Researchers Giuseppe Viterbo and Tobias Buck, affiliated with the University of Cambridge’s Institute of Astronomy, have introduced a novel framework that leverages advanced machine learning techniques to study the gravitational potential of the Milky Way and the properties of stellar streams. Their work, published in the journal Physical Review Letters, offers a new approach to understanding the complex dynamics of our galaxy, with potential implications for the energy sector’s use of space-based assets.
Stellar streams, which are remnants of tidally disrupted star clusters or dwarf galaxies, offer a sensitive probe of the Milky Way’s gravitational potential. The phase-space morphology of these streams encodes information about both the tidal field of the host galaxy and the internal structure of their progenitors. Traditional techniques for analyzing these streams, such as orbit-fitting and action-angle methods, often struggle to capture the full complexity of the interactions between the progenitor and the host galaxy.
Viterbo and Buck’s research introduces a framework that combines Flow Matching and Simulation-Based Inference (SBI) to jointly infer the parameters of the progenitor of the GD-1 stellar stream and the global properties of the Milky Way potential. Flow Matching is a machine learning technique that learns the vector field transporting a base Gaussian distribution into the posterior, enabling efficient and amortized inference directly from stream phase-space data.
The researchers generated a large suite of mock GD-1-like streams using their differentiable N-body code Odisseo, sampling self-consistent initial conditions from a Plummer sphere and evolving them in a flexible Milky Way potential model. They then applied conditional Flow Matching to learn the vector field that transports a base Gaussian distribution into the posterior, enabling efficient, amortized inference directly from stream phase-space data.
The study demonstrates that the method successfully recovers the true parameters of a fiducial GD-1 simulation, producing well-calibrated posteriors and accurately reproducing parameter degeneracies arising from progenitor-host interactions. This approach enables joint inference on progenitor and Galactic parameters, capturing complex dependencies that are difficult to model with classical likelihood-based methods.
For the energy sector, this research could have practical applications in improving the accuracy of space-based assets, such as satellites and other orbital systems, by enhancing our understanding of the gravitational dynamics within the Milky Way. This could lead to more precise orbital predictions and better management of space-based energy infrastructure.
The research was published in Physical Review Letters, a prestigious journal known for publishing significant research in the field of physics.
This article is based on research available at arXiv.