Main

The reddish dust particles suspended in the martian atmosphere have a mean diameter of less than 5 µm (refs 4–6). Airborne particles are collected on the magnets on board the Mars Exploration Rovers2,3,7,8. Three types of magnets will be discussed here: the filter and capture magnets, and the sweep magnet. The sweep magnet is designed to allow only non-magnetic particles to enter into an area within the centre of a cylindrical magnet. The field from this magnet is so strong that it deflects the paths of all particles with any significant magnetic susceptibility3. The filter and capture magnets are designed to capture a layer of dust for investigation by the panoramic camera (Pancam) and the instruments on the robotic arm of the rovers. The capture magnet is very strong and will capture airborne particles with a wide range of magnetic properties, whereas the filter magnet will preferentially capture strongly magnetic particles3.

Throughout the mission, Pancam has been used to monitor the dust on each magnet. Occasionally the dust on the capture and filter magnets are investigated using the α-particle X-ray spectrometer (APXS), the Mössbauer spectrometer, and the Microscopic Imager. Investigation of particles collected by magnets on Mars during previous missions has shown that most of the airborne particles must contain a ferrimagnetic mineral9. Furthermore, on the Mars Exploration Rovers it has been possible to show unambiguously that the dust particles have a wide range of magnetic properties and therefore can be sorted accordingly: the filter magnet has assembled a more-magnetic and darker fraction of the dust particles than the capture magnet10. A similar correlation between magnetic properties and colour was seen in the details of the dust pattern on the capture magnet11. Imaging of the sweep magnet3 has shown that most—if not all—of the airborne dust particles are magnetic10. It is impossible to detect any particles in the centre of the sweep magnet in the Pancam images—presumably because the magnetic force has prevented particles from settling there.

During the mission, the APXS has been deployed on a series of different soils12,13. The overall chemical composition of these soils is rather uniform and indicates a basaltic origin. Variability across the soil units at Gusev crater and Meridiani Planum is discussed elsewhere14. The thin layer of dust on the magnet aluminium surface3 complicates the quantitative analysis of the APXS spectra. The surface is only partially covered by dust, and the thickness of the layer is comparable to the penetration depth of characteristic X-rays from the major elements. A direct comparison of raw spectra would lead to a relative underestimation of heavier elements in the dust on the magnets compared to an infinitely thick layer. Nevertheless, firm general conclusions can be drawn from the APXS results. To a first approximation, the dust on the magnets has an elemental composition that is similar to average bright soil (Table 1). All major element peaks present in rocks and soils are also present in the dust on the magnets. By comparing line areas from the raw spectra for neighbouring elements, the complications associated with the different (energy-dependent) effective thicknesses for different elements are partly avoided. As an example, we have selected the two pairs of elements Na and Mg, and P and Si, and plotted the relative peak areas of these pairs for rocks, soil and the dust present on the magnets (Fig. 1). Figure 1 shows that a classification of samples can be made on the basis of the ratios plotted. The figure is consistent with the hypothesis that the soil is composed of a rock component (of composition as Adirondack and Humphrey) and a dust component, which is enriched in Na and P (among other elements). This suggests that the bright soil is derived from a combination of basaltic sand (originating from the rocks) and a dust component similar to the dust on the magnets.

Table 1 Compositions of martian soil and dust
Full size table
Figure 1: Selected peak area ratios from APX spectra.
figure 1

Shown are data for rock and soil samples from Gusev crater, together with samples of dust from magnets labelled according to spacecraft (A, Spirit; B, Opportunity), measurement sol, and type of magnet (C, capture; F, filter). Error bars represent one standard deviation.

Full size image

Figure 2 shows a Mössbauer spectrum of the dust attracted to the capture magnet on Opportunity. The spectrum was recorded from martian day (sol) 328 to sol 330, and over a total time interval of about 50 h. For comparison, the spectrum of a typical soil is also included in the figure.

Figure 2: Mössbauer spectrum of airborne dust. Mössbauer spectrum of the dust attracted to the capture magnet on Opportunity, sol 328–330; data are from the 14.4 keV channel of the Mössbauer spectrometer.
figure 2

Also shown are (in red) a typical bright soil spectrum from Opportunity (MER B; sol 60) and (in blue) a spectrum of the Compositional Calibration Target (CCT) on Opportunity. These spectra have been scaled to match the intensity of the capture magnet spectrum. The (thin) error bars on each data point in the Mössbauer spectrum show ± √N, where N is the number of counts in each data point.

Full size image

The dust layer is of the order of 1 mg cm-2 thick, and the sampling depth of the Mössbauer spectrometer15 is of the order of 30 mg cm-2. The intensity (and quality) of the spectrum is therefore low. From the striking similarity between the soil spectrum and the dust spectrum at velocities from -2 mm s-1 to +4 mm s-1, where paramagnetic compounds dominate, it is evident that these spectra contain similar paramagnetic components16. As for typical bright soils, these doublets are assigned to the phases olivine, pyroxene and paramagnetic and/or superparamagnetic Fe3+-compounds, possibly including oxyhydroxides and haematite16,17,18. It should be noted that the ferric doublet is more intense in the dust on the magnet than in typical soils.

One could argue that olivine and pyroxene are not unambiguous interpretations of the ferrous doublet components in this spectrum of dust on the magnets. However, the martian surface is known to act as a dynamic source and sink for airborne dust (for example, via dust tracks, dust devils, and local storms), which imposes some relationship between dust and soil mineralogy. Furthermore, given the similarity of soil and dust spectra (both Mössbauer and APXS) and the arguments that have led to the assignment of these minerals for the components of the soil spectra, it would seem somewhat far-fetched to assume that a different set of minerals with similar spectra (and in similar relative abundance) should be the cause of these components in the dust.

The magnetically ordered components in the spectrum are characterized by the following features: the scattered intensity in the velocity ranges around -8 mm s-1 and +9 mm s-1 show a broad scattering peak at -8 mm s-1 and a more narrow and intense peak at +8.5 mm s-1—this pattern is indicative of the mineral magnetite (Fe3O4; see Methods). For comparison, we have included in Fig. 2 a Mössbauer spectrum of the compositional calibration target (CCT) of the rovers (this is a thin rock slab that is mounted under the rover solar panels for calibration of the instruments on the robotic arm). The CCT has a very high concentration of nearly stoichiometric magnetite. Comparing the dust spectrum with the spectrum of the CCT, it is clear that—to a first approximation—most components of the spectrum of the dust on the capture magnet can be accounted for by a sum of the spectra of the bright soil and of the magnetite of the CCT. This shows that the dust is magnetic because it contains the mineral magnetite (approximately 50% of the spectral area), and also that the dust contains major paramagnetic components. We note that in 1979, Pollack et al.6 suggested the presence of magnetite in martian dust, on the basis of the optical properties of the dust as observed by the Viking cameras.

Fitting of the spectrum shows that either the magnetite in the dust is not perfectly stoichiometric magnetite (the line at -7 mm s-1 is of significantly lower intensity than in the CCT), or that some ferric oxide (maghemite and/or haematite) is also present (see Methods for details).

On the basis of the Mössbauer spectrum of the dust on the Opportunity capture magnet (Fig. 2), it is clear that the most probable candidate for the ferrimagnetic mineral in the airborne dust is magnetite. It is possible that the ferrimagnetic mineral is a somewhat oxidized variant of magnetite—perhaps non-stoichiometric magnetite, or a solid solution of magnetite and maghemite. But neither maghemite alone (nor maghemite and haematite) can fit this spectrum adequately. The magnetite may also be isomorphically substituted with titanium. In the APX spectrum of the dust on the Opportunity capture magnet (not shown), the peak areas of all elements continuously increase during the first 150 sols. After a dust removal event (strong wind gust or dust devil around sol 200), we see an increase in peak areas of Fe, Ti and Cr accompanied by a decrease of the area of all other peaks (except Al, which is the material on which the dust is residing).

The Viking biology experiments showed that the martian soil was not only oxidized, but also strongly oxidizing when brought into contact with liquid water19. This led some to believe that the magnetic mineral must also be highly oxidized, and the mineral maghemite was therefore proposed as the most probable candidate for the cause of the magnetic properties of the dust9,20,21. However, the present results show that the magnetic mineral is less oxidized than hitherto believed.

The minerals in the dust on the magnets have about 45 ± 10% of the Fe in the ferric state, which is higher than for average martian soil14. The spectrum contains all the components found in Mössbauer spectra of different soils in both Gusev crater and on Meridiani Planum18. In this sense, the mineralogical composition of the dust on the magnets does not differ substantially from average martian soil.

A comparison of the Mössbauer spectra of dust on the filter magnet (not shown) with that of the capture magnet shows that a significant amount of a magnetic mineral, magnetite, is present in dust particles on both magnets. Furthermore, both magnets hold significant amounts of particles containing paramagnetic minerals. Mössbauer spectroscopy gives no information about the existence of iron-free particles, such as feldspar. However, within the resolution of the Pancam images the centre of the sweep magnet is clean of dust10 (unless the particles here have a reflectance identical to aluminium). This indicates that the airborne dust does not contain more than a few per cent of single-phase non-magnetic particles (for example, feldspar), and therefore suggests that all airborne particles are composite. So the magnets are evidently not just culling a subset of single-phase magnetic particles like, say, magnetite (Fe3O4). Instead, the results allow the firm conclusion that most of the airborne particles are composite.

A possible interpretation of the results described above is that the airborne dust consists of two groups of minerals: one of primary minerals and one of secondary (possibly hydrated) minerals. The first group consists of olivine, pyroxene and magnetite (possibly titanomagnetite) crystallites eroded from rocks. The second group consists of more oxidized crystallites of ferric and nanocrystalline oxides of which many could be secondary minerals. Some of these minerals may have formed by interaction with water. Crystallites from both of these groups seem to have been assembled into dust grains—perhaps just bound by electrostatic forces. Products of palagonitization, commonly used as terrestrial Mars dust analogues, correspond to such a mixture of particles22. Each particle contains several metallic elements (such as Na, Mg, Al, Si, K, Ca, Ti and Fe), and each particle contains several ferrous and ferric mineral phases.

How did such particles originate? The fact that the suspended particles contain the mineral olivine—(Mg,Fe)2SiO4—shows that the particles cannot be solely a result of precipitation in liquid water. The presence of olivine in the dust points towards physical alteration rather than chemical weathering. On the other hand, the dust particles cannot be exclusively unaltered ‘small basaltic rocks’ because of the relatively high abundance of the doublet assigned to nanocrystalline ferric oxides and the possible presence of haematite. Some chemical alteration must have taken place, including oxidation of the Fe(ii) in the rocks. The findings are consistent with the following scenario.

The primary minerals of the dust particles formed by physical processes (diurnal temperature cycles, comminution by meteoritic impacts, wind abrasion) from parent basaltic rocks. These processes have been operating since the formation of the planet. Some very slow alteration processes, possibly driven by diurnal condensation of thin films of atmospheric water23, may have occurred simultaneously. The secondary minerals could also (or in addition) be remnants from an early water-rich period in the history of the planet. In that case they would not be an integral part of the bulk of the dust particles, but would rather be bound electrostatically to their surface. As long as no information on the morphology of the dust particles is available on a microscopic scale, the time of origin of the secondary minerals remains unknown. (We note that a microscope with a resolution better than 10 µm will be on board the Phoenix lander, Mars Scout Mission 2007; ref. 24.) In any case, the secondary Fe(iii) compounds appear to be the dominant chromophore component of the dust. The paramagnetic or superparamagnetic iron oxides (probably haematite and maybe oxyhydroxides), which provide the dust with its characteristic reddish colour, are thus different from the mineral (magnetite) that causes the magnetism of the dust.

Methods

The Mössbauer spectrum of pure stoichiometric magnetite, Fe(iii)[Fe(ii)Fe(iii)]O4, consists of two superimposed sextet components: one originates from Fe(iii) ions tetrahedrally coordinated in the close-packed oxygen structure of the magnetite. The other sextet, originating from octahedrally coordinated Fe ions (indicated by square brackets), has an intermediate isomer shift, between typical values for Fe(iii) and Fe(ii). The fact that these two oxidation states of iron give rise to only one sextet is caused by the process of rapid electron exchange between the Fe(ii) and the Fe(iii) in magnetite. The two sextets in stoichiometric magnetite have area ratios of 2:1 for octahedrally to tetrahedrally coordinated iron, respectively. The two sextets coincide at high velocities (about +5.0 mm s-1 and +8.5 mm s-1), causing the characteristic asymmetric appearance of the Mössbauer spectrum of pure stoichiometric magnetite.

Maghemite (γ-Fe(iii)2O3), which holds tetrahedrally and octahedrally coordinated iron (both in the ferric oxidation state), also has two sextet components, but they differ very little in isomer shift, and therefore appear as one slightly broadened sextet. To the naked eye, a content of maghemite would appear in the Mössbauer spectrum as an increase of the first of the two sextets described for magnetite. This is practically indistinguishable from a partial oxidation of magnetite into what we refer to as non-stoichiometric magnetite. In this case, the oxidation is expressed by an area ratio between the two sextets in magnetite that differs from the stoichiometric ratio of 2:1. Also, substitution of other elements (primarily at the octahedral site) can cause a change in the apparent ratio between the two sextets of magnetite. This change is accompanied by a broadening of the second sextet (in pure magnetite assigned to octahedrally coordinated iron).

Haematite has slightly larger values for the magnetic hyperfine field and the isomer shift as compared to maghemite. In addition, owing to its non-cubic structure, it has a non-zero quadrupole shift, leading to a small asymmetry in the distance between the pairs of lines 1 and 2 and between 5 and 6. The fit of the Mössbauer spectra of the dust on the magnets is improved by the inclusion of a sextet haematite component, indicating a minor content of haematite in the dust.