Main

Earth’s earliest fossil assemblages are important for understanding the origins of life on Earth and, by analogy, how and where to search for signs of primitive life in the rock record of other planets2. The oldest widely accepted evidence of life on Earth is in marine metasedimentary rocks of the Pilbara Craton, Australia, in the form of a microbial stromatolite reef3,4 and fossil biofilms5 of the 3,450-million-year-old (Myr) Strelley Pool Formation. Putative microfossils6 (Strelley Pool Formation) and stromatolites (3,490-Myr-old Dresser Formation)7 also occur in the Pilbara Craton, but their biogenicity is equivocal8. In Greenland, geochemical features compatible with microbial activity exist9,10, but their interpretation has been questioned11,12,13. The presence of 3,700-Myr-old stromatolites in Greenland’s Isua supracrustal belt (ISB), if true, would represent an entirely new and compelling type of biosignature in Earth’s oldest rocks and establish the start of the fossil record 200 Myr earlier than previously thought1.

The putative stromatolites, discovered approximately 150 km northeast of Nuuk (Extended Data Fig. 1), were described1 as elongate cones and domes 1–4 cm high, with apices pointing upward relative to overturned sedimentary bedding. Combining those attributes with: (1) internal stromatolitic lamination that is continuous across the crests of the structures; (2) diverse morphologies similar to younger stromatolites; (3) associated shallow-water sedimentary features, including sedimentary onlap; (4) differences in chemical composition inside the structures compared to surrounding sedimentary rock; (5) the presence of low temperature dolomite; and (6) seawater-like rare earth element and yttrium (REE + Y) composition of the dolomite, it was proposed in the previous study1 that the structures are stromatolites produced by microbial communities in a shallow marine, carbonate-platform environment similar to the stromatolites in the Strelley Pool Formation3,4. Preservation of these features was attributed to an approximately 30-m × 70-m low-deformation lacuna in the hinge of an anticline1.

We located the discovery outcrops using data from the previous study1. Site A consists of brownish-grey layered dolomitic rock, with light-grey triangular features mostly oriented apex-up relative to the overturned layering. However, some are apex-down (Fig. 1a and Extended Data Fig. 2c), which is inconsistent with upward growth of the structures from a palaeo-seafloor. Dolomitic breccia nearby (site C) was previously interpreted as a tempestite (storm deposit), which in turn was taken as evidence of a shallow-water, ice-free sedimentary environment1. However, a wider view of the outcrop shows ductile and brittle deformation of the clasts, including extreme elongation when viewed from the side (Extended Data Fig. 3), indicating that the breccia has a tectonic origin and has no bearing on water depth, ice or other sedimentary conditions.

Fig. 1: Putative stromatolites of Greenland.
figure 1

a, Seven structures in outcrop (arrows, white dashed line). b–d, Sample from site A. b, Face 2 shows even, parallel layering. c, d, Face 1 shows irregularly layered fabric with planar discontinuities (arrows in c) and convex-up features (two yellow arrows in d). Yellow dashed boxes indicate panels expanded in g, h as indicated. e, Oblique view of the sample. f, Sample from the previous study1, equivalent to face 1. Lines indicate the path of X-ray fluorescence scans in the previous study1. The ‘d’ denotes their scan through the stromatolite. g, PIXL element maps of stromatolite and matrix (yellow box in c). Dashed line marks the edge of the structure, below which the composition shows a gradient from a Ca–Mn–Fe-rich rim to a Si-rich interior. h. PIXL maps of the light-grey layer (yellow box in b) show elemental composition, including  Ti and K depletion, identical to the ‘stromatolite’.

Full size image

A sample was acquired approximately 0.5 m from the original ‘stromatolite’ sample site of site A14, including one of the triangular structures (Extended Data Fig. 2). Cut parallel to the weathered face (face 1, Fig. 1c–e), the sample shows irregularly layered light- and medium-grey quartz–dolomite layers with dark micaceous layers and foliation. An array of millimetre- to centimetre-scale convex-up features, of which the triangle structure is the largest, all have subparallel axial planes (Fig. 1c, d). Notably, the base of the triangular structure is also convex-up and conformable with small convex-up features in the underlying quartzose layers. The fabric is extensively disrupted by planar discontinuities, or spaced cleavage, subparallel to the axial planes of the convex-up features (Fig. 1c). By contrast, when viewed orthogonally (face 2, Fig. 1b), the rock shows flat, even layering without any stromatolites, bumps or irregularities (Fig. 1b, e and Extended Data Fig. 4).

Such orientation-dependent, contrasting fabrics are inconsistent with sedimentary processes. Rather, they are typical deformation fabrics found in a multi-layered rock that has been shortened in one direction (parallel to layering), producing minor folds, cleavage and other compressional features similar to those observed on face 1; and substantially lengthened in an orthogonal direction, producing extensional rod-like features such as those observed on face 2. The type of deformation indicated is consistent with the structural setting of the rock, within the hinge of an anticline1.

Accordingly, cuts parallel to face 1 show that the ‘stromatolites’ are not cones or elongate cones, but ridges extending at least 10 cm (our sampling depth) into the rock, aligned with the lengthening direction. The ridges probably extend further, given the extreme elongation of the rock fabric observed in the outcrop. Photographs published in the previous study1 suggest that the structure that they sampled is also ridge-shaped. Although ridge morphology alone does not preclude biological origins, it is easier to produce ridges abiotically than cones3,4. More importantly, a deformational origin is more plausible given the alignment of the ridges with the lengthening direction indicated by the rock fabrics.

The previous study1 included outcrop photographs of thin, recessive laminae that tangentially truncate against a structure, which the authors use as evidence of seafloor growth of a stromatolite. However, similar truncation occurs in our sample where micaceous foliation terminates against the triangular structure—an observation that is consistent with the presence of a rigid object (the quartzose ridge) in a ductilely deforming rock, leading to deflection and pressure solution of the mica and carbonate foliation on the shortening side. Tangentially truncated laminae observed in the thin section1 occurred at site B. However, the putative stromatolitic structures from site B illustrated in the previous study1 are very different from those at site A: the published image only shows an undulose lithologic contact (Fig. 2b of the previous study1).

Fig. 2: Relative abundance of REE + Y for different components of the Isua rock samples.
figure 2

Blue and red lines, our sample, which was separated into silicate and carbonate fractions. Grey lines, sample analysed in the previous study1, combining carbonate and silicate. All have light rare earth elements (LREE) element depletion (Pr/Yb of \(\ll 1\)), Y/Ho ratio of ≥30, positive La and Eu anomalies, broadly consistent with Archean seawater origins. However, carbonate and silicate fractions are different in abundance (due to mica) and pattern, including a more pronounced positive Eu anomaly in the silicate fractions. PAAS, post-Archean Australian Shale composite.

Full size image

PIXL (planetary instrument for X-ray lithochemistry) micro-X-ray fluorescence maps of elemental composition cast new light on putative evidence for biological activity1. First, maps of the distribution of the elements calcium, iron and manganese show that ‘stromatolitic’ lamination internal to the structures is actually a dolomitic alteration rim on a quartzose interior (Fig. 1g) and that there is no other compositional relict of internal lamination in the structures. Second, titanium and potassium are depleted not only in the ‘stromatolites’ but also throughout the quartzose layers (Fig. 1h), owing to the fact that they have considerably less potassium- and titanium-bearing mica compared to the dark layers. Finally, the iron and silicon maps show iron-rich/silicon-poor dark layers and silicon-rich/iron-poor light layers, suggesting that the rock may have originally consisted of intercalated cherty and iron-rich strata, which is consistent with previous studies and our observations of carbonate-altered banded iron formation and cherty metasediments in nearby outcrops15.

In the previous study, low-temperature dolomite formation was inferred from C and O isotopes, and this was interpreted as evidence of biogenic dolomite formation in the sedimentary environment1. However, the role of microbes in low-temperature dolomite formation is equivocal16. Furthermore, low temperature does not preclude secondary origins—an hypothesis supported by the presence of dolomite alteration rims observed in PIXL maps (Fig. 1g).

The REE + Y geochemistry of dolomite was mentioned as evidence of primary marine carbonate sedimentation in the previous study1. However, synchrotron X-ray fluorescence element maps show that the 800-μm-wide laser-ablation inductively coupled plasma-mass spectroscopy scans collected by Nutman et al.1 would have sampled a mixture of dolomite, quartz and micas. Therefore, their REE + Y patterns cannot be attributed to dolomite alone. Micas, in particular, are important trace element carriers in these rocks (Extended Data Fig. 6 and Supplementary Information).

To resolve this uncertainty, we separated carbonate and silicate (quartz and mica) fractions (Extended Data Fig. 5) by acid digestion and measured the REE + Y by mass spectroscopy (Fig. 2 and Extended Data Table 1). Both have REE + Y patterns broadly consistent with the properties of Archean to Paleoproterozoic seawater17,18,19. However, the overall abundance is higher in the silicates than carbonates, and higher in the mica-rich silicate sample than in the mica-poor silicate sample. These observations can be attributed to a high REE + Y concentration in micas. The silicate REE + Y pattern also has a larger positive Eu anomaly than the carbonate, indicating a different origin of the silicates compared to carbonate. Given the observed carbonate alteration of quartz (Fig. 1g), the most plausible interpretation is that the carbonate REE + Y composition was inherited from diagenetic and/or metasomatic fluids. In summary, the texture-specific distribution of major and minor elements, and the REE + Y composition of the rocks are consistent with original deposition in a marine environment, followed by secondary carbonate alteration any time between early diagenesis20 and late carbonate metasomatism—the latter process having been well-documented in nearby ISB meta-sedimentary rocks15.

Therefore, we propose that none of the previously published results support the interpretation of the ISB structures as stromatolites: they lack internal lamination and we found no evidence of synsedimentary growth. Their triangular ridge shape is not an indicator of biogenicity and they do not exhibit unique chemical compositions that indicate a localized microbial influence on the sedimentary processes1. We agree that the host rock protolith formed in a marine environment; however, there is no evidence for shallow water depth, and there is no unambiguous evidence that carbonate was part of the primary sedimentary assemblage. The inherent attributes of the structures, their geological setting in a fold hinge, the deformation fabrics observed in the host rock, and the shape and alignment of the structures within the overall rock fabrics—all indicate non-biological origins. In our view, it is very reasonable to interpret the ISB structures as products of structural deformation and carbonate alteration of layered rocks. On the other hand, we believe that the current evidence does not support the interpretation of these structures as 3,700-Myr-old stromatolites.

Methods

Data reporting

No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment.

PIXL analysis at the Jet Propulsion Laboratory

PIXL (planetary instrument for X-ray lithochemistry) is a microfocus X-ray fluorescence instrument developed at the Jet Propulsion Laboratory to fly aboard NASA’s Mars 2020 rover mission. The PIXL engineering prototype used in this study employs a Moxtek 12-Watt 60-kV MAGPRO Rh X-ray tube, in 180° emission geometry, mated with an XOS (6319) glass polycapillary focusing optic. The optic delivers a sub-millimetre (around 100 μm diameter at 7 keV) focused X-ray beam spot at a nominal 2-cm standoff distance from the target. Two Ketek Vitus H50 AXAS-D detectors (model: D5C2T0-H50-ML9BEV), oriented at 20° relative to the beam axis, provide a near-backscatter geometry for optimized X-ray detection. Analogue-to-digital conversion of X-ray signals and multi-channel binning of pulses is performed by the Ketek-built electronics. Data were acquired using an in-house-designed acquisition software built using a National Instruments LabVIEW platform.

Measurements of the Greenland rock were performed in air using 28 kV/100 μA X-ray tube-operating conditions. The rock sample was cut, polished and cleaned. The data of Fig. 1g were acquired by rastering the X-ray beam across the sample surface in x and y directions in 100-μm steps, with 15-s integration at each step, to produce a 10 × 15-mm2 image containing 15,000 data points. For Fig. 1h, a 200-μm step size was selected and a 10 × 20-mm2 area, containing 5,000 data points, was imaged.

To accurately identify the elements present, all of the spectra in each map were summed together to produce a whole-map summed spectrum. Principal elemental peaks of the major elements of interest were identified and peak areas under the characteristic X-ray peaks were derived using the in-house software package PIQUANT21, designed to process and fit spectral data generated by PIXL. PIQUANT uses a rigorously applied linear least-squares spectrum peak-fitting approach to ensure robust identification of elements, the fitting routine of PIQUANT applies Gaussian functions to each of the X-ray lines that describe a peak. The analytic integral of the combined Gaussians produces the net peak intensity. Peak intensity is one of the fitted variables. Also variable are two parameters associated with the peak widths and two more to describe the channel bin-to-peak energy conversion. The ‘noise’ corresponding primarily to the bremsstrahlung background that underlies the X-ray peaks is fitted separately using a SNIP17-fitting algorithm. The background contribution is subtracted as part of deriving the net peak intensities. This approach enables accurate distinction of element peaks even when the peaks overlap. In the Greenland rock maps, the Ba Lα X-ray peak (4.47 keV) and Ti Kα X-ray peak (4.51 keV) are an example of this. Both appear almost as part of one peak, broadened by the contribution and with 40-eV separation of the two elemental lines. With PIQUANT, the individual lines are distinguished, given that the line energies are constrained to a fixed energy and the peak widths and channel energy calibration parameters are constrained by values derived from the dominant lines of neighbouring elements (for example, Ca and Fe). Harnessing these constraints enables separation of Ti from Ba.

One challenge that persists through these spectra is that coherent scatter diffraction peaks appear in the energy range of spectra in which Ti and Ba peaks reside. Their presence has the potential to be registered falsely as a characteristic X-ray response. The PIQUANT software does not yet possess a treatment process that would allow for correction of this contribution. Therefore, a very small amount of data from this region may represent diffraction scatter instead of a X-ray response.

PIQUANT utilizes the databases from a previous study22, although a number of those databases have been, or are currently being, modified.

REE + Y analyses at Stony Brook University

Four samples were prepared and analysed by inductively coupled plasma mass spectrometry (ICP-MS) for their rare earth element (REE) and Y concentrations (REE + Y). A slab of the sample shown in Fig. 1b, c was sectioned with a tile saw (Extended Data Fig. 5) to separate a portion of the carbonate-rich rim on the light-grey part of the rock (subsample ‘C’) and a portion of the overlying dark-grey part of the rock (subsample ‘M’). These two subsamples were hand crushed in a ceramic mortar and then powdered in an agate shatterbox. Half a gram of each powder was weighted out and sonicated with 2% nitric acid for 30 min to leach carbonate from the samples. The supernatant was then removed, diluted and analysed by ICP-MS; these are samples C3C and M3C. About 40 mg (dry weight) of the remaining leached sediment was then dried and dissolved in a mixture of hydrofluoric and nitric acid for 12 h in sealed Teflon vials on a hotplate at around 120 °C. They were then dried and dissolved in aqua regia for 12 h in sealed Teflon vials on a hotplate at around 120 °C. Once completely dissolved, the aqua regia was dried off and the samples were reconstituted in nitric acid, which was then diluted and analysed by ICP-MS; these are samples C3M and M3M.

Elemental concentration analyses were performed in the FIRST (Facility for Isotope Research and Student Training) Laboratory in the Department of Geosciences at Stony Brook University on an Agilent 7500cx quadrupole ICP-MS. Samples were diluted to match to the signal of mixed calibration standards and unknown concentrations were calculated based on standard calibration curves, with standards run frequently between unknowns to monitor for drift in signal intensity. The U.S.G.S. Cody shale standard, SCO-1, was used to calculate the elemental concentrations for these samples. The REEs and Y concentrations in parts per million (p.p.m.) are shown in Extended Data Table 1. Concentrations were calculated for the carbonate leach of each sample assuming only carbonate was dissolved in 2% nitric acid using Ca and Mg concentrations to calculate the carbonate mass dissolved.

High-resolution X-ray fluorescence mapping at NSLS-II

High-resolution synchrotron X-ray fluorescence (XRF) spectra were collected on the Sub-micron Resolution X-ray spectroscopy (SRX) beamline at the National Synchotron Light Source 2 (NSLS-II) at the Brookhaven National Laboratory. The capabilities of the SRX beamline have previously been described23,24. In brief, SRX is a hard X-ray microprobe that performs scanning micro-fluorescence microscopy (μ-XRF) and X-ray absorption near-edge structure (μ-XANES) analysis using the high-brightness NSLS-II as a source of incident radiation. The SRX optics allow investigation of elemental distribution and chemical speciation at the sub-micrometre scale. For our analyses, we used an incident beam energy of 12 keV. The beam was focused to a spot size of 1 μm and XRF spectra were collected for 0.6 s at each spot. Motorized stages were used to move the sample under the beam with a 2-μm step size, thus generating a two-dimensional map of the elemental composition of this sample. Fluorescent X-rays were detected using an energy dispersive X-ray detector (Hitachi Vortex silicon drift detector). We analysed two areas, called ‘map 1’ and ‘map 2’ (Extended Data Fig. 6, Supplementary Information), on a cut and polished slab sample from locality A. This slab contains one of the putative stromatolite-like features, bounded above and below by alternating light and dark layers. XRF spectra for map 1 were collected on a 1–2 mm thick dark-black layer bounded above and below by thicker grey coloured layers. XRF spectra for map 2 were collected from within the core of the stromatolite-like feature. The dimensions of the maps are 200 × 200 μm2 and 125 × 125 μm2, respectively. To generate element maps, the individual X-ray spectra from each map were first summed into single bulk spectra (that is, one for map 1 and another for map 2) and fitted using the PyXRF analysis package developed at NSLS-II25. PyXRF uses a nonlinear least-squares method to determine global parameters such as peak width, energy calibration values, parameters related to the Compton and elastic scattering peaks, and element identities. Once elements were identified in the summed spectra, individual spectra were searched for those elements, their peak areas were fitted, and maps of elemental distribution and fluorescence intensity were generated.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Code availability

PIQUANT software used in the study is available from the corresponding authors upon reasonable request.