Background
The NSF-led Vera C. Rubin Observatory (VRO) is the newest and most powerful astronomical facility for searching for moving and time-variable objects in space. After initial commissioning work in 2024, final commissioning was conducted in summer 2025, with the full 10-year Large Survey of Space and Time (LSST) expected to start in fall 2025. The initial commissioning data has proven the expected performance of the telescope, with automated recoveries of solar system targets down to magnitude 24. Because of the huge number of observations that VRO is expected to obtain, the VRO project is focused on automated single-night recoveries.
However, several ongoing and planned spacecraft missions could benefit from focused searches with VRO for much fainter solar system objects, down to magnitude 28. Notably, this could be substantially enabling for a potential extended mission for the SwRI-led NASA Lucy spacecraft, as well as opening up opportunities for SwRI to compete on significant NASA planetary defense and other science contracts. In order to reach magnitude 28 (equating to an object ~100 meters across in the asteroid belt or ~10 km in the Kuiper Belt), several nights of data must be collectively searched together to maximize the signal-to-noise ratio of the target. Historically, multi-night searches have been avoided as being too computationally complex, but both faster computers and better techniques can now enable these searches.
Approach
The goal of this project is to develop the tools to perform deep multi-night searches in VRO data, and to perform calibrations of its performance on synthetic data. The tools and techniques draw on the SwRI expertise developed in the search for New Horizons extended mission targets and are notably different (and more computationally intensive) than the standard VRO search but can search much deeper. A key tool is a convolutional neural net machine learning layer which eliminates most of the false positives produced by the initial search algorithm, which in turn allows the search algorithm to test the much wider amount of possible trajectories required for multi-night stacking. Since full data is not yet available to actually test the techniques on, the tools will be tested on synthetic implanted objects demonstrate their performance and guide future proposed investigations.
Accomplishments
Initial work was focused on developing the tools for implanting known synthetic solar system objects into the purely synthetic DP0.2 preview dataset. This was successfully accomplished over the summer, using “smeared” Point Spread Functions (PSFs) that enable much more precise simulations of faint targets than the default unsmeared PSFs available from the VRO project tools. Work is now proceeding on testing synthetic object injection on the initial VRO commissioning data released in July (DP1), which provides the first test on real data with real noise and optical artifacts. Progress is ongoing to search for the implanted data. While the initial plan had been to use the pure-synthetic DP0.2 data, DP1 may be used instead to maximize the realism of the simulations. The search tools are on track to be finalized by the end of the calendar year, followed by intensive searches of the implanted data.