- 1Department of Earth and Planetary Sciences, Weizmann Institute of Science, Rehovot, Israel (maayan.ziv@weizmann.ac.il)
- 2Institut für Astrophysik, Universität Zürich, Zürich, Switzerland
- 3Université Côte d’Azur, Observatoire de la Côte d’Azur, Nice, France
Understanding Jupiter's internal structure is crucial for uncovering its formation and evolutionary history, providing valuable constraints that have broader implications for other giant planets and the Solar System. The primary observational data used to constrain Jupiter’s interior come from precise gravity field measurements by NASA's Juno mission, atmospheric data from both Juno and the Galileo entry probe, and Voyager radio occultations. However, these observations are limited compared to the vast range of plausible interior configurations and their associated parameters, making it challenging to reconcile the data with theoretical models.
In this study, we use NeuralCMS, a deep learning model based on the concentric Maclaurin spheroid (CMS) method, coupled with a self-consistent wind model to efficiently explore a wide range of interior models without prior assumptions. This integrated approach allows us to identify models consistent with the available measurements. We apply it to determine the permissible range of the dynamical contribution to the gravity field from Jupiter's dilute core model, demonstrating that our modeling approach provides tighter constraints on recently published, widely considered interior models.
Using clustering techniques on the full multidimensional dataset of plausible interior structures, we identify four charachteristic interior structures distinguished by their envelope and core properties (dilute and compact). Our results show that Jupiter’s interior can be effectively described using only two key parameters, significantly reducing the dimensionality of the problem. We also highlight the most observationally constrained interior structures and show that they might be confined to one of the identified key structures.
Our framework establishes a baseline for using deep learning models to constrain planetary interiors based on gravity data, offers a self-consistent approach to coupling interior and wind models, and provides valuable insights into the multidimensional nature of the problem. This approach can also enable meaningful comparisons with interior models of other giant planets.
How to cite: Ziv, M., Galanti, E., Howard, S., Guillot, T., and Kaspi, Y.: Characterizing Jupiter's interior using machine learning reveals four key structures, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-8235, https://doi.org/10.5194/egusphere-egu25-8235, 2025.