The identification of an eroded fore-arc crust component in arc magmas is challenging due to the combined effects of mantle metasomatism and crustal assimilation–fractional crystallization. In this study, molybdenum (Mo) isotope compositions are used in conjunction with Sr-Nd-Hf isotopic and elemental data to identify eroded fore-arc crust components in adakites from the Cuyapo and Balungao volcanoes of the northern Bataan segment of the Luzon arc (Philippines). The Mo isotopic ratios (δ98/95Mo, relative to the NIST SRM 3134 standard) of these adakites increase with increasing εNd (+4.3 to +5.6) and Ba/Nb (206–286). The low δ98/95Mo (−0.36‰ to −0.26‰) in the Cuyapo adakites coupled with low Sr-Nd-Hf isotopic ratios suggests contributions from the residual slab, which lost isotopically heavy Mo during dehydration. Interestingly, the high δ98/95Mo (−0.18‰ to 0.00‰) Balungao adakites have Mo-Sr-Nd-Hf isotopic ratios similar to those of the Luzon basement. Fractionated Nb/Ta (16–18) and high Sr/Y indicate the coexistence of melt with residual rutile and garnet ± amphibole assemblages, corresponding to a source region (>~45 km) below the present Luzon crust (~33 km). This thus suggests an origin of heavy Mo from partial melting of eroded crust in the mantle wedge rather than in the upper-plate crust. Our work not only demonstrates that Mo isotopes may be a potential tracer of eroded crust but also highlights that lavas with combined high δ98/95Mo, εNd, and Ba/Nb emplaced at subduction zones with juvenile arc crust may be a result of subduction erosion.

Arc magmatism plays a key role in the growth of continental crust at convergent margins (Rudnick and Gao, 2014). In contrast, sediment subduction and subduction erosion (i.e., the transfer of components derived from the fore-arc crust into the mantle wedge) play a destructive role in maintaining the dynamic balance of continental crust growth (Stern and Scholl, 2010; Gómez-Tuena et al., 2018; Stern, 2020; Straub et al., 2020). However, distinguishing eroded crust from components derived from the subducted basaltic crust and marine sediments in arc magmas is challenging because of source (e.g., mantle metasomatism) or crustal contamination (e.g., crustal assimilation–fractional crystallization, AFC) (Straub et al., 2020). The study of subduction-related adakites provides an opportunity to investigate the components involved in arc magma genesis given that recent research shows that they may represent direct melts of eroded fore-arc crust (Goss and Kay, 2009) or the subducted remnant slab (Defant and Drummond, 1990). Crustal AFC may likewise produce adakitic magmas (Castillo, 2012). The recently developed study of Mo isotope systematics may help to identify the various components because Mo may be fractionated dramatically during dehydration melting and fractional crystallization, given that the heavier Mo isotopes preferentially enter the liquid phase whereas the lighter ones are retained in the solid phases (e.g., rutile) (Voegelin et al., 2014; Freymuth et al., 2015; Chen et al., 2019). It is thus predicted that (1) if the studied adakites represent melts of eroded fore-arc crust, they should have heavy Mo iso-tope compositions that were inherited from the previous cycles of subduction dehydration; (2) if they represent the remnant slab at greater depth that underwent progressive dehydration, they should have inherited the light Mo; and (3) if they represent intracrustal differentiation of normal basaltic magmas, they should have heavy Mo isotope compositions due to fractional crystallization. We report Mo isotope compositions coupled with Sr-Nd-Hf isotopic and elemental data for adakites from the Cuyapo and Balungao volcanoes and basalts from the Amorong volcano of the northern Bataan segment of the Luzon arc (Philippines), with the aim of distinguishing contributions from eroded fore-arc crust versus components derived from the subducted slab.

The Luzon arc, forming añ1200 km chain of Miocene to recent stratovolcanoes and volcanic necks extending from Taiwan (24°N) to Mindoro, Philippines (12°N), is a result of the eastward subduction of the South China Sea along the Manila Trench (Fig. 1) (Defant et al., 1990). Miocene to Pleistocene adakitic lavas in the northern Bataan segment and western Northern Luzon (Yumul et al., 2003) are attributed to the subduction of the Scarborough Seamount Chain (ca. 13–7 Ma; Zhang et al., 2017) and the fossil South China Sea spreading center (Fig. 1; ca. 17–15 Ma; Li et al., 2014a) beneath the Luzon crust. Based on gravity data, the crustal thickness under the region is <~33 km (Parcutela et al., 2020). The Luzon arc comprises two belts in the study area: a Miocene-Pliocene western volcanic chain and a Quaternary eastern volcanic chain (Balungao, Cuyapo, and Amorong; Fig. 1) (Yumul et al., 2003). The basement of both arcs is thought to be composed mainly of Cretaceous mafic-ultramafic rocks similar to the Zambales ophiolites (Fig. 1) (Encarnación et al., 1999).

Bulk-rock element and isotope compositions were measured for a suite of adakites, basalts, and lower-crustal xenoliths from northern Bataan and Northern Luzon that were previously partly studied for geochemistry (Liu et al., 2020). Detailed descriptions of the geologic setting and the petrography of the samples are provided in Item S1 in the Supplemental Material1.

Bulk-rock element and Mo-Sr-Nd-Hf isotope compositions are presented in Table S1 in the Supplemental Material, and in Figures 24. Mo isotope compositions were measured following the method described by Li et al. (2014b) at the State Key Laboratory of Isotope Geochemistry, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences. Details of the analytical methods are provided in Item S2.

The basalts of Amorong are relatively primitive with 49–52 wt% SiO2 and Mg# (Mg/[Mg + Fe]) = 60–70 (Fig. 2; Table S1). Sr/Y values are in the range typically observed for arc magmas (<40; Defant and Drummond, 1990), Nb/Ta values (10–16) are generally close to those of mid-ocean ridge basalt (MORB; 11–16; Goss and Kay, 2009), and Ba/Nb values (102–134) are higher than that of the depleted mantle (5.71; Salters and Stracke, 2004) (Figs. 2 and 3). Amorong basalts have low δ98/95Mo (-0.31‰ to -0.25‰, relative to the NIST SRM 3134 standard) and a limited range of Ce/Mo (25–27), compared to the elevated δ98/95Mo values (>-0.20‰) and wide range of Ce/Mo (6–47) found for lavas of the Izu-Mariana arc (Fig. 3A; Freymuth et al., 2015). Amorong Sr-Nd isotope compositions (87Sr/86Sr = 0.70355–0.70362, εNd = +4.7 to +5.7) are similar to those of the Scarborough Seamount Chain basalts but Amorong εHf (+12.4 to +13.1) is higher (Fig. 4).

The Cuyapo and Balungao adakites have 61–67 wt% SiO2, with high Sr contents (655–992 ppm) and Sr/Y ratios (61–763), low Y (1.1–11.0 ppm) and Yb contents (0.1–1.2 ppm), variable Mg# (37–58), and fractionated Nb/Ta ratios (16–19) with respect to MORB (Fig. 2; Table S1). The δ98/95Mo values (-0.36‰ to 0.00‰) of northern Bataan adakites essentially correlate negatively with Ce/Mo (38–10) but positively with εNd (+4.3 to +5.6) and Ba/Nb (206–286) (Fig. 3). Specifically, Balungao adakites have low Ce/Mo (10–38) and high δ98/95Mo (-0.18 to 0.00‰) within the ranges of the Izu-Mariana lavas and lower crustal xenoliths of Luzon (Fig. 3A). Figure 4A illustrates a positive correlation between the Sr and Nd isotopes for the northern Bataan adakites. The low-εNd end member falls into the field of the Scarborough Seamount Chain basalts whereas the high-εNd end member points toward the Luzon basement rocks (Fig. 4A). The Cuyapo lavas have lower εHf (+12.4 to +12.9) comparable to that of the Amorong basalts, whereas the Balungao lavas have higher εHf (+13.6 to +13.9) closer to that of the basement rocks (+14.1 to +16.0) (Fig. 4B).

It has been suggested that the adakitic signatures, i.e., high Sr and low Y and heavy rare earth element (REE) contents, may not be the result only of melting of subducted materials but also of crustal AFC (Castillo, 2012). In the latter case, a negative correlation between δ98/95Mo and Mg# would be expected because fractionation of amphibole and/or garnet, required to explain the low Y and heavy REE contents, would enrich the residual liquid in heavy Mo (Voegelin et al., 2014). This is not observed (Fig. 2B). The Cuyapo lavas have low Mg# (<48), indicative of fractional crystallization. However, their δ98/95Mo values are comparable to those of the Amorong basalts, the most primitive lavas in the region (Mg# >60) (Fig. 2). This suggests that the Mo isotope composition of the Cuyapo lavas is not significantly affected by fractional crystallization. Thus, the Mo isotope composition of northern Bataan adakites may reflect the nature of their sources, and the variation might be the result of mixing between isotopically different magmas. Figures 3 and 4 show that the northern Bataan adakites can be modeled by two components: (1) a high-Ce/Mo component having low Mo-Sr-Nd-Hf isotopic ratios as in the case of Cuyapo, and (2) a low-Ce/Mo component with high Mo-Sr-Nd-Hf isotopic ratios represented by the Balungao lavas (Items S3 and S4).

Detecting the Subducted Components

It has been proposed that the dehydrated slab and partial melts and/or fluids derived thereof would become progressively depleted in 98Mo and higher in Ce/Mo as subduction progressed (Fig. 3A) (Freymuth et al., 2015; Chen et al., 2019). Such fractionation is observed in the Izu-Mariana arc lavas, which exhibit lower Ce/Mo and higher δ98/95Mo than MORB (Ce/Mo = ~32, δ98/95Mo = -0.21‰) (Fig. 3A), as a result of the addition of fluids and/or melts released from the subducted slab (Freymuth et al., 2015), while eclogites derived from subducted crust show higher Ce/Mo (mainly >30) and lower δ98/95Mo (mainly <-0.33‰) than MORB, suggesting loss of heavy Mo during dehydration (Chen et al., 2019).

The descending slab is located ~120 km below the Cuyapo and Amorong volcanoes (Fig. 1) and is therefore subjected to eclogite-facies metamorphism. Thus, the low δ98/95Mo and high Ce/Mo with respect to MORB (Bezard et al., 2016) of Cuyapo and Amorong lavas (Fig. 3A) may result from an input of the remnant slab. Compared to the depleted mantle, the Cuyapo and Amorong lavas have enriched Sr-Nd-Hf isotopes (Fig. 4), and thus a second component is required. Fluids and/or melts derived from the subducted Scarborough Seamount Chain could be this component given that the Cuyapo and Amorong lavas have Sr-Nd isotopes similar to those of the Scarborough Seamount Chain basalts (Fig. 4). However, the decoupling of Nd-Hf isotopes does not support this hypothesis. Specifically, the Cuyapo and Amorong data are parallel to the “seawater array” (Fig. 4B) (Albarède et al., 1998) in Nd-Hf space, which strongly suggests the involvement of marine sediments. Calculations show that a mixture of South China Sea MORB and sediment melts (85:15) could essentially reproduce the Sr-Nd-Hf isotopes of the Cuyapo and Amorong lavas (Fig. 4). Because Nd and Hf are melt-mobile elements (Kessel et al., 2005), the high Ce/Mo and low Mo-Sr-Nd-Hf isotope compositions of the Cuyapo and Amorong lavas are thus interpreted as a result of basaltic oceanic crust plus sediment melt input in their source. The lack of “garnet signature” in the Amorong and Cuyapo lavas, i.e., low (Dy/Yb)N (normalized to chondrite values) (<1.2) (Fig. 2C), suggests their source region is much shallower than the descending slab. This may result from a meltperidotite interaction forming pyroxene and amphibole veins, which then melt to produce basaltic to andesitic melts at shallower levels; during this process, the middle to heavy REEs are buffered by the mantle wedge (Castillo, 2012; Straub et al., 2020).

In contrast, the elevated δ98/95Mo and low Ce/Mo of the Balungao adakites cannot be inherited from the remnant South China Sea slab, which is argued to release light Mo as noted above. Because crustal differentiation likewise is not a viable option, it is suggested that the heavy Mo may be inherited from the upper-plates crust, given the isotopic similarity between the Balungao adakites and Luzon basement rocks (Figs. 3 and 4). Crustal melting in the upper plate, however, is unlikely given that the Balungao ada-kites have elevated Nb/Ta ratios with respect to MORB (Fig. 2C), which is likely caused by residual high-Mg amphibole or residual rutile, given that both minerals have DNb/DTa < 1 (D is the partition coefficient between mineral and melt) during partial melting (Münker et al., 2004). Residual high-Mg amphibole cannot account for the fractionated Nb/Ta because it would yield melts with low (Dy/Yb)N (<1) due to its high DDy/DYb (>1) (Tiepolo et al., 2007), which is not observed for the Balungao lavas (Fig. 2C). It is thus argued that residual rutile accounts for the fractionated Nb/Ta. This, combined with the elevated Sr/Y ratios of the Balungao lavas, indicates the coexistence of the melt with residual rutile and garnet ± amphibole assemblages. These assemblages stabilize only at pressure >1.5 GPa (Xiong et al., 2005), corresponding to a >~45 km depth, greater than the depth of the present Luzon lower crust (<~33 km). Thus, partial melting of basement rocks within the mantle wedge is required. Such an interpretation is supported by the high Mg# of the adakites (mainly >48) (Fig. 2B; Table S1), suggestive of acquisition of primitive-mantle signatures via melt-mantle interaction (Rapp et al., 1999). The fore-arc crust materials were most likely dragged down into the mantle wedge via subduction erosion, given that a thickened crust, which is required for delamination, is absent in Luzon.

The presence of crustal signatures in the northern Bataan lavas suggests melting of subducted crust (Gómez-Tuena et al., 2018). Whether this is the result of melting of mélangediapir in the hot mantle wedge (Nielsen and Marschall, 2017; Parolari et al., 2018) or of fluid-fluxed melting induced by dehydration of the subducting lithosphere serpentinites (Chen et al., 2019; Li et al., 2021) is still a matter of debate. The key to distinguishing mélange melting from fluid-fluxed melting is finding out whether the resulting lavas have recorded the severe heavy Mo extraction during subduction dehydration, because in the former scenario, significant partitioning of elements and stable isotopes occurs during the melting of mélangediapir, whereas in the latter case, significant elemental and isotopic partitioning is confined to the subducting slab (Ryan and Chauvel, 2014; Nielsen and Marschall, 2017). The Balungao adakites might be generated by mélange melting, given that their maximum δ98/95Mo (0.00‰), representative of the sample least affected by magma mixing, suggests insignificant heavy Mo loss. On the other hand, the crustal signatures of Cuyapo and Amorong lavas were most likely generated by fluid-fluxed slab melting, given that they exhibit lighter Mo isotope composition with respect to MORB, consistent with severe heavy Mo extraction (Item S4).

Implications for Subduction Erosion

Our interpretation that the high-δ98/95Mo Balungao adakites indicate subduction erosion provides new insights into mass transfer in subduction zones with juvenile arc crust. The positive correlations between δ98/95Mo versus εNd and Ba/Nb documented for the Izu-Mariana arc lavas are widely considered as the signatures of aqueous fluids. These fluids are more significant when they are added to a more depleted and refractory mantle wedge (Villalobos-Orchard et al., 2020). Compared to the Cuyapo and Amorong lavas, the Balungao lavas have higher δ98/95Mo, εNd, and Ba/Nb (Fig. 3), which cannot be due to increasing aqueous fluid input to the depleted mantle. Instead, it is best interpreted as the result of partial melting of eroded fore-arc crust in the mantle wedge. Specifically, melting of eroded crust in the presence of rutile and amphibole would also produce melts with higher Ba/Nb, given that both minerals have DBa/DNb < 1 (Tiepolo et al., 2007; Xiong et al., 2005). Moreover, the “fluid signatures” can be enhanced if the eroded crust already has elevated δ98/95Mo and Ba/Nb. Thus, caution must be taken when estimating the flux of these elements (e.g., Mo, Nd, Nb, and Ba) in subduction zones with juvenile arc crust.

This work was supported by the National Natural Science Foundation of China (NSFC; grants 41890812, 41973011, and 41625007) and the Key Special Project for Introduced Talents Team of Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou) (GML2019ZD0202). This paper has been greatly improved by constructive comments from editor Marc Norman, Mattia Parolari and an anonymous reviewer. This is contribution No. IS-3241 from the Guangzhou Institute of Geochemistry, Chinese Academy of Sciences (GIGCAS).

1Supplemental Material. Geological background, sample information, methods, and supplemental discussion and tables. Please visit to access the supplemental material, and contact with any questions.
Gold Open Access: This paper is published under the terms of the CC-BY license.