A star can return to its original velocity without returning its planets to their original orbits.
I needed a small thought experiment to make that sentence concrete. Phil Plait’s column on the distant future of the solar system had sent me to a 2026 paper by Konstantin Batygin, Jim Fuller, and Fred C. Adams. Their subject is the Sun losing its outer layers in discrete, asymmetric ejections. The ejecta carry momentum away; the star recoils. Thousands of small recoils can disturb its planets.
But why should the sequence matter, rather than just the final change in the star’s velocity?
Two shoves
Set the dying Sun aside. Give an idealized star a fixed mass and one planet of negligible mass, initially on a circular orbit. We will apply two instantaneous shoves to the star. This is a toy about recoil, not a stellar-evolution model.
For the first shove, push the star against the direction in which the planet is travelling. The planet’s own velocity does not change at that instant. The star’s does. Relative to the star, the planet is now moving faster.
Its separation from the star has not jumped, so its gravitational potential energy has not changed at that instant. Its orbital kinetic energy has increased. Make the shove small enough that the planet stays bound: its circular orbit becomes an ellipse, and the point where we shoved the star is the ellipse’s closest approach.
Wait until the planet reaches the opposite end of that new ellipse. It is now travelling in the opposite direction.
Give the star an equal shove in the opposite direction to the first. The two imposed changes in the star’s velocity cancel exactly. The star is back to its original velocity—not its original position.
The planet gains relative speed again.
Both shoves were against the planet’s motion when they arrived. Both raised its orbital energy. Adding the two stellar velocity changes gives zero; adding the two changes in the planet’s orbital energy does not.
The accounting is simple. If the planet’s velocity relative to the star is v, and the star receives a velocity change u, the planet’s new relative velocity is v − u. At fixed separation and stellar mass, the change in specific orbital energy is:
Δε = −v·u + |u|²/2
The planet’s velocity at the moment of the shove is part of the calculation. A final sum of the stellar kicks leaves that information out.
Back to the Sun
The paper does not arrange two perfectly opposed kicks. It models many independently directed recoils during mass loss, along a stellar-evolution track. Nor is recoil the only change: as the Sun loses mass, the gravitational attraction itself weakens.
The controls help separate these effects. Smooth mass loss produced an enlarged, stable version of the outer solar system. Randomly timed but spherically symmetric ejections also left the modeled system stable for three billion years. In these experiments, large excitation needed both discrete events and asymmetric ejection that recoiled the star.
At the authors’ fiducial parcel mass, roughly one ten-thousandth of a solar mass per event, about 90 percent of runs were classified unstable within three billion years after white-dwarf formation. That is a result conditional on their recoil model and its calibration—not a measured probability for the actual future Sun. Extending episodic ejection throughout the giant-branch phases, rather than restricting it to the better-calibrated asymptotic giant branch, is an additional assumption; they test both variants.
Their Appendix A names the crucial inference explicitly: observations constrain accumulated net recoil, while orbital excitation depends on the sequence of impulses. Connecting those quantities statistically requires assumptions about the events’ directions and independence.
The two-shove toy cannot supply that calibration. It only makes one limitation visible. Knowing how fast the star ends up moving does not tell us everything that happened to the planets while it got there.