Harmonic $1$-forms on real loci of Calabi-Yau manifolds

Michael R. Douglas, Daniel Platt, Yidi Qi, Barbosa, RodrigoarXiv 2024 · DOI · arXiv

TL;DR

This paper uses machine learning to numerically search for a specific type of mathematical object – nowhere vanishing harmonic 1-forms – on the real parts of complex Calabi-Yau manifolds. They found evidence suggesting such forms exist on certain manifolds, particularly when they are close to a singular state, which could lead to new examples of exotic $G_2$-manifolds.

What Problem Does This Paper Address?

Calabi-Yau manifolds are fundamental objects in string theory and pure mathematics, known for possessing a special type of metric called a Ricci-flat metric. While billions of examples of these manifolds exist, finding explicit descriptions of their Ricci-flat metrics is notoriously difficult. A specific application in differential geometry requires a "harmonic 1-form" that is nowhere zero on the "real locus" (the real points) of a Calabi-Yau manifold. This harmonic 1-form is crucial for constructing $G_2$-manifolds, which are important in M-theory and for understanding fundamental mathematical questions. However, proving the existence of such a nowhere vanishing harmonic 1-form has been a significant challenge, as the Calabi-Yau metric itself is unknown. This paper tackles the problem of numerically detecting the existence of these elusive harmonic 1-forms.

Methods

The authors employ a two-step machine learning approach:

  1. Approximate Calabi-Yau Metric: They first train bihomogeneous neural networks to approximate the unknown Calabi-Yau metric ($g_{CY}$) on the manifold. This is a challenging task, as it involves solving a complex geometric equation (the complex Monge-Ampère equation).
  2. Approximate Harmonic 1-Form: With the approximate metric in hand, they then use another neural network to compute an approximate harmonic 1-form ($\lambda$) with respect to this metric on the real locus of the manifold.

The core idea is that if they can find an approximate harmonic 1-form that is "well-behaved" (e.g., doesn't vanish) with respect to their approximate metric, and if this approximate metric is sufficiently close to the true Calabi-Yau metric, then it implies the existence of a true harmonic 1-form for the actual Calabi-Yau metric. This is inspired by techniques used in computer-assisted proofs in geometric analysis.

Datasets

The paper uses four specific Calabi-Yau manifolds as test cases:

  • Fermat quintic: A standard example, used as a control.
  • Quintic: A smoothed version of a singular cubic, designed to have a non-zero first Betti number on its real locus.
  • CICY1 and CICY2: Complete intersection Calabi-Yau manifolds, which are diffeomorphic to $S^1 \times S^2$ and thus have a non-zero first Betti number. CICY2 is specifically constructed to be close to a singular limit.

These are not "datasets" in the traditional sense of pre-existing collections of data, but rather carefully constructed mathematical objects on which the numerical methods are applied.

Key Insights and Ideas

  • Machine Learning as a Tool for Geometric Existence Proofs: The paper demonstrates the potential of using neural networks not just for approximation, but as a stepping stone towards proving the existence of mathematical objects that are hard to find analytically.
  • Singularity Proximity Matters: The quality of the approximate Calabi-Yau metric and the likelihood of finding a nowhere vanishing harmonic 1-form are strongly correlated with how "close" the manifold is to a singular state. Manifolds closer to singularities tend to have larger "Monge-Ampère loss" (indicating a poorer metric approximation) but also show more promise for admitting a non-vanishing harmonic 1-form.
  • The "Long Neck" Phenomenon: For manifolds that are smoothings of singular Calabi-Yau manifolds, the numerical results suggest the formation of "long necks" near the smoothed singularity, a phenomenon predicted by theoretical results.
  • $G_2$-Manifold Construction Potential: The most exciting finding is the evidence for a nowhere vanishing harmonic 1-form on the real locus of CICY2, which is close to a singular limit. This suggests CICY2 could be a candidate for constructing new $G_2$-manifolds.

Key Results

  • Fermat Quintic: As expected, no non-zero harmonic 1-form was found, consistent with its real locus having a first Betti number of zero.
  • Quintic: The numerical results suggest that while a harmonic 1-form exists, it must have zeros, aligning with theoretical predictions for its real locus.
  • CICY1: A harmonic 1-form was found, but it appears to be approximately zero in one direction ($S^1$) and non-trivial on the other ($S^2$).
  • CICY2: The most significant result. The data strongly suggests the existence of a nowhere vanishing harmonic 1-form on the real locus of CICY2. This 1-form appears to be approximately constant in the $S^1$ direction and approximately zero in the $S^2$ direction, consistent with theoretical expectations for manifolds near a singular limit with $U(1)$ symmetry.

Limitations and Future Work

  • Numerical Verification, Not Proof: The paper presents numerical evidence and suggests potential avenues for computer-assisted proofs, but it does not provide a rigorous mathematical proof of the existence of the harmonic 1-forms. The authors acknowledge that the quality of their approximate solutions might not be sufficient for such proofs yet.
  • Computational Cost: Training the neural networks for both the metric and the 1-form is computationally intensive, requiring significant hyperparameter tuning and epochs.
  • Generalizability: While the methods are applied to specific examples, extending them to a wider range of Calabi-Yau manifolds and proving their robustness remains a future direction.
  • Understanding the $U(1)$-Symmetry Connection: Further theoretical work is needed to fully understand how the $U(1)$-symmetry of the singular limit of CICY2 leads to the observed properties of the harmonic 1-form.
  • Constructing $G_2$-Manifolds: The ultimate goal of using these harmonic 1-forms to construct new $G_2$-manifolds is a significant future undertaking.

Summary updated

Abstract

We numerically study whether there exist nowhere vanishing harmonic $1$-forms on the real locus of some carefully constructed examples of Calabi-Yau manifolds, which would then give rise to potentially new examples of $G_2$-manifolds and an explicit description of their metrics. We do this in two steps: first, we use a neural network to compute an approximate Calabi-Yau metric on each manifold. Second, we use another neural network to compute an approximately harmonic $1$-form with respect to the approximate metric, and then inspect the found solution. On two manifolds existence of a nowhere vanishing harmonic $1$-form can be ruled out using differential geometry. The real locus of a third manifold is diffeomorphic to $S^1 \times S^2$, and our numerics suggest that when the Calabi-Yau metric is close to a singular limit, then it admits a nowhere vanishing harmonic $1$-form. We explain how such an approximate solution could potentially be used in a numerically verified proof for the fact that our example manifold must admit a nowhere vanishing harmonic $1$-form.
BibTeX
@misc{douglas2024harmonic,
  author = {Michael R. Douglas and Daniel Platt and Yidi Qi and Barbosa, Rodrigo},
  title = {{Harmonic $1$-forms on real loci of Calabi-Yau manifolds}},
  year = {2024},
  eprint = {2405.19402},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2405.19402},
}

Loading related content...

Figure