The Mars Exploration Rovers each carry a set of Magnetic Properties Experiments designed with the following objectives in mind: (1) to identify the magnetic mineral(s) in the dust, soil and rocks on Mars, (2) to establish if the magnetic material is present in the form of nanosized (d < 10 nm) superparamagnetic crystallites embedded in the micrometer sized airborne dust particles, and (3) to establish if the magnets are culling a subset of strongly magnetic particles or if essentially all particles of the airborne dust are sufficiently magnetic to be attracted by the magnets. To accomplish these goals, the Mars Exploration Rovers each carry a set of permanent magnets of several different strengths and sizes. Each magnet has its own specific objective. The dust collected from the atmosphere by the Capture magnet and the Filter magnet (placed on the front of each rover) will be studied by the Mössbauer spectrometer and the Alpha Particle X-ray Spectrometer, both of which are instruments located on the rover's Instrument Deployment Device. The captured dust particles will also be imaged by the Pancam and Microscopic Imager. The Sweep magnet will be imaged by Pancam and is placed near the Pancam calibration target. The four magnets in the Rock Abrasion Tool (RAT) are designed to capture magnetic particles originating from the grinding of Martian surface rocks. The magnetic particles captured by the RAT magnets will be imaged by Pancam.
 Through the Magnetic Properties Experiments (MPE) on the Viking landers and the Pathfinder lander, it has not been possible to identify the ferrimagnetic mineral (or minerals) responsible for the magnetism of the Martian fines. Therefore the Mars Exploration Rovers (MERs) each carry a set of Magnetic Properties Experiments (MPE), with the aim to identify the ferrimagnetic mineral(s) in the Martian soil. The dust on the magnets will be studied by visible and near-infrared spectroscopy using the panoramic camera, Pancam [Bell et al., 2003]. Furthermore, the MPE on the MERs include comparison of Mössbauer spectra [Klingelhöfer et al., 2003] of dust collected by two magnets of different strength with Mössbauer spectra of dust on the Martian surface. The elemental composition of the dust on these magnets will be studied also by means of Alpha Particle X-ray Spectroscopy (APXS) [Rieder et al., 2003]. The comparison of the elemental composition of the dust collected by the magnets and the elemental composition of the dust on the surface of Mars may, as elaborated below, disclose essential aspects of the formation of the Martian dust.
 The following pages describe the MPE on the MERs. The experiments are described in terms of 1) science goals, 2) description of hardware, 3) operations, and 4) data analysis.
2. Science Goals of the Magnetic Properties Experiments
 A major science goal for the MPE on the rovers is to identify the magnetic mineral(s) responsible for the magnetism of the Martian soil and dust. In addition an effort will be made to obtain information on the crystallite size of the magnetically ordered Fe(III) compound(s) in the dust. In the following we shall briefly describe the significance of the science goals of the MPE.
2.1. Iron in the Martian Soil and Dust
 From elemental analysis by means of x-ray fluorescence on previous missions, it is known that Martian soil contains about 13% (by weight) of the element iron [Toulmin et al., 1977; Rieder et al., 1997]. The mineralogical composition of the Martian soil is unknown, but by means of visible spectroscopy it has been established that the majority of the iron in the soil is present in the oxidation state 3+, Fe(III) [Bell et al., 1990].
 The particles suspended in the Martian atmosphere have linear dimensions of about 3 micrometers [Smith et al., 1997; Tomasko et al., 1999], and the particles appear to be composite, consisting mainly of silicate phases covered with iron oxides or cemented by iron oxides (and/or sulfates). The particles must contain a few percent of a strongly magnetic (ferrimagnetic) mineral, which may be maghemite (γ-Fe2O3). The saturation magnetization of pure maghemite is 70 A m2 kg−1. The saturation magnetization σS of the Martian soil and dust is not known with high precision, but has been estimated to be within the following limits:
 It has been suggested that the particles suspended in the atmosphere all have a similar composition [Hargraves et al., 1977, 1979; Madsen et al., 1999]. This implies that the particles are all somewhat magnetic. The composition may of course vary from particle to particle and the magnetic properties of a single particle are determined mainly by the amount of ferrimagnetic material in the particle. It should, however, be emphasized that it has not been established with certainty if all particles suspended in the Martian atmosphere contain a ferrimagnetic mineral, or if the magnets on the Viking and Pathfinder landers were culling only a subset of the suspended particles.
 Some of the Fe(III)-ions appear to be present in nanosize (∼10 nm) grains. These tiny grains are embedded in the composite particles, which, as mentioned, consist mainly of silicate phases [Morris et al., 1989; Morris and Lauer, 1990; Bell et al., 1990; Madsen et al., 1999]. The fact that some of the Fe(III)-ions seem to occur in nanosized, possibly superparamagnetic particles, will be significant for the MPE on MERs.
 In “large” grains (d > 30 nm) of a magnetically ordered material the magnetization vector M will be locked into a so-called easy direction by the forces of crystal anisotropy for a long time compared to the timescale of Mössbauer spectroscopy (∼10−9 s). For smaller particles (d < 10 nm) and above a certain temperature (depending on the size of the particle) the anisotropy forces are not able to hold M fixed in an easy direction for a long time compared to 10−9 s. For such particles, M fluctuates stochastically in response to the thermal noise in the particle. This fluctuation of M will strongly influence the shape of Mössbauer spectra. For more details, see below in section 5.1.3.
 The dominant iron type present in the rock forming minerals on Mars is Fe(II). The iron in the Martian dust and soil is mainly present as Fe(III)-ions. An oxidation of the iron in the surface rocks therefore has taken place. The main process (processes) that led to the widespread oxidation of the Martian surface rocks is unknown. An important goal in the study of the evolution of the Martian surface therefore is to establish how the widespread oxidation of the iron on Mars took place. There are several conceivable pathways for the formation of magnetic dust on the surface of Mars: (a) Volcanism, (b) meteoritic influx and impacts, (c) precipitation from liquid water and (d) gas-solid reactions (slow and dry weathering). These processes are briefly discussed in sections 2.2–2.5. Focus will be on the magnetic properties of the expected end products. Finally, the iron oxides could also (hypothetically) be traces of biological activity, likely to have been encountered in a water-rich environment.
2.2. Volcanic Processes
 Volcanism on Mars has been extensive throughout a significant part of the history of the planet [Hartmann et al., 1999]. Some calculations indicate that the surface atmospheric pressure was approximately constant and close to the present value (typically 10 mbar) during the last 3.5 Gy [Fanale et al., 1992]. Thus a considerable part of the volcanic eruptions on Mars have occurred at low surface pressure, and many of these eruptions may have been of explosive nature, distributing large amounts of volcanic ash over a large area and up to considerable atmospheric altitudes. It is likely that these particles contain a ferrimagnetic (e.g., titanomagnetite) phase. Once these particles are exposed to the Martian environment, they are subjected to long-term weathering processes, such as the interaction with solar ultraviolet radiation and chemically active molecules (e.g., radicals) of the Martian atmosphere. Much has been speculated, if such alteration processes can account for the observed redox and magnetic properties of Martian dust. In any case, the number of volcanoes on the surface of Mars does favor the idea that a significant part of the surface dust actually represents a deposited volcanic aerosol [Hargraves, 1999].
2.3. Meteoritic Influx and Meteoritic Impacts
 Two other possible pathways for the formation of Martian dust are influx of micrometeorites and metereoritic comminution of surface rocks and subsequent weathering. The contribution of meteoritic material to indigenous Martian surface material is not known, but has been estimated to less than 30 wt% of Martian soil [Flynn and McKay, 1990; Bland and Smith, 2000]. Magnetite is found in micrometeorites and in many types of meteorites (both bulk and crust), although with highly varying abundances (up to 20 wt%). Hematite is generally absent in meteorites, but could have formed from meteoritic magnetite in the Martian environment.
 The Martian dust may also have acquired its physico-chemical characteristics through the action of liquid water, during a period (or periods) when the planet was wetter and warmer than at present. This may imply a broad spectrum of different processes: Alteration of basaltic glass (also termed “palagonitization”), alteration of meteoritic material, and/or alteration of bedrock through leaching (sub-surface liquid water action) and subsequent precipitation from solution.
2.4. Precipitation of Iron Compounds From Liquid Water
 The processes of precipitation of iron oxides from water are highly complex, and we will limit ourselves to a few brief statements relevant for the study of magnetism on Mars. For more details see Banin, Hargraves and collaborators [Banin et al., 1993, 1997; Hargraves et al., 2000]. Figure 1 presents a crude survey of possible pathways to the formation of the various precipitates. It is assumed that the Fe(II) in the surface rocks [(Mg, Fe)2SiO4, (Mg, Fe)SiO3] is dissolved in abundant liquid water. If the highly soluble Fe(II) is oxidized, precipitation of Fe(III) oxides and Fe(III)-oxyhydroxides will follow. The various pathways are not known in detail, not even for precipitation of iron oxides in laboratory experiments, and therefore much less for precipitation in natural waters. The type of precipitate formed is dependent on many factors, including iron concentration, pH, Eh, anion type and concentration.
 Very generally it can be stated that if oxygen is comparatively scarce, the oxidation will be slow, and the iron-ions will pass through the phase known as green rust (Fe62+Fe23+(OH)16CO3 · H2O). Green rust contains iron in the oxidation state +2 (Fe(II)) as well as in the oxidation state +3 (Fe(III)). The iron in the green rust will often be further oxidized and the end products of the precipitation process in case of moderate availability of oxygen will be magnetite (Fe3O4, ferrimagnetic, σS = 92 A m2 kg−1) or maghemite (γ-Fe2O3, ferrimagnetic, σS = 70 A m2 kg−1), both strongly magnetic minerals.
 However, if oxygen is available in abundance, as it is on Earth, the oxidation of Fe(II) will be swift, ferrihydrite will form, and the end product will often be hematite (α-Fe2O3, canted antiferromagnetic, σS = 0.4 A m2 kg−1) or goethite (α-FeOOH, antiferromagnetic). The precipitates following rapid oxidation are thus much less magnetic than the materials following slow oxidation. The iron oxides or oxyhydroxides formed by precipitation in natural waters are often rather pure iron oxides or oxyhydroxides, often containing some aluminum.
 The red soils that form in tropical areas of the Earth often contain a substantial amount of superparamagnetic (nanophase) particles. The conditions leading to the formation of superparamagnetic particles of iron oxides or oxyhydroxides in nature are unknown, but on Earth it seems to be commonly associated with precipitation in water.
 In summary and for further reference: A magnetic mineral forming via dissolution of Fe(II) in water on Mars, oxidized to Fe(III) and followed by precipitation in the water will be a (nearly) pure iron oxide, perhaps containing some aluminum. Specifically, the mineral will not contain the element titanium as an impurity. If superparamagnetic crystallites are present in high abundance in the dust particles on Mars, it will indicate that the dust is a result of precipitation in liquid water, followed by drying.
 A remote possibility for the source of a ferrimagnetic phase is the formation of fine-grained magnetite from thermal decomposition of siderite as has been suggested for Martian meteorites. Decomposition of siderite depends critically on heating temperature, time, crystallinity, porosity and the availability of O2 [Pan et al., 2000]. Under oxidizing conditions the alternation products are magnetite, maghemite and hematite.
2.5. Dry Gas-Solid Reactions
 Recently strong magnetic anomalies have been identified on Mars, by means of the magnetometer on Mars Global Surveyor [Acuña et al., 1998, 1999]. These results prove that at least some of the rocks on Mars contain ferrimagnetic minerals. Furthermore, the SNC meteorites, widely accepted to be surface rocks from Mars [McSween, 1985], are basaltic rocks containing up to 2% of a ferrimagnetic mineral, titanomagnetite (Fe3-xTixO4, 0.1 < x < 0.6). The content of magnetic minerals in SNC meteorites is insufficient to explain the strength of the Martian magnetic anomalies [Cisowski, 1986; Collinson, 1997], suggesting that the SNCs came from other parts of the planet where the magnetic anomalies are not present. But overall, there are strong reasons to expect that the basaltic surface rocks on Mars contain ferrimagnetic minerals, and that such ferrimagnetic minerals contain the element titanium.
 Therefore, if weathering on Mars has been less intense, i.e., if the dominant Fe(II) ions in the bedrock were never completely released into aqueous solution, the primary titanomagnetite grains in the rocks may have survived the weathering process more or less intact, perhaps being oxidized to titanomaghemite [Hargraves et al., 2000; Coey et al., 1990]. Magnetic grains inherited from the basaltic surface rocks will be attracted to permanent magnets as those carried by the Viking and Pathfinder landers. In this case the magnetic grains will, with near certainty, contain the element titanium. See also the work by Bell, Morris and collaborators [Bell et al., 1993; Morris et al., 1993, 2001].
 From X-ray fluorescence spectroscopy performed during the Viking and Pathfinder missions the element titanium is known to be present in the Martian soil in an abundance of nearly 1% by weight [Toulmin et al., 1977; Rieder et al., 1997]. An important goal of the MPE on MERs will therefore be to determine if the titanium on Mars is located in the magnetic minerals. If this is the case the airborne dust collected by the magnets will show an enhancement of titanium relative to the dust on the Martian surface.
 In conclusion: The identification of the magnetic phase(s) in the soil and dust on Mars, and the determination of the distribution of grain size of the magnetic mineral, will contribute to our understanding of the role of liquid water in the formation of the ubiquitous reddish soil and airborne dust on the planet.
3. Description of the Magnets
 Each Mars Exploration Rover carries a set of permanent magnets of various strengths and dimensions. The magnets are given the following names: Capture magnet, Filter magnet, Sweep magnet, and Rock Abrasion Tool (RAT) magnets. Dust particles are attracted to the magnets by the magnetic force. The force (i.e., the magnetic field and field gradient) is controlled by the shape and position of the magnetic material. Each type of magnet has its individual design and purpose, which is briefly described below.
3.1. Capture Magnet and Filter Magnet
 The Capture and Filter magnets are located on the front of the rovers (see Figure 2). When enough dust has been attracted to the magnets, this dust will be studied by the three instruments on the Instrument Deployment Device (IDD): 1) The Mössbauer Spectrometer (MB), 2) the Alpha Particle X-ray Spectrometer (APXS), and 3) the Microscopic Imager (MI). The IDD is able to bring these three instruments as close to the Capture and Filter magnets as desired.
 The Capture and Filter magnets (see Figure 3) are basically constructed in the same way. The magnets have circular symmetry and the active magnetic material used is Sm2Co17, which has a magnetization of M = 8.75 · 105 A m−1 (at room temperature) and a Curie temperature of TC = 1190 K. The magnets are embedded in a structure made of aluminum, and the active surface (where the dust is attracted) is covered with a foil of high purity aluminum. For each magnet, the Sm2Co17 part has a diameter of 25 mm, while the aluminum structure has a diameter of 45 mm. All excessive material has been removed from the aluminum structure.
 The main difference between the two magnets is their strengths, i.e., the magnitude of the magnetic field B and of the magnetic field gradient ∇∣B∣ on the active surface of the magnet. B and ∇∣B∣ are controlled by the size and shape of the active magnetic material (the Sm2Co17 part). The dust will be collected mainly on the part of the active surface where the field gradient is largest, because the force on the magnetic grains is proportional to the field gradient. The force density is given by f = (M·∇) B, where M is the magnetization of a given grain. (Unit of f: N m−3). The shape and size of the two magnets are chosen with the following objectives in mind. The Capture magnet is designed to be as strong as possible, and therefore B and ∇∣B∣are as large as possible. For the Filter magnet we have chosen the magnitude of B and ∇∣B∣ such that particles with a high magnetic susceptibility will be attracted much more efficiently than weaker magnetic particles (hence the “filtering” of dust particles according to magnetic properties). For the sake of the interpretation of the Mössbauer spectra obtained of the dust on the magnets, the magnetic field B should not vary too much across the active area of the magnet. The magnetic field gradient should also not vary drastically to assure a nearly homogeneous layer of dust on the magnets. These objectives cannot all be achieved at the same time, so tradeoffs have been made. A further complication is the fact that we do not know the distribution of magnetic properties of the airborne dust particles on Mars. The design of the magnets was optimized according to the objectives mentioned above, and the final design of the MER Capture and Filter magnet is described below.
 The Sm2Co17 part of the Capture magnet is a small cylinder surrounded by 3 concentric rings, which all have a height (thickness) of 5.0 mm. The dimensions and locations (depth below the active surface) of the 4 components are given in Table 1. The central cylindrical magnet is magnetized with the North pole near the active surface. The innermost ring magnet has the South pole outward, and so forth; the magnetic poles alternate. Figure 4 shows the magnitudes of the magnetic field and the magnetic field gradient across the active area for the Capture magnet. The maximum value of B is 0.46 T and for ∇∣B∣ the corresponding value is 550 T m−1. The direction of the field and field gradient is also shown on Figure 4.
Table 1. Dimensions of Capture Magnet Components
Inner Radius, mm
Outer Radius, mm
Depth Below Surface, mm
 The shape of the Sm2Co17 part of the Filter magnet is a slightly modified ellipsoid, with a height of 4 mm. The lower part, 0.5 mm high is a cylinder with radius 12.5 mm. The upper part, 3.5 mm high, is a rotated modified ellipse described by the formula
where a = 12.5 mm and b = 3.5 mm. This curve is rotated with respect to the z axis (symmetry axis of the magnet). The Sm2Co17 part is located 2 mm below the active surface. Figure 5 shows the magnitudes of the magnetic field and the magnetic field gradient across the active surface for the Filter magnet. The maximum value of B is 0.2 T and for ∇∣B∣ the corresponding value is 34 T m−1. Figure 5 also shows the direction of B and ∇∣B∣ across the Filter magnet.
 The Filter and Capture magnets are both equipped with surface markings (indentations) of depths 5 μm, 10 μm and 20 μm. The indentations were produced by gently pressing a spherical steel ball into the surface until the required depth was reached. The resulting diameters of the spherical indentations are between 100 and 200 micrometers, which should be easily observable using the MI [Herkenhoff et al., 2003].
3.2. Sweep Magnet
 The Sweep magnet is placed on the upper surface of the rover (on the –X solar panels) near the Pancam Calibration Target. The Sweep magnet is therefore in view of Pancam. The purpose of the Sweep magnet is to detect non-magnetic atmospheric dust grains. Figure 6 shows drawings of the Sweep magnet. The magnet itself (Sm2Co17) is a ring magnet with inner radius of 2.0 mm and outer radius of 4.5 mm. The thickness (height) of the ring is 5 mm. The ring is embedded in an aluminum structure. For reasons explained below the magnet should be strong and therefore the Sm2Co17 magnet is placed only 0.4 mm below the active surface. The maximum value of the magnetic field at the surface of the magnet is B = 0.42 T and the maximum value of the field gradient is 450 T m−1.
Figure 7 shows the direction of the magnetic field gradient across the Sweep magnet. At a very small distance above the surface the magnetic field gradient points upward from the magnet in the center, and the magnetic force will therefore prevent magnetic particles from reaching the central part of the surface of the Sweep magnet.
3.3. RAT Magnets
 Each rover carries a Rock Abrasion Tool (RAT) [Gorevan et al., 2003] as one of the payload elements on the Instrument Deployment Device (IDD) [Squyres et al., 2003]. The purpose of the RAT is to remove dust and weathering products from selected rocks to provide a pristine surface suitable for investigation of the interior of rocks by the Microscopic Imager, the APXS and the Mössbauer spectrometer.
 The RAT will remove material from the rock surface by using a fine grinding tool moving rapidly across the surface with an applied force of several tens of newtons. The grinding head rotates at about 3000 RPM, and as it does a planetary gear sweeps it in an arc, grinding away a circular region with a diameter of 45 mm. For more details, see Gorevan et al. . The RAT is equipped with two brushes with the purpose of keeping the worked area clean of dust particles.
 Four RAT magnets are integrated into the RAT as shown in Figure 8. They are placed behind the grinding tool and the small brush. The purpose of the RAT magnets is to detect magnetic minerals in the abraded rock material. The four magnets are grouped in pairs marked with numbers: 1, 2, and 1, 3. Figure 9 shows the variation of the magnetic field and the magnetic field gradient across the active area for the two pairs of RAT Magnets. Two of the RAT magnets, ‘type 1’, are identical, and the maximum value of B and ∇∣B∣ is 0.28 T and 350 T m−1. The remaining 2 magnets are weaker: For ‘type 2’, B = 0.1 T and ∇∣B∣ = 120 T m−1, while the maximum values for ‘type 3’ are B = 0.07 T and ∇∣B∣ = 80 T m−1.
 Special rover operations will be conducted for each component of the Magnetic Properties Experiment. The simplest are those for the Sweep magnet. This magnet is located directly adjacent to the Pancam calibration target; each image of the target will include the Sweep magnet in it. Images of the target will be included for calibration purposes with every major Pancam imaging sequence, typically through a number of color filters. A good time baseline of high-resolution color imaging of the Sweep magnet will therefore be acquired in the course of normal operations.
 The RAT magnets will be imaged before and after each use of the RAT. In a normal grinding sequence, the IDD will first be used to position the IDD so that the RAT magnets can be viewed by Pancam at the highest possible spatial resolution. Images will be acquired through each Pancam color filter, and then the RAT will be used to remove material from the rock. After grinding, the IDD will move the RAT back to the same imaging position, and the multispectral imaging sequence will be repeated.
 The most complex magnet operations will be those involving the Filter and Capture magnets. These will first be imaged through all Pancam color filters as soon as possible after landing to establish a baseline for their dust-free appearance under Martian illumination. They will subsequently be imaged every few sols through at least two band-pass filters that are widely separated in wavelength to monitor the buildup of dust on their surfaces.
 To further substantiate the evaluation of the amount of dust that has accumulated on the surfaces of the magnets, the previously described surface markings (indentations) on the magnets may be used. From the way the indentations become increasingly masked by the accumulated dust layer, we may be able to estimate the thickness of the dust layer. Observation of these markings using the MI may allow us to obtain more precise information of the thickness of the dust layer, i.e., more precise information than would be possible using Pancam alone. We anticipate that the results of few MI observations could be used to calibrate the contrast versus time of the dust on the magnets as observed by Pancam, so that we will know the connection between contrast and layer thickness. There is some risk, however, that such MI observations could result in inadvertent disruption of the accumulating dust by the MI contact sensor, so we will make the decision whether or not to use the MI to assess dust buildup only after we have assessed the behavior of the IDD under Martian conditions.
 Once Pancam and perhaps MI images have shown that a thick enough layer of dust has accumulated, one or more sols of operations will be devoted specifically to magnet observations. The magnets will be examined carefully by Pancam, the Mössbauer Spectrometer, the APXS, and the MI. Pancam imaging will be conducted at full spatial resolution through all color filters, with the rover positioned for optimal illumination of the magnets. Mössbauer spectra will be obtained throughout the Martian diurnal cycle, in order to acquire data over a wide range of temperatures. APXS and MI data will be obtained as well. The Mössbauer Spectrometer requires direct physical contact with the magnets in order to minimize velocity noise at the sensor head. Because this contact is likely to disturb the dust on the magnet, the Mössbauer spectrometer will be the last instrument used.
 Mini-TES cannot be used to view the magnets, because magnets lie outside the range of elevation angles (+30° to −50°) viewable by Mini-TES.
 It is difficult to predict when during the mission an adequate accumulation of dust will have taken place. The Sojourner rover carried a materials adhering experiment (MAE) monitoring the amount of dust settling on the solar array during the Mars Pathfinder mission [Rover Team, 1997]. The dust accumulation rate was 3 μg cm−2 per sol, which is in agreement with the globally averaged sedimentation rate calculated by Pollack [Pollack et al., 1995, and references therein]. Accordingly, after 80 sols (i.e., the duration of the Pathfinder mission) the amount of dust settling on all surfaces of the Mars Pathfinder lander would have been 0.24 mg cm−2. We estimate that the strongest magnet in the magnet arrays on Mars Pathfinder should have been at least one order of magnitude more efficient in capturing magnetic dust than a surface with no magnetic field. If correct (and if most of the particles were magnetic), this implies that there were at least a couple of milligrams of dust per cm2 on the strongest magnet toward the end of the Mars Pathfinder mission [Madsen et al., 1999]. Using other methods Gunnlaugsson estimated the amount of dust on the strongest Mars Pathfinder magnet to be somewhat less, i.e., about 0.7 mg cm−2 [Gunnlaugsson, 2000].
 The strongest magnet on the MER rovers is somewhat stronger than the strongest magnet on Mars Pathfinder. Our best estimate based on Pathfinder experience is that the amount of dust collected by the strongest magnet will be between 1 mg cm−2 and 5 mg cm−2 after a nominal accumulation interval of 30 sols. As we shall see below the amount of dust on the magnets will be decisive for the quality of the Mössbauer spectra obtained.
4.1. Capture and Filter Magnets
 A Mars simulation chamber has been built at the University of Århus in Denmark. The chamber contains a re-circulating wind tunnel, which is 0.3 m wide, 0.4 m high and 2 m long. Temperature, pressure and wind velocity can be controlled and the atmosphere in the chamber can be loaded with dust. The chamber has been described elsewhere [Merrison et al., 2002].
 The ability of the magnets to capture dust has been tested in the simulation chamber using a series of Mars sample analogues. The Mars analogue dust used in the example discussed here was red soil found on a location in Denmark named Salten. The iron-containing minerals in the Salten soil are hematite (α-Fe2O3, ∼10%), maghemite (γ-Fe2O3, ∼10%) and goethite (α-FeOOH, ∼60%). The soil also contains other salts and a small amount of quartz. Figure 10 shows the magnetization of the Salten soil as function of an applied external magnetic field. The saturation magnetization of the Salten soil appears to be similar to the saturation magnetization of the Martian dust [Madsen et al., 1999]. The saturation magnetization of the Salten soil is about 3 A m2 kg−1 (the increase of the magnetization with an applied field above 200 mT is due to a paramagnetic contribution to the magnetization). It should be stressed that the Salten soil is a reasonable Mars Dust Analogue with respect to the saturation magnetization. Other properties of the Salten soil are very different from Martian soil.
Figure 11 shows the result of the experiment in the Mars Simulation chamber. The Capture and Filter magnets were placed in the chamber and dust of Salten soil was introduced into the chamber. The wind velocity was about 1 m/s. The Salten soil is rather homogeneous, and there is only a slight (if any) difference between the dust attracted to the two magnets in the simulation experiments.
 An important question is if there (on Mars) will be a difference in the composition of the dust attracted by the two magnets. Will the magnetic particles in the airborne dust on Mars be “picked out” (“filtered”) by the magnets? The stronger Capture magnet will certainly be able to attract particles with smaller magnetic susceptibility than the Filter magnet. We shall return to these properties of the magnets when we discuss the Alpha Particle X-ray Spectroscopy experiment.
4.2. Sweep Magnet
 As previously stated: From the Mars Pathfinder results it has not been possible to determine if the magnets were able to attract every particle in the airborne dust that approached the magnet, or if the magnets were attracting only a subset of the particles suspended in the Martian atmosphere.
 The strong, ring shaped Sweep magnet will be able to keep a certain area inside the ring (the “sweep area”) free of magnetic particles. Magnetic particles heading for the “sweep area” will be deflected by the magnetic force on the particle. By “magnetic particles” we here include also paramagnetic particles. All particles having a magnetic susceptibility χ ∼ 10−3 and higher will be unable to settle in the sweep area (χ ∼ 10−3 corresponds to a specific susceptibility of about κ = χ/ρ ∼ 0.3 · 10−6 m3 kg−1; ρ ∼ 3000 kg m−3). By this process the Sweep magnet experiment tries to answer the following question: Are there any non-magnetic particles present in the Martian airborne dust, i.e., particles containing little or no iron?
 During the simulation experiment with the Capture and Filter magnet in the Mars simulation chamber at Århus University a Sweep magnet was also present in the chamber. The result is shown in Figure 12a. Salten soil is evidently swept away from the sweep area, i.e., the area inside the ring. The specific magnetic susceptibility of Salten soil is κ = 20 · 10−6 m3 kg−1.
 A series of simulation experiments with the Sweep magnet have been performed in a more primitive simulation chamber (terrestrial atmosphere and atmospheric pressure). Various types of Mars analogue dust were used. Figure 12b shows results obtained with hematite (α-Fe2O3). Hematite has a specific magnetic susceptibility of κ = 5 · 10−6 m3 kg −1. The hematite particles are nearly all swept away from the area inside the ring. Figure 12c shows the results obtained by using the clay mineral Riverside nontronite. The specific magnetic susceptibility of Riverside nontronite is low, κ = 0.5 · 10−6 m3 kg−1 (For further discussion of Riverside nontronite, see section 5.1.3). It is seen that the ring pattern of the low susceptibility clay mineral is slightly different from the ring patterns of Salten soil and hematite, and a small amount of Riverside clay is trapped in the sweep area. The Riverside particles are less effectively attracted by the magnet.
4.3. Rock Abrasion Tool (RAT) Magnets
 During the process of RAT operations, rock material will be removed in the form of fine particles. Grinding experiments have shown that nearly all of the particles will be smaller than 100 μm in diameter, and that several tens of percent by mass will be smaller than 20 μm [Gorevan et al., 2003]. Some of the abraded particles will pass the RAT magnets. If part of the abraded rock material is magnetic, this material will be attracted to the magnets, where it can be trapped and subsequently imaged. A complicating factor is that the magnets will be partly hidden beneath the RAT's grinding heads; this means that viewing conditions may be better for imaging of one of the pairs of magnets than for the other pair.
5. Data Products
 The data products for the Magnetic Properties Experiment will consist of 1) Mössbauer Spectra of dust collected by the magnets, 2) Alpha Particle X-ray Spectra of dust collected by the magnets, 3) Images, and visible and infrared spectra of dust collected by the magnets (MI and Pancam).
5.1. Mössbauer Spectra of Dust Collected by the Magnets
 The Mössbauer spectroscopy will be discussed under the following headings: 1) Mössbauer spectroscopy and the amount of dust on the magnets, 2) Polarization effects, 3) Superparamagnetic relaxation effects.
 It should be clearly understood that the description below does not include any detailed analysis of the Mössbauer spectra presented. The purpose is to illustrate which data products we can expect, and which problems we may encounter in the Mössbauer spectroscopy study of Martian magnetic dust. An introduction to Mössbauer spectroscopy will not be given here, but can instead be found in Greenwood and Gibb  and Wdowiak et al. . From the Mössbauer spectra of the magnetic dust on Mars, we hope to identify the mineral(s), which causes the high magnetization of the dust. Together with other mission results, the Mössbauer spectra will give clues to understanding the processes that formed the minerals in the dust.
5.1.1. Mössbauer Spectroscopy and the Amount of Dust on the Magnets
 In our simulation experiments we use a backscattering Mössbauer spectrometer similar to the MIMOS-II spectrometer [Klingelhöfer et al., 2003]. A significant difference is that our spectrometer has one detector while the MIMOS-II spectrometer uses four detectors. The geometry used in the simulation experiments is shown schematically in Figure 13. The essential aspect is that the center of the circular Mössbauer source is axially above the center of the circular magnet. The shield/collimator is circular with an inner radius of 8 mm. The distance from the collimator to the surface of the magnet is 10–15 mm. The γ-rays impinge on a circular area of approximately 2 cm2 of the active surface of the magnet.
 In order to find the minimum amount of dust we can detect with the Mössbauer spectrometer a series of simulation experiments have been performed. Small amounts of hematite (α-Fe2O3) powder were weighed and placed on a magnet. A backscattering Mössbauer spectrum was then recorded.
 The signal to background ratio depends on several parameters, but we expect the magnitude of the Mössbauer resonance signal to be a linear function of the thickness of the dust layer on the magnet. This is certainly true for small thicknesses.
 Two such Mössbauer spectra are shown in Figure 14a. The amount of material (α-Fe2O3) on the magnet was 0.7 mg/cm2 (lower spectrum) and 2.9 mg/cm2 (upper spectrum). One more spectrum was recorded (1.4 mg cm−2) and Figure 14b shows a linear fit of the signal to background (the 'area' of the spectrum, unit: mm s−1·%) versus amount of material used. By inspection alone it is clearly seen that a change from 0.7 mg cm−2 to 2.9 mg cm−2, i.e., a change in the dust loading by about a factor of 4, makes a significant improvement in the quality of the spectrum obtained.
 The spectra shown in Figure 14a were recorded with a Mössbauer source with a strength of about 3.7 · 108 Bq (10 mCi). The counting time was 24 hours and in this case only one detector was used. The Mössbauer spectra to be recorded on Mars, using a stronger source and four detectors, will be better in the sense that the signal to noise (background) ratio will be substantially improved. However, our simulation experiments clearly show that the amount of dust on the magnets for which a useful Mössbauer spectrum can be obtained is of the order of a few mg cm−2, depending somewhat on the number of iron-containing phases in the dust.
Figure 15a shows a Mössbauer spectrum of Salten soil without an external magnetic field. The scatterer was a layer of soil with a thickness of about one cm, i.e., the scatterer was, from the point of view of Mössbauer spectroscopy, infinitely thick. Figure 15b shows a backscattering Mössbauer spectrum of Salten soil on the Capture magnet. The amount of dust on the magnet was 6 mg cm−2.
 The spectra are shown with the purpose of illustrating the difficulties met when several iron-containing phases are present simultaneously. As previously mentioned the Salten soil has about the same saturation magnetization as the dust on Mars, but in other respects the magnetic properties of the Martian dust and the Salten soil are probably very different. For instance: In the Salten soil the magnetic minerals are very close together and there are thus significant interactions among the crystallites. Such interactions will strongly influence the Mössbauer spectrum of the soil (superferromagnetism [Mørup, 1983]). For further discussion of these problems we refer the reader to section 5.1.3.
 Before proceeding to such discussions we shall first describe the polarization effects in connection with Mössbauer spectroscopy of dust on the magnets.
5.1.2. Polarization Effects
 The following discussion will show that for a variety of reasons, the detailed interpretation of the Mössbauer spectra of dust collected by the magnets may turn out to be rather complicated. We shall not treat all the problems related to a proper interpretation of the spectra but limit ourselves to a discussion of some essential aspects.
 Any magnetically ordered mineral (ferromagnetic, ferrimagnetic or antiferromagnetic) will have a magnetic hyperfine field acting on the iron nuclei. The hyperfine field gives rise to a Mössbauer spectrum showing the nuclear Zeeman effect, reflected in the well known six line spectrum of for example metallic iron (Figure 16) or maghemite. When a ferrimagnetic mineral is placed in a sufficiently strong external magnetic field B the magnetization vector M will be parallel to B. The magnetic hyperfine field is parallel (or antiparallel) to M. In this case, and using the usual transmission geometry, there is a simple expression for the relative intensity of the six Mössbauer absorption lines as a function of the angle θ between the propagation direction of the γ-radiation and the magnetic hyperfine field [Greenwood and Gibb, 1971].
 For backscattering Mössbauer spectroscopy the situation is somewhat more complicated because as well the incident angle θ1 relative to the magnetic field as the scattering angle θ2 relative to the field must be taken into account [Balko and Hoy, 1974].
 In the appendix we have given the angular dependence of the relative intensity of the Mössbauer lines in backscattering geometry and in a magnetic field. Countless Mössbauer spectra have been recorded in transmission geometry and with an applied external magnetic field. In transmission geometry the applied magnetic field is usually strong, constant and homogeneous across the absorber. The applied magnetic field is usually applied either parallel to (θ = 0), or perpendicular to (θ = π/2) the direction of propagation of the γ-rays.
 In backscattering geometry and for dust on a permanent magnet, the applied magnetic field will vary across the magnet, i.e., the field changes from grain to grain in magnitude as well as direction. The same is true for θ1 and θ2. These facts make the detailed interpretation of the backscattering Mössbauer spectra substantially more complicated than the corresponding spectra in transmission geometry.
Figure 16 shows a backscattering Mössbauer spectrum of a block of multicrystalline metallic iron without an applied external magnetic field. In the iron block the small metal grains constituting the block are stochastically distributed and consequently the magnetic hyperfine fields are likewise stochastically distributed. From formulas given in the appendix it is easy to demonstrate that also in backscattering geometry the relative intensity of the 6 Mössbauer lines for the iron block should be 3:2:1:1:2:3. This is what the spectrum in Figure 16 shows.
Figure 17a shows the result of backscattering Mössbauer spectroscopy of a powder of maghemite γ-Fe2O3. The spectrum to the left is a spectrum of maghemite in zero applied magnetic field and the spectrum to the right is of maghemite on a Filter magnet. The relative intensity of the six lines in the left spectrum (without magnetic field) is, like in the case of metallic iron, 3:2:1:1:2:3. This is not the case for the spectrum to the right. Qualitatively this can be understood as follows.
 If the magnetic field on the dust were constant and everywhere perpendicular to the surface of the magnet and strong enough to orient the magnetization of the grains, and if all the γ-rays were impinging along the applied magnetic field (θ1 = π), then lines 2 and 5 would disappear. As the spectrum shows, a polarization effect is present. Lines 2 and 5 have not disappeared, but they are clearly diminished relative to the corresponding lines in the spectrum to the left, i.e., the spectrum of maghemite with no applied external magnetic field. There are several factors, which give rise to the fact that line 2 and 5 do not disappear completely. The incident angle θ1 and the scattering angle θ2 vary from grain to grain. The applied magnetic field from the magnet vary in both magnitude and direction from grain to grain, and the applied field is not everywhere strong enough to orient the magnetization of all the grains along the field. Qualitatively: In the central part of the Filter magnet the magnetic field is strong and perpendicular to the surface of the magnet, while close to the rim the field is weaker and mainly parallel to the magnet surface. A large part of the Filter magnet has the field B perpendicular to the magnet surface (see Figure 5). The lines 2 and 5 are diminished, but they have not disappeared.
Figure 17b shows backscattering Mössbauer spectra of hematite (α-Fe2O3) in zero applied field (left) and a spectrum of hematite on a magnet (right). Here the polarization effect causes lines 2 and 5 to increase. This is due to the fact that hematite has a canted antiferromagnetic structure, where the magnetic hyperfine field on each iron nucleus is almost perpendicular to the total magnetic moment of the hematite grains. The hematite grains are partially oriented by the magnetic field.
 The magnetism of Martian dust may be caused by more than one magnetic phase; the soil may for instance contain a mixture of maghemite and hematite. A further complication is that the results of previous missions [Hargraves et al., 1977; Madsen et al., 1999] show that the ordered magnetic phase (or phases) constitutes only part of the iron-containing minerals in the soil and dust. The minimum number of iron-containing minerals (phases) present in the soil of Mars are two: One (magnetically ordered) ferrimagnetic phase and one non-ferrimagnetic phase. We expect, however, that the Martian soil contain several different iron-containing minerals, and therefore we will receive data corresponding to a Mössbauer spectrum, which is a superposition of spectra of several iron-containing phases.
 From the discussion above it appears that Mössbauer spectra of the soil and dust on the magnets may turn out to be rather difficult to interpret. In connection with this statement one essential aspect should be emphasized: A Mössbauer spectrum of dust collected by the magnets may disclose if superparamagnetic particles are present in the dust on Mars. We therefore turn to a brief discussion of superparamagnetism in relation to Mössbauer spectroscopy.
5.1.3. Superparamagnetic Relaxation Effects
 The phenomenon of superparamagnetism is associated with fluctuation of the magnetization vector M in magnetically ordered microcrystals (d < 20 nm).
 The magnetization vector M in a ferrimagnetic crystal has certain preferred orientations, called easy directions. The easy directions are determined by the crystalline magnetic anisotropy energy. In the simple case of uniaxial anisotropy, the anisotropy energy of the crystallite is given by:
where K is the anisotropy energy constant [J m−3], V is the volume of the crystallite, and θ is the angle between the magnetization direction and the easy direction. According to (1) the two easy directions θ = 0 and θ = π are separated by a potential barrier of KV.
 For temperatures T, where kT ≪ KV, the magnetization vector M will be kept essentially along the easy direction by the forces of anisotropy (k is Boltzmann's constant).
 For kT ≈ KV the magnetization vector has a finite probability of surmounting the energy barrier that separates the two minima.
 For kT ≫ KV the magnetization vector will fluctuate rapidly back and forth between the two minima, in response to the thermal noise in the microcrystal. This phenomenon is referred to as superparamagnetic relaxation.
 The rate of the process of change of the magnetization vector is characterized by a relaxation time τ. Stated differently, the relaxation frequency (τ−1) is defined as the rate at which the magnetization vector surmounts the energy barrier separating the minima.
 The relaxation time is strongly dependent on the temperature and the volume of the particle. For an isolated microcrystal with uniaxial anisotropy, τ is given approximately by
 The Mössbauer spectrum of a ferrimagnetic microcrystal is strongly dependent on τ [Mørup and Topsøe, 1976]. The critical timescale of Mössbauer spectroscopy that should be compared to τ is the Larmor precession time of the nuclear magnetic moment in the given magnetic hyperfine field. In the hyperfine fields encountered in the magnetic iron oxides the Larmor periods are of the order of 10−9 s.
 If a superparamagnetic particle has a relaxation time τ ≫ 10−9 s, a magnetically split Mössbauer spectrum will be observed. In this case the magnetic hyperfine field on the iron nuclei will be constant over many Larmor periods and the various Mössbauer levels are thus well defined. In Mössbauer spectroscopy the microcrystal will appear as a macroscopic crystal.
 If, however τ ≪ 10−9 s, the magnetization vector fluctuates swiftly. Consequently, the magnetic hyperfine field will fluctuate so rapidly that the various Mössbauer levels are ill defined. To put it differently, for τ ≪ 10−9 s the average value of the magnetic hyperfine field, as seen with Mössbauer spectroscopy, is zero. The Mössbauer spectrum will in this case collapse to one or, if a quadrupole splitting is present, two lines.
 The blocking temperature (TB) of a microcrystal is defined as the temperature below which a given microcrystal behaves, in Mössbauer spectroscopy, like a magnetically ordered macrocrystal. The blocking temperature TB can be given approximately as the temperature for which kTB ≈ KV.
 According to (2), the relaxation time at a given temperature is very sensitive to the volume V of the particle. A Mössbauer scatterer, like the soil on Mars, will contain a distribution of particle sizes, and the observed spectrum will be a sum of spectra corresponding to different relaxation times. For a relaxation time of the order of τ ≈ 10−9 s the Mössbauer spectra will be rather complicated. Due to the exponential function in equation (2), τ will be of the critical order (∼10−9 s) for only a very narrow range of particle volumes. Therefore only a tiny fraction of the particles in the soil and dust will have volumes, which, for a given temperature, corresponds to the critical region of τ ∼ 10−9 s. The relaxation time will generally be either much larger or much smaller than the relaxation times in the critical region. Therefore the Mössbauer spectrum of a powder of naturally occurring particles will consist of two components. For the “large” particles (τ ≫ 10−9 s) a magnetically split spectrum will appear. Such particles are, at the given temperature, large enough to block the fluctuations of the magnetization vector. For “small” particles (τ ≪ 10−9 s) a paramagnetic spectrum, often with two lines due to quadrupole splitting, will be the result. For the “small” particles the average value of the magnetic hyperfine field over a time period corresponding to a characteristic Mössbauer timescale (Larmor period ∼10−9 s), will be zero. Examples are widespread in the literature and may be seen in books on Mössbauer Spectroscopy [e.g., Greenwood and Gibb, 1971]. See also Morris et al. .
 Below the blocking temperature TB the Mössbauer spectrum of magnetically ordered microcrystals show the characteristic, magnetically split 6-line spectrum. Such spectra may, however, still be affected by fluctuations of the magnetization vector.
 For a magnetic microcrystal the probability f(θ) that the magnetization vector forms an angle between θ and θ + dθ with the easy direction is
If the magnetic anisotropy energy KV is very large compared to kT the magnetization vector M is locked along the easy direction. For somewhat smaller values of KV the magnetization vector M will fluctuate in directions close to the easy direction, following the thermal noise in the small crystal. These kinds of fluctuations are called collective magnetic excitations, and they are fast compared to the timescale of Mössbauer spectroscopy. If the magnetic hyperfine field in a given macrocrystal is observed to be B0, the magnetic hyperfine field B, observed in a corresponding microcrystal will be
For KV ≫ kT a linear approximation can be used:
The magnetic hyperfine field for such crystallites are thus smaller than the hyperfine field observed in corresponding macrocrystals.
 If an external magnetic field Bext is applied to a sample of superparamagnetic particles, the magnetic energy of a single particle is given by
where m is the total magnetic moment of the given particle. The magnetic moment of the particle is thus more likely to be parallel to the applied external field Bext than antiparallel to the field. For very large external fields, the energy barrier will become so large that the magnetic moment will be locked in the direction of the external field, and no superparamagnetic relaxation will be observed. The Mössbauer spectrum of a sample of superparamagnetic particles may therefore change drastically if an external magnetic field is applied to the sample, even if the external field is not strong enough to completely lock the magnetization vector of all the particles. This simple fact will be of great importance for the study of the Martian soil and dust by means of Mössbauer spectroscopy. The spectra of the general surface soil and dust will be carefully compared with the Mössbauer spectra obtained of the dust collected by the magnets. The temperature dependence of the spectra will be studied by comparing the day and night integrations.
 We have performed simulation experiments to observe the significance of the magnets using samples containing superparamagnetic particles. As an example we show here the results of one such experiment.
 The clay mineral Riverside nontronite has been studied as a Mars sample analogue [Moskowitz and Hargraves, 1982]. We have used Riverside nontronite from the same batch as used by Moskowitz and Hargraves. By heating the nontronite to 950°C for 20 minutes, the clay is decomposed and becomes reddish and magnetic with a saturation magnetization of about 10 Am2 kg−1. The magnetism of the decomposed clay mineral is caused by the presence of small crystallites of the mineral maghemite.
Figure 18a is a Mössbauer spectrum of a sample of thermally decomposed Riverside nontronite without applied magnetic field. The spectrum is complicated and we shall not analyze it in detail here but call attention to the following pertinent facts: The spectrum has been fitted with two ferric doublets, lying between −1.0 mm s−1 and +1.5 mm s−1, and a broad poorly resolved sextet. The shape of the sextet is due to a distribution of hyperfine fields. The distribution has a most probable value of 47 T and an average value of 41 T. The isomer shift of the components is similar and close to 0.27 mm s−1. These values are consistent with poorly crystalline maghemite of nanosize dimensions.
 The essential feature of the simulation experiment is the following. A Mössbauer spectrum of a sample of decomposed Riverside nontronite placed on a magnet (B ≈ 0.15 T) is drastically different from a Mössbauer spectrum of the same sample in zero magnetic field. See Figure 18b. The rather weak magnetic field is evidently strong enough to substantially suppress the fluctuations of the magnetization vector in the superparamagnetic maghemite particles. The magnetic hyperfine splitting is therefore restored.
 The area of the sextet component in Figure 18b has grown significantly, relative to the magnetic component in Figure 18a, at the expense of the intensity of one of the ferric doublets. Another effect of the external magnetic field is a drastic narrowing of the hyperfine values. The outer lines of the sextet are steeper on the outer side than on the inner side, testifying to collective magnetic excitations in some of the particles. Note moreover the polarization effects: Lines 2 and 5 have a rather low intensity.
 The described change of the Mössbauer spectrum as a result of the application of an external magnetic field is typical of the behavior of microcrystals of ferrimagnetic materials. Therefore, if superparamagnetic particles are present in the Martian soil and dust it is evident that a Mössbauer spectrum of the dust collected by the magnets may turn out to be helpful for the interpretation of the Mössbauer spectra obtained of the soil on the surface of Mars.
 The particle size of the maghemite crystals in the Salten soil and the maghemite in heated Riverside nontronite is approximately the same (d ∼ 10–20 nm). The Mössbauer spectrum of the heated Riverside nontronite changes drastically when placed on a magnet (Figures 18a and 18b). The difference between the Mössbauer spectra of Salten soil with and without a magnetic field is more subtle (Figures 15a and 15b). The crystallites of maghemite in heated Riverside nontronite are so far apart that they do not interact, while there are strong interactions among the magnetic crystallites in Salten soil. The superparamagnetic relaxation of the Riverside sample in respect to Mössbauer spectroscopy is therefore much more pronounced than the corresponding properties of Salten soil.
 A detailed analysis of the Mössbauer spectra of Salten soil shown in Figure 15 is rather involved. We here limit ourselves to the following remarks:
 The spectrum of Figure 15a was fitted with four components. Two of the components are simple six line components. One component has a magnetic hyperfine field of 49 T, an isomer shift of 0.37 mm s−1, a quadrupole shift of −0.1 mm s−1, corresponding to small crystallites of imperfect hematite. Another sextet component has a magnetic hyperfine field of 47 T, an isomer shift of 0.32 mm s−1, a quadrupole shift of 0 mm s−1, corresponding to crystallites of imperfect maghemite. Additionally a sextet with split-lorentzian line profiles were included to model the relaxation of the third magnetically split component. Finally a paramagnetic doublet component was present.
Figure 15b shows the spectrum of the Salten soil as collected by a Capture magnet. The main effect of the field from the permanent magnet is that part of the area of the doublet is moved into the broad relaxed component. The simpler sextet components seem virtually unaffected by the magnetic field. This behavior is in accordance with an assignment of the broad relaxed component to goethite, which is not far below its apparent transition temperature, and in this range goethite is very sensitive to the presence of a low intensity magnetic field.
 The dust particles on Mars are probably mainly silicate particles containing some iron oxides. Among these iron oxides there is at least one ferrimagnetic phase, possibly maghemite [Madsen et al., 1999]. The maghemite particles will be distributed in the weakly magnetic silicates, and therefore the interaction between the maghemite crystallites in the Martian dust particles will be extremely weak. Therefore, from the point of view of Mössbauer spectroscopy it is highly probable that the heated Riverside nontronite is a better Mars soil analogue than the Salten soil.
5.2. Alpha Particle X-Ray Spectroscopy
 The elemental composition of the magnetic dust on the Filter and Capture magnets will be studied by the Alpha Particle X-ray Spectrometer (APXS) [Rieder et al., 2003], which is located on the Instrument Deployment Device (IDD). The elemental composition of the dust on the magnets will be compared to the elemental composition of the soil on the surface of Mars.
 As noted above, questions related to the element titanium are of particular significance. If the dust collected by the magnets is found to be enriched in titanium relative to the average dust on the surface, it will be nearly certain that the magnetic phase is titanomagnetite or titanomaghemite, i.e., the magnetic mineral has been inherited more or less intact from the Martian surface rocks.
 As previously mentioned, the particles suspended in the Martian atmosphere seem to be composite (multiphase) particles [Madsen et al., 1999]. The magnetic properties may therefore differ from particle to particle. The Filter magnet will not attract weakly magnetic particles as efficiently as the Capture magnet. By comparing the APX spectra of the dust collected by the two magnets we shall be able to decide if a filtering of the dust according to magnetic properties has taken place. Should the magnetism of the soil, contrary to what we now think, be caused mainly by discrete (single phase) particles of a strongly magnetic oxide (for example titanomagnetite), the Filter magnet will attract these particles with nearly the same efficiency as the Capture magnet. For example: If all the titanium (∼0.7% Ti of soil by weight) is present in titanomagnetite (or titanomaghemite) the grains will have a high saturation magnetization, ∼30 A m2 kg−1, and such grains will easily be attracted by both magnets. The question of filtering is especially important if single-phase titanomagnetite as well as multiphase particles are present in the airborne dust on Mars.
 The sensitivity of the APXS investigations is substantially higher than the corresponding Mössbauer investigations. While there is a risk that the amount of dust on the Filter magnet will not be sufficient to obtain a useful Mössbauer spectrum, we are confident that there will be enough dust also on the Filter magnet to give a useful alpha particle X-ray spectrum.
5.3. Imaging of the Capture and Filter Magnets
 Occasionally, the Filter and Capture magnets will be imaged using the full spectral capability of Pancam, providing visible and infrared spectra of the material that has been attracted to the magnets. Any difference in color and/or brightness will contain information of the spread in properties of the particles suspended in the Martian atmosphere. Particularly, we will look for differences between the dust accumulated on the Filter magnet and dust imaged as it is on the ground around the rovers. From a comparison of these spectra we may be able to derive information about the magnetic phase(s) of the dust. This information will be an important supplement to similar observations by the APXS and the Mössbauer spectrometer.
5.4. Sweep Magnet and RAT Magnets
 The Sweep magnet will be imaged regularly by Pancam. Analysis of the data obtained may allow us to establish if dust particles are present in the inner part of the ring shaped magnet (the “sweep area”). As previously stated only dust particles with a low specific magnetic susceptibility (κ < 0.3 · 10−6 m3 kg−1) can settle in this area. The detailed shape of the circular pattern of dust on the magnet is also somewhat dependent on the specific magnetic susceptibility of the particles.
 Analysis of data from images recorded through the various filters of the Pancam will allow studies of the optical properties of the captured magnetic dust particles. The optical properties of the dust on the Sweep magnet will be compared with the corresponding properties of the soil on the Martian surface.
 From Pancam images of the RAT magnets, it will be possible to estimate the amount of material adhering to at least one of the pairs of magnets. From such images we shall be able to establish if the rock contains ferrimagnetic minerals in any substantial amount. From an analysis of the images we may be able to derive a value for the average magnetization of the rock material.
 Mössbauer spectra, at day and night temperatures on Mars, of the dust collected by the Capture magnet will with near certainty disclose if nanosized ferrimagnetic crystallites are present as part of the composite particles suspended in the Martian atmosphere. By comparing the Mössbauer spectra of the soil on the surface of Mars with the spectra of the dust on the Capture magnet we may be able to identify the ferrimagnetic mineral in the dust. If enough dust is present on the Filter magnet to record a Mössbauer spectrum, we shall be able to establish if the difference in strength of the two magnets has allowed a magnetic “filtering” of the airborne dust particles.
 Obtaining a useful alpha-particle X-ray spectrum demands substantially less dust on the magnets than a corresponding Mössbauer spectrum. A comparison of the elemental analysis of dust on the Filter magnet with a similar analysis of dust on the Capture magnet will also help disclose if a filtering of dust according to magnetic properties has taken place. Furthermore, by comparing the elemental composition of the dust on the surface of Mars with the composition of the dust on the magnets we shall be able to establish if the magnetic dust is enriched in the element titanium. If so, the magnetic mineral in the soil and dust is not a result of precipitation of iron oxides from liquid water; the magnetic mineral has in this case been inherited (more or less) intact from the surface rocks.
 Placed close to the Pancam calibration target, the Sweep magnet will regularly be imaged by Pancam. The images of (and the spectroscopy of) the dust on the Sweep magnet will allow a study also of Martian dust particles containing little or no iron-containing phases, if such particles exist.
 The particles imaged on the RAT magnets will allow a study of the magnetic properties of the abraded rocks, showing whether or not the rocks contain any ferrimagnetic minerals.
Appendix A:: Relative Intensities of Mössbauer Lines in Backscattering Geometry
 In transmission geometry the relative intensities of the six lines are
where θ is the angle between the propagation direction of the γ-rays and the magnetic hyperfine field. The Mössbauer transitions (lines) are numbered as shown in Figure A1.
 In backscattering Mössbauer spectroscopy as well the incident angle θ1 relative to the field as the scattering angle θ2 relative to the field have to be taken into account. The relative intensities in backscattering geometry become:
When a photon is absorbed as “line 1”, the emitted photon must also be “line 1” with the same angular distribution. The intensity of line 2 is calculated as follows: the probability that the nucleus is excited via transition number two is proportional to 2 sin2 θ1. After that the nucleus may return to either the state with magnetic quantum number mg = +1/2 or the state with mg = −1/2. The angular distribution is different for these two transitions.
 This paper is dedicated to Robert B. Hargraves (1928–2003) in grateful memory.
 The Danish Natural Science Research Council and DELTA Danish Electronics, Light and Acoustics are acknowledged for support of the project. Through several years the Danish Mars group has benefited from a close collaboration with professor Robert B. Hargraves at Princeton University. We are grateful for the extraordinary dedication to this project from the personnel at the workshop of the Ørsted Laboratory.