HLRS: Particle Scattering Model Could Improve Low-Orbit Spaceflight

Key Takeaways

• 225k particle impacts simulated on HLRS Hawk supercomputer
• ML kernel interpolates reflections across VLEO velocity‑angle space
• Integrated into DSMC, improves drag and attitude predictions
• Enables material selection for better reflection control
• Supports concepts for atmospheric‑oxygen‑based propulsion

Summary

Scientists at the University of Stuttgart’s ATLAS center used HLRS’s Hawk supercomputer to run 225,000 molecular‑dynamics simulations of oxygen atoms striking satellite materials in very low Earth orbit (VLEO). The data trained a machine‑learning scattering kernel that can predict particle‑surface reflections far more accurately than traditional DSMC models. Integrated into the PICLas DSMC code, the kernel enables realistic drag and attitude simulations for VLEO satellites. The breakthrough opens pathways for material optimization, drag‑counteracting propulsion concepts, and longer, cheaper satellite missions.

Pulse Analysis

Very low Earth orbit (200‑450 km) is becoming the sweet spot for next‑generation satellites, offering lower launch costs and higher imaging resolution. However, the thin residual atmosphere—dominated by atomic oxygen—creates drag that shortens mission life and complicates attitude control. Traditional computational fluid dynamics fail at these altitudes, forcing engineers to rely on simplified Direct Simulation Monte Carlo (DSMC) models that assume idealized particle reflections, leading to inaccurate lifetime forecasts.

The ATLAS team tackled this gap by marrying high‑performance computing with data‑driven methods. Using HLRS’s Hawk supercomputer, they performed 225,000 molecular‑dynamics simulations of oxygen atoms colliding with aluminum oxide surfaces, covering five speeds and nine incident angles. A month of runtime on 128 cores generated a rich dataset, which fed a generative machine‑learning algorithm. The resulting scattering kernel captures nuanced reflection behaviors, bridging the microscopic physics of molecular dynamics with the mesoscopic scale of DSMC simulations, and delivering predictions that closely match experimental observations.

The implications for the satellite industry are profound. With a reliable drag model, designers can optimize surface materials and geometries to harness atmospheric forces for attitude control or even develop intake‑based propulsion that harvests ambient oxygen. This could extend satellite operational periods, reduce the need for costly propellant, and mitigate space‑debris accumulation by enabling controlled de‑orbiting. As VLEO constellations grow, the new model positions operators to achieve higher performance at lower cost, accelerating the commercial viability of ultra‑low‑orbit missions.