Calabi-Yau Metrics with Kähler Moduli Dependence

1. Problem Statement

The central challenge in string phenomenology is deriving Standard Model parameters from the geometry of extra dimensions (Calabi–Yau manifolds). While physical observables like Yukawa couplings and kinetic terms depend on the Ricci-flat Kähler metric, obtaining these metrics analytically is notoriously difficult.

• Current Limitations: Existing numerical methods (e.g., Donaldson's algorithm or neural networks) can approximate Ricci-flat metrics at specific points in the moduli space with high precision. However, they produce implicit representations (e.g., trained neural networks) that do not provide explicit analytic formulas.
• The Gap: There is a lack of closed-form analytic expressions that retain explicit dependence on Kähler moduli (parameters controlling the size and shape of the manifold). Without these, it is difficult to systematically study how physical quantities vary across the moduli space or to analyze moduli stabilization.
• Specific Difficulty: Donaldson's algorithm relies on integral Kähler classes tied to specific line bundles. Varying Kähler moduli continuously requires changing the line bundle and the entire ansatz, making it impossible to parametrize the metric continuously as a function of moduli.

2. Methodology

The authors propose a hybrid strategy combining numerical machine learning with symbolic regression to construct approximate analytic expressions. The workflow consists of four main steps:

1. Numerical Data Generation:

   • They use the cymetric package to numerically compute the Ricci-flat Kähler potential ($K$) at selected points in the Kähler moduli space using neural networks.
   • The Kähler potential is decomposed as $K = K_{FS} + \phi$, where $K_{FS}$ is the Fubini-Study potential and $\phi$ is a correction term.
   • Calculations are performed for various ratios of Kähler parameters ($t_1/t_2$) while fixing the total volume to unity to isolate shape dependence.

2. Analytic Ansatz Construction:

   • They propose an explicit analytic form for the correction term $\phi$. This ansatz is based on a finite-dimensional space of sections of an ample line bundle $L = \mathcal{O}_X(k)$.
   • The ansatz takes the form:
     $$\phi(x, t) = V(t)^{1/3} \frac{\sum_{I,J} \alpha_{IJ}(t/t_m) s_I(x) s_J(x)}{\prod_r (\sum |x^{(r)}_a|^2)^{k_r}}$$
   • Crucially, the coefficients $\alpha_{IJ}$ are promoted from constants to functions of the Kähler moduli ratios (dimensionless parameters), ensuring the correct scaling behavior under rescaling of the Kähler form.

3. Symmetry Exploitation:

   • The authors identify discrete symmetry groups ($G$) acting on the specific Calabi–Yau manifolds.
   • They verify numerically that the Ricci-flat metric respects these symmetries (even for non-freely acting symmetries).
   • They constrain the ansatz by grouping monomials into $G$-invariant combinations (singlets), significantly reducing the number of independent coefficients to determine.

4. Symbolic Regression:

   • The numerical data for the coefficients $\alpha_{IJ}$ (fitted from the neural network results) are fed into PySR (a symbolic regression tool).
   • PySR searches for the simplest analytic expressions (using building blocks like $+$, $\cdot$, $\exp$, $\log$, $\sin$, $\cos$) that fit the data.
   • This yields closed-form formulas for the coefficients as functions of the Kähler moduli.

3. Key Contributions

• First Explicit Moduli-Dependent Metrics: The paper presents the first known examples of approximate Ricci-flat Kähler metrics on Calabi–Yau threefolds with explicit, closed-form dependence on Kähler moduli.
• Hybrid Framework: It establishes a robust pipeline: Neural Networks (for high-precision data) $\to$ Analytic Ansatz (for structure) $\to$ Symbolic Regression (for closed-form coefficients).
• Validation of Symmetry: It demonstrates that even for non-freely acting discrete symmetries, the Ricci-flat metric respects the symmetry of the underlying manifold, justifying the use of symmetry-constrained ansätze.
• Efficiency: The method achieves high accuracy using relatively low-order truncations (small line bundle degrees), suggesting that the dominant geometric variations are captured by low-frequency modes.

4. Results

The method was applied to two Calabi–Yau threefolds with $h^{1,1}=2$:

Case A: Bicubic Hypersurface in $\mathbb{P}^2 \times \mathbb{P}^2$

• Symmetry: A large symmetry group of order 972.
• Ansatz: Used line bundle $\mathcal{O}(2,2)$. The symmetry reduced the 36x36 coefficient matrix to 8 independent singlets.
• Accuracy: The resulting analytic expression reproduces the numerically learned Kähler potential with an average deviation of ~1.8% to 2.2%.
• Outcome: The coefficients were successfully fitted to analytic functions involving logarithms and polynomials of the moduli ratio $t_1/t_2$.

Case B: Bi-degree (2,4) Hypersurface in $\mathbb{P}^1 \times \mathbb{P}^3$

• Symmetry: A smaller symmetry group ($\mathbb{Z}_2 \times S_4$).
• Ansatz: Used line bundle $\mathcal{O}(2,2)$. The symmetry reduced the 30x30 matrix to 30 singlets, but numerical analysis showed that only the top 12 contributed significantly.
• Accuracy: The analytic expression achieved an average deviation of ~1.9% to 2.8%.
• Geometric Insight: The study revealed how the accuracy of different truncation orders depends on the moduli ratio (e.g., when one factor is squeezed, the degree associated with the other factor becomes dominant).

General Findings:

• Pure Fubini-Study metrics (no correction) have errors of ~17-18%.
• Including the correction term $\phi$ with low-degree truncations ($k=(2,2)$) reduces errors to the percent level.
• The symbolic regression successfully captured the complex functional dependence of the coefficients on the moduli ratios.

5. Significance and Future Outlook

• Bridge Between Numerics and Analytics: This work bridges the gap between purely numerical results and analytic constructions, enabling systematic studies of moduli dependence that were previously impossible.
• Phenomenological Impact: Explicit analytic metrics allow for the direct computation of physical quantities (Yukawa couplings, masses) as functions of moduli, facilitating the search for regions in moduli space that match the Standard Model.
• Future Directions:
  • Extending the framework to include complex structure moduli dependence.
  • Attempting to bypass the neural network stage by applying symbolic regression directly to the Monge-Ampère equation loss function.
  • Investigating whether the analytic metrics are sufficiently accurate for computing derived quantities (like metric derivatives) where errors might propagate.

In summary, the paper provides a powerful new tool for string phenomenology, transforming the "black box" of numerical metric reconstruction into a transparent, analytic, and moduli-dependent framework.