these "multiple fibers." It's as if, instead of one thin slice of orange, you suddenly found a slice that was actually two or three stuck together. The big question for a long time was: Can these special, perfectly balanced shapes ever have these weird, double-layered slices? For simple shapes like elliptic curves (which look like donuts), mathematicians already knew the answer was no. But for the giant, high-dimensional versions, no one was sure.

This paper by Yoon-Joo Kim, Keiji Oguiso, and Evgeny Shinder acts like a master detective solving a mystery about these high-dimensional shapes. They prove that for a very important class of these shapes—specifically, the "hyper-Kähler" ones and the "Calabi–Yau" ones that are simply connected (meaning they have no holes or loops you can't shrink)—it is impossible to have these double-layered slices in the most common, visible parts of the shape. They show that if you try to build such a shape with a double slice, the math simply falls apart, like a house of cards collapsing.

The authors used a clever mix of old and new mathematical tools to crack the case. They imagined taking the shape and wrapping it around itself in a loop (a "cyclic covering trick") to see what would happen. They also used a special kind of "mathematical ruler" called the Beilinson norm to measure the complexity of the shape's layers. Their calculations revealed a contradiction: if a double slice existed, the shape would have to be both simple and impossibly complex at the same time. Therefore, for these specific shapes, the slices must always be single and clean.

However, the story has a tiny, fascinating twist. While they proved that double slices are impossible for most cases, they found one very specific exception. If the shape has an even-dimensional slice (the "relative dimension" is even) and the base is a simple line (like a one-dimensional string), a double slice can exist, but only if the shape contains a special, slightly "twisted" sub-shape inside it (called an Enriques–Calabi–Yau manifold). The authors didn't just say this was possible; they showed how to build examples of these rare shapes, proving that their rule is as tight as it can possibly be.

In short, this paper puts a definitive stop to the idea that these high-dimensional, perfectly balanced shapes can have messy, double-layered slices in their main structure. It confirms that nature, in this mathematical sense, prefers clean, single layers for these specific types of geometric worlds, with only one very narrow, exotic loophole. This discovery helps mathematicians understand the fundamental rules of how these shapes are built, which is crucial for anyone trying to map out the geometry of the universe or solve deep problems in string theory.

Problem Statement

The paper addresses the existence of codimension one multiple fibers in fibrations of higher-dimensional complex varieties, specifically focusing on Lagrangian fibrations of hyper-Kähler manifolds and Calabi–Yau fibrations on simply-connected $K$ -trivial varieties. While it is a classical result that elliptic K3 surfaces over $\mathbb{P}^1$ possess no multiple fibers, the behavior of these fibers in higher dimensions was not fully established. The authors investigate whether this non-existence property extends to:

Lagrangian fibrations $f: X \to \mathbb{P}^n$ where $X$ is a hyper-Kähler manifold.
Calabi–Yau fibrations $f: X \to \mathbb{P}^n$ where $X$ is a simply-connected, smooth, projective, $K$ -trivial variety.

A codimension one multiple fiber is defined as a prime divisor $D \subset B$ such that the preimage $f^{-1}(D) = mE$ for an integer $m > 1$ and an effective divisor $E$ . The paper also explores the limits of these results when the total space $X$ is singular.

Methodology

The authors employ a combination of classical algebraic geometry techniques and new computational tools involving the Grothendieck ring of the base projective space. The core strategy involves a proof by contradiction: assuming the existence of a codimension one multiple fiber of multiplicity $m > 1$ , they derive a contradiction using the following steps:

Cyclic Covering and Divisibility: Utilizing the cyclic covering trick and the simply-connectedness of $X$ , they deduce that the pullback of the hyperplane class from the base, $f^*\mathcal{O}_{\mathbb{P}^n}(1)$ , must be uniquely divisible by $m$ in the Picard group of $X$ . This allows the construction of a specific line bundle $\mathcal{O}_X(1/m)$ .
Higher Direct Image Sheaves: They analyze the higher direct image sheaves $R^i f_* \mathcal{O}_X(j/m)$ . Using Kollár's vanishing theorems for higher direct images and the assumption that $X$ is $K$ -trivial with rational singularities, they establish that these sheaves are locally free on the base $\mathbb{P}^n$ .
Horrocks's Criterion: By applying Horrocks's splitting criterion, they show that these locally free sheaves decompose into direct sums of line bundles on $\mathbb{P}^n$ .
Beilinson Norm and Hilbert Polynomials:
- For Lagrangian fibrations, the authors introduce a "Beilinson norm" on the Grothendieck group $K_0(\mathbb{P}^n)$ , defined with respect to the Beilinson basis $[\mathcal{O}_{\mathbb{P}^n}(-n)], \dots, [\mathcal{O}_{\mathbb{P}^n}]$ . They compute the norm of the class $[R f_* \mathcal{O}_X(1/m)]$ in two ways: first using the Euler characteristic (via the Huybrechts–Riemann–Roch formula) and second using the triangle inequality on the direct sum decomposition. The resulting inequality yields a contradiction.
- For Calabi–Yau fibrations, they utilize integrality constraints on the Hilbert polynomial of the class $[R f_* \mathcal{O}_X(1/m)]$ . They demonstrate that the polynomial derived from the geometry of the fibration cannot satisfy the required integrality conditions for $m > 1$ in most cases.

Key Contributions and Results

The paper establishes the following main theorems:

Theorem 1.1 (Lagrangian Fibrations): A Lagrangian fibration $f: X \to \mathbb{P}^n$ from a hyper-Kähler manifold $X$ has no codimension one multiple fibers.
- Significance: This result is tight; the authors note that multiple fibers can exist in non-compact settings or in higher codimensions, but they are strictly forbidden in codimension one for this class of manifolds.
- Consequences: This result renders several previous results in the literature unconditional. It implies the existence of local sections in codimension one, confirms the primitivity of the embedding $f^*(\text{Pic}(\mathbb{P}^n)) \subset \text{Pic}(X)$ , and simplifies the analysis of Tate–Shafarevich groups in boundedness problems. It also supports Sawon's conjecture regarding the structure of general singular fibers (showing they are reduced and of types I, II, III, or IV under specific Picard number conditions).
Theorem 1.3 (Calabi–Yau Fibrations): Let $X$ be a simply-connected, smooth, projective, $K$ -trivial variety with a Calabi–Yau fibration $f: X \to \mathbb{P}^n$ of relative dimension $r$ .
- If $n \ge 2$ or $r$ is odd, $f$ has no codimension one multiple fibers.
- If $n = 1$ and $r$ is even, $f$ has no multiple fibers of multiplicity greater than 2, and there is at most one fiber with multiplicity 2.
- Optimality: The authors prove this bound is sharp by constructing examples (generalizing Borisov–Nuer examples) of $K$ -trivial varieties containing an Enriques–Calabi–Yau divisor, which admit a fibration over $\mathbb{P}^1$ with exactly one multiple fiber of multiplicity 2. This exceptional case arises specifically when the base dimension is $n=1$ and the relative dimension $r$ is even.
Generalizations to Singular Spaces:
- The authors show that the results do not automatically extend to varieties with Gorenstein canonical singularities (citing a singular Kummer K3 surface as a counterexample).
- However, they prove that for factorial irreducible symplectic varieties (a singular generalization of hyper-Kähler manifolds), the non-existence of codimension one multiple fibers still holds.
- For elliptic fibrations on varieties with terminal singularities where $K_X$ is base-point-free, the non-existence of multiple fibers is also preserved.

Significance and Claims

The paper claims to provide "optimal generalizations" of the classical non-existence of multiple fibers for elliptic K3 surfaces to higher dimensions. The primary significance lies in:

Unconditional Results: Removing the need to assume the non-existence of multiple fibers in prior works concerning Lagrangian fibrations, thereby solidifying the structural theory of these fibrations.
Structural Insights: Clarifying the relationship between the primitivity of the Picard embedding and the existence of multiple fibers.
Sharpness of Bounds: Precisely delineating the boundary between existence and non-existence of multiple fibers in Calabi–Yau fibrations, identifying the specific numerical case ( $n=1, r$ even) where a single double fiber is possible, and providing explicit constructions for this case.

The authors maintain a modest tone regarding singular cases, acknowledging that while their methods require smoothness or factoriality, the results hold in specific singular contexts (factorial irreducible symplectic varieties) but fail in others (canonical singularities), leaving the general singular case as an open area for further investigation.

More like this