Introduction

Water is essential for partial melting of the continental crust (CC) to produce granites1,2. There are two dominant mechanisms for water supply in this process: decomposition of hydrous minerals in the CC and external water fluxing into melting system3. Recent studies suggest water-fluxed crustal melting is the primary drive to produce granites in the sub-arc region, in which water primarily originates from the dehydration of subducting slabs2. However, a significant volume of granites, formed through water-fluxed crustal melting also occurs within the interior of continent, well beyond the scope of an arc system4,5. A typical example is the North China Craton (NCC) which hosts a large volume of Late Mesozoic granites that formed far distance (>1000 km) from the Paleo-Pacific plate margin (Fig. 1)4,5,6. More interestingly, these intracontinental granites, particularly the Early Cretaceous ones, contain abnormally higher water contents in zircon than typical continental arc magmas7. However, the water origin of these intracontinental granites remains poorly constrained.

Fig. 1: Distribution of Mesozoic granitic rocks in the NCC and Liaodong Peninsula.
figure 1

a Simplified geological map of eastern China. The red front represents the present location of the western Pacific trench. Yellow star marks sampling location of basalt. b Distribution of Mesozoic granites in the Liaodong Peninsula, showing the sampling locations of granites.

Full size image

Hydrogen isotopes are useful to constrain origin of water. Continental arc granites typically exhibit relatively enriched deuterium (D), with primary δD values (~−80‰ to ~−30‰; Supplementary Data 3) greater than that of depleted mantle (δD = −80 ± 10‰; parts per thousand relative to the Standard Mean Ocean Water, SMOW)8 and comparable to metasomatized enriched mantle (−94‰ to −30‰)9. This is due to the release of high δD water by dehydration from subducting slab at sub-arc depth (<~200 km) to overlying mantle wedge and crust. Complementally, the dehydrated slab has low δD9,10,11,12 and accordingly water from this source is isotopically light. Apatite usually crystallizes early in deep-emplacement granitic magmas13,14,15,16 and tends to be trapped as inclusion in other igneous minerals (e.g., feldspar, amphibole, biotite, zircon, etc.)16,17,18,19. It can thus record the original volatile compositions of the magma13,14,15,16,17,18,19.

Here, the in situ hydrogen and oxygen isotope compositions, along with water content, of apatites in Jurassic and Early Cretaceous granites and basalts from the NCC (Fig. 1) were simultaneously measured by secondary ion mass spectrometry (SIMS). The H isotope compositions of biotites and amphiboles in some granites were also determined by TC/EA technique and SIMS, respectively. Major and minor element concentrations (F, Cl, Na, Mn, etc.), and Nd isotope compositions were analyzed by electron probe microanalysis (EPMA) and laser ablation multi-collector inductively coupled plasma mass spectrometry (LA-MC-ICP-MS), respectively, to identify the influence of magma sources and post-crystallization processes. The results show significant variability in δD from −203‰ to −47‰ and contrasting δD values in Jurassic and Early Cretaceous, which are attributable to slab dehydration at different depths. In conjugation with geological records and geophysical investigation, the extremely light δD values suggest a deep origin of water from the previously dehydrated slab, probably stored in the mantle transition zone. This study introduces the use of apatite H isotope analysis as a method to investigate the origin of water in magmatic systems.

Results

Intracontinental granites and basalts formed during the destruction of the NCC

The NCC is bordered by the Central Asia Orogen to the north, while the Dabie-Sulu Orogen separates it from the Yangtze Craton in southeastern China (Fig. 1a). The eastern NCC consists of Archean to Paleoproterozoic basement rocks overlain by unmetamorphosed Mesoproterozoic to Paleozoic sediments. During the Late Mesozoic, the NCC lost much of its lithospheric root20,21, nowadays known as cratonic destruction4. The destruction of the NCC is evidenced by widespread basalts and extensive granitic magmatism over an intracontinental area of ~70,000 km2 (refs. 4,7). Two major sub-provinces of Jurassic (200‒145 Ma) and Early Cretaceous (145‒110 Ma) granites have been identified4, with the latter accounting for ~75% of the entire province. These Early Cretaceous granitic magmatisms are attributed to lithospheric thinning and crustal extension in the eastern NCC4.

About 20,000 km2 of intrusive rocks are distributed in Liaodong Peninsula, along with minor volcanic rocks (Fig. 1b). The apatite phenocrysts, apatite inclusions in zircons (AIZ), and hydrous minerals (i.e., biotite and amphibole) were selected from nine I- and A-type granites of eight well-characterized plutons from this region (Jurassic, 155‒178 Ma; Early Cretaceous, 119‒129 Ma) (Fig. 1b; Supplementary Table 1). Early Cretaceous granites have been subject to more pronounced weathering than those of Jurassic (Supplementary Fig. 1). Additionally, apatite grains from two Early Cretaceous (~120‒124 Ma) fresh basalts of Yixian formation in the adjacent area (Chaoyang and Sihetun) were analyzed (Fig. 1a). The Sihetun basalt (12.1 wt% MgO) that collected from the lower unit is more primary than upper-unit Chaoyang basalt (1.3 wt% MgO), both of which originated from the sub-continental lithospheric mantle (SCLM)22.

Compositions of apatite phenocrysts, inclusions, and hydrous minerals

Most apatite phenocrysts separated from granites and basalts are 100 μm to 200 μm in length with homogeneous or oscillatory zoning in the cathodoluminescence (CL) images (Supplementary Fig. 2). The AIZs that are sufficiently large for micro-probe analysis are only found in Early Cretaceous granites (Supplementary Fig. 2). All apatite grains in granites and basalts from the NCC are fluorapatite (Supplementary Fig. 3a) and have similar variation of volatiles (F, Cl, H2O) (Supplementary Fig. 3). Water contents calculated from oxide contents based on chemical formula of apatite match with those measured by SIMS (0.03‒0.55 wt%) (Supplementary Fig. 3b). The AIZs and apatite phenocrysts from the same samples have indistinguishable F and Cl contents (Fig. 2a). The Early Cretaceous apatites are also characterized by elevated \({{{{\rm{\varepsilon }}}}}_{{{{\rm{Nd}}}}}({{{\rm{t}}}})\) (−14.9 to −8.2) compared to Jurassic grains (\({{{{\rm{\varepsilon }}}}}_{{{{\rm{Nd}}}}}({{{\rm{t}}}})\) −43.6 to −21.3) (Supplementary Data 5).

Fig. 2: Volatile concentrations and hydrogen isotope compositions in apatite phenocrysts, AIZs, and hydrous minerals.
figure 2

a Relationships of volatile contents between AIZs and phenocrysts. Error bars denote statistical 2 SD for each sample. Volatile contents in apatite phenocrysts are nearly identical to that of AIZs in this study. In contrast, AIZs in porphyry deposit44 and complex plutons45 contains more Cl and H2O, but lower F than apatite phenocrysts in matrix. b Apatite δD of granites and basalts from the NCC against zircon U-Pb age. Error bar indicates internal precision (2SE). Gray band denotes the δD range expected for apatites crystallized from granites produced by dehydration melting of continental crust of the NCC; c Comparison of δD of apatite and hydrous minerals obtained in this study with those for continental arc, oceanic arc, stable craton, and SCLM from worldwide. Circles and triangles in Early Cretaceous and Jurassic granites in the NCC represent δD values of hydrous minerals analyzed by SIMS and TC/EA technique, respectively. Because of the modification by meteoric water, weathered biotites and hornblendes display higher δD values than apatites in Early Cretaceous samples. N denotes the analysis number. Abbreviations are as follows: Ap apatite, Amp amphibole, Bt biotite, MI melt inclusion, Ol olivine, Cpx clinopyroxene Opx orthopyroxene. Data sources are provided in Supplementary Data 3.

Full size image

The Jurassic granitic apatite phenocrysts have a δD range of −137‰ to −47‰ with a median of −107‰ (Fig. 2b, c). In contrast, the apatite phenocrysts in Early Cretaceous granites have δD ranging from −203‰ to −127‰ with a median of −160‰ (Fig. 2b, c), which are similar to those contemporaneous basalts (δD = −197‰ to −107‰, median −158‰) but significantly lower than those of Jurassic granites (Fig. 2b, c). The ∆DJurassic-Cretaceous (∆DA-B = δDA-δDB) for all apatites is as large as 150‰, well beyond the range of the external precisions of SIMS analyses (<37‰, 2 SD). The δD values of Jurassic biotite and hornblende range from −54‰ to −107‰, comparable to those of apatites (Fig. 2c). Early Cretaceous hydrous minerals show δD values of −117‰ to −147‰, higher than those of co-existing apatites (Fig. 2c). δD values for Jurassic granitic apatites and hydrous minerals from the NCC overlap with those of typical island arc granites and SCLM peridotites (δD from ~−120 to ~−70‰) (Fig. 2c). The low δD of apatite in Early Cretaceous granites and basalts are similar to those of melt inclusions in olivines of picrite and komatiite (Fig. 2c), which are supposed to represent the hydrogen isotope compositions of dehydrated slab or the mantle transition zone (MTZ)11,12 (Fig. 2c).

The δ18O values of apatite grains from Early Cretaceous granites (3.7–9.1‰) are slightly lower than those of Jurassic granites (4.7–9.9‰) (Supplementary Data 1), with the external precision better than 1‰ (2 SD). Jurassic hydrous minerals analyzed by TC/EA technique have δ18O values of 4.3 ± 0.6‰ to 6.1 ± 0.6‰. Two Early Cretaceous biotite samples have low δ18O values of −0.4 ± 0.6‰ and 1.6 ± 0.6‰, while one biotite sample have δ18O of 5.7 ± 0.6‰ (Supplementary Fig. 4). The low δ18O values of Early Cretaceous biotites, together with their comparatively higher δD relative to co-existing apatite, indicates the incorporation of meteoric water in biotites and hornblende during weathering processes for their layered structure and well-developed cleavages.

Resistance of apatite to water-rock interaction

The fresh basalts, characterized by their rapid solidification and absence of hydrothermal signature, exhibit low δD values that cannot be attributed to water-rock interactions (Fig. 2c). But the hydrogen isotope composition in granites could be altered by hydrothermal fluids. Previous studies have documented markedly reduced δD values (<-230‰ or even <-300‰), often attributable to the involvement of low-δD glacial meltwaters in Antarctica or during syn-Snowball Earth23,24. The paleolatitude of the NCC at Late Mesozoic is 40°–60°N and the average δD for meteoric water at this time in North China is approximately −120‰ (ref. 25). The hydrogen isotopic fractionation between mineral and aqueous fluids, ∆Dmineral-water, can theoretically reach as low as −100‰10. If this extreme fractionation applies to apatite-water system, the exceptionally low δD value (as low as −203‰) in Early Cretaceous granitic apatite (Fig. 2b) could potentially result from interactions with Mesozoic meteoric water.

To assess the viability of this hypothesis, we performed water-rock interaction simulations to emulate such scenarios (Supplementary Fig. 5). The results indicate that, assuming an initial δD of −130‰ for the granite, a water-to-rock ratio (W/R) exceeding 1 wt% is necessary to produce a decrease in δD to values below −150‰ (Supplementary Fig. 5). A lower δD in apatite (or a higher initial δD for granite) requires higher W/R ratios (Supplementary Fig. 5). A W/R ratio of 1 wt% signifies a water-rich environment compared to the typical water content of apatite in Early Cretaceous (~0.04 wt% to ~0.4 wt%). Such conditions would increase water content in both apatite and whole rock26. However, all the studied Early Cretaceous apatites exhibit low δD values, irrespective of their water contents or F/H2O (Fig. 3a; Supplementary Fig. 6a), indicating that water-rock interactions are unlikely predominant mechanism responsible for the low δD characteristics in Early Cretaceous apatites.

Fig. 3: Evaluation of influence of water-rock interaction on apatite.
figure 3

Apatite δD versus (a) H2O and (b) Na2O content. c Zircon \({{{{\rm{\varepsilon }}}}}_{{{{\rm{Hf}}}}}({{{\rm{t}}}})\) plot against apatite \({{{{\rm{\varepsilon }}}}}_{{{{\rm{Nd}}}}}({{{\rm{t}}}})\). The data of unaltered and altered granites is from ref. 32. Solid regression line is for unaltered granites. d, e H isotopic variation from rim (r) to core (c) in Jurassic and Early Cretaceous individual zircon grains. Error bar of δD represents internal precision of single analysis (2SE) for all panels.

Full size image

Furthermore, it is important to emphasize that although the lack of H isotope fractionation coefficient between apatite and water currently hinders us to quantitatively evaluate the influence of hydrothermal fluids on δD of apatites, it is difficult obtain a ∆Dapatite-water as low as −100‰. Such extreme fractionations have only been reported in epidote (400 °C)27 and smectite (25 °C)28. For most mineral-water systems, ∆D is larger than −60‰ (refs. 10,27). Achieving low δD values in Early Cretaceous apatites through water-rock interactions is thus challenging. Moreover, apatite is readily to be altered within days to several tens of days in aqueous fluids at temperature >100 °C29,30, a process that can profoundly change the texture, element concentrations, and Nd isotope composition in apatites16,31,32. Specifically, altered apatites are characterized by jagged zones of varying brightness in CL images, high F/Cl ratios, reduced \({{{{\rm{\varepsilon }}}}}_{{{{\rm{Nd}}}}}({{{\rm{t}}}})\) values and Na2O content16,31,32. The absence of expected correlations between F/Cl ratios and δD, the negative correlation between Na2O contents and δD in apatites, and coupled apatite \({{{{\rm{\varepsilon }}}}}_{{{{\rm{Nd}}}}}({{{\rm{t}}}})\) with zircon \({{{{\rm{\varepsilon }}}}}_{{{{\rm{Hf}}}}}({{{\rm{t}}}})\) (Fig. 3b, c; Supplementary Fig. 6b) hence preclude hydrothermal alteration of the studied apatites.

Kinetic fractionation arising from H and D diffusion during water-rock interactions could result in lighter H isotope composition in minerals, generating an extremely low δD rim33. If it occurs at temperatures below 100 °C, an apatite grain of 200 μm would preserve a measurable diffusion profile of H isotopes over a time scale of 1–10 Ma (Supplementary Fig. 7). Because H diffuses faster than D, kinetic fractionation is expected to lead to substantially lower δD values at the rim compared to the core (∆Drim-core < −300‰) (Supplementary Fig. 7). However, such expected core-rim δD profiles are not observed in individual apatite grains (Fig. 3d).

Hydrogen isotope fractionation in magmatic processes

Experiments suggest that the diffusion rates of F, Cl, and OH in apatite are relatively rapid in magma chambers34, with H-D exchange occurring even more quickly35. This would typically result in a homogeneous distribution of volatiles and hydrogen isotopes within granitic apatites. However, natural samples often deviate from these experimentally defined trends. The halogen and water content of individual apatite grains from granite exhibit a compositional gradient of volatile contents and δD from core to rim, indicative of volatile content variation during magmatic processes13,17. The chemical isolation is a possible mechanism to keep apatite being pristine34,35. In this study, we observed uniform volatile contents and H isotope compositions in intra- and inter-apatite grains within the same granite sample, suggesting equilibration of the volatiles between apatites and granitic melts.

Furthermore, the hydrogen isotopic fractionation between minerals and melts is temperature independent and rather small, with δD variation generally less than 15‰ (refs. 8,26,36). This is corroborated by the overlapping δD values between Jurassic apatite and unweathered biotite and hornblende (Fig. 2c). Consequently, magma differentiation processes were not responsible for the large difference in δD (∆D > 150‰) between Jurassic and Early Cretaceous apatites from the NCC.

In contrast, degassing of magma chamber has been proved to have a substantial influence on hydrogen isotope composition of melt, especially in open systems37. D tends to fractionate into water (∆Dwater-melt > 0), resulting in a D-depleted residual melt after magma degassing37,38. Changes in element content of melts due to magma degassing, similar to water-rock interactions, are characterized by decreasing Na2O, and increasing F/Cl and F/H2O ratio39,40,41. As a result, Na2O and F/Cl (or F/H2O) ratio would positively and negatively correlate with δD, respectively. The data, however, do not reflect these expected trends (Fig. 3a, b; Supplementary Fig. 5).

The AIZs provide more insights into the preservation of original concentration of volatiles that were incorporated at the time of entrapment19,42,43. For example, apatite grains in the matrix contain lower Cl and H2O contents and higher F content in porphyry deposits and plutonic complex than AIZs (Fig. 2a), because Cl and H2O were discharged from matrix apatite during magma degassing whereas AIZs remain intact44,45. Virtually identical volatile contents in apatite inclusions in zircon and in apatite phenocrysts underscore that the magma degassing has not significantly influenced the volatile contents in apatite phenocrysts (Fig. 2a). Early crystallization of apatites and great emplacement depth of granites may be favorable for the preservation of original volatile contents and H isotope compositions in apatite phenocrysts13,14,15,16. Another contributing factor may be that the apatite phenocrysts are also enclosed in other rock-forming minerals such as feldspar and amphibole, preventing them from subsequent modifications16,17,18.

Variable mantle contribution to two suites of granites from the NCC

Given the consideration of the outlined factors, the divergence in δD characteristics between granites of Jurassic and Early Cretaceous is more convincingly explained by distinctions in their sources. Petrological and chemical compositions have revealed the crust-mantle mixing in the source of Late Mesozoic granites in the NCC46. As a result, both the crust and mantle components contribute to the water abundance in these granites. Water contents in the studied granites are estimated to be higher than or comparable to that of continental arc granites7. Due to the deficiency of hydrous minerals in the lower continental crust beneath the NCC47,48, dehydration melting through the breakdown of hydrous minerals in the crust cannot produce such hydrous melts. Therefore, most of water in granitic melts may come from mantle-derived hydrous fluids/melts which underplated in the base of the crust7,49. Positive correlation between zircon \({{{{\rm{\varepsilon }}}}}_{{{{\rm{Hf}}}}}({{{\rm{t}}}})\) and water contents, and the high water contents in the lithosphere mantle beneath the NCC substantiate this proposal7,22,50. Therefore, distinct δD values of two suites of granites from the NCC most likely reflect the temporal changes in the H isotope composition in the source region.

As shown in Fig. 4, the Jurassic apatites define a negative correlation between δD and δ18O, whereas the Early Cretaceous apatites do not show any apparent correlation (Fig. 4). The different correlations are also observed on the plot of apatite δD versus zircon δ18O, \({{{{\rm{\varepsilon }}}}}_{{{{\rm{Hf}}}}}({{{\rm{t}}}})\), and water content (Supplementary Fig. 8). Both suites of granites share a common endmember with CC-like high δ18O (~9‰ to ~10.5‰; Fig. 4) and low \({{{{\rm{\varepsilon }}}}}_{{{{\rm{Hf}}}}}({{{\rm{t}}}})\) (ca. −30; Supplementary Fig. 8b), δD values of which vary from ~−140‰ to ~−110‰ (Fig. 4).

Fig. 4: Plot of δD versus δ18O of apatite from Jurassic and Early Cretaceous granites in the NCC.
figure 4

Shaded fields depict the isotope composition of continental crust, altered oceanic crust10,54, metasomatized mantle I (SCLM)9,10,86, and metasomatized mantle II (MTZ). The δD range of continental crust (−110‰ to −140‰) is derived from the data in this study. The δD range of the MTZ (−200‰ to −289‰) is estimated based on δD values of primitive melt inclusions in komatiites (−181‰ to −289‰), which are proposed to originate from the MTZ11, and the lowest δD observed in granites in this study (~−200‰). Blue curves represent mixing trends between continental crust and metasomatized mantle I for the Jurassic granites. Orange curves represent mixing trends among continental crust, altered oceanic crust, and metasomatized mantle II for the Early Cretaceous granites. Dots on the curves represent 10% increments, and the ratio of hydrogen concentration in A to B (HA/HB) is indicated (cc: continental crust, ac: altered oceanic crust, mm1: metasomatized mantle I, mm2: metasomatized mantle II). Oxygen concentrations in all endmembers are assumed to be identical. Error bars of δD and δ18O represent internal precision of single analysis (2SE).

Full size image

Moreover, the other components involved in Jurassic and Early Cretaceous apatites are characterized by different δD but similar mantle-like δ18O (4.7–5.9‰) of apatite (Fig. 4). Due to very limited D/H fractionation during magmatic differentiation and melting8,36, the similarity of apatite δD in Early Cretaceous granites and contemporaneous basalts further argues for the ingress of external water in granite source, probably from mantle-derived hydrous fluids/melts (Fig. 2c). Specifically, the mantle component involved in the Jurassic granites is more enriched in D (Metasomatized mantle I, δD > ~ −80‰) than in Early Cretaceous granites (Metasomatized mantle II, δD < ~ −200‰) (Fig. 4). Hydrogen isotope composition of the former is consistent with that of SCLM, while that of the latter is close to the dehydrated slab and MTZ values (Fig. 2c). The two metasomatized mantle components also have higher zircon water contents than the CC component (Supplementary Fig. 8c)7, which is in accordance with the occurrence of Late Mesozoic hydrous basalts (aver. 3.4 wt% H2O) in the NCC22,50,51,52.

In addition to two metasomatized mantle components, low-δ18O (<4.7‰) apatites in Early Cretaceous granites may suggest the involvement of minor amount of altered oceanic crust (AOC) in the source of granites53,54. What is more, the presence of apatite with intermediate δD (ca. −130‰) but mantle-like δ18O in Early Cretaceous granites also suggests the involvement of isotopically heavy AOC or metasomatic mantle I in their sources (Fig. 4).

Extremely D-depleted and hydrous MTZ beneath the NCC

The heterogeneous Late Mesozoic mantle beneath the NCC was likely related to metasomatism by hydrous fluids/melts55,56. The metasomatic fluids/melts most likely come from a deeper source, attributable to the dehydration of slab at variable depths, given the pressure-dependent D/H fractionation during slab dehydration10. The average δD of the subducting oceanic plate and the depleted mantle are estimated to be −50 ± 20‰ and −80 ± 10‰8,10, respectively. During subduction, dehydration of the slab at sub-arc depths liberates D-enriched water, which has δD of about −30‰ given for the ∆Dwater-mineral = ~20‰ (ref. 10). This process leaves behind a D-depleted dehydrated slab, which may go beyond the zone of magma generation beneath the volcanic arc. Fluids released by an already dehydrated slab at a greater depth are expected to have isotopically lighter hydrogen10,12,57. The dehydrated slab in the MTZ is predicted to have even lower δD than −230‰10.

Basaltic lavas generated from a source containing dehydrated, recycled oceanic crust usually show arc-like δD values9. Extremely low primitive δD (<−200‰) are only found in melt inclusions of olivines in picrites and komatiites from Lesser Antilles volcanic arc and stable cratons (Fig. 2c)12,58, which may have been sourced from a greater depth, i.e., dehydrated slab and the MTZ. One possible mechanism for the scarcity of low-δD lavas is the rehydration of dehydrated slab surface by fluids from dehydrating isotopically heavy serpentine in cooler, deeper parts of the slab in the depth of >~200 km (ref. 9). Alternatively, upwelling fluids released from dehydrated slab are inevitably contaminated by relatively high-δD water in depleted mantle or metasomatized mantle. The extremely low δD in Early Cretaceous granites and basalts indicate the presence of dehydrated, isotopically light slab beneath the NCC. Although primitive δD of oceanic arc magma from Grenada, Lesser Antilles, can be lower than −140‰ (down to −202‰)12, due to the long distance (>1000 km) of the study area from island arc in Early Cretaceous4,5,6 and the absence of low-δD rocks in Japan arc59,60, the water in the NCC magmas is most likely to come from the MTZ.

Whether the deep mantle can accommodate a large amount of water is a matter of hot debate. Thermodynamic modeling showed that majority of water of subducting slab is lost by sub-arc dehydration61. As a result, dehydrated slab going further down become essentially dry and carry only small amount of water62. However, the MTZ and the lower part of the upper mantle beneath the NCC have the highest electrical conductivity in the world63, consistent with an exceptionally hydrated mantle in this region22,50,51,52. Such a peculiar scenario is probably related to the continuous release of water from the stagnant slab within the MTZ to overlying upper mantle50,64. A hydrous MTZ is further suggested by experiments that show major constituent phases of the MTZ can accommodate much higher water than shallow mantle minerals65, and by hydrous ringwoodite found in deep-originated diamonds66. A similar scenario can be envisaged for the eastern NCC during the late Mesozoic, as a big mantle wedge structure may have initiated as early as in the Early Cretaceous67,68,69,70,71.

Water from Paleo-Pacific slab dehydrated at different depths

The recently reconstructed Paleo-Pacific subduction history4,67,68,69,71 provides a plausible tectonic framework that helps us understand the different apatite δD values in Jurassic and Early Cretaceous granites. It has been suggested that progressive shallowing of the subducting Paleo-Pacific Plate in the Middle-Late Jurassic was followed by subsequent slab rollback in Early Cretaceous4,6,67,69. Relatively flat subduction of the Paleo-Pacific Plate beneath the NCC in Jurassic expelled the asthenospheric mantle between the lithosphere and subducting slab. The cooling effect caused by the cold slab beneath the lithospheric mantle limited the dehydration of the subducted slab at sub-arc and underneath the lithosphere (Fig. 5a)4. As a result, a metasomatized mantle with arc-like δD was generated, and only a small amount of water with intermediate δD fluxed into the crust (Fig. 5a). This explains the relatively small volume of Jurassic granites and their arc-like δD values (Fig. 2c). Although water that existed in the mantle prior to subduction could potentially have H isotope compositions (δD −94‰ to −30‰)9 similar to those of arc magmas, its low water content (<1 wt% in basalts)9 makes it unlikely to be dominant in the water that fluxed into the crust from the mantle.

Fig. 5: Schematic illustration explaining the variation in δD in Jurassic and Early Cretaceous granites from the NCC and its relationship with the Paleo-Pacific Plate subduction processes.
figure 5

Blue arrows denote water transportation. a Moderate dehydration of the cold paleo-Pacific plate at shallow depth when it subducts flatly along the craton’s lithospheric root make the deeper slab not markedly D-depleted, thus the water released from deeper slab, and the resultant metasomatized mantle and granites are moderately depleted in D. b In Early Cretaceous, hot subduction generate extremely D-depleted dehydrated slab in the MTZ due to substantial dehydration at shallow levels, resulting in an extremely isotopically light metasomatized mantle and granites.

Full size image

In Early Cretaceous, Liaodong Peninsula was located farther from the trench due to trench retreat4,5,67,69. Meanwhile, the Paleo-Pacific Plate started to roll back widening the asthenospheric window beneath the NCC (Fig. 5b). Subducting slab became stagnated within the MTZ, which released a large amount of extremely D-depleted water into and metasomatized overlying SCLM. Water fluxing triggered the melting of the SCLM generating the Early Cretaceous basalts22. Underplating of hydrous mafic magmas at the crust-mantle boundary introduced a large amount of water into the lower crust, producing voluminous granites with an extremely low δD (Fig. 5b).

While this study favors that the extremely low δD values of apatites in Early Cretaceous granites of the NCC ultimately reflect a deep mantle origin of water, alternatives such as water-rock interaction and magmatic degassing cannot be fully ruled out at this stage. The role of these processes in D/H isotopic fractionation deserves further investigation. Following research strategies are desirable. (1) Comparing δD values of hydrothermally altered apatites against those of pristine magmatic apatites in granites is necessary to elucidate the influence of hydrothermal alteration on δD. (2) If the interpretation presented in this study is valid, Late Mesozoic granites and basalts further west from the trench should have similar low hydrogen isotopic compositions. Consequently, hydrogen isotopic analyses on a much broader selection of Late Mesozoic granites and mantle-derived rocks from diverse geological settings will help build a comprehensive perspective on δD as a geochemical tracer. (3) Examining the hydrogen isotopic composition of hydrous minerals (such as hornblende) in Cenozoic primitive mantle xenoliths and their host lavas, with a particular focus on those influenced by the subducted Pacific slab, could provide insights to the fluid processes in the big mantle wedge underneath east Asia. Some mantle components with extremely low δD are expected from this region.

Methods

SIMS analysis of Apatite H, O isotope composition and water contents

Apatite and zircon grains with apatite inclusion were embedded in random orientations to Sn-Bi alloy mounts for SIMS analysis, following the method of Zhang et al.72. All separated apatite phenocrysts and apatite inclusions in zircon were documented with CL images to reveal their internal textures and choose appropriate areas (free of inclusions, fractures, and surface contaminations) for SIMS analysis. Analyses were performed using a CAMECA IMS 1280-HR at the Guangzhou Institute of Geochemistry, Chinese Academy of Sciences (GIGCAS). The vacuum of the analysis chamber remained at ~1.7 × 10−9 mbar. A primary Cs+ beam current of about 4.0 nA was used. A uniform flat-bottom sputter crater was ensured by using a 15 μm raster73. The water background can be reduced to <10 ppm or even to 1.2 ppm74,75.

The 16O, 16O1H, 18O, 17O1H, 16OD signals were simultaneously counted in a single acquisition which was divided into 3 sequences. The value of mass resolving power (MRP) used to separate 16OD from 17O1H was set at 15,000. The measuring of 17O1H was to check whether 16OD was measured correctly or 17O1H and 16OD are completely separated. Each measurement consisted of 16 cycles, with a total analysis time of approximately 12 min, including pre-analysis sputtering and peak centering73.

The water content of apatite was determined from the measured 16O1H/16O ratio, and a calibration curve ([H2O] = 121.78×[16O1H/16O]; R2 = 0.9991) was constructed by the analyses of previously characterized two apatite standards, Kovdor apatite (H2O = 0.98 wt%) and Durango apatite (H2O = 0.0478 wt%)73,76. The two relative standard errors (2RSE) is ~4% that include internal precision of single analysis and uncertainties in linear regression. Measured 16OD/16O and 18O/16O values were normalized to the Vienna Standard Mean Ocean Water composition (SMOW, 16OD/16O = 155.76 × 10-6, 18O/16O = 2.0052 × 10-3). Corrections for instrumental mass fractionation (IMF) on hydrogen and oxygen isotopic compositions of apatite were performed using the Durango apatite standard (δD = −86 ± 4‰, δ18O = 9.8 ± 0.3‰)76 and monitored by analyzing Qinghu apatite with recommended δ18O of 5.6 ± 0.2‰77. We measured the H isotope composition of Qinghu apatite using TC/EA technique, and its δD value is −110 ± 3‰ (Supplementary Data 2). The internal precision of δD for single analysis of SIMS ranges from 5‰ to 34‰ (2SE). High water content apatite usually shows better precision. The external precisions (i.e., reproducibility) of δD for Durango apatites are 21‰ (2 SD, n = 16) and 37‰ (2 SD, n = 18) in two alloy mounts. Both the internal and external precisions of δ18O are <1‰ (2SE or 2 SD). δD and δ18O of Qinghu apatite that calculated from total 19 analysis spots across two alloy mounts are −114 ± 32‰ (2 SD) and 5.7 ± 1.3 (2 SD), in agreement with recommended values.

EPMA analysis of apatite element concentrations

In situ major element analyses were obtained using a CAMECA SX Five field emission Electron Probe Microanalyzer (EPMA) at GIGCAS. The mounts were coated with carbon before analysis. Concentrations were quantified using the PAP (Pouchou and Pichoir) matrix correction procedure78. Point analysis of Si, Mg, Fe, Mn, Ca, Na, P, S, F, and Cl of separated apatite phenocrysts were acquired using an electron beam of 15 kV accelerating voltage, 20 nA beam current, and 20 μm beam size. Si, Ca, P, F and Cl of apatite inclusions in zircon were acquired using an electron beam of low energy (15 kV accelerating voltage, 5 nA beam current) with a beam size of 5 μm. The detailed configurations of instrument and analytical procedures were described in He et al.79. Stoichiometry calculation for H2O in apatite is based on F and Cl contents from EPMA analysis, according to the structural formula of apatite80, i.e., 26 oxygen for the OH-poor apatite in this study (mole fraction of OH < 0.5). Calculated water contents match well with SIMS-measured water contents except for the high water contents (> ~0.4 wt%) which may be affected by analysis error that propagated from major elements (e.g., Ca, P, F) and simplification algorithm in stoichiometry calculation80.

LA-MC-ICP-MS analysis of apatite Nd isotope

In situ Nd isotope compositions of apatites were obtained by a Neptune Plus multi-collector inductively coupled plasma mass spectrometry (MC-ICP-MS) equipped with a NWR 193 nm ArF Excimer laser-ablation system at the Guangzhou Tuoyan Analytical Technology Co., Ltd., Guangzhou, China. The spot size is 70 μm dependent on Nd signal intensity. Other parameters are the same as analyses of trace elements. Eight Faraday cups receive 142Nd, 143Nd, 144Nd, 145Nd, 146Nd, 147Sm, 148Nd and 149Sm at the same time. The mass discrimination factor for 143Nd/144Nd was determined using 146Nd/144Nd (0.7219) with the exponential law. The 149Sm signal was used to correct the remaining 144Sm interference on 144Nd, using the 144Sm/149Sm ratio of 0.22332. The mass fractionation of 144Sm/149Sm was calibrated by the 147Sm/149Sm ratio of 1.08680 and exponential law. Data reduction was conducted using “Iso–Compass” software81. Two natural apatite megacrysts, Durango and MAD, were used as the unknown samples to verify the accuracy of the calibration method for Nd isotope analysis of apatites. The 147Sm/144Nd and 143Nd/144Nd obtained for Durango and MAD standards were 0.0902 ± 0.0055 (2 SD) and 0.512401 ± 0.000077 (2 SD); 0.0867 ± 0.0066 (2 SD) and 0.511235 ± 0.000053 (2 SD), respectively, consistent with their reference values82.

Bulk analysis of H, O isotope, and water content of biotite, hornblende and Qinghu apatite

Samples were weighed into silver capsules, degassed for 1 h at 100 °C, then crushed and loaded into a zero-blank autosampler. The hydrogen isotopic composition was measured using a MAT 253 Stable Isotope Ratio Mass Spectrometer coupled to a Thermo Scientific TC/EA High Temperature Conversion Elemental Analyzer, with a precision for δD of 3‰ (2 SD).

Oxygen isotope analysis were conducted at the Queen’s Facility for Isotope Research, Queen’s University, Canada. Samples were acidified prior to oxygen extraction to remove carbonates using 20% HCl at room temperature until no reaction was observed. Samples were rinsed 3x using 18.2 MW deionized water and dried in an oven overnight at 100 °C. Oxygen was extracted from ca. 5 mg of treated silicate sample at 550–600 °C according to the conventional BrF5 procedure of Clayton and Mayeda83. Samples and standard reference materials (NBS28 and an in-house basalt) were loaded into Ni bombs with excess BrF5 and heated overnight. The liberated oxygen was converted to CO2 via sublimation of a carbon rod (EMS, CVP grade) and analyzed using the multiport of a dual inlet system on a Thermo-Finnigan Delta Plus XP Isotope Ratio Mass Spectrometer (IRMS). Sample gas was measured against a reference gas calibrated using a suite of international standards (NBS18, NBS19, NBS20, and NBS23) for 8 cycles, at 2500 mV intensity, in CO2 gas configuration using Isodat 3.0. A 17O correction using the SSH algorithm was applied84 to the measured values. The precision for δ18O is 0.6‰ (2 SD).

SIMS analysis of hornblende H isotope composition

The hornblende grains were also embedded in Sn-Bi alloy mounts. The in-situ hydrogen isotope analysis of hornblende was conducted by using the CAMECA IMS 1280-HR at GIGCAS. A Cs+ primary beam of 1.4–2.4 nA with an impact energy of 10 keV was used to sputter secondary ions from the samples. A normal-incidence electron gun (diameter ∼ 100 μm; 0.3–0.8 μA current flow from the sample holder) was used to compensate for the charge. Negative secondary ions were extracted and accelerated through a 10 kV with a 400 μm contrast aperture. The size of the analytical area was 30 × 30 μm (15 μm spot size + 15 μm rastering). The energy slit was closed to a bandwidth of 50 eV width and shifted 5 eV below the maximum transmission. A 122 μm entrance slit, 608 μm exit slit, and 100× transfer optical magnification were used to guide the secondary ions. A peak jump model sequentially measures the mass of H, D, and 28Si, and 10 cycles were collected, making a single analysis last 11 minutes. The Faraday cup with a 1012 Ω resistor (mono-collector system) was used to detect H and 28Si, while the electron multiplier (EM) (mono-collector system) was used to detect D. Before each analysis session, the Faraday cup and the electron multiplier high voltage were calibrated to ensure their consistent yielding. The waiting time and counting time for H, D, and 28Si in each mass scan cycle are 6, 1.2 and 2 s and 10, 15 and 2 s, respectively. Peak centering was performed before each analysis by centering the peak of 1H, which was also used as the internal reference signal. The pre-sputtering process lasted for 180 s with a 25 × 25 μm raster area, which was larger than the analysis area, to minimize the water background signal. Appropriate pre-sputtering in an ultra-high vacuum can attain the limit of detection of less than 1 ppm. Instrument mass fractionation (IMF) for H isotopes are corrected by standard hornblende samples, Kipawa (δD = −88‰)85. Other additional hornblende standards, Bamble (δD = −61‰) and Seljas (δD = −64‰)85, were repeatedly analyzed during data acquisition for quality control. Internal precision is better than 15‰ (2SE). The external precision of δD for Kipawa is 15‰ (2 SD, n = 13). The δD values obtained for Bamble and Seljas are −78 ± 13‰ (2 SD, n = 12) and −75 ± 15‰ (2 SD, n = 12), which are consistent with recommended values within the error.