Microfossils are the microscopic remains of organisms. The organisms may be prokaryotic cells of the Bacteria or Archaea domains, unicellular eukaryotes (protists), whole multicellular eukaryotes, or parts of multicellular microscopic or macroscopic eukaryotes. Virus can be included, although so far only very few reports suggesting their presence in the fossil record exist e.g. [46]. The size of a microfossil ranges from the smallest living cell (250 ± 50 nm constitutes a reasonable lower size limit for life as we know it, [39]) to larger sizes that are not visible with the naked eye.

Microfossils can have a variety of morphology and chemical composition, depending on their original properties and the conditions in which they are preserved (Fig. 1).

The Rio Tinto river (Spain) is an acid mine drainage characterized by a very low pH (frequently near or below 1) and the intensive precipitation of iron oxides and sulphates that can entomb cell remains [19] and [20]. This environment has been proposed as a geochemical and mineralogical analogue to Terra Meridiani on Mars, even though the physical processes of rock formation differ significantly between the two sites. Another interesting feature of the Rio Tinto is the presence of 1 or 2 Ma old terraces permitting a direct comparison between past and recent geobiological processes acting in the river. Coccoidal and filamentous bacteria, algae and fungi are preserved locally as casts and molds in laminated ironstones, even when they experienced diagenetic transformations, such as the transformation of goethite into hematite [19] and [20]. More generally, bacteria can trigger mineral formation in any environment that is saturated with respect to a specific mineral phase, at least locally and even temporarily around the bacterial cells. Mineral precipitation by microorganisms can be achieved in a wide range of physico-chemical conditions, and by various passive or active processes that are reviewed in details by Konhauser [43]. In oxic sediments forming in neutral-pH freshwater lakes, natural bacterial exopolymers and cell walls can become covered with very thin nanoscale elongated crystals of poorly ordered iron oxides [22] and [53] leading eventually to the preservation of iron-organic carbon assemblages with approximately the same morphology as the original microbes.

Environments such as glacial lakes, melt waters or rocks in the Dry valleys of Antarctica are also inhabited by diverse microbes including cyanobacteria, algae, or fungi. Cyanobacteria and fungi live in pores of the McMurdo Dry valleys sandstones. After death, their cells decay and their cell walls and sheaths are covered by allochtonous clay minerals and sulfate-rich salts filling the sandstone pores while the empty moulds of cells are filled by minerals. The organic cellular structures serve as templates for diffusing mineral elements and give rise to a characteristic distribution pattern of precipitates inside the fossilized cells. In the absence of organic matter preservation, the former presence of these cells and their mineralization impart a special texture to the rock with formation of cell-shaped structures [70].

In Iceland hot springs, intra- and extra-cellular silicification is due to the polymerization of silica from hydrothermal fluids. Only microbial sheaths are usually preserved. Although silicification can alter the morphology of the microbes beyond identification, it has been proposed that microorganisms impart an influence on the fabric of the siliceous sinters that form around hot spring vents. Indeed, this fabric consists in alternating laminae of flat-lying and upright filamentous microorganisms resulting from the behaviour and motility of living cells [44] and [45].

In Greenland, Neoproterozoic carbonate tidal flats preserve mats bearing the cyanobacteria Entophysallis, which have a mammillate (pustulate) surface similar to modern mat surfaces built by the same organisms in Abu Dhabi sabkha. At the microscopic level, microfossils are preserved by the permineralization of their cell walls by silica, which leaves some remains of organic matter along the cell walls. Some colonies even show patterns of pigmentation in the most external layer of cells. These are interpreted (by comparison with modern analogues) as the fossil remains of pigmented cells most exposed to the light and UV [30], [31] and [41]. Remarkably well-preserved coccoids and sheaths of microbes are abundantly present in stromatolitic and non-stromatolitic cherts (i.e. silica rich sedimentary rocks) of the 1.9 Ga Gunflint iron formation (Canada). It has been shown that these cherts precipitated directly on the seafloor, and did not result from secondary precipitation of silica nodules within carbonates. The Gunflint stromatolites resemble hot spring sinters like those existing presently in Yellowstone, and differ from past and modern tidal flats stromatolites in which the microfossils are usually mat builders. This can be in particular inferred from the fact that the fossils in Gunflint stromatolites are jumbled together without preferential orientation (no dense intertwined bundles of filaments). The fossils are coated by iron oxides and permineralized by silica, with partial preservation of organic walls [40] (Fig. 1).

Proterozoic shales from peritidal to basinal marine environments show exceptional preservation of organic-walled microfossils e.g. [14], [32], [33], [34] and [42]. The fossils are preserved in 2D (flattened) rather than in 3D as in chert, but exquisite details of the ornamentation and the ultrastructure of the fossil walls can be observed which may provide key information on their taxonomy [34] and [65].

Conducting experiments in the laboratory offers a complementary approach to investigate taphonomic processes (i.e. processes occurring after the death of the organisms) and decipher what can be preserved in microfossils, regarding their chemistry as well as their ultrastructure. Although these studies may not all reflect the full complexity of natural environmental conditions, they reveal biotic patterns and preservable properties that have been overlooked so far. Fossilization, i.e. the stabilization or replacement of organic structures by mineral deposits, is indeed a very quick process that can be achieved in few hours or days. For example, exposing bacterial cells to a solution rich in Ca, Fe or Si can lead to their encrustation in few hours [4]. If precipitation occurs in close connection with the microbial structures and if the minerals are small enough, very fine cellular details can be preserved. For example, the 40 nm thick cell wall of Gram-negative bacteria can be fossilized by calcium phosphates e.g. [4] or iron minerals e.g. [53] (Fig. 3). Embryos of eukaryotes could be preserved as well in the laboratory at different development stages by phosphatization, simulating what likely occurred during the fossilization of the famous ∼630 Ma old Doushantuo embryos e.g. [73]. While some organic carbon might be degraded during these processes, it has been shown that most of it can be preserved within the resulting mineralised microfossils e.g. [4]. After the precipitation of minerals on the organic structures, further degradation of the morphology and degradation/maturation of the organic remains can occur. This stage is influenced by mechanical stresses, circulation of fluids, and metamorphism. Observations of natural samples have shown that microfossils can sometimes be preserved even after high grade metamorphism e.g. [7] (Fig. 2). Such processes take place over much longer timescales, so it is more difficult to simulate them in the laboratory. However, some recent studies provide an interesting way to address this issue, in particular by noting that time and temperature are inherently linked and that aging at low temperature over long timescales can be simulated by shorter aging at a higher temperature (see [63]).

Many experimental taphonomy simulations can be found in the literature. Cyanobacteria have received a particular attention. Intra- and extra-cellular silicification was observed. It has been shown that sheaths are preferentially preserved to walls and cytoplasms e.g. [1]. Some morphological changes could be observed including cell collapse, shrinkage and disappearance, trichome disarticulation, varying terminal cell morphology, and shrivelling, constriction/swelling, discoloration, rupture of sheath e.g. [66]. Although virus may have been major players in life evolution e.g. [21], only little is known about their geological record. Interestingly, some taphonomy experiments have been performed on these “organisms”. Biomineralisation experiments on viruses show that dissolved iron ions are able to penetrate the virus capsids and bind to internal sites [17]. As a result, virus capsids can serve as nuclei for the growth of iron oxide particles. The resulting morphology differs from abiotic iron oxides and organic molecules composing originally the capsids can be efficiently preserved by the iron minerals that armour them offering the possibility to look for them e.g. [37].

One key problem encountered in the study of microfossils is that relatively complex morphologies can be produced by purely abiotic processes (e.g. [28], Livage, this issue) and can thus make their identification difficult. Laboratory experiments can be of interest in order to look for possible specific features (if any) that might be used to discriminate abiotic from biological objects. In addition to the morphology, the composition of the organic matter produced by reactions such Fischer-Tropsch has been carefully scrutinized and compared with mature kerogens e.g. [52]. For example, a combination of infra-red spectroscopy (providing an aliphaticity index of the organic matter) and microscale measurements of the carbon isotopic compositions was used by Sangely et al. [58] to distinguish between biology and Fischer-Tropsch-type reactions as genetic processes for the bitumen found in the Cretaceous uranium deposits of Athabasca. Known abiotic products that can mimic life morphologies or chemistries include vesicles made in the laboratory from meteoritic kerogen or in other prebiotic chemistry experiments e.g. [18], fluid inclusions, carbonaceous filamentous shapes resulting from migrating organic matter (with carbon isotopic fractionation resembling life patterns) around minerals casts in hydrothermal environments [11] and [12], aggregates of silica spheres and rods in silica-rich waters of hydrothermal springs, migration of carbonaceous materials along microfractures [68] or around silica spheres formed in silica-saturated water (these are less than 5 μm in diameter and formed in hydrothermal conditions, e.g. [36]). Mineralised pseudo-fossils have been produced using a mixture of barium carbonate and silica in laboratory experiments [28]. The resulting auto-assembling segmented filaments were rigid tri-dimensional objects, not hollow originally, but made hollowed and filled by organic matter during further experiments. In principle, these artificial objects could form in nature and be confused with fossils or biominerals. However detailed observations of the septae morphology, the population morphology and distribution in the rock may alert the micropaleontologist. For example, such objects would never be found in fine-grained siliciclastic sediments where filaments are preserved as carbonaceous objects flattened in two dimensions.

Further studies are much needed to investigate the range of abiotic processes mimicking life properties, to define unambiguous traces of life for paleobiology but also for the search of life beyond Earth.

The first step in the recognition of past traces of life is the determination of the paleoenvironmental conditions of preservation. The samples should come from rocks of known provenance, of established age and demonstrating geographic extent. Moreover, the possible traces should occur in a geological context plausible for life: these criteria apply mostly for sedimentary environments. It has been traditionally inferred that magmatic and metamorphic rocks were not appropriate for the search of fossils as the former form under conditions incompatible with life and the latter have experienced temperature and pressure conditions that might have erased any trace of life. However, recent studies have suggested that even basalts can be altered by life e.g. [67] and might retain traces of this bio-activity although there is still some discussion e.g. [6], [23], [24] and [25]. Moreover, it has been shown that fossils of plant spores could be very well preserved in rocks that have experienced high grade metamorphism, ∼14 kbar and 360 °C, i.e. a burial of ∼35 km [7].

Rocks even present at depth can be colonized secondarily by recent microorganisms through fractures for example. It is thus important to show that microfossils are clearly endogenous to the rock, i.e. enclosed within the minerals composing the rock or between the grains before cementation of the rock. They should be observed in thin sections cut in the rock perpendicular to bedding. If the microfossils are organic-walled and preserved in two dimensions in fine-grained siliciclastics, following sedimentation with other fine particles, additional thin sections parallel to bedding will show them more clearly (in cross-cutting sections, they appear as a fine brown/black streak). The repeated use of in situ analytical techniques to study microfossils within rock samples allow ascertaining that those objects do not result from contamination/artefacts produced during sample collection or preparation. Mineralised microstructures should also be shown in situ in thin sections, and their primary occurrence should be demonstrated by petrologic analyses proving early cementation as other grains making up the rock or active building of the rock structure (as in some stromatolites; e.g. [57]). Finally, the possibility to observe the interface between microfossils and their hosting minerals down to the few nanometer-scale can help unravelling possible chemical interactions between them and prove further the endogenicity of the objects [8]. Endogenicity is however not sufficient to ascertain that the object that is observed is a microfossil. Indeed, some microorganisms, called endolithic microorganisms can bore into rocks and live inside. They have been in particular abundantly documented in carbonate deposits and stromatolites e.g. [16]. Hence, one can potentially find a microbe younger than its hosting rock. Their actual significance in the geological record still remains to be evaluated.

Most of the time, microfossils cannot be dated directly and their age is thus inferred from the age of the host rock. It is thus important to prove that microfossils have the same age as the rock itself, i.e. they occurred in the sediments prior to their diagenesis and lithification (syngeneity). Some criteria can be proposed: there should be no hydrothermal veins or fractures cutting through the studied samples as they are preferential path for the input of exogenous abiotic or biotic organic compounds in the rock. Endolithic microorganisms are another issue as explained above. One way to test this is to perform Raman spectroscopy analyses on isolated microfossils to assess their thermal maturity [9] and check that it is consistent with the thermal maturity of other carbonaceous particles found within the rock and the metamorphic grade recorded in the minerals composing the hosting rock [48] and [50]. The hypothesis of contamination by younger material may then be discarded.

As mentioned above, abiotic processes can form relatively complex morphologies similar to those produced by life. In addition to morphology, some criteria have thus been proposed in the literature e.g. [15]. The microstructures should be relatively abundant (a population), have a biological morphology (comparable to that exhibited by modern microorganisms or well-documented fossils from younger successions) and have biological size ranges (larger than the smallest extant free-living organisms (>0.01 μm³). The past debate on the existence of purported nanobacteria was of particular interest regarding the size criterion. It has been clearly shown that those objects, too small to be bacteria, were indeed not microorganisms despite their biological-like shapes e.g. [3] and [54]. Mineralized microstructures should exhibit morphologies that are relatively complex, even though we note that this notion of complexity has not been yet clearly defined in the literature. For example, it is difficult to distinguish silica spheres representing mould of bacterial cells or abiotic chemical precipitates, unless they are hollow and contain traces of endogenous kerogen. Moreover, some abiotic processes can mimic so-called complex biological morphologies e.g. [27]. Thus, microstructures with simple morphologies, i.e. without ornamentation (decoration of the wall such as processes, striation, polygonal networks, pustules, vesicles, chagrinate pattern…) should ideally show a structurally distinctive carbonaceous cell wall. Organic-walled microstructures may show cell lumina (originally cytoplasm-filled cell cavities) although this is difficult to observe in flattened specimens in thin sections through compacted mud or silt but more easily seen in 3-D preserved fossils. They should show evidence of taphonomic degradation, such as thin concentric folds or lanceolate folds, or folding over, and, if preserved in fine-grained sediments, flattening in two-dimensions, demonstrating the flexibility of the original wall. Several techniques provide key information on the composition of possibly biogenic, carbonaceous particles: for example, Raman microspectroscopy e.g. [35], [38], [61], [71] and [74], Infra-Red microspectroscopy e.g. [49], Scanning Transmission X-Ray microscopy e.g. [5] and [48] or carbon isotope spectrometry e.g. [26]. However, none of these techniques gives a definitive answer to the biogenicity of carbonaceous matter. For example, it has been clearly mentioned by Pasteris and Wopencka [55] that Raman spectroscopy of carbonaceous material does not provide definitive evidence of biogenicity by itself. Analyses of the carbon isotope fractionation of the bulk kerogen should reveal negative values that are consistent with a biogenic interpretation. However, such values can also be produced by abiotic processes [51]. The biogenicity of microstructures is further strengthened by the observation of their distribution in the rock and fossilized behaviour: orientation and distribution caused by mobility and interaction with the environment, the presence of pigmentation at colony surfaces, and cellular division e.g. [13], [31], [41] and [47].

Noting the difficulty of discriminating unambiguously biogenic from abiotic objects, Brasier et al. [12] propose a different approach consisting in demonstrating that an object cannot be abiotic instead of trying to find biogenic signatures: this is the falsification approach. As underlined above, abiotic processes can produce microstructures mimicking biological morphologies and sometimes chemistries (see section 3.4). They include migration of abiotic (or biotic) carbonaceous material along microfractures, around mineral casts, or around silica spheres formed in silica-saturated water. The formation of filaments in which abiotic carbonaceous matter is intimately associated with minerals has been shown in the laboratory [27]. These structures may be recorded in the rock record, and they may be preserved in three dimensions in carbonaceous chert (as the SOS-self-organizing structures – described by Brasier et al. [12]). In this regard, the taphonomic processes associated with preservation in fine-grained siliciclastics such as flattening in two dimensions, folding and other soft wall deformation of carbonaceous microstructures help the paleobiologist to determine their biogenicity.