The Sun May Shake the Outer Planets Loose as It Dies

Isaac Newton thought the planetary order was mortal. Three centuries of celestial mechanics answered him, and kept answering: the four giant planets are so securely spaced that chaos alone would need something like a billion billion years to unseat Uranus, far longer than the universe has existed. A paper published Sept. 21 in The Astrophysical Journal Letters reopens the question by changing one thing that sounds like bookkeeping: it assumes the Sun will not die tidily.

Konstantin Batygin at the California Institute of Technology, with Jim Fuller at Caltech and Fred C. Adams at the University of Michigan, simulated the Sun's last act and the four giant planets riding it out. The standard picture has the Sun shedding nearly half its mass as a smooth leak, which expands every orbit by about the same factor while leaving the architecture otherwise intact. The new runs replace the leak with a stutter: thousands of separate parcels of gas, each thrown off in its own direction, each giving the Sun a shove of a few meters per second. At the parcel size the team takes as its reference case, the Sun receives about 4,600 such shoves.

Individually those are nothing. What matters is that they arrive at random times and in random directions, so their effects never average out. Each one tugs the Sun away from the center of every planet's orbit, and the orbits take a random walk. The team's own null tests pin the cause. Zero the recoils and keep the mass loss, and the planets simply expand into a frozen copy of the system we have; make the ejections random in time but symmetric in direction, and everything stays quiet. Both the lumpiness and the lopsidedness are needed. With both in place, the period ratio of Jupiter and Saturn scatters far beyond anything the smooth case allows, Neptune's near-circle is pumped into a long oval, and the neat concentric rings that conventional stability calculations start from are gone.

The instability begins while the Sun is still a red giant

That is the result the authors put first, and the timing is the whole of it. In the reference case, with parcels of about a ten-thousandth of a solar mass and ejections running through all of the Sun's swollen phases, orbits begin to cross before the mass loss has even finished in three-quarters of the runs. The earliest crossings happen while the Sun is still a red giant weighing 0.89 solar masses. By the moment the white dwarf forms, 19 of those 48 runs, about 40%, have already flung a planet onto a wild orbit or lost one outright. Nothing in this concerns Earth: the simulations include the Sun and the four giant planets and nothing else, and the paper states that the rocky planets are left out.

Coming apart here means losing planets rather than crashing them together. Most of the runs that first showed crossing orbits went on to eject a planet, because the expanded giants are massive enough and far enough out that a close pass can throw a planet clear instead of colliding with it. The individual histories read like demolition notes: in one, Saturn is scattered and ejected within 2.5 million years of the white dwarf forming, leaving Uranus and Neptune bound together; others hold their shape for a billion years and then unravel.

Every time quoted in the paper is counted from white dwarf formation, not from today. Pushed forward three billion years from there, with no passing stars invited to help, 96% of the reference runs go unstable in the variant where kicks operate throughout the giant phases and 83% in the more cautious variant that confines them to the last one. That bracket is the roughly 90% the abstract quotes. Onsets spread from a few million years to 2.6 billion years, with no preferred moment in between. The paper attaches no date to any of it and quotes no figure for how long the Sun has left.

The answer swings on a number nobody has measured

All of that is conditional on one quantity: how coarsely the Sun breaks up its envelope. The grid of simulations was built to explore exactly that, spanning a thousandfold range of parcel sizes with 48 random starting conditions at each, in two variants, for 672 runs. At the gentle end, parcels of three ten-millionths of a solar mass or less, fewer than 2% of runs go unstable and the old picture survives intact. Across the whole grid, 448 of the 672 systems are still standing after three billion years. The violent numbers belong to the coarse end.

So which end is the Sun's? The evidence splits in two here, and the paper is careful about the seam. The kick itself is measured: observations of wide double stars by the Gaia spacecraft show newborn white dwarfs recoiling at about 0.75 kilometers per second. El-Badry and Rix inferred those kicks from changes in the separations of thousands of pairs, and a second study has since reported a similar result. The explanation that the recoil is the sum of thousands of discrete ejections is a model, set out in a companion preprint by Fuller that has not yet been peer reviewed, and the parcel mass for the Sun is not measured at all. It is the axis of the grid, not an observation. "Present evidence favors the violent branch," the authors write, pointing out that even the low end of the permitted kick range maps onto parcels coarse enough to destabilize the system. Their reference runs accumulate 0.33 to 0.58 kilometers per second, below the measured value, so the headline cell is not the aggressive edge of what the data allow.

A new rung at the bottom of a very long ladder

The outer Solar System's stability has long been quoted as a ladder of ever longer times. The billion billion years sits at the top of it, Murray and Holman's 1999 estimate of how long the planets' own mutual tugging would need to unseat Uranus, with no stellar death and no passing traffic included. Zink, Batygin and Adams added both in 2020 and got 30 to 100 billion years. This paper adds a rung underneath, approximately a gigayear after white dwarf formation: a billionfold cut against chaos alone and a hundredfold against the estimate that already allowed for the Sun's death. Only the second comparison replaces a number anyone actually held.

Two of that estimate's three authors are authors here, and the new paper also revisits part of it, arguing that the locked Jupiter and Saturn rhythm the old flyby story depended on was a coordinate-handling error in their own earlier pipeline. That is their account, in their own appendix, and nobody outside the group has checked it.

If the parcel picture holds, none of this is unique to the Solar System. Nearly every star ends as a white dwarf, ice-giant-class planets are common products of planet formation, and lighter planets on wider orbits are easier to shake loose than Jupiter and Saturn are. The ones thrown clear could drift off as free-floating planets, while debris that falls onto white dwarfs, which astronomers already detect in their atmospheres, could arrive on timescales like those in these simulations. "Newton's envisioned instability is real after all," the authors write. "He was mistaken only about the perpetrator."