Introduction

Volcanic arcs are key localities for element cycling and mass transfer between Earth’s surface and interior, which greatly affects crust-mantle evolution and Earth’s long-term climate change. Compared to MORB (mid-ocean ridge basalts), arc magmas are enriched in large ion lithophile elements (LILE; e.g., Ba, Cs) and depleted in high field strength elements (HFSE; e.g., Th, Nb). The enrichment of LILE is generally attributed to the incorporation of slab-derived materials into the mantle wedge1. However, the nature and transport mechanisms of these slab-derived materials remain debated. Generally, three scenarios have been proposed: (1) melting of mantle wedge metasomatized by aqueous fluids from altered oceanic crust (AOC) and/or hydrous melts from sediment2,3; (2) melting of mantle wedge metasomatized by hydrous melts from both AOC and sediment4,5,6,7,8,9; (3) melting of mélange diapir, which is formed by the mechanical mixing between slab materials and mantle peridotites in subduction channels and diapirically rise into the mantle wedge10,11,12,13. Thus, delineating the applicability of such mechanisms is crucial for understanding the genesis of arc magma.

A re-evaluation of the long-standing interpretation regarding the variation of Ba or Ba/Th in arc lavas could help resolve these controversies. Traditionally, ratios of fluid-mobile to fluid-immobile elements like Ba/Th in arc magmas were regarded as a typical indicator of the contribution from AOC-derived fluids3,14. However, significant variations of Ba/Th in arc magmas were recently ascribed to differences in melting temperature or composition of subducted sediment among different arcs6,7,8. For example, high-pressure/high-temperature experimental results showed that the Ba/Th ratio of sediment melt decreases with increasing temperatures (750–1050 °C)15. As such, Ba/Th variation in arc magmas was ascribed to sediment melts formed at different temperatures6. However, arc lavas with high Ba/Th (e.g., Izu arc; up to 1000) do not show slab-surface temperatures beneath the arc significantly different (<100 °C) from those with low Ba/Th (e.g., Lesser Antilles arc; <200)16,17,18. Moreover, based on significant Ba/Th variations in different types of subducted sediments, recent studies suggested that the Ba/Th variation in arc magmas reflects spatial variation of sediments with varying Ba/Th8. However, it is difficult to constrain the spatial variation in the bulk composition of subducted sediments.

Alternatively, the Sr-Nd-B-Mg isotopic variation of arc lavas was proposed to be incompatible with mixing between mantle and sediment melts and/or AOC fluids, but is consistent with the mixing of bulk sediment or serpentinite and mantle (i.e., mélange)11,12. The overlap of trace element characteristics (e.g., Ba/Th) between mélange and arc magmas further supports that direct melting of mélange could produce such signatures of arc magmas10. However, the eclogitized oceanic crust and the underlying hydrated mantle may be too dense to ascend as diapirs into the mantle wedge. Instead, these materials might remain at subarc depths and then release fluids to hydrate the overlying mantle. Therefore, elucidating the mechanisms driving the variation of Ba or Ba/Th in arc magmas is critical for constraining the nature and transport mechanisms of slab-derived materials (fluid, melt, or mélange diapir).

Barium content and ratios to other elements are susceptible to magma differentiation, such as partial melting and fractional crystallization, while Ba isotopes exhibit minimal fractionation during these processes19,20 and are distinguishable in different geological reservoirs. The Ba isotopic compositions are expressed as δ138/134Ba (δ138/134Ba = [(138Ba/134Ba)sample/(138Ba/134Ba)SRM3104a − 1] × 1000‰). δ138/134Ba of low-temperature AOC (0.24 ± 0.21‰, 2 SD)21 are generally higher than those of depleted MORB mantle (DMM; 0.03 to +0.14‰)13,22,23. Fluids released from AOC dehydration are also suggested to have heavier Ba isotope compositions24,25. In contrast, seafloor sediments typically exhibit lighter Ba isotopic compositions (δ138/134Ba = 0.02 ± 0.10‰, 2 SD)22,26,27. Consequently, Ba isotopes serve as a potentially excellent tool for tracing the source of Ba in subduction zones28.

The Izu-Bonin-Mariana (IBM) arc system serves as a typical example of an intra-oceanic subduction zone with a cold slab17. The Izu arc is located in the northern section of the IBM system and exhibits marked across-arc volcanic distribution29,30. Studies on chemical compositions of the across-arc volcanic chains could minimize the effects of spatial variations in subducted components, making the Izu arc an ideal location for constraining the nature and transport mechanisms of slab-derived materials.

In this study, we conducted whole-rock Ba isotope analyses on volcanic rocks from the Izu arc with varying volcano-to-slab depths, as well as the offshore sediments near the Izu trench (IODP Site 1149). The results indicate that the high Ba/Th and δ138/134Ba in global arc magmas (including those in the Izu arc) reflect a dominant contribution from AOC-derived fluid rather than sediments. By integrating Ba-Sr-Nd isotopic and trace elemental data from global subduction zones, we propose a model for the generation of arc magma in cold subduction zones, i.e., mélange ascend as bulk diapirs into the mantle wedge, which are subsequently metasomatized by AOC-derived fluids to form the source of arc magmas. This model highlights the combined role of AOC-derived fluids and the mélange melting.

Results and discussion

Results

The Izu arc, located at the northeastern edge of the Philippine Sea Plate (PSP), is formed by the subduction of the Pacific Plate beneath the PSP and stretches over 480 km between 30.5°N and 34.5°N (Fig. 1). It is a typical intra-oceanic arc system with negligible inputs from sub-arc crust to the lavas and is also a non-accreting convergent margin with a thin sedimentary cover on the downgoing slab29,30,31,32,33. The across-arc chain studied here extends to the southwest from Oshima and includes three rear-arc volcanoes behind the volcanic front, Toshima, Niijima, and Kozushima (Fig. 1b)29,30,32.

Fig. 1: Geological setting and sampling map of the Izu arc, and location of IODP Site 1149 (modified from ref. 35).
Fig. 1: Geological setting and sampling map of the Izu arc, and location of IODP Site 1149 (modified from ref. 35).
Full size image

a Overview of the Izu-Bonin-Mariana arc. b Volcanoes of the Izu arc from which samples were collected. Frontal arc volcanic islands are denoted by blue markers, rear-arc volcanic islands by orange markers30. Bathymetric data are plotted with GeoMapApp software (www.geomapapp.org).

This study analyzed δ138/134Ba of twenty-six samples from the frontal-arc regions and eleven samples from the rear-arc regions. The δ138/134Ba of these arc lavas range from −0.05 to +0.15‰ (Table S1). Samples from frontal-arc regions were collected from volcanic islands from Oshima to Torishima (Fig. 1), which are mainly late Quaternary tholeiitic basalts or basaltic andesites. These samples exhibit δ138/134Ba values ranging from 0.03‰ to 0.15‰, which almost overlap with mantle range (0.03‰ to +0.14‰) (Fig. 2a). No significant δ138/134Ba variations were observed along the arc front (Figure. S1). Samples from the rear-arc regions were collected from volcanic islands extending from Oshima toward the rear-arc direction (e.g., Toshima; Fig. 1b), which are late Quaternary basaltic to rhyolitic lavas. Their δ138/134Ba values are between −0.05‰ and 0.07‰, overlapping with mantle range but extending to lower values (Fig. 2a). Notably, δ138/134Ba values and Ba/Th decrease with increasing depths to the Wadati-Benioff zone (Fig. 2).

Fig. 2: Across-arc distribution of δ138/134Ba and Ba/Th of the Izu arc lavas.
Fig. 2: Across-arc distribution of δ138/134Ba and Ba/Th of the Izu arc lavas.
Full size image

Plots of δ138/134Ba (a) and Ba/Th (b) versus depths to Wadati-Benioff zone71. The range of depleted mantle (DMM; 0.03 to 0.14‰) is from ref. 22,23. The error bar in Fig. 2a represents the long-term external precision of δ138/134Ba (±0.04 ‰).

Barium isotope data of offshore sediments from IODP Site 1149 in the western Pacific Plate are presented in Table S2. The core is located at a proximal geographic position to the Izu-Bonin trench (Fig. 1), likely rendering it an ideal representation of the Ba isotope input of sediments to the Izu arc34,35,36. The core is approximately 440 m long, and the studied sediment samples are mainly volcanic ash, pelagic clay, radiolarian chert, and nannofossil marl (Table S2)34,36,37. Those samples show high and variable Ba concentrations (14.7–2717 µg/g, Table S2) compared to DMM (ca. 1.2 µg/g)38. δ138/134Ba of these sediments are −0.02 to 0.08‰ (except a radiolarite sample of −0.12‰; Fig. S2 and S3 and Table S2), with a weighted mean of +0.02‰ (weighted to interval mass and Ba concentrations).

Tracing the source of Ba in the Izu arc

All volcanic samples are fresh, with no clay minerals30. Low loss-on-ignition (LOI) values (<2 wt.%) and chemical index of alteration (CIA) values (40–52)39,40, as well as no significant correlation of these indices with δ138/134Ba (Fig. S4), indicate the negligible effect of alteration on Ba isotopes. Compared with data from subaerial arcs that do not record significant alteration13, we do not observe higher Th/Rb and Th/Cs ratios for any samples in our study (Fig. S5). Therefore, all samples investigated here are unlikely to have experienced significant alteration. Moreover, δ138/134Ba values show no systematic correlation with SiO2 or Mg# [molar ratios of Mg2+/(Mg2++Fe2+)], suggesting that magmatic differentiation has minimal impact on Ba isotope composition (Fig. S6). Therefore, the Ba isotopic characteristics of the Izu arc lavas primarily reflect the compositions of mantle source affected by slab-derived materials. Traditionally, Ba/Th ratios in arc magmas were regarded as a typical indicator of the contribution from AOC-derived fluids3,14. However, recent studies attributed Ba/Th variation in arc magmas to sediment melts formed at different temperatures6, or spatial variation in composition of sediments with varying Ba/Th8. Therefore, clarifying the source of Ba in the Izu arc lavas, especially the causes for large Ba/Th and δ138/134Ba variations, could help resolve these controversies.

The frontal arc samples exhibit remarkably high Ba/Th that decrease toward the rear-arc region (Fig. 2b). Given that slab temperature increases with greater slab depth, it may lead to the fractionation of Ba from Th (i.e., a decrease of Ba/Th) in sediment melts with increasing slab-depth6,15. Therefore, we need to evaluate whether the variation in Ba/Th results from the fractionation of Ba and Th in sediment melts at varying temperatures. First, at the slab-depth range from the frontal arc to the rear-arc regions (100–130 km; Fig. 2), the temperature variations of the slab are limited (<100 °C)16,17,18, which would result in a minor decrease in Ba/Th (<2 times)15. Therefore, it cannot account for the wide range of Ba/Th (100–1000; Fig. 2b) observed in the Izu arc lavas, which is comparable to the global range for arc magmas. Moreover, although the increase in melting temperatures of sediment may reduce Ba/Th in sediment melts, partial melting at high temperatures is anticipated to have a limited effect on Ba isotope fractionation in sediment melts19. Therefore, the variation in melting temperatures of sediment cannot account for the remarkable decrease in both δ138/134Ba and Ba/Th with slab depths observed in the Izu arc lavas.

On the other hand, the sediments subducting into the Izu arc are characterized by terrigenous sediments in the upper layers with low Ba/Th and pelagic/carbonate sediments in the lower layer with high Ba/Th (Fig. S2 and Table S2)36,37. Their mixing may affect the Ba/Th variation in Izu arc samples, as previously proposed to explain the Mariana arc data8. However, both upper sediments and lower sediments have similar δ138/134Ba despite their largely variable Ba/Th (17–2813; Figs. S2 and S3), therefore, their mixing cannot explain the covariation of δ138/134Ba and Ba/Th in the Izu arc lavas. Moreover, the decrease in both δ138/134Ba and Ba/Th with increasing slab depth is difficult to explain by mixing between the high-Ba/Th lower sediments and the low-Ba/Th upper sediments, as this scenario would require preferential incorporation of the lower sediments in the frontal arc, followed by the incorporation of the upper sediments at greater depths in the rear arc. Similarly, the across-arc decrease in Ba/Th in the Mariana arc41 cannot be explained by the mixing of lower pelagic sediments with high Ba/Th and upper volcaniclastic sediments with low Ba/Th, given the similarity of sedimentary lithologies subducting beneath the Mariana arc to those subducting beneath the Izu arc34,36.

The Izu arc samples with high δ138/134Ba and Ba/Th (>500) exhibit notably high 143Nd/144Nd (>0.51306; Fig. 3a), indicating that their high Ba/Th cannot be attributed to sediment-dominated components. Global data show that arc lavas with low Ba/Th tend to exhibit lower 143Nd/144Nd and higher 87Sr/86Sr (Fig. 3), clearly indicating sediment contributions37. In contrast, arc lavas with high Ba/Th show relatively higher 143Nd/144Nd and lower 87Sr/86Sr (Fig. 3), consistent with AOC-derived fluid signatures34.

Fig. 3: Ba/Th versus Sr-Nd isotope compositions for global arc lavas.
Fig. 3: Ba/Th versus Sr-Nd isotope compositions for global arc lavas.
Full size image

Plots of Ba/Th ratios versus 143Nd/144Nd (a) and 87Sr/86Sr (b). Arc lavas from cold and hot subduction zones are shown in circular and square symbols, respectively. The data are from the GEOROC database, and the compiled dataset are provided in Supplementary Data S1.

Thus, we conclude that AOC-derived fluids, rather than sediment melts, constitute the primary source of the high Ba/Th and δ138/134Ba characteristics in arc magmas. In most subduction zones, the oceanic crust component dehydrates to release hydrous fluids, rather than partially melting1,2,3.

Tracing the transfer mechanism of Ba in the Izu arc

Boron isotope ratios (i.e., δ11B) and B/Nb in the Izu and many other arc lavas show an across-arc decrease, with δ11B showing roughly linear correlations with B/Nb as well as Sr or Pb isotopes29,42,43,44,45,46,47. Therefore, the across-arc decrease in δ11B was considered to result from an across-arc decrease in relative proportion of slab-derived fluids compared to the mantle29,42,43,47, coupled with B isotope fractionation during progressive fluid loss from the slab44,45,46. Similar to B isotopes, Ba/Th and δ138/134Ba of the Izu arc lavas exhibit across-arc decreases, with δ138/134Ba showing positive correlations with 87Sr/86Sr and 143Nd/144Nd (Fig. 4). The high δ138/134Ba end-member has a Sr-Nd isotope composition similar to AOC, which most likely reflects the contribution of AOC fluids34. As boron and Ba are both fluid-mobile elements, the across-arc decrease in δ138/134Ba of the Izu arc lavas also suggests a reduction in the contribution of AOC-derived fluids with increasing slab depths, which is consistent with the positive correlations between δ138/134Ba and 87Sr/86Sr, Ba/Th, Cs/Nb, and Pb/Nb (Fig. 4a and S7a–c).

Fig. 4: Barium-Sr-Nd isotope variations of the Izu arc lavas and mixing modeling between DMM (deleted-MORB mantle), metasediment, and AOC fluid.
Fig. 4: Barium-Sr-Nd isotope variations of the Izu arc lavas and mixing modeling between DMM (deleted-MORB mantle), metasediment, and AOC fluid.
Full size image

Plots of δ138/134Ba versus 87Sr/86Sr (a) and 143Nd/144Nd (b). Yellow areas represent calculated mixing fields between DMM-type mantle and bulk metasediment; Numbers denote the proportion of bulk sediment in mélange by weight; Green areas represent the calculated mixing fields between mélange and AOC fluid. Details of the modeling calculations can be found in Supplementary Material Text S1 and Table S3. The inset figures are horizontal-axis extensions of their respective main plots and present the Ba isotope fractionation of sediments during the subduction process. The pre-subduction sediment data are from IODP Site 1149 near the Izu arc34,36,37. Error bars for δ138/134Ba represent 2 standard deviations (SD) of each sample.

Previous studies have linked the decrease in δ11B to more mantle-like Sr and Pb isotope ratios, suggesting that the low δ11B end-member points to a mantle source with B isotopic compositions consistent with a DMM signature29,42,43. However, a pure DMM component cannot explain the Ba isotopic composition of the mantle end-member for the Izu arc lavas, as it has a significantly lighter Ba isotopic composition (ca. –0.10‰) than DMM (0.03–0.14‰) (Fig. 4). To address this problem, we propose an alternative scenario. The low δ138/134Ba end-member shows slightly enriched signatures in both Sr and Nd isotope compositions relative to DMM, most likely representing a mantle source hybridized by subducted metasediments. In other words, a small amount of sediment initially mixed with DMM, imparting it with a slightly enriched Sr-Nd isotope composition and sediment-like δ138/134Ba values (Fig. 4). This is likely because Basediment/BaDMM ratio is much higher than the corresponding ratios for Nd and Sr37,38. This hypothesis is supported by the positive correlation between δ138/134Ba and 143Nd/144Nd and the negative correlation between δ138/134Ba and Th/Yb. That is, lower δ138/134Ba samples exhibit lower 143Nd/144Nd and higher Th/Yb (Fig. 4b and S7d).

The δ138/134Ba of offshore sediments near the Izu trench (Site 1149) range from –0.02 to +0.08‰ (Fig. S2 and Table S2), which is higher than the expected metasediment component (ca. –0.1‰). However, previous studies on other arc lavas and OIB-type rocks suggested that the significant Ba isotope fractionation may occur during the subduction of sediments, resulting in lower δ138/134Ba (<0‰) of the metasediments at subarc depths compared to the protolith13,20,27,48,49. Such a scenario is also supported by the low δ138/134Ba (–0.16 to –0.05‰) of phengites, a primary host of Ba in metasediments, from eclogites in the Dabie ultrahigh-pressure metamorphic belt, China24,25. Therefore, metasediments beneath the Izu arc with low δ138/134Ba (ca. –0.1‰) are likely to be formed owing to the Ba isotope fractionation during slab subduction. AOC-derived fluids are subsequently added to such sediment-hybridized mantle source, forming the observed positive correlation trends. Our mixing modeling indicates that the addition of <0.8 wt. % bulk sediment into the mantle source could account for the characteristics of the low δ138/134Ba end-member (Fig. 4).

Barium isotope data from two other classic cold subduction zones (Tonga-Kermadec and Mariana arcs) also support this conclusion50,51. In δ138/134Ba vs. 87Sr/86Sr and 143Nd/144Nd plots, the mantle–sediment mixing line passes through the data array defined by most Kermadec lavas and North Mariana lavas (Fig. 5), which suggests that these lavas could be explained by mixing between mantle peridotite and sediments. In contrast, the data arrays defined by samples from the Tonga arc and the South Mariana arc require a more dominant contribution of AOC fluid with high δ138/134Ba (Fig. 5). Critically, the other end of that array extends to their intersection with the mantle-bulk sediment mixing lines (Fig. 5), implying that sediment components are incorporated into the mantle prior to the addition of AOC fluids.

Fig. 5: Barium-Sr-Nd isotope variations of the Tonga-Kermadec and Mariana arc lavas, and mixing modeling between mantle, metasediment, and AOC fluid.
Fig. 5: Barium-Sr-Nd isotope variations of the Tonga-Kermadec and Mariana arc lavas, and mixing modeling between mantle, metasediment, and AOC fluid.
Full size image

Plots of δ138/134Ba versus 87Sr/86Sr and 143Nd/144Nd for the Tonga-Kermadec arc lavas (a, b) and the Mariana arc lavas (c, d). Ba isotope data of the Tonga-Kermadec and Mariana arc lavas are from ref. 50,51. Yellow areas represent calculated mixing fields between DMM-type mantle and bulk metasediment; Numbers denote the proportion of bulk sediment in mélange by weight; Green areas represent the calculated mixing fields between mélange and AOC fluid. Details of the modeling calculations can be found in Supplementary Material Text S1 and Table S3.

The decoupling of Sr-Nd isotopes and their systematic relationships with Ba and Th in arc lavas from cold subduction zones further supports this model (Fig. S8). Backarc lavas (the lavas formed in association with a spreading center) are generally distributed along a “backarc” array (the DMM-bulk sediment mixing line), whereas arc magmas deviate from this array (Fig. S8). In contrast, the array defined by arc magmas lies between a point on the DMM-bulk sediment mixing line (backarc array) and AOC fluid. Notably, as arc lavas deviate farther from the “backarc array”, their Ba/Th increase (Fig. S8). This further indicates that an AOC fluid is added separately to a mantle source that has been hybridized by sediments. Therefore, both Ba isotope and Ba elemental evidence support the idea that, in cold subduction zones, bulk sediments are incorporated into the mantle prior to the addition of AOC fluids.

Consequently, the most likely model to explain the arc magma composition is as follows: mélange containing sediment components formed at the slab-mantle interface migrate into the mantle at depths shallower than subarc regions (i.e., forearc regions), and subsequently AOC fluids are added individually at subarc depths. Previous studies suggest that mélanges have lower densities than mantle rocks at depths <3 GPa (ca. 100 km), allowing them to migrate as diapirs into the mantle wedge10,52,53,54. The quantitative modeling of this process by considering the mélange formed by the hybridization of bulk sediment and DMM, which then experienced the metasomatism of AOC-derived fluids and partial melting, can explain not only the Ba-Sr-Nd isotope compositions but also the trace element compositions of the Izu arc lavas (Fig. S9), lending further support to the proposed model.

Partial melting of sediment-bearing mélange or sediment at the slab-mantle interface is another possibility that needs to be evaluated to explain these arc magma compositions. However, the slab surface temperature at depths shallower than subarc regions (<90 km) in cold subduction zones like the Izu arc (<700 °C)16,17 is probably insufficient to induce the melting of mélange or sediment (>750–800 °C)15,55. If AOC fluids need to be added to the mantle in a separate form, it also requires that mélange or sediment does not melt at the subarc depths and shallower depths. Moreover, mixing of sediment melts with a large range of partial melting (1% and 20%) with the mantle is inconsistent with the Sr-Nd isotope data of backarc lavas (Fig. S8). While some Sr-Nd isotope data for arc lavas may be explained by sediment melts (Fig. S8), the differences between arc and backarc lavas cannot be accounted for by the addition of sediment melts and bulk sediments, respectively. This is because: (1) the slab surface beneath the backarc regions has higher temperatures than that beneath the arc, which favors sediment melting; and (2) the density of bulk sediments at >4 GPa depth could be too high for diapirism to occur10,52,53,54. Therefore, although the melting of mélange or sediment along the subduction channel may occur at forearc depths, the mélange diapir is the most feasible scenario to explain our Ba isotope data.

A model for arc magma formation

The formation of arc magmas has long been attributed to the metasomatism of the subarc mantle by AOC-derived fluids and/or sediment melts. However, two competing models have been proposed: (1) melting of mantle metasomatized by AOC melts and sediment melts4,5,6,7,8,9, and (2) melting of mélange diapirs10,11,12. One key to resolving the debate lies in elucidating the origins of Ba or Ba/Th variation in arc magmas. This study reveals that the high δ138/134Ba and Ba/Th in Izu arc and other arc magmas originate from the contribution from AOC fluid, rather than from sediment melt. Moreover, the synchronous positive correlation between δ138/134Ba and 87Sr/86Sr and 143Nd/144Nd requires the incorporation of mélange diapirs and AOC fluid into mantle at different stages. Integrating the coupling between Ba/Th and Sr-Nd isotopes in global arc magmas, we propose a revised model. In cold subduction zones, mélanges ascend as diapirs at shallower depths (probably at the slab-mantle interface) into mantle wedge, which are subsequently metasomatized by AOC-derived fluid at subarc depths, triggering the formation of arc magmas (Fig. 6). Such mélanges might have experienced melting to some extent in the hot mantle wedge prior to the addition of AOC-derived fluids11,56. However, this process may not produce a melt that is significantly different in composition from the melt we envisioned. In hot subduction zones, AOC largely dehydrates at shallower depths, which may make mélange/sediment-derived signals predominant in Ba isotopic features of arc lavas, such as the Aleutian and Ryukyu arcs27. Previous numerical simulations predicted the occurrence of mélange at the slab-mantle interface, which can ascend to the hot mantle wedge in a diapiric form57,58,59,60,61. In addition, seismological studies, especially through the method of high-resolution seismic tomography, revealed the occurrence of zones with low-velocity, high-attenuation, and strong seismic anisotropy just beneath active arc volcanoes62,63, which is likely a mélange consisting of sediment and peridotite. Such a process is supported by the investigation of a deeply buried mélange “package” in Kyrgyz Tianshan, southern Altaids, which are formed at subarc depths, and subsequently propagate and dynamically mix with overlying mantle64.

Fig. 6: A schematic diagram of the model showing the transfer mechanism of slab-derived materials in cold subduction zones.
Fig. 6: A schematic diagram of the model showing the transfer mechanism of slab-derived materials in cold subduction zones.
Full size image

(1) Bulk sediments incorporate other components such as mantle peridotite and AOC to form mélanges, which ascend as diapirs into overlying mantle at shallower depths. (2) Subsequently, AOC (and/or slab serpentinite) releases aqueous fluids beneath the arc, triggering the melting of mélanges and the formation of arc magmas.

This model breaks through the limitations of traditional single-stage melt/fluid processes by coupling mélange diapirs with AOC fluid contributions, thereby supporting the “multi-stage, multi-mechanism” paradigm of subduction-zone mass transfer. The mélanges in the subduction channel are enriched in volatiles and may act as a temporary sponge of water, carbon, and sulfur at the slab-mantle boundary. Therefore, the separate dehydration of these slab materials could control the pathways and budgets of the volatiles into the mantle wedge or the deep mantle. Quantitative modeling of the spatial distribution and mixing efficiency between AOC fluids and mélanges in the mantle wedge is helpful for further validation through thermodynamic simulations and geophysical evidence.

Methods

Barium isotope measurements were conducted at the Metal Stable Isotope Geochemistry Laboratory of the University of Science and Technology of China (USTC). Approximately 10–100 mg of rock powders were decomposed using a 1:3 mixture of concentrated HNO3 and HF for one to two days. Subsequently, they were dried down and fluxed with aqua regia overnight to remove fluorides. Samples were then dried down and refluxed with 2 mL 3 mol·L−1 HCl to convert nitrate compounds to chloride. Finally, the samples were dried down and dissolved in 1 mL of 3 mol·L−1 HCl for further purification. The above digestion and evaporation procedures were conducted on a hotplate with temperatures of 120 to 140 °C. Barium was separated from the matrix using a Bio-Rad 200-400 mesh AG50W-X12 resin and two column procedures according to an ion-exchange chromatographic method of ref.65,66. Matrix elements were eluted with 3 mol·L−1 HCl, and Ba was then concentrated through 3 mol·L−1 HNO3. After purification, Ba concentrates were dried and then dissolved in 2% (m/m) HNO3. The yield of this purification procedure was >99%. The total procedural blanks were <1 ng. Barium isotopes were analyzed on a Neptune Plus multi-collector inductively coupled plasma mass spectrometer (MC-ICP-MS). Barium isotopic ratios were measured using the double-spike (135Ba-136Ba) method in low-resolution mode under “dry” plasma conditions using an Aridus III desolvator (CETAC Technologies)65,66,67. The results are reported in δ138/134Ba values relative to the National Institute of Standards and Technology (NIST) Standard Reference Material 3104a [δ138/134Ba = 1000 × (138/134Basample − 138/134BaSRM-3104a)/138/134BaSRM-3104a]. All samples were analyzed three times, yielding 2 SD uncertainties of 0.00‰ to 0.05‰. The long-term external precision is better than ±0.04‰ for δ138/134Ba (2 SD) based on the measurement of two in-house standards (USTC-Ba and ICPUS-Ba)65,66. The reference materials were processed with every set of unknowns and displayed Ba isotopic compositions in agreement with previous work (Table S5)65,66,68,69,70.