Using Hydrocode Simulations and Machine Learning to Constrain β and the Material Properties of Dimorphos After the DART Impact

NASA's Double Asteroid Redirection Test (DART) mission is the agency's first attempt to change the trajectory of an asteroid using a kinetic impactor. The target of the DART mission is Dimorphos, the small moonlet of the Didymos binary asteroid system. Prior to impact, very little was known about Dimorphos aside from an estimate of its size. However, asteroid mass, internal structure, porosity, strength, and other material properties all affect the magnitude of deflection. In this study, we use simulations to constrain the material properties of Dimorphos using information about the DART impact.

We ran three-dimensional impact simulations using Spheral, an open-source Adaptive Smoothed Particle Hydrodynamics code. We varied several material properties, including but not limited to the yield strength and shear modulus for intact material, the yield strength and shear modulus of damaged material, and the bulk porosity of the asteroid. From each simulation, we extracted several pieces of information, including the velocity of the asteroid after the impact, the size and morphology of the ejecta cone, the size and morphology of the resulting crater, and the momentum enhancement factor, β. After running a training set of simulations with randomized material parameters, we used several machine learning algorithms to predict parameter combinations that would reproduce the velocity change observed in the DART impact, including decision tree, random forest, and neural network regression.

Constraining β and the material properties of Dimorphos based on the results of the DART impact is a degenerate problem, with a wide variety of possible material parameter combinations. The machine learning algorithms we utilized describe several families of solutions, with porosity and the strength of intact material playing major roles in the general response of the asteroid. We analyze these families of solutions in terms of geological realism and likelihood in order to suggest the most probable material parameter combinations describing Dimorphos. We also examine what additional observations could further constrain these parameters, like crater morphology and asteroid mass.

This work is supported by the DART mission, NASA Contract No. 80MSFC20D0004. Prepared by LLNL under Contract DE-AC52-07NA27344. LLNL-ABS-837537