🔢 mathematics

Lower bounds on fibered Yang-Mills functionals: generic nefness and semistability of direct images

This paper establishes lower bounds for fibered Yang-Mills functionals on polarized fibrations by relating them to the Harder-Narasimhan slopes of direct image sheaves, thereby providing an analytic characterization of generic nefness and refining obstructions for metrics with constant horizontal mean curvature through the study of semiclassical limits.

Original authors: Siarhei Finski

Published 2026-03-25

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Siarhei Finski

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 trying to build a massive, multi-story library. This library isn't just a single building; it's a complex structure where every floor (the "fibers") is a unique, intricate room, and all these rooms are stacked on top of each other to form a tower (the "fibration").

Your goal is to make sure the entire structure is perfectly balanced and stable. In the world of mathematics, this stability is measured by something called Yang-Mills functionals. Think of these as "stress tests" or "balance scales" that tell you if the building is leaning too much to one side or if the tension in the beams is uneven.

This paper, written by Siarhei Finski, is about finding the lowest possible stress this building can have. It turns out that the answer to "how stable can this building be?" is hidden in the algebraic DNA of the structure itself.

Here is the breakdown of the paper's big ideas using everyday analogies:

1. The Two Languages: Geometry vs. Algebra

The paper connects two different ways of looking at the same problem:

The Geometric View (The Physical Building): This looks at the actual shape, curves, and tension of the building. It asks: "Is the roof flat? Are the walls straight?" In math terms, this is about curvature and metrics (how we measure distance and angles).
The Algebraic View (The Blueprint): This looks at the building as a collection of data points and rules. It asks: "Is the blueprint balanced? Do the parts fit together without conflict?" In math terms, this involves Harder-Narasimhan filtrations (a way of sorting the building's components from "strongest" to "weakest") and slopes (a measure of how much "stuff" is in a specific section).

The Big Discovery: Finski proves that the physical stress on the building (Geometry) is directly determined by the balance of the blueprint (Algebra). You can't fix the physical building without understanding the algebraic blueprint.

2. The "Semiclassical Limit": Zooming In with a Microscope

One of the paper's main tricks is a concept called the semiclassical limit.

The Analogy: Imagine you have a digital photo of a painting. If you zoom in too far, you just see pixels. If you zoom out, you see the whole picture. But what if you could look at the painting through a special microscope that shows you the painting as a collection of billions of tiny, vibrating dots?
The Math: The author looks at the library not as one big building, but as a collection of layers (tensor powers) that get finer and finer. As he zooms in infinitely (letting a number $k$ go to infinity), the complex, wiggly physics of the building starts to look like a smooth, predictable algebraic pattern.
The Result: By looking at these "microscopic" layers, he can predict the "macroscopic" stress of the whole building.

3. The "Stress Test" and the "Blueprint"

The paper establishes a Lower Bound. This is like saying, "No matter how you try to design this building, you cannot make the stress lower than X."

The Blueprint's "Slope": The author calculates the "slope" of the building's components. Imagine the building is made of bricks. Some bricks are heavy (high slope), some are light (low slope). The "Harder-Narasimhan filtration" is a way of sorting these bricks from heaviest to lightest.
The Connection: The paper proves that the minimum stress the building can ever have is exactly equal to the difference between the heaviest and lightest bricks in the blueprint.
- If the blueprint is perfectly balanced (all bricks are the same weight), the building can be perfectly stable (zero stress).
- If the blueprint is unbalanced (some bricks are huge, some tiny), the building must have some stress. You can't fix it; it's mathematically impossible.

4. "Generic Nefness": The "Good Enough" Test

The paper also introduces a concept called Generic Nefness.

The Analogy: Imagine you are checking if a library is safe. You don't need to check every single bookshelf in the entire building. You just need to check a few random, representative shelves. If those are stable, the whole building is likely "generically" stable.
The Math: The author shows that if the "stress test" passes on these random, representative slices of the building, then the whole structure is "nef" (a fancy math word for "positively curved" or "stable"). This gives mathematicians a shortcut: they don't need to solve the impossible equation for the whole building; they just need to check the "generic" parts.

5. Why Does This Matter?

Why should a non-mathematician care?

Optimization: In physics and engineering, we often want to find the "most efficient" state of a system (like the most stable shape for a bridge). This paper gives a rule for the absolute limit of that efficiency.
Predicting Failure: If the "blueprint" (the algebraic data) shows a huge imbalance, we know immediately that the physical structure will fail or be unstable, no matter how much we try to tweak the materials.
Unifying Worlds: It bridges the gap between the messy, continuous world of shapes (calculus/geometry) and the discrete, logical world of numbers and rules (algebra). It shows they are two sides of the same coin.

Summary

Think of this paper as a master key. It tells us that to understand the physical stability of a complex, multi-layered structure, we don't need to measure every inch of it. Instead, we just need to look at its algebraic blueprint, sort its components by weight, and calculate the difference between the heaviest and lightest. That difference tells us the minimum amount of stress the structure must endure. If the blueprint is unbalanced, the structure will never be perfectly calm.

1. Problem Statement and Context

The paper addresses the relationship between differential-geometric invariants of holomorphic fibrations and algebraic invariants of the associated direct image sheaves. Specifically, it investigates the behavior of the horizontal mean curvature of a polarized fibration $\pi: X \to B$ in the semiclassical limit ( $k \to \infty$ ).

The Setting: Let $\pi: X \to B$ be a holomorphic submersion between compact complex manifolds ( $\dim X = n+m, \dim B = m$ ). Let $L$ be a holomorphic line bundle on $X$ that is relatively ample.
The Geometric Object: The paper studies the fibered Einstein equation, which requires the horizontal mean curvature $\wedge_{\omega_B} \omega_H$ to be constant. This generalizes the Hermite-Einstein equation for vector bundles to the setting of fibrations.
The Algebraic Object: The paper considers the direct image sheaves $E_k := R^0\pi_* L^k$ (for large $k$ ) and their Harder-Narasimhan (HN) filtrations and slopes with respect to a Gauduchon metric $\omega_B$ on the base $B$ .
The Core Question: How do the lower bounds of the fibered Yang-Mills functionals (measuring the deviation from the fibered Einstein condition) relate to the asymptotic distribution of the HN slopes of the direct images $E_k$ ?

2. Methodology

The author employs a synthesis of Toeplitz operator theory, Hermitian Yang-Mills (HYM) theory, and asymptotic Riemann-Roch-Grothendieck theorems.

Semiclassical Limit via Toeplitz Operators:
The central technical tool is the work of Ma and Zhang, which establishes that the curvature of the direct image bundle $E_k$ (equipped with the $L^2$ -metric induced by $L^k$ ) is asymptotically approximated by Toeplitz operators acting on the fiber. Specifically, the curvature term $\frac{\sqrt{-1}}{2\pi} \wedge_{\omega_B} R_{E_k}$ converges to $k \cdot T_k(\omega_H)$ , where $T_k$ is the Toeplitz operator associated with the horizontal curvature form $\omega_H$ .
Reduction to Vector Bundle Theory:
By leveraging the asymptotic expansion of the curvature of $E_k$ , the problem of bounding the fibered Yang-Mills functional on the total space $X$ is reduced to bounding the standard Hermitian Yang-Mills functional on the vector bundles $E_k$ over the base $B$ .
Harder-Narasimhan Slopes:
The author utilizes the unique Harder-Narasimhan filtration of the coherent sheaves $E_k$ . The slopes of the graded pieces of this filtration define a probability measure $\eta_k^{HN}$ on $\mathbb{R}$ . The paper proves the weak convergence of these measures to a limit measure $\eta^{HN}$ as $k \to \infty$ .
Asymptotic Riemann-Roch-Grothendieck:
To connect the analytic slopes to algebraic geometry, the paper establishes an asymptotic version of the Riemann-Roch-Grothendieck theorem in the Bott-Chern cohomology setting. This allows for the calculation of the asymptotic slopes of direct images even in the presence of singular subspaces, extending previous results that required flatness assumptions.

3. Key Contributions and Results

A. Lower Bounds on Fibered Yang-Mills Functionals (Theorem 1.7)

The paper establishes rigorous lower bounds for the fibered Yang-Mills functional $FYM_{p,c}(\pi, h_L)$ in terms of the limit measure $\eta^{HN}$ of the Harder-Narasimhan slopes.

$L^\infty$ Bound: The supremum and infimum of the horizontal mean curvature are bounded by the essential supremum and infimum of the limit measure $\eta^{HN}$ :
$\inf \wedge_{\omega_B}\omega_H \leq \eta^{HN}_{\min} \quad \text{and} \quad \sup \wedge_{\omega_B}\omega_H \geq \eta^{HN}_{\max}$
$L^p$ Bound: For $p \in [1, \infty)$ , the functional is bounded below by the $L^p$ -distance between the constant $c$ and the measure $\eta^{HN}$ :
$FYM_{p,c} \geq \int_{\mathbb{R}} |x-c|^p d\eta^{HN}(x) \cdot \int_X [\omega^n] \cdot \pi^*[\omega_B^m]$
This result generalizes the classical Atiyah-Bott lower bounds for vector bundles to the fibered setting.

B. Asymptotic Semistability and Fibered Einstein Metrics (Theorem 1.1)

The paper proves that the existence of approximate fibered Einstein metrics (metrics where the horizontal mean curvature is arbitrarily close to a constant) implies the asymptotic semistability of the direct image bundles $E_k$ .

If approximate fibered Einstein metrics exist, then for large $k$ , the bundles $E_k$ are "almost" semistable (the difference between the slope of any quotient and the slope of $E_k$ vanishes as $k \to \infty$ ).
This establishes a tight correspondence between the fibered Einstein equation on $L$ and the Hermite-Einstein equation on $E_k$ in the semiclassical limit.

C. Numerical Obstructions and Generic Nefness (Theorems 1.3 & 1.12)

The author derives numerical obstructions for semistability using the slopes of irreducible analytic subspaces $Y \subset X$ .

Numerical Semistability: If $E_k$ are asymptotically semistable, then the fibration is "numerically semistable," meaning the slope $\mu(Y)$ of any subfamily $Y$ projecting surjectively to $B$ satisfies $\mu(Y) \geq \mu(X)$ .
Analytic Characterization of Generic Nefness: The paper provides a novel analytic characterization of stably generically fibered nef line bundles. A relatively nef line bundle $L$ is stably generically fibered nef if and only if there exist metrics with arbitrarily small negative horizontal curvature (Theorem 1.12). This bridges the gap between algebraic positivity (nefness) and analytic curvature bounds in the fibered context.

D. Conjecture on Tightness

The paper conjectures that the lower bounds derived in Theorem 1.7 are tight (i.e., the infimum of the functional over all metrics equals the integral against $\eta^{HN}$ ). This is proven for the case $p=1$ (in a recent related work) and for the $L^\infty$ case regarding the infimum of the horizontal curvature (Theorem 1.10).

4. Significance and Impact

Unification of Geometric and Algebraic Invariants: The paper successfully unifies the study of fibered Einstein metrics (differential geometry) with the stability of direct images (algebraic geometry) via the semiclassical limit. It extends the well-known correspondence between constant scalar curvature metrics and K-stability to the fibered setting.
Generalization to Gauduchon Metrics: Unlike many previous results that required the base $B$ to be Kähler, this work operates in the broader context of Gauduchon metrics (where $\partial\bar{\partial}\omega_B^{m-1}=0$ ). This is crucial for applications where the base manifold is not Kähler.
New Obstructions for Metrics: The derived lower bounds provide new, sharp obstructions for the existence of fibered Einstein metrics. If the algebraic slopes of the direct images are too spread out (i.e., $\eta^{HN}_{\min} < \eta^{HN}_{\max}$ ), no fibered Einstein metric can exist.
Extension of HYM Theory: By utilizing Toeplitz operators and asymptotic expansions, the paper effectively lifts the classical Atiyah-Bott and Donaldson theory of Hermitian Yang-Mills functionals to the realm of fibered manifolds, offering a powerful toolkit for studying moduli spaces of fibrations.
Resolution of Singularities: The development of the asymptotic Riemann-Roch-Grothendieck theorem in the singular setting (Theorem 4.1) allows these results to apply to subfamilies with singularities, broadening the scope of applicability in complex geometry.

In summary, Finski's work provides a rigorous framework for understanding the stability of fibrations through the lens of semiclassical analysis, offering precise lower bounds on curvature functionals and deepening the connection between analytic metrics and algebraic stability conditions.