A nuclear burst in orbit could disable satellites long after the flash, model suggests

A nuclear explosion in low Earth orbit could damage far more satellites through lingering radiation than through its initial burst, according to a new modeling study. In its central scenario, about 98% of the modeled population becomes disabled within one week. The resulting loss of collision-avoidance capability then raises the expected number of orbital impacts.

The study, by Zachary Petersen, Hong Qin, Sarah Thiele and Sébastien Philippe, was posted to arXiv on November 6, 2026. It is a preprint, not established evidence of peer-reviewed findings. This account draws on the complete paper text, including its supplementary methods and results, rather than the abstract alone.

Why the danger lasts beyond the flash

The mechanism has a historical precedent. The 1962 Starfish Prime nuclear test created an artificial radiation belt that damaged satellites, including spacecraft launched afterward. Radioactive products of nuclear fission release electrons as they decay. Some electrons become trapped in Earth’s magnetic field, exposing spacecraft repeatedly as their orbits cross the affected region.

The researchers connected several models to follow that chain: initial X-ray exposure, radioactive electron production, magnetic trapping, radiation-belt evolution, accumulated spacecraft dose, collisions and atmospheric drag. Their radiation calculation followed 13,306 catalogued objects selected using their positions in January 2026. The collision inventory also included other payloads, rocket bodies and existing debris.

Their central scenario used a one-megaton explosion at 500 kilometres altitude over the South Atlantic. This was a deliberately severe case, selected because its magnetic geometry produces an intense belt overlapping heavily populated orbital altitudes. It should not be read as the expected outcome of every explosion in low Earth orbit.

Repeated exposure drives the modeled losses

In that scenario, prompt X-rays disable approximately 2.9% of the modeled population. Repeated exposure to trapped electrons produces much larger cumulative losses: about 25% after one day, 78% after three days and 98% after seven days.

Those figures assume two millimetres of aluminium-equivalent shielding and a uniform radiation-failure threshold of 40 kilorads in silicon. The threshold is an illustrative modeling choice, not a measured specification for every spacecraft. Raising it threefold to 120 kilorads reduces modeled losses to about 72% after a week, but approximately 97% are still disabled by day 15.

An important uncertainty is how long radioactive debris remains above the atmosphere, continuing to supply electrons. The authors tested short release periods and an intentionally extreme six-month upper-bound case. These assumptions change the pace of failure substantially. Nevertheless, all three source cases produce widespread losses in the selected scenario.

Location also matters. Electrons that remain trapped near their release point can be lost later if their path around Earth takes them into the atmosphere. The study finds that some burst locations do not sustain a persistent trapped belt under its assumptions.

Disabled satellites remain collision targets

The model assumes radiation-disabled spacecraft can no longer maneuver. It estimates approximately 20 collisions within two years and 26 over a decade, compared with about 12 over a decade in a matched calculation without the explosion. These are expected population-level totals, not predictions of particular spacecraft hitting one another.

The additional hazard is concentrated early: roughly three-quarters of the detonation case’s decade-long collision total occurs in the first year. Atmospheric drag subsequently removes many disabled spacecraft from lower orbits, reducing the number of remaining collision targets. Higher objects persist longer.

That later decline is not recovery. It reflects the removal of spacecraft, while lost services remain lost and debris continues to pose hazards.

What the calculation cannot establish

The collision results depend strongly on assumptions about spacecraft geometry and atmospheric removal. Alternative collision-area prescriptions produce roughly 16–88 expected impacts over ten years. These sensitivity cases are not statistical confidence intervals.

The model also excludes collision hazards from fragments smaller than 10 centimetres, assumes an optimistic 99% avoidance benefit when an operational spacecraft is involved, and includes neither new launches nor scheduled disposal. Its simplified belt evolution and uniform spacecraft tolerances add uncertainty to failure timing.

Finally, the replacement-satellite calculation assumes the new spacecraft survives radiation. It therefore cannot identify a safe date for restoring operations. The study’s central warning is conditional but consequential: orbital infrastructure could remain vulnerable after the initial flash and even after collision activity begins to subside.