Both these claims for fossil life have been disputed. Buseck et al. (7) questioned the identification of magnetite by Friedmann et al. (6) and noted that intact magnetite chains from magnetotactic bacteria were unlikely to survive in rocks. Buseck et al. (7) found that magnetites from three different strains of bacteria had different shapes and criticized the techniques used by Thomas-Keprta et al. (4, 5) to select magnetite crystals and to study their shapes. They also argued that the crystal size distributions of bacterial and meteoritic magnetites appeared to be different. Although Buseck et al. (7) recognized that the similarities between the Martian and bacterial magnetites are intriguing, they, like Treiman,§ inferred that strong evidence for biogenic Martian magnetites was lacking. Buseck et al. (7) called for more detailed studies of the shapes of the Martian magnetite crystals to resolve this issue.

Arguments against a biogenic origin for magnetites in ALH 84001 were previously advanced by Bradley et al. (8, 9). Their TEM studies showed that whisker-like magnetites had grown epitaxially on a carbonate fracture surface so that two crystal lattices were closely aligned at their interface. They concluded that these magnetites had condensed from a vapor above 120°C. However, Thomas-Keprta et al. (4) pointed out that the whisker-like magnetites, which comprised 7% of their sample, were found only in the interiors of carbonate grains. They inferred that the elongated prismatic magnetites, which they suggested were biogenic, are largely located in the optically opaque rims where epitaxial relationships are not observed. Four other origins have been suggested for some or all of the magnetite in ALH 84001: deposition of abiogenic grains that formed elsewhere (4, 5), thermal oxidation of sulfides (9), precipitation from aqueous solutions,¶ and in situ formation because of shock-induced, thermal decomposition of carbonate (‖, 10–12).

We have performed TEM studies of the ALH 84001 magnetites and their associated carbonate, silicate, oxide, and sulfide minerals in situ. Here, we focus solely on results that bear on the origin of the magnetite.

Minerals and glasses in typical locations were characterized by TEM imaging, x-ray microanalysis, and electron diffraction, taking note of their morphologies, internal microstructures, compositions, and spatial interrelationships. Our specimen was a doubly polished thin section (ALH 84001,353) supplied by the National Aeronautics and Space Administration (NASA) Johnson Space Center. We approached the controversial origin of the carbonates in ALH 84001 from a perspective gained from TEM studies of carbonates of many types (e.g., refs. 13–16). Carbonate-bearing areas were identified by optical microscopy and examined by using Zeiss DSM-962 and Hitachi (Tokyo) S-800 field emission scanning electron microscopes. The section was then demounted, and pieces of it were glued to coarse 3-mm-diameter copper grids with epoxy resin. To make TEM specimens, we utilized an Ion Tech (Teddington, U.K.) beam-milling system with FAB (fast atom beam) sources, operating at 5 kV with low intensities and at low angles of beam incidence to avert risk of thermal alteration. Selected areas were examined with Philips (Eindhoven, The Netherlands) CM20, JEOL 200-CX, and JEOL 2010 transmission electron microscopes, all operated at 200 kV and equipped with detectors for energy dispersive spectrometry. These microscopes are able to form small electron probes for x-ray microanalysis (1 nm diameter is the minimum possible for 2010; typically, we used ≈5–10 nm), thus enabling reliable compositional analyses to be obtained. The 2010 microscope was equipped with a Gatan (Pleasanton, CA) slow-scan charge-coupled device (CCD) camera, which facilitated lattice imaging. Moderate electron beam intensities were used in TEM, and, whenever possible, short exposures of carbonate and sulfide phases to electrons were used to minimize radiation damage and avoid thermal volatilization. A minimum exposure method (17) was used when lattice imaging.

The orthopyroxene rock-mass is strongly deformed and unrecovered. More heterogeneously, it exhibits signs of alteration because of shock, which take several forms: breakdown to fibrous morphologies, formation of ≈100-nm-wide lamellae of glass of orthopyroxene composition, (glass of this composition was reported by Bell et al.**), and transformation to clinopyroxene (18). The orthopyroxene glass indicates that shock pressures of 30–40 GPa were attained locally (19). Carbonate crystals show only local mild deformation, and they clearly crystallized around the orthopyroxene grains after the latter were strongly deformed and fractured. In fractured orthopyroxene, we found carbonate grains containing lenses and stringers of orthopyroxene glass, suggesting that the carbonate was fluid when the orthopyroxene melt was emplaced.

In magnesian regions of Mg-Fe-Ca carbonate grains (those with >80 mol % MgCO₃), we discovered periclase, MgO, occurring in ≈1-μm-sized patches of fine grains ≈3 nm in size. These patches usually exhibit evidence of a weak texture in electron diffraction (i.e., a tendency to be preferentially oriented) consistent with {111}_per//(0001)_carb, as found with the decomposition of calcite to lime, CaO (ref. 20) (Fig. 1 a and b). Less commonly, periclase occurs as individual crystals (typically 30–50 nm diameter) either enclosed within, or growing into, tiny voids in the carbonate (Fig. 1c). The MgO ring pattern (e.g., Fig. 1b) somewhat resembles that for Fe₃O₄ (e.g., Fig. 1d) because the lattice parameter of Fe₃O₄ is double that of MgO. Thus, positive identification of periclase requires diffraction rings at the 200_MgO and 220_MgO radii when iron is absent from the x-ray spectrum. In magnesian regions of massive carbonate, the patches of periclase are most easily recognized adjacent to fractures and open porosity. The nature and locations of the periclase are clear signs of in situ crystal growth during decomposition of the solid carbonate.

Figure 1 TEM images of periclase and magnetite in carbonate and their electron diffraction patterns: (a) nano-crystalline aggregate of periclase (P) in foamy region of magnesian carbonate; (b) diffraction pattern from a: 200 and 220 rings for MgO are prominent (more ...)

Magnetite crystals are found at numerous locations and in various forms in ferroan carbonate (carbonate with <80 mol % MgCO₃). The magnetite has four distinct occurrences: in optically opaque rims where it is agglomerated and intermixed with iron sulfide (which in our section appears to be mostly amorphous); as single crystals or small groups of crystals in cracks; as individual crystals in small voids in carbonate; or as crystals within solid carbonate. In the last two categories, the crystals are usually euhedral in form. Examples of these two categories are visible in Fig. 1e. As in the magnesian carbonate, the oxide inclusions are often associated with voids, and they are abundant adjacent to fractures and open porosity. Where magnetites and voids are especially abundant, both magnetite and fine-grained periclase are found.

We confirmed that magnetites growing in small voids and in microfractures frequently have epitaxial relationships with the carbonate (¶, 8). However, for the nanocrystals that are fully embedded in the ferroan carbonate, the crystallographic relationships are more extensive and very significant. These crystals constitute the most pertinent group to the mechanism of the formation of magnetite. We studied some of this group carefully by using lattice imaging methods to investigate the relationships between these magnetites and their host. The magnetites were too small to give good diffraction patterns, but we could simulate these from the lattice images by Fourier transforms. Our first discovery was that these magnetite crystals give good lattice images (e.g., Fig. 2), whichever zone axis of the carbonate is chosen for lattice imaging. This finding implies topotaxy (three-dimensional lattice continuity) rather than epitaxy. Therefore, this possibility was examined more systematically by imaging each crystal for two different carbonate zone axes. The first of these pairs of images invariably showed good continuity of the lattice planes across the magnetite-carbonate interfaces, although with some distortion. The second image was usually inferior in quality because of the damaging effect of the electron irradiation on the carbonate, rendering it less crystalline and more amorphous with time, even though we used a minimum dose technique (17). Nonetheless, we were able to confirm that the second image of each pair also exhibited lattice continuity at the magnetite-carbonate contacts, i.e., the interfaces are coherent or at least semicoherent. Such coherency is characteristic of the early stages of exsolution (e.g., ref. 21).

Figure 2 Lattice image of part of a small embedded magnetite that was topotactic with the surrounding carbonate lattice (the white lines indicate the periphery of the magnetite). The carbonate close to the magnetite crystal is electron damaged but still shows (more ...)

Fig. 3 a and b shows a pair of images from a magnetite crystal, with their corresponding diffraction patterns as Insets (actual patterns from the carbonate, not simulations derived from the lattice images). Because the lattice continuity is general (i.e., it does not depend on the direction of imaging), we became convinced that the fully embedded magnetites are topotactic with their carbonate host. To determine relationships between the lattices of the two phases, Fourier transforms of some of the high resolution images of magnetites in carbonate were used to simulate diffraction patterns. From several sets of data, we established the relationships between the two phases:{111}_mag//(0001)_carb and {110}_mag//{110}_carb. Although these relationships do not apply in every case, they are identical to those reported for the decomposition of calcite into lime (20) and are also consistent with our results for periclase. We have yet to identify any lime although we suspect it is present.

Figure 3 Lattice images demonstrating the topotaxy of magnetite crystals that are fully embedded in ferroan carbonate. (a) Image of a magnetite crystal, larger than that of Fig. 2, viewed along a direction very close to the carbonate [100] (more ...)

Periclase, unlike magnetite, which forms in a variety of igneous, metamorphic, and sedimentary rock types, is relatively uncommon in meteorites and earth rocks, and in nearly all cases it forms by thermal decomposition of dolomite or magnesian carbonate (e.g., refs. 22 and 23). In ALH 84001, many of the periclase crystals are preferentially oriented with their close-packed layers of oxygen atoms oriented approximately parallel to those in the surrounding magnesian carbonate (Fig. 1b), as found in the decomposition of calcite to lime (20). Thus, the orientation and location of periclase crystals in ALH 84001 are strong evidence for in situ growth as a result of thermal decomposition of solid carbonate and loss of CO₂. In view of the localized shock melting of orthopyroxene and chronological evidence for major impact heating event around 4.0 Gyr (24), it is more likely that impact was the heat source for carbonate decomposition rather than deep burial and thermal metamorphism. Shock-induced formation of periclase from dolomite to yield a nanometer-sized crystalline product has previously been reported (25). Lime is seldom apparent as a decomposition product of dolomites, although undoubtedly present in low volume fraction.

Our lattice imaging results show that many of the small magnetites outside the magnetite-rich rim (and probably all those that are entirely embedded in carbonate) have crystal lattices that are coherent with those of the host (Figs. 2 and 3). The crystallographic relationships between the magnetite and carbonate are characteristic of second phase crystals that have exsolved (i.e., precipitated in the solid state by means of diffusion). There are good examples in both metallic and nonmetallic systems (21, 26).

Magnetite and periclase are often associated with voids and microfractures in carbonate (Fig. 1a). In some regions, they are intermixed, and both oxides appear to have nucleated both homogeneously and heterogeneously. We suggest that their occurrence is entirely compatible with solid-state diffusion and exsolution as a result of thermal decomposition of carbonate. Because many oxides are not associated with voids, we infer that CO₂ was lost before oxide precipitation. Solid-state exsolution in crystals that are much larger than the effective diffusion length favors formation of numerous tiny single crystals and a tendency (although not an imperative) for them to grow in crystallographic relationships to the host carbonate lattice.

In our view, the species, sizes, and locations of oxide crystals in carbonate are simply functions of the Fe/Mg ratio, the abundance and nature of available nucleation sites, and variations in diffusivity because of the proximity of internal surfaces, vacancies, and other defects. The lower abundance and smaller grain size of periclase in magnesian carbonate compared with magnetite in the more ferroan carbonates is consistent with differences in stability.‡‡ Precipitation processes in supersaturated cooling solids commonly generate crystallites with characteristic faceted morphologies. Heterogeneous nucleation on internal surfaces and other defects is also common. These properties describe the observed carbonate-magnetite intergrowths very well.

At the optically opaque rims of carbonate crystals, where the most Fe-rich carbonate is located (27), there are agglomerates of magnetite particles, and the carbonate exhibits a subgrain structure. We infer that strain fields around the abundant magnetites interacted to produce a microstructure of small (≈100-nm-diameter) subgrains. (This dimension was determined by dark-field imaging.) Magnetites in the rims tend to be more diverse in size and more irregular in shape than other magnetites. Topotaxy with the decomposing host lattice is not favored because of its defective nature and probably also because the magnetites grew in a less controlled manner, as their diversity indicates.

Thomas-Keprta et al. (4, 5) argued that the elongated prismatic magnetites, which they inferred to be biogenic, are located mainly in the magnetite-rich rim where they found the carbonate to consist of randomly oriented crystals as small as 10 nm. We find to the contrary that carbonate in the magnetite-rich rim neither is fine grained nor lacks long-range order, as claimed, but rather that it appears to have been highly strained by the presence of the large volume-fraction of magnetite, imparting a microstructure of slightly misoriented subgrains ≈100 nm in size. Moreover, rim magnetite consists mostly of irregularly shaped crystals, consistent with growth in a highly strained matrix. We are convinced that euhedral magnetites largely occur as individual crystals in the carbonate, rather than in the densely populated magnetite-rich rim. Whereas the population outside the rim includes the rare whisker-like types, which are undoubtedly abiogenic (8, 9), it is mostly more-equiaxed faceted crystals with morphologies like the elongated prismatic crystals reported by Thomas-Keprta et al. (4).

The possibility that the whisker-like magnetites on microfractures (e.g., Fig. 1e) grew directly from a fluid or hot vapor, as Bradley et al. (9) suggested, is not excluded, but these magnetites may have grown by surface diffusion. Where chains of magnetite occur in ALH 84001, it is possible that they grew on ledges and kink sites on microfractures. Nucleation on such sites will be preferred above nucleation on defects like dislocations, or clustering on lattice sites. (Deposition on surfaces readily yields strings of precipitates that may appear curved at the microscopic level, because ledges and terraces commonly side-step on the atomic scale.)

Thomas-Keprta et al. (4, 5) offered several arguments against formation of magnetite in ALH 84001 by thermal decomposition of carbonate. They concluded that magnetites that formed this way should contain Mg and Mn, contrary to their analyses. However, experiments by Golden et al. (12) show that thermal decomposition of Mg-Fe-Ca carbonates can yield magnetite with undetectable Mg. Thomas-Keprta et al. (4) argued further that carbonate adjacent to magnetites that formed in situ by thermal decomposition should show chemical evidence for Fe mobilization, which they did not observe. However, our finding of topotactic magnetites and faceted voids (negative crystals) provides clear evidence for solid-state diffusion of ions and vacancies on scales of hundreds of nanometers. According to Thomas-Keprta et al. (4), magnetites that formed from the ALH 84001 carbonate should not contain the small concentrations of Al and Cr that they detected in irregular and whisker-shaped magnetite, because these elements have not been reported in the Martian carbonate. However, the ALH 84001 carbonates are notoriously impure: electron microprobe analyses of these carbonates show 0.01–1 wt. % Cr, Na, K, and Si (see ref. 28) and traces of Al. Even if these elements are located in minor phases, there are clearly sources of Al and Cr available to growing magnetite crystals. The lack of Al and Cr in elongated prismatic magnetites (4) is entirely consistent with our proposed formation mechanism. We should expect euhedral crystals that grow with coherent boundaries to be purer than irregular crystals growing with incoherent boundaries because the latter, unlike the former, readily accommodate impurities. Gibson et al. (29) argued against in situ formation of magnetite because Raman spectroscopic studies failed to detect CaO and MgO. However, our identification of periclase shows that earlier studies were not sensitive enough to detect these oxides. We suspect that CaO is also present sometimes, as an admixture; as yet, we have no certain proof. Given that the occurrences of magnetite in ALH 84001 carbonate can be explained perfectly well by established physico-chemical principles, there is no good reason why biogenic origins should be invoked and applied to a fraction of the isolated magnetites.

Our results provide strong evidence that most and probably all of the magnetite grains in ALH 84001 are abiogenic in origin and that they formed because of shock-heating of the meteorite. Their occurrence is matched by the formation of periclase from the more magnesian fraction of the carbonates. The group of magnetites that are fully embedded in carbonate (which includes the size range of the putative biogenic magnetites) are topotactic with their host. We find {111}_mag//(0001)_carb and {110}_mag//{110}_carb [the same as for lime, which is a decomposition product of calcium carbonate (20)]. This topotaxy is characteristic of exsolution in the solid state. We have failed to find any sites where a special group of magnetites, like those identified by Thomas-Keprta et al. (4, 5) as probably biogenic, might be isolated or concentrated. Because a large proportion of the magnetite crystals are clearly attributable to decarbonation and decomposition of Fe-rich carbonate, it is highly improbable that a significant fraction was produced by a totally different, biogenic process and then intimately mixed with the former. The observations of chains of magnetite crystals (6) that have been offered as evidence of bacterial sources of magnetite can be explained by the well-documented mechanism of nucleation at terraces on internal carbonate crystal surfaces (e.g., microfractures).

We thank K. Z. Baba-Kishi for access to the JEOL 2010 microscope, G. K. H. Pang for invaluable assistance, and J. F. Kerridge for a useful discussion. We thank the National Aeronautics and Space Administration (NASA) Johnson Space Center Curatorial Facility and the Meteorite Working Group for the sample. We also thank three anonymous reviewers for constructive criticisms. This work was partially supported by NASA Grant NAG5-11591 to K. Keil and National Science Foundation Grant OPP-9714012 to E.R.D.S.