Poster
2025 workshop program
When Stars Collide: Letting Neural Networks Pick Up the Pieces
The work
Abstract
Recent HST and JWST observations of dense star cluster cores have renewed interest in stellar collisions and their role in producing exotic objects such as blue stragglers, millisecond pulsars, and X-ray binaries. Motivated by these findings, we perform 3D hydrodynamic simulations to study the outcomes of such collisions. However, the parameter space—spanning stellar ages, masses, velocities, and impact parameters—is too large to explore with brute-force methods. To overcome this, we apply machine learning techniques for the first time in this context, enabling efficient prediction of collision outcomes and rapid exploration of high-dimensional interaction spaces.