A fundamental assumption in the model of Rosing and others1 is that the coexistence of the mineral phases siderite (FeCO3) and magnetite (Fe3O4) in banded iron formations (BIFs) represents an assemblage that is nearly in thermodynamic equilibrium with the atmosphere, placing stringent constraints on the partial pressures of both CO2 and H2. The mineralogy of some BIFs is dominated by FeCO3 (ref. 2), whereas others have mineralogies dominated by ferric oxide phases3. However, in well-studied BIFs with mixed Fe3O4–FeCO3 phases (such as the Kuruman, Hamersley and Old Wanderer iron formations), carbon and iron isotope work4,5 indicates that these minerals did not precipitate in isotopic equilibrium with the ocean. Therefore, diagenetic conditions controlled mineral formation; there is evidence for an initial rain of ferric oxides and secondary formation of reduced and mixed-valence iron minerals during early or later-stage diagenesis4,5. This is important given that dissimilatory iron-reducing bacteria generate Fe3O4 even at very high aqueous carbon dioxide concentrations and headspace values of [∑CO2] ≈ 50 mM (ref. 6) and ≈ 0.2 atm (ref. 7). Mixed Fe3O4–FeCO3 assemblages develop as a result of an imbalance between rates of non-reductive ferric oxide dissolution, Fe2+ transport, and rates of Fe3+–Fe2+ conversion6,7,8. Therefore, values very much above the modern value do not preclude the formation of magnetite.

In addition, the authors assume that both the atmospheric partial pressure and aqueous concentration of H2 will be controlled by hydrogenotrophic methanogenesis. However, the preservation of ferric oxides, together with a paucity of organic carbon in most BIFs, strongly suggests that H2 pressures or concentrations would instead be buffered most often by dissimilatory iron-reducing bacteria in BIF diagenetic environments. This would decrease the aqueous concentration of H2 significantly9, changing the relevant stability boundary to Fe2O3–FeCO3. For instance, a threshold [H2] of about 0.1 nM would yield estimates between about 30–100 PAL (present atmospheric level) at 25 °C–35 °C (Fig. 1a of ref. 1). We view this as a non-trivial difference given the attendant implications for climatologically plausible CH4/CO2 ratios and atmospheric CH4 concentrations10.

Finally, even if we were to assume a simple thermodynamic control regulated by (via methanogenesis) and , it is unlikely that the coexistence of Fe3O4 and FeCO3 in BIFs provides any direct evidence regarding the chemistry of a surface ocean that is nearly in equilibrium with the overlying atmosphere. There is overwhelming petrographic evidence that the Fe3O4 observed in many BIFs is metamorphic or late-stage alteration product11,12,13. Indeed, the co-existence of Fe2O3–Fe3O4–FeCO3 assemblages and significant mineralogical variability on small spatial scales in many BIFs suggest either the formation and preservation of mineral phases out of thermodynamic equilibrium with the ambient ocean–atmosphere system or secondary/metamorphic alteration. Although we do not suggest that a single depositional model applies for all BIF occurrences throughout the Earth’s history, we do contend that many of the basic processes operating during their formation (microbial Fe oxidation/reduction, organic-matter remineralization in sediments, and metamorphism) are pervasive if not ubiquitous and should be considered in any attempt to relate BIF mineralogy to atmospheric composition.

Rosing et al.1 rightly point out that some previous work (see references in ref. 1) based on mineral stability in palaeosol profiles has also suggested low values. However, this work contrasts with experimental studies14, indicating that mineral stability as a function of during the formation of ancient soil profiles is not fully understood. Nonetheless, the numerical model proposed by Rosing et al.1 has a number of important implications. For example, by implicitly linking the Earth’s planetary albedo to the oxidation state of the atmosphere, the model may provide a new mechanism for explaining the inception of widespread glaciation following changes in Proterozoic atmospheric oxygen content. Such considerations are novel and will be fruitful to examine, and it is likely that any explanation for the clement early Earth will involve a complex interplay of forcings and feedbacks. However, we conclude that early Precambrian atmospheric levels remain poorly constrained and that increased levels of atmospheric gases (CO2, CH4, C2H2) remain a compelling solution to the ‘faint young Sun paradox’.