🔢 mathematics

On the Localization of the Bergman Kernel and applications to Toeplitz theory

This paper establishes the localization of Bergman kernels for big line bundles on compact complex manifolds with Bernstein-Markov measures, confirming a conjecture by Zelditch and demonstrating that the resulting Toeplitz operators form an algebra with equidistributing spectra.

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

The Big Picture: The "Spotlight" Effect

Imagine you are in a large, dark concert hall (the Complex Manifold). You have a massive choir of singers (the Sections of a Line Bundle). Each singer can stand anywhere in the hall, but they are only allowed to sing if they are part of a specific group defined by the rules of the room.

Now, imagine you want to focus a spotlight on just one specific person in the audience (a specific point $x$ ). You ask the choir to arrange themselves so that the combined sound is loudest exactly at that person's ear, while being as quiet as possible everywhere else.

The Bergman Kernel is the mathematical description of this "perfect spotlight." It tells you how the energy of the choir is distributed when you try to focus on one spot.

The Main Discovery:
The paper proves a fascinating phenomenon called Localization. As the choir gets larger and larger (mathematically, as the power $k$ goes to infinity), the spotlight becomes incredibly sharp.

The Intuition: If you try to focus the light on person A, the light doesn't just stay on A; it spills over a little bit.
The Result: The paper proves that as the choir grows, that "spill" disappears. The light becomes so concentrated that if you look at any two people who are not standing right next to each other, the connection between them vanishes. The energy is entirely "localized" on the diagonal (where person A looks at person A).

The "Ghost" Problem (Measures and Pluripolar Sets)

Usually, in math, we assume the audience is spread out nicely (like a smooth volume of air). But what if the audience is weird? What if they are clustered on a thin wire, or a scattered set of points that looks like dust?

In complex geometry, there are "ghostly" sets called Pluripolar Subsets. Think of these as invisible, zero-volume islands. If your audience (the Measure) is entirely on these ghosts, the math breaks down.

The Paper's Breakthrough:
Finski shows that even if the audience is weird, as long as they aren't completely stuck on these invisible ghosts, the "Spotlight Effect" still works. The light still concentrates perfectly on the diagonal. This is a huge generalization because it works for very messy, irregular distributions of points, not just smooth, perfect spheres.

The "Bernstein-Markov" Super-Condition

The paper also looks at a special, very well-behaved type of audience called Bernstein-Markov measures.

Analogy: Imagine a crowd where the loudest singer is never too much louder than the average singer, no matter how big the choir gets. They are "well-behaved."
The Result: For these well-behaved crowds, the paper confirms a guess made by a famous mathematician named Zelditch. It proves that not only does the light concentrate, but the shape of the light distribution converges to a perfect, predictable "Equilibrium Measure." It's like the spotlight eventually settles into a perfect, stable pattern that mathematicians can calculate exactly.

The Application: The "Toeplitz" Machine

Why do we care about this spotlight? Because it powers a machine called Toeplitz Operators.

The Analogy:
Think of a Toeplitz operator as a filter or a mixing board.

You have a song (a function $f$ ).
You run it through the machine (the operator $T_k(f)$ ).
The machine processes the song based on the rules of the choir.

The "Algebra" Discovery:
In the real world, if you mix two songs together, the result is usually messy. But this paper proves that for these specific Toeplitz machines, as the choir gets huge, the mixing becomes perfectly predictable.

If you run Song A through the machine, then Song B, it is almost exactly the same as running "Song A times Song B" through the machine once.
The Metaphor: It's like having a blender. If you blend apples, then blend bananas, you get a mess. But this paper says: "If you have a super-blender with infinite speed, blending apples then bananas is exactly the same as blending a pre-mixed apple-banana smoothie." This allows mathematicians to treat these complex operators like simple numbers.

The "Spectrum" and the "Equidistribution"

Finally, the paper looks at the Spectrum of these machines.

The Analogy: Every machine has a set of "natural frequencies" or "notes" it can play (its spectrum).
The Result: For the well-behaved crowds (Bernstein-Markov), the paper proves that these notes spread out perfectly evenly across the range of possible sounds. They don't clump together; they equidistribute.
Why it matters: This is a modern, high-dimensional version of a famous 100-year-old theorem by Szegő about polynomials. It tells us that even in the most complex, multi-dimensional geometric spaces, the "notes" of these operators follow a simple, fair distribution law.

Summary in One Sentence

This paper proves that in complex geometric spaces, even with messy or strange distributions of points, the mathematical "spotlight" (Bergman kernel) eventually focuses so tightly on a single point that it ignores everything else, allowing us to predict the behavior of complex mathematical machines (Toeplitz operators) with perfect accuracy.

1. Problem Statement and Context

The paper addresses the asymptotic behavior of the Bergman kernel (also known as the Christoffel-Darboux kernel) associated with a big line bundle $L$ over a compact complex manifold $X$ , equipped with a continuous Hermitian metric $h_L$ and a general Borel measure $\mu$ .

Core Problem:
Traditionally, the localization of the Bergman kernel (the concentration of mass near the diagonal $x=y$ as the tensor power $k \to \infty$ ) has been studied under "regular" assumptions, such as $\mu$ being a smooth volume form, $L$ being ample, and $h_L$ being smooth and strictly positive. In these settings, off-diagonal decay is often exponential.

Finski investigates whether this localization phenomenon holds under much weaker assumptions:

$L$ is only big (not necessarily ample).
$\mu$ is a general Borel measure that does not give full mass to pluripolar subsets (a condition weaker than being non-pluripolar).
The metric $h_L$ is merely continuous.

The specific questions are:

Does the off-diagonal Bergman measure $\mu_k^{\text{Berg}}$ asymptotically vanish away from the diagonal for such general measures?
Does the diagonal Bergman measure converge to a specific equilibrium measure?
Do these localization properties imply that Toeplitz operators form an asymptotic algebra and exhibit spectral equidistribution?

2. Methodology

The author employs a combination of complex geometry, pluripotential theory, and functional analysis. The methodology is structured into several key steps:

A. Reduction and Truncation

Subvariety Reduction: The proof of the general localization theorem (Theorem 1.1) is reduced to the case where the measure $\mu$ is supported away from a fixed subvariety $Y$ . This is achieved by showing that truncating the measure near subvarieties (or more generally, closed sets that do not carry mass in the limit) does not significantly alter the off-diagonal behavior of the kernel.
Comparison of Kernels: The author utilizes a comparison lemma (inspired by Lubinsky) to bound the difference between Bergman kernels associated with different measures. Specifically, if $\mu_1 \leq \mu_2$ , the $L^2$ -difference of their kernels is controlled by the difference in their diagonal values.

B. Quantitative Decay via Operator Theory

Hilbert-Schmidt Norms: To establish decay for measures supported away from divisors, the author constructs a sequence of operators $S_k$ involving the commutator of the Bergman projection and multiplication by a rational function $f = s_1/s_2$ .
Codimension Argument: By analyzing the kernel of these operators, the author shows they have a codimension bounded by $n_k - n_{k-k_0}$ . This allows bounding the Hilbert-Schmidt norm of the kernel, leading to a quantitative estimate of the off-diagonal mass.

C. Pluripotential Theory and Equilibrium Measures

Partial Bergman Kernels: The paper introduces partial Bergman kernels associated with vanishing along subvarieties (or pluripolar sets). It proves that these kernels decay exponentially in neighborhoods of the singular sets.
Volume and Energy Functionals: The proof relies on the continuity of the volume functional for big line bundles (Fujita's theorem) and the properties of the Monge-Ampère energy functional (Berman-Boucksom theory).
Bernstein-Markov Property: For measures satisfying the Bernstein-Markov property, the author leverages existing results (Berman-Boucksom-Witt Nyström) regarding the convergence of diagonal measures to the equilibrium measure $\mu_{\text{eq}}$ .

D. Toeplitz Operator Theory

Asymptotic Algebra: The localization results are translated into the language of Toeplitz operators $T_k(f) = B_k \circ M_k(f)$ . The off-diagonal decay of the kernel is shown to imply that the commutator $[T_k(f), T_k(g)]$ and the difference $T_k(f)T_k(g) - T_k(fg)$ vanish in the Schatten $p$ -norm as $k \to \infty$ .
Spectral Equidistribution: Using the trace formula and the convergence of diagonal measures, the spectral distribution of Toeplitz operators is linked to the equilibrium measure.

3. Key Contributions and Results

Main Theorems

Theorem 1.1 (General Localization):
For any Borel measure $\mu$ that does not give full mass to pluripolar subsets, the associated Bergman measures $\mu_k^{\text{Berg}}$ do not asymptotically place mass away from the diagonal.
$\lim_{k \to \infty} \int_K d\mu_k^{\text{Berg}} = 0$
for any compact $K \subset X \times X$ disjoint from the diagonal. This confirms a conjecture by Zelditch in a broad setting.
Theorem 1.3 (Bernstein-Markov Convergence):
If $\mu$ is a Bernstein-Markov measure with non-pluripolar support, the measures $\mu_k^{\text{Berg}}$ converge weakly to the diagonal embedding of the equilibrium measure:
$\mu_k^{\text{Berg}} \rightharpoonup \Delta_* \mu_{\text{eq}}(K, h_L).$
This resolves a question left open by Zelditch regarding the limiting off-diagonal measure.
Theorem 1.5 (Weight Variation):
The asymptotic value of the diagonal Bergman measure is locally determined. If two measures $\mu_1$ and $\mu_2$ differ by a continuous weight factor $e^f$ , their associated diagonal Bergman measures converge to the same limit (weakly).
Theorem 1.7 (Asymptotic Algebra):
For the general class of measures described above, Toeplitz operators form an asymptotic algebra. Specifically, for $f, g \in C^0(K, \mathbb{R})$ :
$\lim_{k \to \infty} \| T_k(f) \circ T_k(g) - T_k(f \cdot g) \|_p = 0$
where $\|\cdot\|_p$ is the Schatten $p$ -norm. This holds even when the operator norm convergence fails.
Theorem 1.9 (Spectral Equidistribution):
For Bernstein-Markov measures, the spectral measures of Toeplitz operators equidistribute according to the equilibrium measure:
$\lim_{k \to \infty} \frac{1}{n_k} \sum_{\lambda \in \text{Spec}(T_k(f))} g(\lambda) = \int_X g(f(x)) d\mu_{\text{eq}}(x).$
This generalizes the Szegő limit theorem and results by Boutet de Monvel-Guillemin to singular and big line bundle settings.

Technical Extensions

Singular Spaces: The results are extended to compact complex reduced irreducible analytic spaces (singular spaces) by using resolutions of singularities and pull-backs of metrics.
Orthogonal Polynomials: The paper recovers and refines classical results on orthogonal polynomials (e.g., on the unit circle or interval $[-1, 1]$ ) as special cases of the general manifold theory.

4. Significance and Impact

Generality: The paper significantly broadens the scope of Bergman kernel asymptotics. It moves away from the restrictive "smooth/ample" regime to include big line bundles and singular measures, provided they are not concentrated on pluripolar sets.
Confirmation of Conjectures: It confirms Zelditch's conjecture regarding the limiting off-diagonal measure for Bernstein-Markov measures, providing a rigorous proof where only heuristic or partial results existed.
Unification of Fields: It creates a unified framework connecting complex geometry (pluripotential theory), approximation theory (orthogonal polynomials), and operator theory (Toeplitz operators).
Robustness of Toeplitz Theory: A major insight is that the "asymptotic algebra" property of Toeplitz operators is robust. It does not require the strong spectral gap properties of the Kodaira Laplacian (which fail for semi-positive metrics) but relies instead on the localization of the kernel in an averaged (Schatten norm) sense.
New Tools: The introduction of the "forbidden region" concept for partial Bergman kernels and the use of Hilbert-Schmidt norm estimates for off-diagonal decay provide new technical tools for future research in complex geometry and spectral theory.

In summary, Finski's work establishes that the localization of the Bergman kernel is a fundamental phenomenon driven by the geometry of the line bundle and the non-pluripolarity of the measure, rather than the smoothness of the metric or the specific nature of the measure, thereby unifying diverse areas of mathematical analysis.