Given the scarcity of reliable paleoclimate record, the surface temperatures of the first half of Earth’s history remain poorly constrained. Here we show how the climate-sensitive δ18O value of surface precipitation recorded in Archean igneous and hydrothermal formations can help to resolve the state of early Earth climate. The Keivy complex, Kola craton (Fennoscandian Shield), formed via the intrusion of granitic and mafic magmas in the shallow crust at 2.67 Ga, where circulation of meteoric water created a distinct archive of the contemporaneous water cycle. Using whole rock data, mineral separates, and in situ zircon δ18O measurements, we disentangle the reaction mechanisms between the shallow magma and local precipitation. Syn-emplacement hydrothermal alteration produced near-contact lithologies with δ18O values as low as −8‰ recorded in amphiboles, while igneous zircons from granites crystallized from melts with δ18O from +1‰ to +4.5‰. High-precision U-Pb geochronology constrains the granite intrusion at 2673.5 ± 0.3 Ma. Using the Δ17O approach, these rocks reveal that the precipitation had a δ18O value 18‰ lower than the hydrosphere, providing one of the earliest quantitative records of continental precipitation generally compatible with a cold climate at high latitudes.

It is increasingly evident that the exposure of continental crust represents a major shift in Earth history sometime around the end of Archean, facilitating new geochemical pathways driven by subaerial hydrosphere-rock interactions (e.g., Bindeman et al., 2018; Wang et al., 2021). However, the timing, extent, and climate conditions of subaerially exposed crust remain poorly constrained. A unique feature of emerged subaerial continental crust is exposure to the hydrological cycle. Over the course of atmospheric vapor transport, precipitating liquid water becomes progressively depleted in the heavy isotope 18O relative to unevaporated seawater, which today has δ18O = 0‰. Consequently, δ18O values track atmospheric moisture and are controlled primarily by the temperature at which precipitation takes place, with colder temperatures resulting in more extreme depletions (Dansgaard, 1964). Modern-day δ18O gradients in precipitation range from as low as -60‰ at the South Pole to minimal depletion relative to seawater in tropical latitudes (IAEA, 2022). Silicate minerals with δ18O < 0‰ are produced through interaction with meteoric water (Taylor, 1977), so investigating low-δ18O continental crust and reconstructing the δ18O values of meteoric waters that shifted them away from their original compositions is a powerful way to obtain paleogeographic information for subaerially exposed fragments of old crust.

Here we use a diverse set of δ18O measurements, including in situ and bulk techniques applied to a suite of magmatic and hydrothermal rocks from an intact continental magmatichydrothermal system of the Kola craton (Fennoscandian Shield). Following the recently developed Δ17O approach (see Herwartz et al., 2015), we reconstruct the δ18O value of precipitation, extending previous meteoric water reconstructions back to the Neoarchean. These rocks, collectively termed the Keivy complex (Fig. 1), crop out over an area of 3000 km2 and are located in the northeastern part of Fennoscandian Shield. An assembly of 2.67 Ga peralkaline granites and their host rocks from the Keivy complex were described to be pervasively low in δ18O due to hydrothermal cycling of meteoric waters and magmatic assimilation (Zakharov et al., 2022). Moreover, the low-δ18O host rocks bear field, petrographic, and geochemical evidence for fluid cycling during emplacement of the peralkaline granites. The common features of alteration include fenitization (gains of Na and K) and the occurrence of alkaline amphiboles, K-feldspars, and fluorite that developed in the near-contact host rocks, suggesting that the hydrothermal alteration of host rocks is closely tied to emplacement of the peralkaline granites. In combination with previous geochronological efforts, we provide a high-precision U-Pb zircon age to pinpoint the timing of Neoarchean precipitation.

Several samples of zircon, quartz, whole rock, and amphibole were measured for δ18O in bulk using the laser fluorination technique. We use samples extracted from peralkaline granites, near-contact altered host rocks and quartzolites (see sample description in the Supplemental Material1). Quartzolites are near-contact quartzdominated (>50 vol%) rocks rich in zircon and other rare-metal accessory minerals. They are irregularly shaped bodies that formed as a product of focused fluid cycling in the apical parts of peralkaline granites and their host rocks. In addition, the samples were measured for triple O isotope values.

We extracted zircon from the low-δ18O granites and quartzolites for δ18O in situ measurements from ~10 μm domains using secondary ion probe mass spectrometry (SIMS; see methods in the Supplemental Material). We targeted the same crystals that were used earlier for U-Pb dating (Zakharov et al., 2022). Collected concurrently with δ18O by SIMS, the 16OH/16O ratio is used here to filter zircons that are least affected by alteration caused by radiation damage and hydration. To ensure the accuracy of the measurement, we used indium sample holders, which induce minimal contamination of the 16OH/16O ratio.

We selected zircon from sample PM8 for detailed geochronology based on consistent SIMS ages and δ18O values, low U, and low 16OH. The zircon crystals were treated by chemical abrasion and dated via bulk dissolution and chemical abrasion isotope-dilution thermal ionization mass spectrometry (CA-ID-TIMS). For this study, ten zircon crystals were annealed and then chemically abraded in hot HF at 210°C for 12 hr to remove domains affected by Pb loss. This process caused grains to break up into small pieces, which allowed duplicate measurements of four of the grains using separate fragments of the same original crystal.

Bulk δ18O and In Situ SIMS δ18O-16OH/16O Measurements

The mineral separates from peralkaline granites and their host rocks display bulk δ18O values between -8‰ and +5‰ (Fig. 1), with the lowest δ18O measured in the near-contact altered rocks. The SIMS δ18O values of zircons from peralkaline granites exhibit a broad range from -1‰ to +7‰ (see Table S1 in the Supplemental material), some of which resulted from damaged and altered domains (Fig. 2). The majority of magmatic zircons have near-zero 16OH/16O (as do the anhydrous grains of standard zircon 91500 mounted in the same indium holder; see Fig. S4) corresponding to a range of δ18O values between -1‰ and +2‰ (Fig. 2). The zircon δ18O values from quartzolites vary between -8‰ and +3‰, with the lowest 16OH/16O corresponding to the lowest δ18O values (Fig. 2). Many domains have 16OH/16O close to that of the anhydrous standard grains, with each sample systematically pointing toward the same y-axis intercept of -8‰ (Fig. 2).

Triple O Isotopes

The triple O data set (Table S3) has δ18O values spanning -8.2‰ to +5.2‰ accompanied by increasing Δ′17O values from -0.06‰ to -0.01‰, expressed as 103·ln(q17O/1000 + 1) – 0.528·103·ln(δ18O/1000 + 1) (see definitions in the Supplemental Material for delta notations). Using the nonlinearized δ18O-Δ17O expressions (Δ17O = q17O – 0.528·δ18O), the data set displays a prominent trend toward the common origin, i.e., Neoarchean meteoric water with low δ18O and elevated Δ17O (Fig. 3).

CA-ID-TIMS Zircon Geochronology

All pairs of zircon fragments from the same crystals were found to be of the same age within analytical uncertainty, and an overlapping population of ten analyses obtained after discarding four younger ages that are affected by Pb loss yielded a weighted mean age of 2673.49 ± 0.32/0.44/6.8 Ma (analytical/tracer calibration/decay constant 2s uncertainties, MSWD = 2.6, n = 10; individual analyses in Table S2 report only analytical uncertainties).

Hydrothermal Alteration by Neoarchean Meteoric Water and Magmatic Assimilation

The zircon SIMS δ18O values represent a resilient archive of magmatic crystallization. Based on the zircons with low 16OH/16O, the peralkaline granites inherited their low δ18O values from magmatic incorporation of pre-existing low-δ18O materials, likely the low-δ18O altered host rocks. The range of δ18O values measured in these zircons corresponds to melt δ18O values between +1‰ and +4.5‰ if we assume a Δ18Omelt-zircon (i.e., δ18O fractionation between melt and zircon) of 2‰ (Bindeman, 2008). This range agrees with silicate bulk δ18O measurements across multiple samples (see Fig. 1) and is interpreted to indicate various degrees of magmatic assimilation of low-δ18O hydrothermally altered rocks.

Concurrent hydrothermal alteration is documented by notably low δ18O values (as low as -8‰ in altered gneiss) in the host rocks altered in the vicinity of the peralkaline granites. Further, we report a new δ18O measurement of -3.7‰ from a quartzolite, emphasizing the low-δ18O nature of fluids that circulated in the near-contact aureoles. At hydrothermal temperatures (T ≈ 300 °C), equilibrium H2O has δ18O of -11‰ (Sharp et al., 2016), while the initial meteoric water must have had an even lower δ18O value prior to reaction in the hydrothermal system. The zircon enclosed in the quartzolite also has low δ18O, as low as -8‰, preserved in the low-16OH/16O domains. These pervasively low δ18O values exemplify the environments where the Neoarchean meteoric water is recorded via high-temperature water-rock exchange reactions. The initial hydrothermal exchange likely occurred contemporaneously or shortly after the intrusion of peralkaline granites at 2673.49 ± 0.32 Ma. The zircons extracted from the quartzolites were previously dated in situ to ca. 1.76–1.83 Ga (Fig. 2), reflecting the age of regional metamorphism that affected many parts of the Kola craton (Balagansky et al., 2021). These zircons are rich in U (as much as 5000 ppm) and were susceptible to radiation damage (see Fig. 2). A metamorphic origin of the low δ18O values is unlikely because it fails to explain the origin of low-δ18O peralkaline magma and 2.67 Ga zircons. Further, the occurrences of the lowest δ18O values are so far found only in the altered host rocks at the contact with the 2.67 Ga granites (Zakharov et al., 2022). Given that O is the major mineral-forming element, the low δ18O value of quartzolites is more likely to have originated due to the high-temperature cycling of low-δ18O water at 2.67 Ga rather than during regional metamorphism.

Oxygen Isotope Record of the 2.67 Ga Precipitation

To decipher the initial precipitation δ18O values, we use the triple O isotope approach (Herwartz et al., 2015). The δ18O and Δ17O values yield systematic differences between the global precipitation and unaltered crustal rocks, with hydrothermally altered rocks occupying the intermediate space offset by temperature-dependent fractionation (Fig. 3). The Keivy complex data set yields the altered end-member composition at δ18O of about -15‰ assuming the modern-day meteoric water line. Given the average mineral-water fractionation of +3‰ ± 1‰, the δ18O value of the meteoric water must have been -18‰ ± 6‰.

However, the δ18O of ancient precipitation is intimately tied to the long-term evolution of the seawater δ18O value because atmospheric vapor originates via evaporation in the oceans. While this has been a controversial topic (see Jaffrés et al., 2007), a recent view of the δ18O of seawater through time suggests that the oceans had a δ18O of ~-2‰ by 2.7 Ga (Herwartz et al., 2021, and references therein), which is similar to the isotope composition of modern ice-free oceans. The 2‰ difference between the Neoarchean and modern seawater does not affect our interpretation significantly, given that the uncertainty is >2‰ (Fig. 3). Alternatively, the Neoarchean hydrosphere could have had a lower δ18O value as a result of diminished hydrothermal and/or increased weathering fluxes (Muehlen-bachs, 1998). Such seawater is expected to have had an elevated Δ17O value (Zakharov et al., 2021; McGunnigle et al., 2022). If we accept a seawater δ18O value of ~-6‰ as a potential intermediate solution between variation of these fluxes and the sedimentary data (e.g., Sengupta et al., 2020), the contemporaneous meteoric water would have reached δ18O values of -24‰ based on our data set (see Fig. 3). Thus, despite the potentially lower δ18O of the hydrosphere, our data are still best explained by a subaerial water cycle that produced precipitation with a δ18O valuê18‰ lighter than the contemporaneous seawater.

Paleoclimate Implications

How much moisture remains in atmospheric air after evaporation and condensation controls the extent of 18O depletion. Most evaporation and precipitation occurs at the equatorial latitudes, where temperatures are high. Previous paleomagnetic data place the Kola and neighboring Karelia cratons at polar latitudes at 2.7 Ga, between 60°N and 90°N (Fig. 4; Salminen et al., 2021). Thus, our study documents 2.67 Ga precipitation at high latitudes, where precipitation must have exhibited the strongest 18O depletions (Dansgaard, 1964). As a crude approximation, modern-day precipitation with δ18O = -18‰ is characteristic for the regions with a mean annual temperature of -6 °C (Yurtsever, 1975). However, modern-day precipitation at high latitudes also exhibits strong δ18O gradients, e.g., the coastal station at Reykjavik, Iceland (latitude 64°N), samples mean annual precipitation with δ18O = -8‰, while central Greenland precipitation ranges between -35‰ and -25‰ (IAEA, 2022). Depletions of -18‰ in the modern world also occur in inland parts of continents and orographically elevated regions at lower latitudes (IAEA, 2022). Evidently, the 18O depletions are also affected by the landmass configuration, yet the details of Neoarchean paleogeography are currently unavailable. However, near-contemporaneous low-δ18O magmatic-hydrothermal systems on Archean cratons can provide a powerful way to learn about the subaerial paleogeography of early Earth.

Combined with the existing evidence on Neoarchean glaciations recorded in 2.7 Ga diamictites from the Dharwar craton in India (Ojakangas et al., 2014), the strong 18O depletions of atmospheric moisture provide an image of the high-latitude climate compatible with mean annual temperatures <0 °C. Meanwhile, the Archean climate is envisaged to have been hot, with ocean temperatures as high as 80 °C based on the sedimentary record (Lowe et al., 2020; McGunnigle et al., 2022). In addition, high CO2 and CH4 concentrations are commonly envisaged for the Neoarchean based on the paleosol evidence and redox state of the atmosphere (Catling and Zahnle, 2020, and references therein). Consequently, our data attest to a dynamic surface temperature regime on the Neoarchean Earth, where temperatures of the oceans may have been 50–80 °C (see Herwartz et al., 2021), while the high-latitude subaerially exposed continents experienced freezing temperatures at least episodically.

We thank Andreas Pack and Tommaso Di Rocco (University of Göttingen, Germany) for carrying out the triple O isotope measurements for this project. We also express gratitude to Maria Ovtcharova (University of Geneva) for assistance with the CA-ID-TIMS measurements, and to the ion probe team at the CRPG, CNRS-Nancy (France) for the in situ O-isotope measurements of zircons. Zakharov expresses gratitude to the Faculty of Geoscience and Environment, University of Lausanne and Prof. Johanna Marin-Carbonne for supporting this study. We are grateful to three anonymous reviewers for their comments and suggestions.

1Supplemental Material. Sample description, methods, oxygen-isotope notations and triple oxygen-isotope fractionation, Figures S1–S5, and Tables S1–S3. Please visit to access the supplemental material, and contact with any questions.
Gold Open Access: This paper is published under the terms of the CC-BY license.