On algebraically coisotropic submanifolds of holomorphic symplectic manifolds

Here is an explanation of the paper "On algebraically coisotropic submanifolds of holomorphic symplectic manifolds" by Ekaterina Amerik and Frédéric Campana, translated into everyday language with analogies.

The Big Picture: The "Perfectly Balanced" Dance Floor

Imagine the universe of this paper is a giant, multi-dimensional dance floor called $M$ . This floor has a very special property: it is holomorphically symplectic.

Think of this property like a magical, invisible grid or a "flow" that runs through the entire floor. This flow dictates how things move and interact. In math terms, this is a "symplectic form." It's like a rulebook that says, "If you move this way, you must balance that way."

Now, imagine a group of dancers (a submanifold $X$ ) standing on this floor. The authors are asking: How can a group of dancers stand on this floor while respecting the flow?

There are two main ways they can stand:

Lagrangian (The Perfect Balance): The dancers stand in a formation where they are perfectly "orthogonal" to the flow. They don't fight the flow; they exist in a state of perfect equilibrium. In math, this is called a Lagrangian submanifold.
Coisotropic (The Flowing River): The dancers stand in a formation where the flow runs along their group. They are like a river flowing down a valley. The flow doesn't push them sideways; it pushes them forward. This is called a coisotropic submanifold.

The paper focuses on a specific type of coisotropic dancer: the Algebraically Coisotropic one. This means the "flow" inside their group isn't chaotic; it's organized into neat, predictable lines (like a school of fish swimming in perfect rows).

The Big Question: Are They Just a Product?

The authors noticed something interesting about these organized groups. When the dancers are not "uniruled" (a fancy way of saying they aren't made of simple, straight lines that can be easily swept away), they seem to follow a very specific pattern.

The Question: If you zoom out and look at the whole dance floor, is it actually just two separate floors glued together?

Floor A: A smaller dance floor where the dancers are doing the "Perfect Balance" (Lagrangian) move.
Floor B: A separate, empty dance floor where nothing is happening.

The authors ask: Is every complex, organized group of dancers actually just a "Perfect Balance" group on one floor, sitting next to an empty floor?

The Findings: What They Proved

The authors didn't solve the puzzle for every possible dance floor, but they cracked it for several important cases.

1. The "Abelian" Dance Floor (The Grid)

First, they looked at a specific type of floor called an Abelian variety. Think of this as a floor that is shaped like a giant, multi-dimensional donut (a torus) or a perfect grid.

The Result: On these grid-like floors, the answer is YES.
The Analogy: If you have a group of dancers on a giant donut-shaped floor that is organized into neat rows, you can always cut the donut open and flatten it out. You will find that the dancers are actually just standing on a smaller, perfect donut (where they are balanced) while the rest of the space is just empty space.
The Catch: They also proved that on a very generic, random grid floor, you can't even find these "Perfect Balance" groups at all! It's like trying to find a flat spot on a perfectly round sphere; sometimes, the geometry just doesn't allow it.

2. The "General Type" Dancers (The Complex Ones)

Next, they looked at dancers who are "complex" (mathematically, having a "big" canonical bundle). These are the dancers who aren't simple lines or curves; they have a lot of internal structure.

The Result: If the dancers are complex enough, they must be "Perfect Balance" (Lagrangian).
The Analogy: Imagine a group of dancers who are so complex and tightly knit that they can't possibly be part of a flowing river. The only way they can exist on this magical floor is if they are in a state of perfect, static balance. The authors proved that if the group is "heavy" enough (complex enough), the "river" idea collapses, and they must be the "balanced" type.

3. The "Hyperkähler" Dance Floor (The Fancy One)

Finally, they looked at Irreducible Hyperkähler manifolds. These are the most fancy, complex, and "simply connected" dance floors (no holes, like a sphere).

The Result: Here, the answer is NO.
The Analogy: On these fancy floors, the dancers can form complex, flowing rivers that cannot be cut apart into a "balanced group" and an "empty floor." The geometry is too rich and twisted. The authors showed that there are many examples of these complex, flowing groups that don't fit the simple "product" pattern.

Why Does This Matter?

In the world of mathematics, we love to break complex things down into simple building blocks.

The Goal: The authors wanted to know if all these complex, organized groups of dancers could be explained as just a simple "balanced" group sitting next to some empty space.
The Conclusion:
- On Grid floors (Abelian): Yes, they are always just a simple group + empty space.
- On Fancy floors (Hyperkähler): No, they can be much more complex and twisted.
- On Complex groups: If the group is complicated enough, it forces itself to be a "balanced" group.

Summary in One Sentence

The paper investigates how organized groups of points can sit on special geometric surfaces, proving that on "grid-like" surfaces, these groups are always just simple balanced groups sitting next to empty space, but on "fancy" surfaces, they can form much more complex and twisted shapes.

A Note on the "Non-Projective" Example

At the very end, the authors mention a weird, non-projective example (a torus that isn't a grid). They show that if you have a magical "time-traveling" dance move (an automorphism) that scales the flow by a weird number, you can create a "Perfect Balance" group out of thin air. It's like a mathematical magic trick that works only in the shadows, not in the bright light of standard projective geometry.

Here is a detailed technical summary of the paper "On algebraically coisotropic submanifolds of holomorphic symplectic manifolds" by Ekaterina Amerik and Frédéric Campana.

1. Problem Statement and Context

The paper investigates algebraically coisotropic submanifolds $X$ within a projective holomorphic symplectic manifold $M$ (where $M$ has dimension $2n $and carries a holomorphic symplectic form$ \sigma$).

Definitions:
- A submanifold $X$ of codimension $c$ is coisotropic if the restriction $\sigma|_X$ has corank $c$ everywhere. This implies the kernel of $\sigma|_X$ defines a regular foliation $\mathcal{F}$ of rank $c$ , called the characteristic foliation.
- $X$ is algebraically coisotropic if this characteristic foliation is algebraic (its leaves are algebraic submanifolds), inducing a fibration $f: X \to B$ .
- If $c=n$ , $X$ is Lagrangian (the restriction $\sigma|_X$ vanishes identically).
The Central Question:
Motivated by previous results on hypersurfaces (Theorem 1.3), the authors ask whether a similar structural decomposition holds for higher-codimension submanifolds that are not uniruled.
Question 1.4: If $X \subset M$ is a non-uniruled algebraically coisotropic submanifold, does there exist a finite étale cover such that the pair $(X, M)$ decomposes as a product $(Z \times Y, N \times Y)$ , where $N, Y$ are holomorphic symplectic manifolds and $Z \subset N$ is a Lagrangian submanifold?

Essentially, the authors seek to reduce the classification of algebraically coisotropic submanifolds to that of Lagrangian submanifolds.

2. Methodology

The authors employ a combination of complex differential geometry, algebraic geometry (specifically the theory of minimal models and Iitaka fibrations), and Hodge theory.

Characteristic Fibration Analysis: They analyze the fibration $f: X \to B$ induced by the characteristic foliation. They utilize the fact that the base $B$ inherits a symplectic structure (on its smooth locus) and has Kodaira dimension $\kappa(B) = 0$ .
Isotriviality Proofs: A key technique involves proving that the fibration $f$ is isotrivial (all fibers are isomorphic) under specific conditions on the canonical bundle $K_X$ . This relies on adapting arguments from their previous work on hypersurfaces and incorporating recent results by Taji regarding minimal models.
Abelian Variety Structure: For the case where $M$ is an Abelian variety, they utilize Ueno's classification of subvarieties of Abelian varieties. This allows them to decompose $M$ into a product of tori and analyze the submanifold $X$ via its projection to a variety of general type.
Hodge Theory: To address the existence of Lagrangian submanifolds on Abelian varieties, they use the theory of Hodge groups. They demonstrate that for "sufficiently general" (Hodge-general) Abelian varieties, the Hodge structure is too rigid to allow for the vanishing of holomorphic 2-forms on submanifolds of dimension $\geq 2$ .

3. Key Contributions and Results

A. General Results on Algebraically Coisotropic Submanifolds

Theorem 1.7 & 1.8 (Isotriviality): If $X$ is algebraically coisotropic and $K_X$ is semiample (or more generally, if the fibers of the characteristic fibration have good minimal models), then the characteristic fibration $f: X \to B$ is isotrivial. Furthermore, the Kodaira dimension satisfies $\kappa(X) = \kappa(F)$ , where $F$ is a general fiber.
Corollary 1.9 (Lagrangian Criterion): If $K_X$ $K_{X}$ is nef and big (or $X$ $X$ is of general type), then $X$ $X$ must be Lagrangian. This generalizes the Hwang–Viehweg theorem (which dealt with hypersurfaces) to higher codimensions.
- Reasoning: If $X$ is of general type, the fiber $F$ is also of general type. By the isotriviality result, $X$ is essentially a product of $F$ and the base. However, the geometry of the symplectic form forces the base to be trivial in this specific context, collapsing $X$ to the fiber, which is Lagrangian.

B. The Case of Abelian Varieties

Theorem 1.11 (Structure on Abelian Varieties): If $M$ $M$ is an Abelian variety and $X \subset M$ $X \subset M$ is algebraically coisotropic, then after a finite étale cover, the pair $(X, M)$ $(X, M)$ splits as:
$M = D \times C \times N \times P$
$X = D \times C \times Z$
Where:
- $N$ and $D$ are holomorphic symplectic (subtori).
- $Z \subset N$ is a Lagrangian submanifold of general type.
- $C$ and $P$ are tori with $\dim(C) = \dim(P)$ .
- The symplectic form $\sigma$ splits accordingly, with $Z$ being Lagrangian with respect to the symplectic form on $N$ .
- Significance: This provides a positive answer to Question 1.4 specifically for Abelian varieties.

C. Non-Existence on General Abelian Varieties

Corollary 5.5: If $M$ $M$ is a Hodge-general Abelian variety (a generic condition stronger than simplicity), then $M$ $M$ contains no Lagrangian subvarieties of dimension $\geq 2$ $\geq 2$ .
- Reasoning: On a Hodge-general Abelian variety, the primitive cohomology is an irreducible module under the Hodge group. Any subvariety of dimension $\geq 2$ would force the pullback of the symplectic form to vanish, which contradicts the irreducibility of the Hodge structure unless the dimension is 1.
- Contrast: This stands in sharp contrast to Irreducible Holomorphic Symplectic (IHS) manifolds, where Lagrangian submanifolds are abundant (e.g., fibers of Lagrangian fibrations).

D. Irreducible Hyperkähler (IHS) Manifolds

The paper surveys known examples of Lagrangian submanifolds in IHS manifolds (e.g., Hilbert schemes of K3 surfaces, fixed loci of antisymplectic involutions).
Proposition 4.1: For a "vertical" subvariety $X = f^{-1}(S)$ inside a Lagrangian fibration $f: M \to B$ , if $X$ is algebraically coisotropic, then $S$ must be a point (i.e., $X$ is a fiber). This suggests that "vertical" coisotropic submanifolds are rare unless they are the fibers themselves.

E. Non-Projective Examples

Section 6: The authors construct a non-projective complex torus $T$ with an automorphism $g$ such that $g^*\sigma = \lambda \sigma$ with $|\lambda|=1$ and $\lambda$ not a root of unity. Using the graph of this automorphism, they exhibit smooth Lagrangian surfaces in $T \times T$ (and associated Kummer surfaces) for specific symplectic forms. This highlights that the projectivity assumption in the main theorems is crucial.

4. Significance

Structural Classification: The paper significantly advances the understanding of coisotropic submanifolds by proving that, under mild positivity conditions on the canonical bundle, they are essentially products of Lagrangian submanifolds and symplectic factors. This reduces a complex classification problem to the study of Lagrangian submanifolds.
Distinction between IHS and Abelian Cases: It rigorously establishes a fundamental difference between Irreducible Holomorphic Symplectic manifolds and Abelian varieties regarding the existence of Lagrangian submanifolds. While IHS manifolds are often covered by Lagrangians, generic Abelian varieties are not.
Generalization of Known Theorems: It extends the Hwang–Viehweg theorem (regarding hypersurfaces of general type) to arbitrary codimensions.
Hodge-Theoretic Obstructions: The application of Hodge theory to rule out Lagrangian submanifolds on general Abelian varieties provides a powerful new tool for studying subvarieties of Abelian varieties.

In summary, Amerik and Campana provide a comprehensive framework for analyzing algebraically coisotropic submanifolds, proving that non-uniruled examples are structurally simple (products involving Lagrangians) and demonstrating that such structures are highly constrained or non-existent in the generic Abelian setting.