Links between the Archean hydrosphere and Earth’s interior Authors: Zheng-Yu Long, Frédéric Moynier, Igor S. Puchtel, Yiheng Li, Tu-Han Luu, Jun Deng, Kun-Feng Qiu Published: 2026 ABSTRACT Earth’s long-term habitability depends on cycling of surface volatiles into and out of the mantle, yet when this cycling began and how the early hydrosphere became connected to Earth’s interior remain debated. Komatiites are ultramafic magmas produced mostly during the Archean by deep, high-degree mantle melting, providing a sensitive archive of early mantle composition. Here we present high-precision K isotope data for komatiites from ten localities. Our data show that Archean komatiites from three localities record heavy potassium isotope compositions, well above values typical of the mantle and most surface reservoirs. This signal is best explained by metasomatism of mantle domains by potassium-rich fluids released from subducted, water-altered oceanic lithosphere, later sampled by komatiite magmas. The data indicate that surface-derived water was transferred into mantle reservoirs by at least 3.6 billion years ago. Komatiites thus link the Archean hydrosphere to Earth’s interior and record early operation of subduction-style volatile recycling. Introduction The timing of Earth’s early onset of plate tectonics remains a topic of intense debate. Since the sustained return of surface volatiles to the mantle occurs via subduction, constraining the timing of the onset of plate tectonics is essential for determining when Earth’s surface-mantle volatile recycling, and its consequent effects on climate regulation and crustal growth, first emerged1–6. However, Earth’s earliest hydrosphere left only subtle imprints in the Archean mantle. Komatiites are ultramafic volcanic rocks representing some of the hottest magmas in Earth’s history, with liquidus temperatures in excess of 1600 °C7. They formed by high degrees of mantle partial melting (>30–50%) and are commonly considered to have been derived from deep mantle plumes. Komatiite magmatism is almost entirely restricted to the Archean, with only a few occurrences in post-Archean times. This nearly complete disappearance of komatiites after ~2.5 Ga is generally ascribed to secular mantle cooling8,9. Komatiites, therefore, provide a distinctive archive of early Earth geodynamics. A long-standing question of komatiite genesis concerns the role of volatiles (especially H2O) in komatiite magma generation8,10,11. Some models have assumed hot, dry komatiite mantle sources, analogous to volatile-poor modern deep mantle plumes12. Consistent with this view, some studies of Archean komatiites have indicated very low water contents in their parental magmas (H2O < 0.2 wt%)13–15, and oxidation states comparable to the present-day dry upper mantle16–18, implying anhydrous melting. In contrast, other models have proposed that komatiites may have originated via wet melting in Archean subduction zones19,20. Analyses of olivine-hosted melt inclusions in komatiites from the Abitibi and Barberton greenstone belts, however, have recorded appreciable H2O contents (~0.5–0.6 wt% or higher; H2O/Ce > 6000)21,22. These studies suggest that hot, initially dry plumes could have entrained water while traversing the mantle transition zone (MTZ; 410–660 km), where a hydrous layer may have formed via partial dehydration of subducted, seawater-altered oceanic lithosphere22. Alternatively, the MTZ water does not need to be subduction-derived; instead, wadsleyite, ringwoodite, and some other even deeper phases may have stored water since very early in Earth history23. Thus, the ultimate sources of water in komatiite genesis and, by implication, in the early Archean mantle, have yet to be established. Potassium (K) stable isotopes provide a sensitive tracer of volatile-driven processes. Potassium has two stable isotopes, 39K and 41K, with isotopic compositions reported as δ41K. As a large-ion lithophile and fluid-mobile element, K is readily transported by aqueous fluids; processes such as seafloor hydrothermal alteration and slab dehydration can significantly fractionate K isotopes24–28. Consequently, geological reservoirs may acquire distinct δ41K signatures24,29–31 (Fig. 1). Mantle xenoliths, for example, display highly heterogeneous δ41K (Fig. 1). Because heavy 41K preferentially partitions into fluids during slab dehydration32, some of the highest δ41K values are reported for subduction zone peridotites metasomatized by slab-derived fluids33,34, providing evidence for infiltration of isotopically heavy K from dehydrated oceanic lithosphere. Here, we report high-precision K isotope data for komatiites from ten localities spanning the full temporal range over which this magma type occurs, from the ~3.6 Ga Schapenburg to the 89 Ma Gorgona komatiites (Fig. 1). We focus in particular on komatiites from the Barberton and Abitibi greenstone belts, which have been proposed to contain recycled oceanic water in their mantle sources21,22.Fig. 1Potassium isotope compositions of komatiites analyzed in this study (left) compared with published values for major reservoirs (right).Shaded bands mark reference ranges for the lithospheric mantle (red; −0.42 ± 0.13‰, 2 SD)31, seawater (blue; +0.12 ± 0.07‰)82, and slab fluids (light blue; +0.1 to +1.4‰)32. Box-and-whisker plots summarize literature δ41K data for different reservoirs (including data for meteorites). Two crustally contaminated groups highlight the effect of crustal contamination on light K isotope signatures. Initial δ41K values, accounting for secondary alteration, are also shown for Schapenburg, Weltevreden, and Boston Creek. Potassium isotope compositions of komatiites analyzed in this study (left) compared with published values for major reservoirs (right). Shaded bands mark reference ranges for the lithospheric mantle (red; −0.42 ± 0.13‰, 2 SD)31, seawater (blue; +0.12 ± 0.07‰)82, and slab fluids (light blue; +0.1 to +1.4‰)32. Box-and-whisker plots summarize literature δ41K data for different reservoirs (including data for meteorites). Two crustally contaminated groups highlight the effect of crustal contamination on light K isotope signatures. Initial δ41K values, accounting for secondary alteration, are also shown for Schapenburg, Weltevreden, and Boston Creek. Results We analyzed 50 komatiite samples from the Barberton Greenstone Belt (Schapenburg, Komati, Weltevreden), the Abitibi Greenstone Belt (Boston Creek, Pyke Hill), the Fennoscandian Shield (Kostomuksha, Vetreny Belt, Finnish Lapland), the Rhodesian Craton (Belingwe), and the Caribbean Oceanic Plateau (Gorgona). Measured δ41K values range from −1.49 ± 0.05‰ to +0.59 ± 0.04‰ (Fig. 1). The Vetreny Belt and Finnish Lapland suites show generally lower δ41K. The Belingwe samples plot at or slightly above the upper end of the mantle range, whereas Gorgona overlaps the mantle range. A chilled-margin sample from Weltevreden has a δ41K (−0.74 ± 0.07‰) that is lower than the other samples from this locality. Notably, the least-altered samples from Schapenburg, Weltevreden, and Boston Creek exhibit particularly high δ41K values exceeding the seawater value. Potassium isotope variation in komatiite systems Potassium isotopes in the studied komatiites were largely unaffected by magmatic differentiation (Fig. S1–S3). Spinifex-textured and olivine-cumulate samples have similar δ41K values. Bulk-rock K and δ41K in olivine-rich cumulates mainly reflect trapped intercumulus melt (mesostasis) and late interstitial phases, rather than olivine. The agreement between spinifex and cumulate samples within individual flows further excludes kinetic isotope fractionation associated with mesostasis-hosted K. Crustally contaminated komatiite suites (e.g., Vetreny Belt and Finnish Lapland) show low Nb/Nb*, elevated K contents, and lower δ41K (Fig. S4), indicating mixing between komatiitic magma and an isotopically light crustal component. Crustal contamination, however, cannot explain the unusually high δ41K in some komatiites. Chemical weathering and hydration-related alteration tend to either leave δ41K largely unchanged or shift it to lower values through preferential loss of isotopically heavy K during fluid-rock interaction35. This is supported by correlations in several suites, in which K content and δ41K both decrease with increasing loss on ignition (LOI; Figs. 2 and 3), indicating net K loss during serpentinization. The key implication is that the alteration drives komatiites toward lower, not higher, δ41K. Thus, the elevated δ41K in the least-altered samples is best interpreted as reflecting a primary magma signature. In this study, the more strongly altered Kostomuksha and Pyke Hill suites were excluded from further source discussion. The subsequent interpretations focus on the better-preserved Schapenburg, Weltevreden, and Boston Creek samples. A full evaluation of magmatic processes, crustal contamination, and alteration is provided in the Supplementary Information.Fig. 2Plots illustrating the effect of serpentinization on K abundances and K isotopic composition in the more strongly altered Pyke Hill and Kostomuksha komatiite suites.a SiO2 vs LOI showing the systematic decrease of SiO2 with increasing alteration. b 100×K/Ti vs LOI for all suites. c, d Pyke Hill: δ41K plotted against 100×K/Ti and LOI. e, f Kostomuksha. 100×K/Ti is used as a mobility index. The blue dashed line indicates the linear regression fit.Fig. 3Plots illustrating the effect of serpentinization on K abundances and K isotopic composition in the better-preserved Boston Creek, Schapenburg, and Weltevreden komatiite suites, with comparison to other alteration systems.a, b Boston Creek. c, d Schapenburg. e, f Weltevreden. g Also shown is a comparison of komatiites with reported alteration systems, e.g., weathered igneous rocks, hydrothermal ore systems, high-T AOC, and low-T AOC, from the literature (Table S3). The isotopic effects of alteration on the Schapenburg and Weltevreden samples are evaluated by subdividing them according to their location within individual flows and examining intra-flow isotopic trends. The blue dashed line indicates the linear regression fit. Plots illustrating the effect of serpentinization on K abundances and K isotopic composition in the more strongly altered Pyke Hill and Kostomuksha komatiite suites. a SiO2 vs LOI showing the systematic decrease of SiO2 with increasing alteration. b 100×K/Ti vs LOI for all suites. c, d Pyke Hill: δ41K plotted against 100×K/Ti and LOI. e, f Kostomuksha. 100×K/Ti is used as a mobility index. The blue dashed line indicates the linear regression fit. Plots illustrating the effect of serpentinization on K abundances and K isotopic composition in the better-preserved Boston Creek, Schapenburg, and Weltevreden komatiite suites, with comparison to other alteration systems. a, b Boston Creek. c, d Schapenburg. e, f Weltevreden. g Also shown is a comparison of komatiites with reported alteration systems, e.g., weathered igneous rocks, hydrothermal ore systems, high-T AOC, and low-T AOC, from the literature (Table S3). The isotopic effects of alteration on the Schapenburg and Weltevreden samples are evaluated by subdividing them according to their location within individual flows and examining intra-flow isotopic trends. The blue dashed line indicates the linear regression fit. Decoupling of K from W and Nd isotope systematics in komatiites Our results indicate that the studied Archean komatiites were derived from Archean mantle domains with unusually heavy K isotope compositions. Earth is known to have been depleted in moderately volatile elements (including K) relative to primitive CI chondrites, consistent with accretion dominated by volatile-poor inner Solar System materials plus a fraction of a carbonaceous chondrite (CC)-like component36. Nucleosynthetic anomalies present in K isotopes indicate that most of Earth’s K inventory was derived from non-carbonaceous (NC) bodies, with only ~10 to 20% CC input37, yielding a Bulk Silicate Earth (BSE) value of δ41K ≈ −0.42‰30, intermediate between NC and CC groups (Fig. 1). Because high-temperature mantle processes such as partial melting and crust-mantle differentiation result in negligible K isotope fractionation30, large-scale crust extraction likely left mantle δ41K composition near its primordial value since the Hadean31. Yet, the studied komatiites display δ41K variations that exceed those of basalts and bulk chondritic meteorite groups (Fig. 1), implying an anomalous source or process that preferentially retained heavy K and drove their mantle source reservoirs to unusually high δ41K. Below, we evaluate two possibilities: (1) volatile loss by evaporation (e.g., early Earth melting) and (2) preservation of an ancient, isotopically distinct mantle reservoir. Evaporation enriches residual melts/solids in the heavier 41K, as the lighter 39K partitions into vapor38,39. However, the models of the Moon-forming impact predict shifts of <0.4‰ in Earth’s δ41K39, and magma-ocean processes result in only modest K isotope fractionation. Consistent with limited planetary-scale effects, the Bulk Silicate Moon (−0.07 ± 0.09‰)40 and Bulk Silicate Mars (−0.28 ± 0.18‰)41 are only slightly heavier than the BSE (−0.42 ± 0.07‰)30, with the higher values correlating with smaller planetary mass and greater volatile loss38,41. Thus, the ~1.0‰ higher δ41K in komatiites compared to the BSE cannot be explained by the evaporation model alone. Komatiites have been shown to preserve Hadean silicate heterogeneities recorded by short-lived isotope systems (Fig. 4). Positive μ182W anomalies in Archean rocks of up to +20 ppm reflect early silicate fractionation in a magma ocean42,43, whereas negative μ182W values may reflect the presence of unradiogenic, low Hf/W material via either late accretion, early silicate differentiation, or core-mantle exchange15,42. Complementary μ142Nd anomalies (–10 to +10 ppm), resulting from Sm/Nd fractionation within the first 500 Ma of Earth history, are attributed to early magma ocean processes that produced complementary enriched and depleted reservoirs15. These reservoirs survived for billions of years despite vigorous convective mantle mixing. If heavy K isotopes found in the studied komatiites were inherited from a single primordial reservoir, δ41K would correlate with the μ182W and μ142Nd signatures. Such a relationship is not observed (Fig. 4). The Boston Creek samples have positive μ182W but negative (terrestrially unresolvable) μ142Nd values43, whereas the Schapenburg samples show negative μ182W and μ142Nd values15, and the Weltevreden samples show no anomaly14, even though all of the komatiites have similarly high δ41K values. This demonstrates that the process generating the K isotope heterogeneity is decoupled from magma-ocean differentiation and core-formation signals recorded by the short-lived 182W and 142Nd isotope systems.Fig. 4Decoupled variations of K, W, and Nd isotopes in komatiites.Plots of a μ182W and b μ142Nd versus δ41K. Each symbol denotes a regional value (locality-scale). The W and Nd isotope data are compiled from ref. 58. Decoupled variations of K, W, and Nd isotopes in komatiites. Plots of a μ182W and b μ142Nd versus δ41K. Each symbol denotes a regional value (locality-scale). The W and Nd isotope data are compiled from ref. 58. Another scenario involves mineral segregation in an early magma ocean. In principle, as a highly incompatible element, K is not appreciably incorporated into lower-mantle phases (e.g., bridgmanite and ferropericlase)44. Although davemaoite can host K45, few mechanisms operating at very high temperatures are capable of enriching residual silicates in isotopically heavy K46. Given the chalcophile behavior of K under highly reducing conditions47,48 and the potential preferential uptake of 39K into sulfide46,49, one possible process is segregation of a K-bearing sulfide melt under S-rich reducing conditions, leaving the residual mantle enriched in heavy K isotopes. However, this mechanism is not viable on thermodynamic and isotopic grounds. First, immiscible sulfide melt segregation requires sulfur saturation, yet the sulfur concentration at sulfide saturation increases sharply with increasing temperature; at a magma-ocean temperature, the BSE sulfur content is far below saturation, rendering global S saturation unlikely50,51. Second, even localized saturation would demand unrealistically high sulfide proportions and/or S contents to sequester appreciable K52,53. Finally, equilibrium isotope fractionation at ultrahigh temperatures is intrinsically small, with predicted sulfide-silicate K isotope offsets far too low to produce the >1‰ δ41K anomalies46,54 observed in the studied komatiites. Collectively, the heavy K isotope signatures in the komatiites are unlikely to be the primordial signatures and, instead, must reflect later processes. A signature of recycled oceanic lithospheric materials in Archean mantle sources Elevated δ41K values in mantle-derived magmas are commonly attributed to slab inputs. Notably, carbonatites (interpreted to sample the deep mantle) display high δ41K (up to 0.17‰) inherited from altered oceanic crust delivered via subducted slabs28. Subducted Archean lithosphere could similarly impart a heavy K isotope imprint28. An additional line of evidence comes from the low-δ18O composition of olivine in Weltevreden komatiites. This signature was initially linked to an isolated lower-mantle domain formed during Hadean magma-ocean crystallization55. However, this interpretation is inconsistent with the minor O-isotope offsets produced by early deep-mantle phase crystallization56. A revised view attributes the low δ18O in komatiitic olivine (down to 3.6‰) to recycling of low-δ18O altered oceanic crust (AOC) into the mantle sources, relative to the typical mantle olivine δ18O value of ~5.2‰57. Komatiite sources often represent depleted residues from earlier melting58 and, thus, are strongly depleted in K, whereas slab components are K-rich. By analogy, the heavy K isotope signatures in Archean komatiites may reflect the contribution of surface-derived materials. In the Archean, vigorous ridge-crest hydrothermal systems could have produced isotopically heavy K in AOC. Modern AOC displays highly variable δ41K values, from lower than −1.0‰ to higher than +0.1‰ (Fig. 3g), with alteration temperature as the dominant control. Low-T alteration forms K-bearing secondary minerals (clays, zeolites) that preferentially incorporate 39K, yielding isotopically light bulk AOC, whereas high-T alteration at ridges either promotes diffusive 39K loss (leaving 41K-enriched residues; e.g., −0.48 to +0.14‰; average −0.22‰ at the East Pacific Rise) or enables isotope exchange with heavy-K seawater24,25,59. However, producing positive δ41K in depleted sources would require unrealistically large AOC contributions (Fig. 5), which would be inconsistent with the depleted elemental and isotopic character of the studied komatiites. Moreover, generating bulk AOC heavy enough to explain δ41K of up to +0.59 ± 0.04‰ is difficult, and the variation of AOC δ41K (from negative to positive) undermines its reliability as a consistently heavy component (Fig. 5a). Thus, a bulk-AOC origin is unlikely.Fig. 5Potassium isotope distributions and slab-fluid mixing models.a Distributions of K isotope compositions for oceanic basalts, altered oceanic crust (AOC), arc lavas, ophiolites, and mantle xenoliths compiled from the literature (data source in Table S3), together with the komatiites from this study. Also shown is the modeled K isotope range for slab-derived fluids32. Heavy K isotope compositions in some arc lavas, ophiolites, and mantle xenoliths have been attributed to metasomatism of their mantle sources by slab fluids33,34,61. b Two-endmember mixing models: depleted mantle mixed with AOC, and depleted mantle mixed with slab-derived fluid. Numbers along the mixing curves indicate the mixing fraction. In this mixing scenario, the AOC end-member represents bulk altered oceanic crust physically incorporated into the depleted mantle source through subduction and convective stirring, rather than AOC-derived partial melts; the large AOC mass fractions required reflect the inefficiency of this bulk-mixing mechanism in raising the source δ41K. Mixing parameters are those specified in refs. 24,32. Potassium isotope distributions and slab-fluid mixing models. a Distributions of K isotope compositions for oceanic basalts, altered oceanic crust (AOC), arc lavas, ophiolites, and mantle xenoliths compiled from the literature (data source in Table S3), together with the komatiites from this study. Also shown is the modeled K isotope range for slab-derived fluids32. Heavy K isotope compositions in some arc lavas, ophiolites, and mantle xenoliths have been attributed to metasomatism of their mantle sources by slab fluids33,34,61. b Two-endmember mixing models: depleted mantle mixed with AOC, and depleted mantle mixed with slab-derived fluid. Numbers along the mixing curves indicate the mixing fraction. In this mixing scenario, the AOC end-member represents bulk altered oceanic crust physically incorporated into the depleted mantle source through subduction and convective stirring, rather than AOC-derived partial melts; the large AOC mass fractions required reflect the inefficiency of this bulk-mixing mechanism in raising the source δ41K. Mixing parameters are those specified in refs. 24,32. A more plausible mechanism to explain the variability is metasomatism by slab-derived fluids. During slab dehydration, equilibrium and/or kinetic fractionation drives 41K into the fluid, leaving an isotopically light, K-depleted eclogitic residue and an isotopically heavy, K-enriched fluid (+0.1 to +1.4‰; Fig. 5a). Consistent with this view, peridotite xenoliths from subduction settings show whole-rock δ41K values of +0.40 to +0.62‰33,34. Likewise, sections of the Troodos ophiolite modified by slab fluids record the most enriched δ41K (up to 1.13‰; Fig. 5a) relative to the unmetasomatized units60. Even a minor addition of K-rich, high-δ41K fluid can raise the source δ41K (Fig. 5b) without strongly affecting the major-element chemistry. Modern arc-lava studies provide independent support for this slab-fluid mechanism. K isotope data from the Izu arc show that both arc-front and rear-arc lavas are heavier than the MORB-source mantle, consistent with isotopic fractionation during slab dehydration61. A recent compilation of arc lavas from circum-Pacific margins reports a broad range (δ41K = −0.86 to −0.01‰)62, but bulk arc lavas do not reach the high δ41K values recorded in our komatiites. This contrast reflects sampling bias rather than a difference in process. Bulk arc magmas integrate slab-fluid signatures over the much larger volume of the mantle wedge; thus, the heavy K isotope signal is substantially diluted by the surrounding mantle. The fractionation factor inferred from arc systems (Δsolid–fluid ≈ −0.40 to −0.50‰)61 is sufficient to drive slab-derived fluids to +0.5‰ or higher. Notably, individual rear-arc lava samples from the Izu arc reach δ41K values as high as +0.8‰61, comparable to or exceeding the highest values we report for the komatiites. This demonstrates that subduction-zone-related systems are fully capable of generating δ41K within the observed komatiite range. Slab-fluid metasomatism, therefore, provides a viable K source for the heavy K isotope signatures recorded in the komatiites, whether such signals were stored in the deep mantle plume source, in the lithospheric mantle traversed during magma ascent, or both (see further discussion below). Implications for early volatile recycling One view holds that a hotter Archean mantle produced a thicker, buoyant oceanic crust, favoring a stagnant lid with vertical recycling via delamination (“drips”) and mantle plumes1,63. Yet, as oceanic lithosphere aged and cooled away from spreading ridges, its density would have increased, enabling intermittent subduction64. Large meteorite impacts may also have helped weaken the lithosphere and trigger episodic subduction initiation65. Paleomagnetic data from the Pilbara Craton (~3.2 Ga) imply latitudinal drift rates of >2.5 cm/yr, comparable to modern rates66; the discovery of ~2.5 Ga eclogite-facies rocks requires Archean slab descent to >65–70 km67; and some Archean volcanic-plutonic suites show enrichments in fluid-mobile large-ion lithophile elements typical of subduction magmas68–70. Other studies of zircon chemistry likewise suggest that some crust-generation processes may have been linked to early subduction by ~4.2–3.7 Ga5,71. Recent models therefore favor a mixed or episodic tectonic regime on the early Earth, with higher mantle temperatures, weaker plate boundaries, frequent slab stalling or dripping, and coexistence of mantle plumes rather than universally modern-style subduction5,72–76. In the Archean, seafloor magmatism produced unusually thick, Mg-rich oceanic crust that underwent extensive hydrothermal alteration by anoxic, Fe2+- and Si-rich seawater, promoting silicification of basalts and hydration of peridotites, and locking water into amphibole, chlorite, and serpentine in the upper oceanic lithosphere77. Higher geothermal gradients likely drove slabs along hotter P-T paths, such that dehydration and/or partial melting may have begun at relatively shallow depths78,79. Thermo-mechanical models further suggest that hotter subduction favored shallow to flat slab geometries, with much of the slab-derived H2O being recycled into the shallow mantle and overriding lithosphere75. Under such conditions, slab-derived agents were likely more compositionally diverse than the dilute aqueous fluids often invoked in modern arc settings4, potentially spanning solute-rich fluids, supercritical liquids, and hydrous silicate melts. K-bearing slab components were therefore likely mobilized at relatively shallow levels, releasing K to ascending melts and fluids before slabs reached the deeper mantle reservoirs80. One way to reconcile the heavy K isotope signatures is by assuming that slab dehydration first introduced K-rich, isotopically heavy metasomatic agents into the shallow mantle or lithosphere. Some of these materials (e.g., dense mafic-ultramafic cumulates or refertilized lithologies) could subsequently have undergone gravitational foundering or dripping, transferring shallow metasomatic signatures into the asthenosphere, where they were later entrained by convection63. Subduction erosion may also remove fragments from the base of the overriding plate, especially where hydration weakens the lithosphere72,74. The recycled fragments could reside for extended periods in the upper mantle or, in some cases, the MTZ before re-entrainment by mantle flow63. This scenario is compatible with previous proposals that some komatiites sampled hydrous, subduction-related mantle reservoirs21,22. Several independent observations are consistent with this deep-storage scenario for the studied komatiite samples. The three komatiite localities recording the highest δ41K values (Schapenburg, Weltevreden, Boston Creek) are precisely those that, on the basis of high H2O contents and low δ18O in olivine, have been argued to have been derived from deep, hydrous mantle sources21,22,57. Petrologic modeling places their melt initiation depths within or below the MTZ (443 km, 605 km, and an MTZ-rooted plume source, respectively)58. Deep storage is not, however, strictly required by the K isotope data alone. Most hydrous minerals (e.g., amphiboles) in metasomatized mantle break down and devolatilize at the upper-mantle levels80; thus, only a fraction of slab-derived H2O would reach the ~410–660 km depths. This suggests that either subduction venting was especially efficient in the Archean or that our understanding of deep recycling processes is incomplete. In this view, the high δ41K values primarily trace slab-fluid metasomatism rather than uniquely identify storage depth. Slab-derived metasomatism and plume sampling need not have been coeval, because recycled signatures can survive and be tapped long after their initial formation. An alternative, or complementary, mechanism is that ascending komatiite magmas interacted with ancient lithospheric mantle domains that had inherited heavy K from earlier slab-derived metasomatism. Because such domains are strongly enriched in K, selective assimilation or leaching during ascent through thick Archean lithosphere could impart high δ41K to the magma without requiring the entire signal to reside in the primary plume source or in the MTZ. Under this interpretation, komatiites record a slab-related fingerprint not necessarily because they were generated directly from a slab-modified deep source, but because they sampled ancient metasomatized mantle domains during ascent. Both scenarios can explain heavy K isotope signatures in plume-related lavas. Importantly, each requires that surface-derived materials be transferred into mantle reservoirs early in Earth’s history. Mantle plumes later re-sampled these reservoirs, producing hydrous and isotopically distinctive komatiites as young as ~2.7 Ga. The heavy K anomalies (Fig. 5), together with elevated H2O/Ce ratios21,22, therefore, point to long-lived preservation of recycled volatile signatures in the mantle over at least ~0.9 Gyr. Overall, the K isotope data indicate that some Archean mantle domains were metasomatized by surface-derived volatiles by at least the Eoarchean, and perhaps even earlier5,65,70. Together with previous evidence for the existence of hydrous komatiite melts, these observations support subduction-like volatile recycling on the early Earth. They are difficult to reconcile with a purely stagnant-lid regime and instead suggest that crustal materials carrying water and K were being transferred into mantle reservoirs by ~3.6 Ga. These data do not require the existence of modern-style plate tectonics at that time, but they do indicate that efficient lithospheric recycling of volatiles into Earth’s interior was already operating by ~3.6 Ga. These results establish komatiites as a unique archive linking the Archean hydrosphere and Earth’s interior and demonstrating how the surface and deep reservoirs may have become coupled through volatile recycling within the first billion years of Earth's history. Methods Potassium isotope measurements were conducted at the Institut de Physique du Globe de Paris (IPGP). Approximately 15 mg of homogenized bulk powder were dissolved in a mixture of concentrated HNO3 (500 μL) and HF (1.5 mL), and the solution was heated at 120 °C for ~48 h. After evaporation of the HNO3:HF mixture, the residues were treated on a hotplate with a mixture of concentrated HCl (1.5 mL) and HNO3 (0.5 mL) to dissolve fluoride complexes. Subsequently, the solutions were dried down and refluxed with concentrated HNO3 (2 mL), followed by two treatments with 0.5 mol/L HNO3 (1 mL). The sample solutions (in 1 mL 0.5 mol/L HNO3) were then loaded onto Bio-Rad Poly-Prep columns containing 2 mL of Bio-Rad AG 50W-X8 cation exchange resin (200–400 mesh). Four passes of column chemistry were conducted for both standards and komatiite samples. Detailed chemical purification and analytical procedures follow established methods81. Potassium concentrations and isotopic compositions were determined using a Nu Sapphire collision/reaction-cell multi-collector inductively coupled plasma mass spectrometer (CC-MC-ICP-MS). Isotopic ratios were measured using the standard-sample-standard bracketing method, and K isotopic compositions are expressed in delta notation relative to the standard NIST SRM 3141a as follows: δ41K (‰) = [(41K/39Ksample/41K/39Kstandard) – 1]×1000. Isotopic analytical uncertainties are reported as 2 SD. Our results for terrestrial standards BHVO-2 (−0.41 ± 0.06‰, n = 7), BCR-2 (−0.45 ± 0.07‰, n = 5), and AGV-2 (−0.44 ± 0.06‰, n = 6) are in agreement with the previously published values81. The measurements of sample replicates and duplicates also show excellent consistency. A replicate refers to repeated sample dissolution, column chemistry, and instrumental analysis. A duplicate refers to repeated instrumental analysis of the same solution. Supplementary information Supplementary Information Transparent Peer Review file Supplementary Information Transparent Peer Review file