Determinantal point processes on complex manifolds: Construction and limit theorems

This paper establishes a coordinate-free probabilistic framework for determinantal point processes on compact complex manifolds by rigorously defining scalar determinants for line-bundle-valued Bergman kernels, proving that finite-dimensional spaces of sections generate such processes, and deriving transfer principles that convert analytic asymptotics into probabilistic limit theorems.

The Gist

The Big Picture: A New Way to Count Points on Curved Surfaces

Imagine you are trying to scatter a specific number of dots randomly across a curved surface, like a sphere or a donut. But these aren't just random dots; they are "repelling" each other. If one dot is here, it makes it very unlikely for another dot to be right next to it. This is a Determinantal Point Process (DPP).

In the world of mathematics, these processes are famous for showing up in random matrix theory (like shuffling cards) and quantum physics (like electrons in a magnetic field). Usually, mathematicians describe these dots using simple numbers (scalars).

The Problem:
This paper tackles a specific, tricky situation: What if the surface you are working on is a complex manifold (a very fancy, multi-dimensional curved shape) and the "dots" are actually sections of a line bundle?

Think of a line bundle like a collection of tiny, invisible strings attached to every point on the surface. The "value" of a dot isn't just a number; it's a value attached to that specific string. Because these strings can twist and turn as you move around the surface, you can't just multiply them together to get a simple number. It's like trying to calculate the volume of a room where the walls are made of shifting, rotating mirrors. The usual math formulas break down because they expect simple numbers, not these twisted, string-based values.

The Solution: The "Intrinsic" Calculator

The author, Thibaut Lemoine, invents a new, coordinate-free way to do the math.

The Analogy:
Imagine you have a group of people standing in a circle, each holding a unique colored ribbon. You want to know the "total pattern" of their ribbons.

• Old Way: You ask everyone to describe their ribbon relative to a specific wall in the room. If you move the wall (change coordinates), everyone's description changes, and the math gets messy.
• Lemoine's Way: Instead of looking at the ribbons relative to a wall, you look at how the ribbons interact with each other directly. You calculate the "pattern" based on the relationships between the people, regardless of where the room is or how the walls are painted.

He defines a special kind of determinant (a mathematical operation usually used to find areas or volumes) that works directly on these twisted strings. This "intrinsic determinant" gives a single, honest number that doesn't depend on how you choose to look at the surface.

The Main Result: The "Bergman Ensemble"

Using this new calculator, the paper proves that if you take a specific collection of mathematical functions (called holomorphic sections) on a complex shape, they naturally form a DPP.

• The Ensemble: Think of this as a "Bergman Ensemble." It's a specific type of random point pattern.
• The Physics Connection: The paper mentions that this is the mathematical description of fermions (particles like electrons) in a magnetic field. In the "Integer Quantum Hall Effect," these particles fill up the lowest energy levels. The "dots" are the positions of these particles. The "twisted strings" represent the fact that the particles' wavefunctions change phase as they move (gauge covariance). The author's new determinant is the "gauge-invariant" way to count them—meaning the answer is the same no matter how you choose to measure the magnetic field.

The "Transfer Principles": A Dictionary for Math

The second half of the paper is like a dictionary or a translator. It shows how to take known facts about the "strings" (the Bergman kernels) and translate them into facts about the "dots" (the probability of where the points land).

The paper creates a list of rules, such as:

1. If the strings get denser in a certain way... $\rightarrow$ Then the dots will spread out evenly across the surface. (This is the "Law of Large Numbers").
2. If the strings wiggle in a specific pattern near a point... $\rightarrow$ Then the dots will look like a specific, universal pattern (like a crystal lattice) when you zoom in very close. (This is "Local Universality").
3. If you remove a few dots from the pattern... $\rightarrow$ The remaining dots rearrange themselves according to a specific rule (Schur complements), which is mathematically the same as forcing the strings to be zero at those removed points.

Why This Matters (According to the Paper)

The paper doesn't claim to discover new physics or solve a medical problem. Instead, it claims to provide a rigorous, clean framework.

• Before: Mathematicians had to do messy calculations by picking a specific "frame of reference" (like choosing a specific wall to measure ribbons against) and hoping the errors canceled out.
• Now: They can use this "intrinsic" method. It's like having a universal translator that works no matter what language (or geometry) you are speaking.

The author emphasizes that this framework allows them to recover known results (like those by Berman) but in a way that is mathematically "pure" and doesn't rely on arbitrary choices. It also sets the stage for future work: if someone discovers a new way the "strings" behave (new analytic input), this "dictionary" can immediately tell us what that means for the "dots" (the probabilistic outcome).

Summary in One Sentence

Thibaut Lemoine has built a new, coordinate-free mathematical tool that allows us to rigorously describe how random points repel each other on complex, curved surfaces, translating deep geometric properties of "twisted strings" into clear predictions about where those points will land.

Technical Summary

Technical Summary: Determinantal Point Processes on Complex Manifolds

Problem Statement
The paper addresses the challenge of formulating a coordinate-free probabilistic framework for Determinantal Point Processes (DPPs) associated with Bergman kernels on compact complex manifolds. In standard DPP theory, correlation functions are defined by the determinant of a scalar kernel $K(x, y)$. However, on a complex manifold $M$ equipped with a Hermitian holomorphic line bundle $L$, the Bergman kernel $B_k(x, y)$ is naturally a section of $\text{Hom}(L^k_y, L^k_x)$, not a scalar function. Consequently, the expression $\det(B_k(x_i, x_j))$ lacks an immediate scalar meaning. While local trivializations allow one to define scalar kernels, these depend on the choice of gauge (local frame), raising the question of whether the resulting point process is intrinsically defined on the manifold or merely an artifact of a specific coordinate choice.

Methodology
The author develops a rigorous framework that extends finite-rank projection DPPs to the setting of line-bundle-valued kernels. The methodology proceeds in three stages:

1. Intrinsic Construction: The paper defines the "intrinsic determinant" of a line-bundle-valued matrix. For points $x_1, \dots, x_m$, the kernel defines an endomorphism on the direct sum of fibers $L_{x_1} \oplus \dots \oplus L_{x_m}$. The correlation function is defined as the determinant of this endomorphism. This scalar quantity is shown to be independent of local frames, thereby establishing the existence of a genuine, coordinate-free DPP.
2. Transfer Principles: The paper isolates a set of finite-dimensional probabilistic identities (transfer principles) that convert analytic asymptotics of projection kernels into probabilistic limit theorems. These principles map specific types of kernel estimates (diagonal, off-diagonal, near-diagonal, Schur complements, and trace expansions) to specific probabilistic behaviors (law of large numbers, local universality, Palm limits, variance bounds, and large deviations).
3. Application to Bergman Ensembles: These abstract principles are applied to the space of holomorphic sections $H^0(M, L^k)$ as $k \to \infty$. The analytic inputs are drawn from established results in Bergman kernel theory, Berezin–Toeplitz calculus, and pluripotential theory.

Key Contributions and Results

• Intrinsic Definition of the Bergman Ensemble: The paper proves that any finite-dimensional Hilbert space of sections of a Hermitian line bundle gives rise to a genuine projection DPP. Theorem 2.3.2 establishes that the correlation functions are given by the intrinsic determinant of the projection kernel. This construction is shown to be the geometric analogue of orthogonal polynomial ensembles and is physically interpreted as the position process of non-interacting fermions filling the lowest Landau level (integer quantum Hall effect).
• Modular Interface: The paper provides a "dictionary" linking analytic inputs to probabilistic outputs:
  • Diagonal Bergman expansion $\to$ Law of large numbers for empirical measures.
  • Near-diagonal expansion $\to$ Local correlation and fluctuation expansions (universality).
  • Off-diagonal localization $\to$ Variance bounds for linear statistics.
  • Schur-complement asymptotics $\to$ Palm limits (conditioning on point presence corresponds to imposing zeros on holomorphic sections).
  • Toeplitz trace expansions $\to$ Cumulant expansions.
  • Gram determinant asymptotics $\to$ Large deviation principles.
• Limit Theorems:
  • Macroscopic Limit: The empirical measure of the Bergman ensemble converges in probability to the normalized curvature volume form (Theorem 4.3.2).
  • Local Universality: Rescaled local point processes converge to the Bargmann–Fock DPP, with complete asymptotic expansions for joint moments and cumulants (Theorem 4.3.6).
  • Cumulants: Linear statistics exhibit a rigidity where the leading order of cumulants of order $\ell \ge 2$ vanishes, with fluctuations appearing only at subprincipal orders (Proposition 4.3.10).
  • Large Deviations: The paper derives a large deviation principle for empirical measures with speed $k N_k$, where the rate function is determined by the equilibrium energy of the associated pluripotential problem (Corollary 4.3.13).

Significance and Claims
The paper claims its primary significance lies not in proving new asymptotic estimates for Bergman kernels (which are taken from existing literature such as Ma–Marinescu and Berman), but in formulating the intrinsic projection-DPP framework that underlies these results.

• Coordinate-Free Rigor: It resolves the ambiguity of defining DPPs on line bundles by providing a gauge-invariant definition of the correlation functions, distinguishing this setting from scalar Bergman-kernel DPPs on domains.
• Modularity: It explicitly separates the analytic input (kernel asymptotics) from the probabilistic mechanism (finite-dimensional transfer principles). This modularity implies that any future improvement in analytic estimates (e.g., for partial Bergman kernels) can be immediately inserted into the same framework to yield new probabilistic limit theorems.
• Physical Interpretation: The work clarifies the connection between complex geometry and fermionic many-body states, identifying the Bergman ensemble as the geometric realization of the integer quantum Hall effect, where the line-bundle-valued nature of the kernel corresponds to gauge covariance.

The author notes that while some consequences (such as large deviations) were previously known under weaker assumptions via Berman's work, this paper provides a unified, intrinsic, and modular perspective that makes the mechanism of the transfer from analysis to probability explicit.