Vol 459 | 11 June 2009 | doi:10.

1038/nature08096

LETTERS
Existence of collisional trajectories of Mercury, Mars
and Venus with the Earth
J. Laskar1 & M. Gastineau1

It has been established that, owing to the proximity of a resonance lunar contributions14 (Supplementary Information). We used the
with Jupiter, Mercury’s eccentricity can be pumped to values large SABA4 symplectic integrator15, which is adapted to perturbed
enough to allow collision with Venus within 5 Gyr (refs 1–3). This Hamiltonian systems. The step size is 2.5 3 1022 years, unless the
conclusion, however, was established either with averaged equa- eccentricity of the planets increases beyond about 0.4, in which case
tions1,2 that are not appropriate near the collisions or with non- the step size is reduced to preserve numerical accuracy.
relativistic models in which the resonance effect is greatly In a first experiment performed without the Moon or relativistic
enhanced by a decrease of the perihelion velocity of Mercury2,3. contributions, we integrated the equations over 5 Gyr for 201 orbits
In these previous studies, the Earth’s orbit was essentially un- with the same initial conditions, except for an offset of 3.8k cm
affected. Here we report numerical simulations of the evolution (k g [2100, 100]) in the semi-major axis of Mercury (its actual
of the Solar System over 5 Gyr, including contributions from uncertainty is of a few metres (refs 16, 17)). In 60% of the solutions,
the Moon and general relativity. In a set of 2,501 orbits with we observed large increases in Mercury’s eccentricity, beyond 0.9
initial conditions that are in agreement with our present know- (Fig. 1a). The statistics (Supplementary Table 1a) are comparable
ledge of the parameters of the Solar System, we found, as in to the results of the non-relativistic secular equations2. Among the
previous studies2, that one per cent of the solutions lead to a large 121 solutions of large eccentricity (e . 0.7), 34 ended in collision with
increase in Mercury’s eccentricity—an increase large enough to
allow collisions with Venus or the Sun. More surprisingly, in one
of these high-eccentricity solutions, a subsequent decrease in a 1
Mercury’s eccentricity induces a transfer of angular momentum
from the giant planets that destabilizes all the terrestrial planets 0.8
3.34 Gyr from now, with possible collisions of Mercury, Mars or
Venus with the Earth. 0.6
Owing to chaotic behaviour of the Solar System4–7, the distance
between two initially close orbital solutions increases by a factor of
0.4
ten every ten million years4,7,8. It is thus hopeless to search for a
precise solution for the motion of the Solar System over 5 Gyr, that
is, over a time comparable to its age or life expectancy (before the Sun 0.2
Maximum eccentricity

becomes a red giant). The most precise long-term solutions for the
orbital motion of the Solar System are not valid over more than a few 0
tens of millions of years9,10. A numerical integration of the Solar
System’s motion over 5 Gyr can thus only be considered as a random b 1
sample of its possible evolution. Statistical studies are then required
to search for possible changes in the planetary orbits that lead to 0.8
collisions or disruption of the system.
The first study of the planetary orbits over several billion years was 0.6
obtained after averaging the equations of motion over the fast orbital
motion of the planets, which allowed a decrease in integration time of 0.4
three orders of magnitude1. This study revealed the possibility of
Mercury’s eccentricity reaching very high values, allowing collisions
0.2
with Venus in less than 5 Gyr. However, this method has the draw-
back that the averaged equations are no longer justified in the vicinity
of a collision. Despite the increase in computational power, recent 0
0 1,000 2,000 3,000 4,000 5,000
integrations on gigayear timescales still neglected the relativistic con- Time (Myr)
tribution2,3,11. The study of the full Solar System using general rela-
tivity and non-averaged equations on gigayear timescale thus needed Figure 1 | Mercury’s eccentricity over 5 Gyr. Evolution of the maximum
eccentricity of Mercury (computed over 1-Myr intervals) over 5 Gyr. a, Pure
to be done.
Newtonian model without the contribution of the Moon, for 201 solutions
The model for the integration of the planetary orbits is derived with initial conditions that differ by only 3.8 cm in the semi-major axis of
from the La2004 model9 that was integrated over 250 Myr for the Mercury. b, Full Solar System model with relativistic and lunar
study of the palaeoclimates of the Earth and Mars9,12. It comprises the contributions, for 2,501 solutions with initial conditions that differ by only
eight major planets and Pluto and includes relativistic13 and averaged 0.38 mm in the semi-major axis of Mercury.
1
Astronomie et Systèmes Dynamiques, IMCCE-CNRS UMR8028, Observatoire de Paris, UPMC, 77 Avenue Denfert-Rochereau, 75014 Paris, France.
817
©2009 Macmillan Publishers Limited. All rights reserved
LETTERS NATURE | Vol 459 | 11 June 2009

the Sun, 86 in collision with Venus and a single one reached the 5-Gyr each with an offset of 0.15k mm in the semi-major axis of Mars. Within
limit before collision. 100 Myr, five cases lead to the ejection of Mars from the Solar System
According to the secular analysis2, we could expect a much smaller (semi-major axis .100 AU) and the remaining 196 solutions end in
number of collisional orbits in the full relativistic model. To estimate collision, with the following distribution: Sun–Mercury, 33; Sun–
the probabilities of large eccentricity deviations in the relativistic Mars, 48; Mercury–Venus, 43; Mercury–Earth, 1; Mercury–Mars, 1;
model, we thus had to increase the scale of the numerical experiment. Venus–Earth, 18; Venus–Mars, 23; Earth–Mars, 29.
Using the JADE supercomputer at the French National Computing The most surprising collision is the one of Venus with the Earth,
({1)
Centre CINES, we integrated 2,501 orbital solutions, Sk, of the com- which occurs in S{468 in a five-stage process (Figs 2 and 3). The first
plete model over 5 Gyr, with the initial semi-major axis of Mercury step is the increase in the eccentricity of Mercury, obtained through
differing by 0.38k mm (k g [21,250, 1,250]) from that in the nominal perihelion resonance with Jupiter2,3 at 3.137 Gyr. This step is essen-
solution, S0, which was adjusted to the planetary ephemeris tial, as it allows a transfer of non-circular angular momentum from
INPOP0617. The results (Fig. 1b and Supplementary Table 1b) are the outer planets to the terrestrial planets20. The eccentricity increase
comparable to those of the relativistic secular equations2, with of Venus, the Earth and Mars, is then obtained through secular reso-
Mercury having a high eccentricity in about 1% of solutions. nances among the inner planets while the eccentricity of Mercury
Among these 2,501 solutions that are compatible with our best decreases between 3.305 and 3.325 Gyr. Once Mars and the Earth
knowledge of the Solar System, in 20 the eccentricity of Mercury acquire large eccentricities, close encounters occur and collisions
increased beyond 0.9. At the time of writing, 14 of these have not ({15) ({1)
become possible, as in S{468 (Fig. 3c). In S{468 , the collision with
yet reached 5 Gyr and may still be running for a few months, as their Mars does not occur, but several close encounters (Fig. 3c) lead to the
step size is greatly reduced. Solution S2947 reached 5 Gyr without
collision, although a close encounter (closest approach, 6,500 km) 0.9
a
occurred between Mercury and Venus at 4.9 Gyr. In S2915, S2210 and 0.8
S33, Mercury collided with the Sun at 4.218, 4.814 and 4.314 Gyr,
0.7
respectively, whereas in S2812, Mercury collided with Venus at
0.6

Eccentricity
1.763 Gyr. The most notable solution is S2468, in which a close
0.5
encounter of Mars with the Earth with a closest approach of only
0.4
794 km occurs at 3.3443 Gyr (Fig. 2). Such a close approach would be
0.3
disastrous for life on the Earth, with a possible tidal disruption of
0.2
Mars and subsequent multiple impacts on the Earth18,19, but we also
0.1
searched for more direct collisions. We integrated 201 different
(k) 0
versions, S{468 (k g [2100, 100]), of S2468, starting at 3.344298 Gyr, b
2
Semi-major axis (AU)

a 1
1.5
0.8
1
Eccentricity

0.6

0.5
0.4

c
0.2
–1
log10[dmin (AU)]

0
–2
b –7

–8 –3
log10(dh)

–9
–4
–10

–11
d –8
log10(dh)

–12
–8.5
c –8
–9
–9
3,344 3,346 3,348 3,350 3,352
log10(dc)

–10 Time (Myr)

–11 Figure 3 | Collisional trajectories for Mars and Venus with the Earth. The
–12 two solutions, S({15) ({1)
{468 (black) and S{468 (colours), are identical until
3.344298 Gyr. At this time, an offset of 0.15k mm is applied to the semi-
–13 major axis of Mars in the solution S(k)
0 500 1,000 1,500 2,000 2,500 3,000 3,500 {468 a, b, Eccentricity (a) and semi-major
Time (Myr)
axis (b) plotted versus time for Mercury (red), Venus (pink), the Earth (blue)
and Mars (green). c, Minimum Earth2Mars (green) and Earth2Venus
Figure 2 | Example of collisional trajectory for Mars and the Earth. (pink) distances in astronomical units, recorded over each 1,000-yr time
a, Evolution of the maximum eccentricity of Mercury (red), Mars (green) and interval. The horizontal lines are the Mars–Earth (6.5322 3 1025 AU, green)
the Earth (blue), recorded over 1-Myr intervals for the solution S({15)
{468 . and Venus–Earth (8.30827 3 1025 AU, pink) distances of collision, dmin,
b, c, Relative variation, dh, of the total energy of the system (b) and total corresponding to the sum of the planets’ radii. For S({15)
{468 , the Earth–Mars
angular momentum, dc (c), on a logarithmic scale. During the whole distance reached at 3.344317 Gyr is 3.7752 3 1025 AU, and for S({1) {468 , the
integration, the relative variations of the total energy and angular momentum Earth–Venus distance reached at 3.352891 Gyr is 3.3791 3 1025 AU.
of the system remain below 2.3 3 1028 and 2.2 3 1029, respectively. d, Relative energy variation.
818
©2009 Macmillan Publishers Limited. All rights reserved
NATURE | Vol 459 | 11 June 2009 LETTERS

diffusion of Mars’s semi-major axis (Fig. 3b) until secular resonances 6. Laskar, J., Quinn, T. & Tremaine, S. Confirmation of resonant structure in the Solar
System. Icarus 95, 148–152 (1992).
produce a decrease in the eccentricity of Mercury together with an
7. Sussman, G. J. & Wisdom, J. Chaotic evolution of the Solar System. Science 257,
additional increase in the eccentricity of Venus and the Earth at about 56–62 (1992).
3.347.3 Gyr (Fig. 3c). At this point, close encounters between Venus 8. Laskar, J. The limits of Earth orbital calculations for geological time scale use. Phil.
and the Earth occur, with several exchanges of the planets’ orbits Trans. R. Soc. Lond. A 357, 1735–1759 (1999).
(Fig. 3b) before a final collision at 3.352891 Gyr (Fig. 3c). 9. Laskar, J. et al. A long term numerical solution for the insolation quantities of the
Earth. Astron. Astrophys. 428, 261–285 (2004).
The essential trigger for the collisional trajectories of Mars and 10. Varadi, F., Runnegar, B. & Ghil, M. Successive refinements in long-term
Venus with the Earth is the great increase in Mercury’s eccentricity. integrations of planetary orbits. Astrophys. J. 592, 620–630 (2003).
If this increase leads rapidly to a collision with the Sun or with Venus, 11. Ito, T. & Tanikawa, K. Long-term integrations and stability of planetary orbits in
as in S33 and S2812, the remaining part of the Solar System is not our Solar System. Mon. Not. R. Astron. Soc. 336, 483–500 (2002).
much affected; these two solutions, after merging of the colliding 12. Laskar, J. et al. Long term evolution and chaotic diffusion of the insolation
quantities of Mars. Icarus 170, 343–364 (2004).
bodies, are actually very stable when extended to 5 Gyr. We expect 13. Saha, P. & Tremaine, S. Long-term planetary integration with individual time
that this will be the case for most of the solutions in which Mercury steps. Astron. J. 108, 1962–1969 (1994).
has high eccentricity. However, in some less frequent events, such as 14. Boué, G. & Laskar, J. Precession of a planet with a satellite. Icarus 196, 1–15 (2008).
S2468, the eccentricity increase of Mercury leads to a total destabiliza- 15. Laskar, J. & Robutel, P. High order symplectic integrators for perturbed
Hamiltonian systems. Celest. Mech. Dynam. Astron. 80, 39–62 (2001).
tion of the inner Solar System. In the simulations described here, we
16. Standish, M. An approximation to the errors in the planetary ephemerides of the
restricted the initial conditions to a very small neighbourhood, Astronomical Almanac. Astron. Astrophys. 417, 1165–1171 (2004).
assuming that the chaotic behaviour of the system will randomize 17. Fienga, A., Manche, H., Laskar, J. & Gastineau, M. INPOP06: a new numerical
the initial conditions. The fact that the probability distribution of planetary ephemeris. Astron. Astrophys. 477, 315–327 (2008).
Mercury’s eccentricity (Supplementary Table 1) is very similar to the 18. Holsapple, K. A. & Michel, P. Tidal disruptions: a continuum theory for solid
bodies. Icarus 183, 331–348 (2006).
results obtained with the averaged equations and a much more widely 19. Asphaug, E., Agnor, C. B. & Williams, Q. Hit-and-run planetary collisions. Nature
distributed set of initial conditions2 leads us to propose that the 439, 155–160 (2006).
probability of a large increase in the eccentricity of Mercury is about 20. Laskar, J. Large scale chaos and the spacing of the inner planets. Astron. Astrophys.
1%. It remains difficult to evaluate the probability of a collision 317, L75–L78 (2007).
involving the Earth within 5 Gyr. Indeed, although we studied varia- Supplementary Information is linked to the online version of the paper at
tions around solution S2468 that lead to collision with the Earth in www.nature.com/nature.
about 25% of the cases, the S2468 orbit is essentially the single event of Acknowledgements This work benefited from support from the Planetology
its kind in our full sample of 2,501 solutions. Programme of the French National Research Centre, from Paris Observatory and
from National Research Agency grant ASTS-CM. The authors thank the computing
Received 17 February; accepted 22 April 2009. centres of Paris Observatory, the Institut de Physique du Globe Paris, the Institute
1. Laskar, J. Large scale chaos in the Solar System. Astron. Astrophys. 287, L9–L12 of Development and Resources in Scientific Computing, the Linear Accelerator
(1994). Laboratory Grid and especially the French National Computing Centre CINES, for
2. Laskar, J. Chaotic diffusion in the Solar System. Icarus 185, 312–330 (2008). providing the necessary computational resources for this work.
3. Batygin, K. & Laughlin, G. On the dynamical stability of the Solar System. Author Contributions J.L. designed the study, performed the simulations and their
Astrophys. J. 683, 1207–1216 (2008). analysis, and wrote the paper. M.G. wrote the computer code.
4. Laskar, J. A numerical experiment on the chaotic behaviour of the Solar System.
Nature 338, 237–238 (1989). Author Information Reprints and permissions information is available at
5. Laskar, J. The chaotic motion of the Solar System. A numerical estimate of the size www.nature.com/reprints. Correspondence and requests for materials should be
of the chaotic zones. Icarus 88, 266–291 (1990). addressed to J.L. ([email protected]).

819
©2009 Macmillan Publishers Limited. All rights reserved