News & Views | Published:

Planetary science

Forming the martian great divide

Nature volume 453, pages 11911192 (26 June 2008) | Download Citation

Early in its history, Mars suffered a convulsion that left a lasting geological and topographical scar. The latest work adds to evidence that the cause was external — a massive impact.

Mars is a divided planet. Its southern highlands cover about two-thirds of the planet and are on average about 4 kilometres higher than the northern plains, a difference that is known as the hemispheric dichotomy1 (Fig. 1). Like an ice cube floating in water, the high topography is held up by the buoyancy of thicker crust (Fig. 2) — the crust is about 25 km thicker in the highlands than in the lowlands2. On the basis of the number of impact craters in both the highlands and lowlands, the dichotomy is thought to have formed more than 4 billion years ago, during the first few hundred million years of martian history3. Moreover, the location of the boundary between the highlands and lowlands may have controlled the subsequent location of Tharsis4, the largest and possibly longest-lived volcanic region in the Solar System.

Figure 1: The martian hemispheric dichotomy.
Figure 1

This image of Mars's eastern hemisphere provides a vivid depiction of the dichotomy, with the southern highlands appearing in red–yellow and the northern lowlands in blue. The hemispheric dichotomy may be due to the impact of a massive asteroid early in martian history7,8,9, producing a proposed 10,000-km-wide impact structure called the Borealis basin as the northern lowlands. The Hellas basin (lower left) is also an impact structure, but at 2,300 km across is much smaller. Another of Mars's geological structures, the Tharsis volcanic region, is on the opposite side of the planet to that shown here. Image: NASA

Figure 2: Crust with a difference.
Figure 2

The southern highlands (left) consist of relatively thick crust and high-standing topography. The northern lowlands (right) consist of thinner crust and lower topography. In the impact model for dichotomy formation, which is supported by the new work7,8,9, the crust also differs in composition. According to this model, the highland crust formed early in martian history, and consists primarily of basalt rock. The lowland crust formed by shock melting of the mantle during the impact event, and so is both younger and different in composition compared with the highland crust.

Unravelling the processes that formed the hemispheric dichotomy is essential to understanding the earliest history of Mars. Previous explanations have invoked either the impact of a large asteroid or comet5, or large-scale convective circulation in the martian mantle6. But observations by spacecraft have not yet permitted a clear choice between these possibilities.

Three papers in this issue7,8,9 provide insight into this problem, and collectively strengthen the plausibility of the giant-impact model. One of the difficulties in testing the different hypotheses is that 30% of the boundary between highlands and lowlands is masked by later Tharsis volcanism. Andrews-Hanna et al.7 (page 1212) used gravity observations to subtract the contribution of Tharsis volcanism from the crust and thus estimated the crustal signature of the dichotomy boundary in this part of Mars.

The key to their approach is that, because Mars has cooled with time, its elastic lithosphere (the strong, outermost layer of the planet) was thicker when the Tharsis volcanoes formed than when the hemispheric dichotomy formed. Assuming that the lithosphere was at least 100 km thick by the time the bulk of Tharsis volcanism occurred, Andrews-Hanna et al. show that the northern lowlands have an elliptical shape that is roughly 10,600 km by 8,500 km across. The statistical uncertainty in the basin dimensions is a few hundred kilometres. The authors conclude that the elliptical basin — known as the Borealis basin — is plausibly the signature of the largest impact in the Solar System. They also argue that this shape cannot be produced by convective flow in the mantle.

A potential problem with the approach of Andrews-Hanna et al.7 is its sensitivity to the thickness of elastic lithosphere. Although some independent evidence10 supports the assumption of a thickness of 100 km or more, other observations11 suggest that the lithosphere was less than 20 km thick during the Noachian period, about 3.8 billion years ago, when Tharsis began forming. If the lithosphere was less than 50 km thick during Tharsis formation, the elliptical shape calculated for the dichotomy lowlands is degraded7, weakening the case for an impact model. Moreover, no calculations have shown that mantle convection is unable to produce an elliptical lowland basin.

A drawback of the original giant-impact model5 for the dichotomy is that the resulting basin was assumed to be circular, as are the vast majority of small impact craters in the Solar System. By contrast, the northern lowlands are clearly elongated, with an eccentricity of 1.2 (ref. 7). Three-dimensional hydrodynamic impact simulations by Marinova et al.8 (page 1216) now show that the original reasoning was incorrect. Small impact craters are essentially formed on a flat surface, and thus are not sensitive to the spherical shape of the planet. But in an impact large enough to form the hemispheric dichotomy, one must consider the interaction of the impactor with the spherical planet. Marinova et al. show that, for planet-scale impacts, an oblique impact angle of 30–60 ° from the horizontal can produce the observed basin eccentricity. This range of impact angles is the statistically most likely range, and the authors' inferred impact velocity of 6–10 km s−1 is reasonable for a body intersecting the orbit of Mars. If the impactor was a rocky or metallic asteroid, it would have had to have been between 1,600 km and 2,700 km in diameter to produce the observed basin size. For comparison, Mars itself is 6,780 km in diameter.

A limitation of the three-dimensional simulations done by Marinova et al.8 is that the vertical resolution is 118 km. This is about twice as large as the estimated average crustal thickness2, and thus details of crustal excavation and the resulting basin shape are not well resolved in these simulations. Nimmo et al.9 (page 1220) have developed a complementary set of numerical models with a vertical resolution of 25 km in the crust and upper mantle. This provides a much better resolution of the behaviour of the crust during an impact, although the models are two dimensional and thus limited to vertical impacts.

In Nimmo and colleagues' models, almost all of the pre-existing crust is excavated from the centre of the impact basin, and a new, thinner crust is generated by shock melting of the martian mantle. The inferred impact energy is a factor of 3–5 lower than in the simulations of Marinova et al.8. This disparity is probably due mostly to the difference between vertical and oblique impacts. Nevertheless, the overall concordance of the two sets of models suggests that we now have a reasonable first-order understanding of the dynamics of the proposed impact event.

These three papers7,8,9 strengthen the plausibility of the impact model for the hemispheric dichotomy on Mars, but they do not rule out mantle convection as the primary cause. One way to further test the impact model would be to gain a better understanding of the topographical variability along the proposed impact basin rim. In some places, the transition between highland and lowland occurs over just a few hundred kilometres, whereas in others a gradual change in elevation occurs over several thousand kilometres. The two boundary types also have distinct patterns in the gravity data. These differences have been proposed to be consequences of post-impact flow in the crust and mantle7,12. Development of quantitative models for these processes would be a great help in understanding the overall geological evolution of the proposed Borealis basin.

Another test relates to crustal composition in the highlands and lowlands. In the giant-impact model, most of the lowland crust formed by shock melting of the mantle9 (Fig. 2). Extraction of the highland crust from the mantle modified the mantle's initial chemical composition. Impact-shock melting of the modified mantle is likely to have produced a lowland crust that differs in composition from that of the highlands. Some differences in composition between highlands and lowlands have been proposed on the basis of spectroscopy measurements made by orbiting spacecraft13. Impact-cratering statistics also suggest that the lowlands are slightly younger than the highlands3. Both observations are consistent with the impact model.

A more precise test of possible differences between highlands and lowlands would involve comparative analysis of bedrock composition in the two regions by future robotic landers. Landers could also measure how seismic-wave velocities vary with depth, which would provide a more comprehensive test of possible variations in composition throughout the entire thickness of the crust. Seismic measurements of crustal structure could also directly probe the proposed location of the dichotomy boundary in southern Tharsis.

Meantime, the main cause of Mars's great divide must remain a story without an ending. But appropriately enough, given the hundredth anniversary of the Tunguska event in Siberia celebrated elsewhere in this issue, these three sets of authors7,8,9 have shortened the odds that it was produced when a huge impactor collided with a young Mars in what would have been an event of awesome proportions.


  1. 1.

    , & Annu. Rev. Earth Planet. Sci. 35, 621–652 (2007).

  2. 2.

    et al. J. Geophys. Res. 109, E08002, 10.1029/2004JE002262 (2004).

  3. 3.

    J. Geophys. Res. 111, E08S91, 10.1029/2005JE002449 (2006).

  4. 4.

    Lunar Planet. Sci. Conf. 39, abstr. 1528 (2008).

  5. 5.

    & Nature 309, 138–140 (1984).

  6. 6.

    & J. Geophys. Res. 111, E06013, 10.1029/2005JE002668 (2006).

  7. 7.

    , & Nature 453, 1212–1215 (2008).

  8. 8.

    , & Nature 453, 1216–1219 (2008).

  9. 9.

    , , & Nature 453, 1220–1223 (2008).

  10. 10.

    et al. Science 291, 2587–2591 (2001).

  11. 11.

    et al. J. Geophys. Res. 109, E07007, 10.1029/2004JE002286 (2004).

  12. 12.

    Geophys. Res. Lett. 32, L22201, 10.1029/2005GL024260 (2005).

  13. 13.

    et al. J. Geophys. Res. 111, E03S05, 10.1029/2006JE002675 (2006).

Download references

Author information


  1. Walter S. Kiefer is at the Lunar and Planetary Institute, 3600 Bay Area Boulevard, Houston, Texas 77058, USA.

    • Walter S. Kiefer


  1. Search for Walter S. Kiefer in:

About this article

Publication history



See Editorial, page 1143.   The researchers involved talk about this work and demonstrate the modelling behind it in this Nature video


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Newsletter Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing