Recycling of Marine Carbonate Induced Calcium Isotope Heterogeneity of Arc Magmas in Subduction Zones

Calcium (Ca) is an essential element constituting sedimentary carbonate in subducting sediments. Ca isotopic characteristics of subduction-related rocks could provide insight into the behavior and budget of carbonate and carbon cycles in subduction zones, due to the distinctive δ 44/40 Ca ranges of sedimentary carbonate with respect to the mantle. Here, we studied the Ca isotopic compositions of arc magmas from the Northern Luzon arc (NLA), which are evolved from a depleted mantle metasomatized by slab-derived fluids and sediment melts. The δ 44/40 Ca values range from 0.76 ± 0.04‰ to 1.01 ± 0.03‰ and cover the typical ranges for bulk silica earth (BSE, ~ 0.94‰) and fresh mid-ocean ridge basalt (MORB, ~ 0.83‰). The Ca isotopes of NLA volcanics are not dominantly determined by the effects of mantle partial melting or fractional crystallization, nor significantly modified by secondary alteration. Instead, the δ 44/40 Ca values of NLA volcanics are controlled by the subduction-related metasomatism. The metasomatism by slab-derived fluids (mainly expelled from altered oceanic crust, AOC) dramatically elevated the contents of fluid-mobile elements (e.g., Ba and Pb) with respect to fluid-immobile elements (e.g., Ce). This process, however, rarely modified the Ca isotopes, possibly ascribed to the δ 44/40 Ca similarity between AOC and the depleted mantle. The δ 44/40 Ca values significantly correlated with subduction indicators (e.g., Sr-Nd isotopes, Ba/Nb, Ce/Pb, and Nb/La), demonstrating the Ca isotopes of NLA volcanics are mainly controlled by the metasomatism of sediment melts subducting from the South China Sea (SCS). Based on the thermal structures and chemical compositions of sediments subducting into global trenches, we propose that carbonate Ca isotopic signals can only be observed in the arcs with high sedimentary Ca fluxes and temperature-pressure conditions well beyond the solidus of H 2 O-saturated sediment melting, e.g., NLA, Nicaragua, Guatemala, Colombia, Peru, South Chile, North Vanuatu, New Zealand, and Kermadec. The absence of such signals in other arcs suggests either limited sedimentary fluxes or much of the subducting sedimentary carbonate has been survived during plate subduction to enter the deep mantle. The melting of DMM was firstly metasomatized by slab-derived fluids at shallow depth, which process changed the δ 44/40 Ca values by a negligible extent but dramatically enriched the magma with fluid-mobile elements (e.g., Ba, Sr, and Pb). Then, the arc magma was metasomatized by sediment melts by different degrees, evidenced by the significant correlations between δ 44/40 Ca values and Sr-Nd isotopic compositions and subduction indicators (e.g., Ba/Nb, Sr/Nb and Nb/La ratios). Our findings contradict with previous works on BABBs and arc magmas, where they reported δ 44/40 Ca values dominantly determined by mantle

of China. The operation conditions were: accelerating potential of 15 kV, beam current of 20 nA, counting time of 20 s, and spot size of 10 μm. Natural mineral standards and ZAF correction scheme were used for calibration.

Major and trace elements
All the geochemical analyses including analyses on major and trace elements, Sr-Nd isotopic compositions, and Ca isotopes are conducted at the Wuhan Sample solution Analytical Technology Co., Ltd., Wuhan, China.
The rock samples were grounded into 200 mesh and dried for 12 h in a temperature-controlled oven at 105 o C. Then 1.0 g dried sample was accurately weighted and heated in a muffle furnace at 1000 o C for 2 h. The weight difference after heating is used to calculate the loss on ignition (LOI). About 0.6 g sample powder was mixed with 6.0 g cosolvent (Li2B4O7 : LiBO2 : LiF = 9:2:1) and 0.3 g oxidant (NH4NO3) in a Pt crucible. The mixture was heated at 1150 o C for 14 min. The obtained melt was quenched with air for 1 min to produce flat discs on the fire brick and was subsequently determined for major elemental contents using an X-ray fluorescence (Primus II, Rigaku, Japan). Measurements on duplicate samples and standard samples (GBW07104, GBW07105, and GBW07111) demonstrate measurement uncertainties of < 5%.
Trace elemental analysis of the rock samples were conducted on Agilent 7700e ICP-MS after the sample powder was fully digested by HNO3 and HF solutions. The measurement quality was controlled by duplicate analyses, blank samples, and analyses on standard samples (AGV-2, BHVO-2, BCR-2, and RGM-2). The overall measurement uncertainties are less than 5%.

Ca isotopes
Ca isotopic analyses of Lutao volcanics were performed on a Neptune Plus MC-ICP-MS (Thermo Fisher Scientific, Dreieich, Germany) equipped with a quartz dual cyclonic-spray chamber (Elemental Scientific Inc., USA) and a 50 μL min −1 PFA MicroFlow Teflon nebulizer (Savillex, USA). Polyatomic interference such as 40 Ar 1 H 2+ and 14 N 3+ was resolved by medium resolution mode. An in-house Alfa Ca standard solution (Lot: 9192737) was used as a bracketing reference standard. All Ca isotope results were reported relative to NIST SRM915a by adding a conversion factor of 0.58 (the δ 44/42 CaSRM915a value of Alfa Ca) (Eq. 1). The instrumental precision and accuracy during routine Ca isotope measurements were monitored by intermediate measurements of NIST SRM915a. The accuracy and reproducibility were controlled by analyses of a calcium carbonate standard NIST SRM915b, a basalt standard BHVO-2, and standard seawater samples. (1) In order to compare our results with previous studies, the δ 44/42 Ca values in this study are converted to δ 44/40 Ca using the following equation (Gussone et al., 2016): The long-term (>3 months) average δ 44/42 Ca of NIST 915a is 0.001 ± 0.058‰ (2 SD, n = 155), indicating the reproducibility of the instrument is better than 0.06‰ (2 SD). The repeatedly analyzed NIST SRM 915b, BHVO-2, and seawater are 0.40 ± 0.04‰ (2 SD, n = 3), 0.37 ± 0.00‰ (2 SD, n = 3) and 0.94 ± 0.04‰ (2 SD, n = 3), respectively. These results are consistent with previous studies within analytical uncertainty, confirming the accuracy of our analytical method for Ca isotopes in geological samples (Amini et al., 2008;Heuser and Eisenhauer, 2008;Feng et al., 2018). In addition, all the measured Ca isotopic compositions follow the theoretical mass-dependent fractionation line between δ 43/42 Ca and δ 44/42 Ca with a slope of 0.506 ( Fig.   S1) , further demonstrating the reliability of our Ca isotopic analyses.
variations of major elements (Table S1) All the volcanic rocks present large ranges for trace elements and REEs (Table S1). Both the trace element and REE distribution patterns are comparable with previously published data on NLA volcanics ( Fig. S3) Lai et al., 2017   .

Discussion
The Quaternary Lutao volcanism occurred within a short period of time, which developed the Ameishan  Chen et al., 1994;Shao et al., 2014). The Lutao volcanic rocks are far more evolved than a direct melt of the mantle wedge beneath the island arc on accounts of the significantly lower MgO contents than the primary magmas and lower Ni and Cr contents than the least-evolved island arc basalts (Perfit et al., 1980). Both the significantly enrichments of LILEs (e.g., Cs, Rb, Ba, and Pb) with respect to HFSEs and the LREEs enrichment over HREEs (Fig. S3) suggest the Lutao volcanics are typical arc magmatic products Lai et al., 2017). Considering the geological setting of NLA ( Fig. 1), the enrichment of LREEs and LILEs could be attributed to the subduction-related metasomatism by sediment melts and/or slab-derived fluids Fourcade et al., 1994).
Therefore, the ~ 0.25‰ variation of Ca isotopic compositions in the Lutao arc magmas could be ascribed to either source effects and/or the evolution of the arc magma during magmatic/post-magmatic processes (e.g., partial melting, fractional crystallization, or seawater alteration) (Huang et al., 2011;John et al., 2012;Antonelli and Simon, 2020;Chen et al., 2020b). In addition to these mass-dependent fractionation processes, the Ca isotopic characteristics could be potentially affected by the radiogenic decay of 40 K(τ1/2 = 1.277 × 10 9 a) (Fantle and Tipper, 2014). However, this process should have provided negligible contribution for the 40 Ca of the Lutao arc magmas because our Ca isotope results were calculated from measured 44 Ca/ 42 Ca ratios (Eq. 1). This statement is further supported by the low K/Ca ratios (mean 0.14) and young ages (< 3 Ma) of the studied arc magmas .

Negligible influence of secondary alteration
The arc magmas may have been secondary altered by seawater or hydrothermal fluids. Because chemical weathering and/or fluid alteration on the igneous rocks would result in the hydration of glasses and formation of hydrous minerals, LOI is an essential indicator of the alteration degree (Banerjee and Chakrabarti, 2018 (John et al., 2012;Blättler and Higgins, 2017;Valdes et al., 2019;Chen et al., 2020a).
The whole rock samples of the Lutao arc volcanics present LOI values of < 2 wt.%, suggesting the secondary alteration is quite limited. Limited secondary alteration is also testified by lacking Ce anomalies (Ce/Ce* in the Lutao volcanics, 0.96 -1.02) and significant correlations between Sm, Sr, and Nd (Polat and Hofmann, 2003;Valdes et al., 2019). The correlation between LOI and δ 44/40 Ca values is not significant, indicating secondary alteration has placed negligible influence on the Ca isotopic characteristics of Lutao arc magmas (Fig. 3). This statement is further supported by the ratios between alteration sensitive elements (e.g., Rb, Pb, U, and Sb) andinsensitive elements (e.g., heavy REE, Y, and TiO2) (Jochum and Verma, 1996). The correlations between δ 44/40 Ca and Rb/TiO2 and U/Th ratios show vague decreasing trends (i.e., increasing degree of alteration) (Fig. 3). The correlation coefficients, however, are not significant. Furthermore, the Lutao volcanics display chemical index of alteration (CIA) values within a narrow range and demonstrate in-significant correlation with the corresponding δ 44/40 Ca values. Therefore, secondary alteration is not responsible for the Ca isotopic variation of Lutao arc magmas.

Fractional crystallization
The Lutao volcanics show SiO2, MgO, and CaO contents covering wide ranges (Table S1), possibly ascribed to fractional crystallization during the cooling of the primary basaltic magma. According to the typical phenocrysts and their chemical compositions in the volcanic rocks (Table S2 and (Fig. S4). Plg is unlikely to be a major accumulating phase because significant Eu anomalies (Eu/Eu*) are absent in all Lutao volcanics, since Eu is strongly partitioned into Pl with respect to other REEs (Vukadinovic, 1993 Heavier Ca isotopes are enriched in phases with shorter bonds and smaller coordination numbers (Chacko et al., 2001;Huang et al., 2010Feng et al., 2014). As a result, Opx and Ol, which possess shorter Ca-O bonds, are isotopically heavier in Ca than co-existing Cpx (Huang et al., 2010;Kang et al., 2016Chen et al., 2019;Valdes et al., 2019), while Hbl and Pl are enriched in lighter Ca isotopes with respect to Ol and pyroxenes Valdes et al., 2019). However, the fractional crystallization of Ol and Opx could not significantly change the δ 44/40 Ca values of the residual magma, due to their extremely low CaO contents with respect to Cpx, Hbl, and Pl Chen et al., 2019). Zhu et al. (2018) suggested the δ 44/40 Ca of the residual melt would be only decreased by < 0.03‰ even when the magma has experienced 50% fractionation of Ol and Opx.
Furthermore, the accumulation of Ol and Opx was not significant during the magma cooling. Cpx, Hbl, and Pl are the major hosts for Ca in the arc magmas from Lutao, Northern Luzon arc ( Cpx has produced limited influence on the Ca isotopic compositions of the Lutao arc magmas, despite both minerals are major accumulation phases (Fig. 4).
These results are consistent with previous works, demonstrating magma differentiation would produce negligible influence on the Ca isotopes of igneous rocks Chen et al., 2019;Zhu et al., 2020a;Wang et al., 2021). The Ca isotopic fractionation factor between Cpx and basaltic melt (αCpx-melt) at magmatic temperatures is close to 1 (Chen et al., 2019(Chen et al., , 2020a, and the residual melt would be only slightly affected by the accumulation of Cpx. This process is especially important when Cpx dominates the Ca contents. Specifically, Zhang et al. (2018) proposed the fractionation factor between Cpx and melt (Δ 44/40 CaCpx-melt) to be (+0.09 ± 0.07) × 10 6 /T 2 and Δ 44/40 CaPlg-melt equals (-0.15 ± 0.08) × 10 6 /T 2 using first-principles calculations and ionic models.
Combined with the calculated Ca isotope fractionation between Hbl and Pl, the estimated Hbl-melt fractionation is proposed as Δ 44/40 Cahbl-melt = (+0.06 ± 0.10) × 10 6 /T 2 . Therefore, the accumulation of Hbl Where f is the fraction of accumulated Cpx. δ 44/40 Camelt,i and δ 44/40 Camelt,a are the Ca isotopic compositions of the melts before and after magma differentiation, respectively. The magma temperature is set at 1000 K ).
The modeling results suggest Cpx accumulation could not significant decrease the δ 44/40 Ca of the residual melt at Δ 44 CaCpx-melt of 0.09 × 10 6 /T 2 , even if the CaO removal was solely ascribed to the crystallization of Cpx (Fig. 5).
By contrast, if the Δ 44 CaCpx-melt were set at 0.16 × 10 6 /T 2 , the δ 44/40 Ca of the residual melt would be decreased from
However, partial melting could have produced negligible influence on the Ca isotopic characteristics of Lutao arc magmas.
(1) the δ 44/40 Ca of the melt would be only slightly lower than the starting material using αOpx-Cpx ranging from 1.00026 to 1.00050. At αOpx-Cpx of 1.00050, partial melting would produce a melt with δ 44/40 Ca of 0.08‰ lower than the starting peridotite Zhang et al., 2018;Chen et al., 2019). (2)

Subduction-related metasomatism
The significant enrichment of LILEs and LREEs indicate the Lutao arc magmas are evolved from DMM metasomatized by enriched components (e.g., aqueous fluids derived from subducted SCS slab and/or sediment melt) Fourcade et al., 1994). In addition to source contamination, whole rock Nd -Hf isotopes and zircon Hf isotope results suggest the NLA volcanics may have been contaminated by continental crust materials Lai et al., 2018). DMM is depleted by 2-3 % melt extraction with respect to the primitive mantle or BSE . Such a low degree of melt extraction would produce negligible Ca isotope fractionation Chen et al., 2019). Therefore, the δ 44/40 Ca value of DMM should inherit the Ca isotopic composition of BSE (0.94 ± 0.05‰).  , 1993). Subsequently, the melting DMM metasomatized by slab-derived fluids (the product is denoted as AMMF) was further metasomatized by sediment melts or contaminated by continental crustal materials Lai et al., 2017).  .
Subduction fluids are dominantly derived from AOC and probably additionally released from subducted sediment at shallow subduction depth (Stern, 2002;Ribeiro et al., 2013). The carbonate solubility could reach up to 5000 ppm at sub-arc depth in the slab-derived fluids (Kelemen and Manning, 2015). Several works suggested the prograde metamorphism dehydration and the alteration of oceanic crust may not significantly modify the δ 44/40 Ca values Lu et al., 2020;Wang et al., 2021 (Fig. 6).
In addition to Sr-Nd isotopes, the influence of slab-derived fluids and sediment melts could be further evaluated by subduction indicators. During subduction, Ba, Cs, and Sr are mobilized in both slab-derived fluids and sediment melts (McCulloch and Gamble, 1991;Pearce et al., 2005), whereas Th and La can only be mobilized with sediment melts (Johnson and Plank, 1999;Spandler et al., 2007). Nb, Yb, and Sm are typical immobile element in slab-derived fluids and sediment melts (Pearce et al., 2005). Therefore, Ba/Nb and Sr/Nb ratios could be used to track the total subduction component. The arc magmas usually inherit the Th, La, and Nb anomalies of subducted sediments, making the Th/La and Nb/La ratios good indicators of recycling sediment (Plank, 2005). In addition, Ce/Pb is a widely used indicator of slab-derived fluids because Ce and Pb show similar incompatibilities during partial melting but different fluid mobilities (Johnson and Plank, 1999).

The δ 44/40 Ca values of Lutao arc magmas show significant correlations with subduction indicators including
Ba/Nb, Ce/Pb, Sr/Nb, and Nb/La ratios (Fig. 7). The decreasing Ce/Pb ratios in Lutao volcanics with respect to DMM could be well explained by the metasomatism by subduction fluids, because Ce/Pb ratios are typically low in fluid-affected rock samples (Johnson and Plank, 1999). The δ 44/40 Ca values, however, increase significantly with Ba/Nb and Sr/Nb ratios, contradicting the fact that the addition of crustal materials would enrich the arc magma with lighter Ca isotopes (Fig. 6). In addition, the Ba/Nb, Sr/Nb, and Nb/La ratios exceed the typical ranges of DMM (both global MORB and DMM show similar ranges for these ratios) and the recycled SCS sediment/UCC , and therefore cannot be illustrated by the simple mixing between DMM and sediment melts/continental crust materials.
This conundrum could be again reconciled by the metasomatism of DMM by slab-derived fluids before metasomatism by sediment melts and/or contamination by continental crustal materials. In this study, the arc  (Fig. S6). Using the proposed Sr-Nd isotopic compositions of AMMF, we estimated the endmember chemical compositions of AMMF (Fig. S6, Table S4). The calculated results manifest fluid-mobile elements (e.g., Ba, Sr, and Pb) that enriched in the subduction fluids have significantly elevated the Ba/Nb, Sr/Nb, and Ba/La ratios of the arc magma. By contrast, the metasomatism of slab-derived fluids have dramatically decreased the Ce/Pb ratios of the arc magma. Subsequently, the sediment melts formed at deep subduction depth were migrated into the magma source. Due to the chemical contrasts between recycled sediments and AMMF, this process simultaneously decreased the δ 44/40 Ca values and the Ba/Nb, Ce/Pb, Sr/Nb ratios, and elevated the Nb/La ratios of the Lutao magma (Fig. 7). Although sediment melting experiments demonstrate that the mobility and compatibility of some elements (e.g., Ba, Th, and Sr) could change significantly when temperature increases through the melting solidus (Johnson and Plank, 1999), favored by the fact that measurable δ 44/40 Ca variation can only be generated when the arc magma has experienced > 75% crustal contamination (Fig. 7), which is far exceeding the previously suggested ranges (< 5%, Fourcade et al., 1994;Lai et al., 2017).

Implications for carbonate recycling in subduction zones
The recycling of carbonate has been proposed to illustrate the Ca isotopic anomalies in mantle-derived basalts and carbonatites (Huang et al., 2011;Liu et al., 2017;Amsellem et al., 2020). However, only limited data have  (Blättler and Higgins, 2017;Chen et al., 2020a;Kang et al., 2021) is slightly higher than that of BSE. Nevertheless, as illustrated by Kang et al. (2021), the Ca isotopes of mantle wedge would be largely unchanged by the addition of AOC carbonate. This result is also observed for the Lutao arc magmas that their Ca isotopic characteristics are remained at BSE level during the metasomatism by slab-derived fluids (Fig. 6, Fig. 7).
The δ 44/40 Ca of Lutao volcanics decreases significantly with subduction indicators, suggesting the metasomatism by sediment melts did affect the Ca isotopes of arc magmas. Such an effect, however, was absent in the reported BABBs and arc magmas (Zhu et al., 2020a;Wang et al., 2021;Kang et al., 2021 sediments (Plank and Langmuir, 1998;Zhu et al., 2020a), or the survival of subducted carbonate during plate subduction (Plank and Langmuir, 1993;Shen et al., 2018;Kang et al., 2021). The arc magmas from these studies present Sr-Nd isotopes and subduction indicators (e.g., Ba/Nb and Nb/La) close to DMM ranges, further supporting the relatively low influence of subducting sediment (Zhu et al., 2020a;Wang et al., 2021;Kang et al., 2021). By contrast, the Lutao volcanics demonstrate Sr-Nd isotopes and subduction indicators away from the DMM values, suggesting the Lutao magma should have received higher proportions of subduction materials than these previously reported BABBs and arc magmas ( Fig. 6 and 7).
Here, we roughly estimate the fluxes of sedimentary Ca subducted into global trenches using the published values on subduction rate, thickness and bulk density, bulk CaO and carbonate contents in the subducted sediments (Table S5;   (Zhu et al., 2020a;Wang et al., 2021;Kang et al., 2021). The sedimentary Ca fluxes and total Ca fluxes in Central America and Manila trenches are orders of magnitude higher than other trenches due to their high carbonate contents in the subducting sediments . However, according to the Sr chemical and isotopic compositions of the subducted sediments in both trenches, the Lutao arc magmas should have received higher proportions of subduction materials with respect to the arc magmas from Central America . Therefore, higher proportions of subducted carbonate in the Central America should have been survived beyond the arc depth during plate subduction. This phenomenon is possibly related to the thermal structures of subduction zones. According to , the slab temperature beneath the arc of Central America (represented by Costa Rica) is 690 o C (D80 model), much lower than that of North Philippines (833 o C). The temperature and pressure condition of the slab underneath the Central America arc is close to but fall out of the solidus of H2O-saturated sediment melting at 2.5 -3.2 GPa and 670 -950 °C . By contrast, the temperature and pressure condition of the NLA slab (represented by North Philippines) allow the H2O-saturated sediment melting (Fig. 8). Therefore, the subducting sediment beneath NLA should have experienced higher degree of melting and induced more intensive metasomatism of sediment melts than that beneath the Central America Arc. Because chemical signatures of subducted sediments are predominantly preserved in the sediment melts rather than subduction fluids (Kawamoto et al., 2012), the limited sediment melting in Central America suggests most of the subducted carbonate is retained on the descending slab. The lower carbonate recycling efficiency in Central America with respect to NLA is  (Johnston et al., 2011).
The melting of subducting sediments during plate subduction is critical for regulating the budget of subducted carbonate recycling back to the arc. The fluxes of sedimentary Ca subducted into global trenches are predominantly constrained by the carbonate contents in the subducting sediments (Table S5) Instead, the Ca isotopic characteristics of the Lutao arc magmas are dominantly controlled by the subductionrelated metasomatism (Fig. 9). The melting of DMM was firstly metasomatized by slab-derived fluids at shallow depth, which process changed the δ 44/40 Ca values by a negligible extent but dramatically enriched the magma with fluid-mobile elements (e.g., Ba, Sr, and Pb). Then, the arc magma was metasomatized by sediment melts by    13 (Zhu and Macdougall, 1998;Jacobson et al., 2015); seawater, hydrothermal fluids, and carbonate (Fantle and