⚛️ high-energy theory

Numerical Sasaki--Einstein metrics and harmonic forms on del Pezzo links

This paper presents a numerical construction of the Sasaki–Einstein metric and harmonic forms on the link of the second del Pezzo surface using polynomial expansions and neural networks to approximate the metric's convex potential, achieving high precision and validating the results against known analytical and numerical benchmarks for applications in holographic IIB backgrounds.

Original authors: Nakwoo Kim, Sejin Kim, Hoseob Shin

Published 2026-09-23

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Nakwoo Kim, Sejin Kim, Hoseob Shin

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

In the vast landscape of theoretical physics, where the fabric of reality is often described as a ten-dimensional tapestry woven from vibrating strings, there exists a specific kind of geometric shape that acts as a bridge between two seemingly different worlds. One world is a universe with gravity, described by the laws of general relativity, and the other is a quantum world without gravity, described by the rules of particle physics. To connect these two, physicists need to understand the shape of the extra dimensions that are curled up so tightly we cannot see them. These shapes are not random; they must satisfy very strict mathematical rules to be stable and to allow for the existence of the particles we observe. Among the most important of these shapes are those that look like cones, with a smooth, five-dimensional surface at their tip. Finding the precise geometry of these surfaces is like finding the perfect blueprint for a building that must stand forever without collapsing, a task that has stumped mathematicians for decades because the equations governing these shapes are incredibly complex and rarely have neat, written-down solutions.

A team of researchers has now successfully mapped the geometry of one such elusive shape, a surface known as the link of a cone over a specific type of polygon called the second del Pezzo surface. For years, this particular shape was a mystery. While mathematicians knew that a perfect, stable version of this shape must exist, no one had ever been able to write down the exact formula that describes its curvature. The difficulty arose because the shape does not possess the kind of symmetry that usually makes these problems solvable by hand. Instead of a smooth, regular pattern, the shape has a more irregular structure, making the equations governing it resistant to traditional algebraic methods. The researchers approached this problem not by trying to solve the equations with pen and paper, but by using powerful computers to approximate the solution numerically. They treated the problem as a search for the perfect surface that minimizes a specific kind of energy, much like a soap film stretching across a wire frame finds the shape with the least surface area.

The team developed a new method to navigate this search, using a combination of advanced polynomial expansions and artificial intelligence networks to build a model of the surface. They started with a rough guess based on known mathematical principles and then let the computer iteratively refine the shape, adjusting the curve millions of times until the equations were satisfied to an extraordinary degree of precision. The result was a highly accurate numerical map of the surface, revealing its exact curvature and structure for the first time. To ensure their map was correct, they tested their method on other shapes where the answers were already known, confirming that their approach could reproduce known results with high fidelity. They also checked their work against a different, previously published numerical study of a related shape, finding that their results matched up perfectly. This validation gave them the confidence to trust their new map of the irregular shape.

What makes this achievement particularly significant is that it opens the door to studying physical phenomena that were previously inaccessible. The shape they mapped is the setting for a specific type of theoretical universe that behaves differently from the standard models used in physics. By having a precise description of this geometry, physicists can now calculate how particles would move and interact within this warped space. This is crucial for understanding "warped" universes, which are models where gravity behaves differently at different distances, potentially explaining why gravity is so much weaker than other forces. The researchers also used their new map to find specific patterns of vibration, known as harmonic forms, that can exist on this surface. These patterns are like the resonant frequencies of a bell, and they correspond to the types of particles that could exist in the associated quantum world.

The study also tackled a related problem that had been a source of confusion. There is a version of this shape where the rules of geometry forbid a stable solution from existing at all. The researchers ran their simulation on this impossible version to see how their method would behave. As expected, the computer failed to find a stable shape, and the errors in the calculation remained large no matter how much they refined the model. This failure was not a bug in their software but a confirmation of a deep mathematical truth: the shape simply cannot exist in a stable form under those specific conditions. This ability to distinguish between a shape that can exist and one that cannot serves as a powerful check on their method, proving that their numerical tools are sensitive enough to detect the fundamental laws of geometry.

The final output of this work is a detailed, high-resolution dataset that describes the geometry of this complex surface and the vibrational patterns it supports. This data is now available for other scientists to use in their own research, particularly in the field of string theory, where such shapes are used to model the hidden dimensions of our universe. By providing a concrete, numerical description of a shape that was previously only theoretical, the researchers have turned an abstract mathematical possibility into a tangible object that can be studied, measured, and used to build new theories about the nature of reality. Their work demonstrates that even when a problem is too complex for a human to solve with a formula, the right combination of mathematical insight and computational power can reveal the hidden structure of the universe.

Technical Summary: Numerical Sasaki–Einstein metrics and harmonic forms on del Pezzo links

Problem Statement
The paper addresses the construction of Ricci-flat Kähler metrics on toric Calabi–Yau cones and their corresponding Sasaki–Einstein (SE) metrics on the compact links. While explicit closed-form solutions exist for specific cases (e.g., the conifold $T^{1,1}$ , the $Y^{p,q}$ family, and the cone over $dP_0 \cong \mathbb{C}^3/\mathbb{Z}_3$ ), metrics for the cones over the second and third del Pezzo surfaces ( $dP_2$ and $dP_3$ ) lack closed-form expressions.

$dP_3$ : The Reeb vector is regular, meaning the transverse geometry is a smooth Kähler–Einstein surface. A numerical solution exists in the literature (Doran, Headrick, Herzog, Kantor, and Wiseman; DHHKW), serving as a positive control.
$dP_2$ : The situation is distinct. The surface $dP_2$ admits no Kähler–Einstein metric (due to a non-reductive automorphism group), and its volume-minimizing Reeb vector is irrational (irregular). Consequently, the transverse geometry is not a manifold or orbifold, and the SE metric on the link has not been constructed numerically or analytically.
Goal: To numerically construct the SE metric on the $dP_2$ link at its irregular Reeb vector and to compute the two primitive harmonic basic $(1,1)$ -forms on it. The authors also aim to validate their methodology against known closed forms ( $T^{1,1}$ , $Y^{p,q}$ ) and the existing $dP_3$ numerical data.

Methodology
The authors utilize the symplectic formulation of toric geometry, where the metric is encoded by a convex symplectic potential $G$ on a moment polygon.

Equation: The Ricci-flat condition reduces to a scalar Monge–Ampère equation on the polygon. The potential is decomposed into a canonical Guillemin potential (fixed by lattice data and boundary conditions) and a smooth correction $\psi(s)$ :
$G = G_{\text{can}} + \ell_b \psi(s)$
where $\ell_b$ is the facet function of the Reeb vector and $s$ are affine coordinates on the slice polygon.
Approximation Classes: The unknown correction $\psi$ $ψ$ is approximated using two distinct function classes:
- Polynomial Expansion: A basis of monomials in the slice coordinates, symmetrized where appropriate (e.g., $D_6$ symmetry for $dP_3$ ). The authors emphasize the necessity of orthonormalizing the basis (via Householder QR) to prevent ill-conditioning and residual plateaus.
- Neural Networks: A multilayer perceptron (MLP) with two inputs and two hidden layers, trained to minimize the Monge–Ampère residual.
Optimization: The objective is the mean squared residual of the Monge–Ampère equation over a set of training samples drawn uniformly from the polygon interior. The constant of integration is fitted alongside the coefficients.
Validation: The authors employ multiple checks beyond the training residual:
- Curvature Invariants: Comparison of $\Phi_1$ and $\Phi_2$ against closed-form references ( $Y^{p,q}$ ).
- Abreu Scalar: Checking the constancy of the scalar curvature of the transverse geometry ( $S=12$ on the slice), a fourth-order quantity not used in the objective.
- Laplacian Spectrum: Computing low-lying eigenvalues of the transverse Laplacian and comparing them with DHHKW's results for $dP_3$ .
- Symmetry Recovery: Testing whether a symmetry-free neural network recovers the $D_6$ symmetry of the $dP_3$ hexagon without explicit architectural constraints.

Key Contributions and Results

Construction of the $dP_2$ Metric: The primary contribution is the first numerical construction of the Sasaki–Einstein metric on the link of the $dP_2$ cone at its irregular, volume-minimizing Reeb vector $b^*$ .
- The polynomial fit achieves a held-out mean-squared Monge–Ampère residual of $\approx 6 \times 10^{-14}$ .
- The Abreu scalar is constant to within $0.0368\%$ (worst case) and $0.0006\%$ (mean) on a $240 \times 240$ grid.
- This metric is irregular, meaning the Reeb orbits are dense in a subtorus, and no quotient surface exists.
Harmonic Forms on $dP_2$ : The paper constructs the two primitive harmonic basic $(1,1)$ -forms on the $dP_2$ link.
- The authors formulate the problem by first seeking torus-invariant, closed, primitive $(1,1)$ -forms $\theta$ on the transverse geometry (represented by a scalar stream function $\chi$ on the polygon).
- These $(1,1)$ -forms are then used to construct the imaginary self-dual harmonic $(2,1)$ -forms $\omega_{2,1}$ required for the flux.
- They define an admissible trial space of rational functions to ensure the correct dimensionality ( $d-3$ ) of the primitive cohomology.
- The solution is obtained by minimizing the self-dual energy, formulated as a generalized eigenvalue problem.
Validation and Benchmarking:
- $Y^{p,q}$ Family: The method reproduces known closed-form metrics with high precision. Orthonormalizing the polynomial basis is shown to be critical; raw monomials lead to residual plateaus (e.g., $10^{-8}$ ) due to ill-conditioning, whereas orthonormal bases achieve residuals down to $10^{-14}$ or lower.
- $dP_3$ (Positive Control): The authors reproduce the DHHKW metric for the $dP_3$ cone. The degree-18 polynomial fit yields a residual of $1.0 \times 10^{-16}$ and matches DHHKW's low-lying Laplacian eigenvalues ( $\lambda_1 \approx 6.3228$ ) and eigenfunction coefficients.
- Symmetry Recovery: A symmetry-free neural network successfully recovers the $D_6$ symmetry of the $dP_3$ solution after training, demonstrating that the symmetry is a property of the solution rather than an artifact of the ansatz.
Negative Control ( $dP_2$ at Regular Reeb Vector):
- The authors attempt to solve the equation at the regular Reeb vector for $dP_2$ , where a Kähler–Einstein metric is known not to exist.
- The residual plateaus at $\approx 4 \times 10^{-2}$ regardless of approximation order or network width, while the in-sample residual continues to drop. This "fingerprint" (overfitting with a large held-out error) serves as a numerical indicator of the non-existence theorem, distinguishing it from the solvable case at $b^*$ .

Significance and Claims
The paper claims to provide the first numerical data for the Sasaki–Einstein metric and harmonic forms on the $dP_2$ link, a geometry previously inaccessible due to the irregularity of the Reeb vector and the lack of a quotient surface.

Holographic Application: The authors state that their data (metric and harmonic forms) enables the study of warped non-conformal holographic IIB backgrounds, specifically analogues of the Klebanov–Tseytlin solution on the conifold. These solutions require the explicit metric to compute the warp factor via the Poisson equation involving the flux $G_3$ .
Methodological Robustness: The work demonstrates that symplectic coordinates combined with orthonormal polynomial expansions or neural networks can robustly solve the Monge–Ampère equation on toric polygons, even in irregular cases where the transverse geometry is singular.
Modesty: The authors explicitly note that their results do not constitute a proof of non-existence for the regular $dP_2$ case (as a minimizing sequence could theoretically degenerate), but rather serve as a numerical confirmation of the obstruction. They also clarify that their comparison with DHHKW validates the implementation and parametrization but does not claim superiority of one method over another at matched computational cost.

The paper concludes by providing the numerical data for the metric and forms, which are intended to be used for Kaluza–Klein spectroscopy and the construction of specific supergravity backgrounds, tasks that were previously impossible without explicit metric data for these del Pezzo links.

More like this