Chromium Isotope Behavior During Serpentinite Dehydration in Oceanic Subduction Zones

Fluids released through the dehydration of serpentinite can be rich in Cl−, which enables the significant mobility of Cr in subduction zones. However, the Cr isotope behavior accompanying the mobility of Cr during serpentinite dehydration is still poorly constrained. Here, we report high‐precision Cr isotope data for a unique suite of serpentinites that represent metamorphic products at different depths in oceanic subduction zones. Low‐grade serpentinites affected by significant Cr loss during serpentinization exhibit remarkably higher δ53Cr, while samples with Cr contents >∼1,800 ppm typically preserve mantle‐like δ53Cr. Antigorite serpentinites have an average δ53Cr value of −0.17‰ ± 0.19‰ (n = 12, 2SD), which is statistically lower than those of low‐grade serpentinite (−0.05‰ ± 0.30‰, n = 80, 2SD) and higher‐grade chlorite harzburgite (−0.10‰ ± 0.27‰, n = 22, 2SD). This suggests that resolvable Cr isotope fractionation occurs during serpentinite dehydration, which is explained by the variability of Cr isotope behavior in the presence of Cl‐bearing fluids at different dehydration stages. No obvious Cr isotope fractionation was found during chlorite harzburgite dehydration, probably related to the limited Cr mobility in a Cl‐poor fluid. Other processes, such as melt extraction, external fluid influx and retrograde metamorphism, have negligible effects on the Cr isotope systematics of meta‐serpentinites. Fluids released by serpentinite dehydration may have a great effect on the Cr isotope heterogeneity of mantle wedge peridotites and arc magmas.

Serpentinites are widely considered key agents in geodynamic and geochemical processes in subduction zones (Deschamps et al., 2013;Scambelluri et al., 2004Scambelluri et al., , 2019;;Spandler & Pirard, 2013).Serpentinite dehydration in subduction zones not only modifies the geochemical composition of the subducting slab, but also transfers unique geochemical signatures to the overlying mantle wedge (e.g., Alt et al., 2012;Harvey et al., 2014;Scambelluri & Tonarini, 2012;Shen et al., 2021;Tonarini et al., 2011) and plays a key role in the oxidation of arc magmas (e.g., Debret et al., 2020;Padrón-Navarta et al., 2023; Y. X. Zhang et al., 2021).As serpentinite dehydration can generate Cl-rich fluids (Kendrick et al., 2011(Kendrick et al., , 2018;;Scambelluri et al., 1997Scambelluri et al., , 2004Scambelluri et al., , 2015)), such a process is expected to not only release volatiles and fluid-mobile elements (FMEs) but also potentially mobilize Cr.Whether Cr mobilization during serpentinite dehydration would fractionate Cr isotopes is of particular interest.This is because serpentinite is a key Cr host in subduction zones, and thus its dehydration has a great bearing on the Cr isotope systematics in both hydrated mantle wedge peridotites and arc magmas, which is still unknown thus far.Given that subducted serpentinites often undergo multistage dehydration during their prograde evolution, a full understanding of Cr isotope behavior at different metamorphic stages requires a comprehensive investigation of samples exhumed from different depths of subduction zones.
In this study, we present high-precision Cr isotope data for a unique sample suite, filling the gap in the δ 53 Cr values of serpentinites metamorphosed under diverse P-T conditions.The samples encompass low-pressure oceanic serpentinite to products of complete dehydration at great depths and thus provide new insights into Cr isotope behavior during various stages of serpentinite dehydration in oceanic subduction zones.

Geological Background and Samples
The samples investigated here include oceanic serpentinite from ophiolites, antigorite serpentinite, chlorite harzburgite, and garnet peridotite from exhumed palaeo-subduction metamorphic terrains (Figure 1).The ultramafic rocks in different massifs represent the remnants of previous lithospheric mantle that was affected by Alpine metamorphism.The eclogite-facies massifs record prograde evolution along the P-T path of subduction zones, providing ideal targets to investigate Cr isotope behavior during the prograde dehydration of serpentinites in subduction settings.The sample localities, mineral assemblages and chemical compositions have been previously reported in detail (Garrido et al., 2005;Halama et al., 2012;Marchesi et al., 2013;Padrón-Navarta et al., 2011;Scambelluri et al., 2001Scambelluri et al., , 2014)), and the chemical compositions for individual samples can be found in Table S1 in Supporting Information S1.Menzel et al., 2019Menzel et al., , 2020) ) showing the peak P-T conditions of samples from the different localities in this work.The numbers on each curve refer to the mineral reactions listed in the text.Yellow, green and blue bands and lighter bands with * are the P-T trajectories of hot, intermediate and cold slabs at the slab-mantle interface and at the Moho depth within the slab after the thermal models of Syracuse et al. (2010), respectively.The blue dashed arrow shows the average metamorphic field gradient of subducted lithologies from exhumed metamorphic terranes (Penniston-Dorland et al., 2015).

Whole-Rock Trace Element Analysis
Whole-rock trace element compositions of MNS-2, MNS-3, MNS-4, Al08-07, and Al08-38 were measured at the Sample Solution Lab, Wuhan, China, using an Agilent 7700e inductively coupled plasma mass spectrometer (ICP-MS).Approximately 50 mg of sample powder was dissolved in a mixture of HNO 3 -HF in a Teflon beaker at 190°C for complete digestion.After evaporation on a hotplate for dryness, it was redissolved in HNO 3 and evaporated.Then, the sample was dissolved in HNO 3 solution, and 1 mL of indium solution (1 ppm) was added as an internal standard.The final solution was diluted with 2% HNO 3 to 100 mL for analysis.To monitor the data quality, two international silicate materials (AGV-2 and BHVO-2) were measured as unknowns, the results of which are consistent with the recommended values (Table S1 in Supporting Information S1).Based on the results of reference materials and duplicate samples, the analytical precision and accuracy were mostly better than ±10%.The procedural blanks for each element are mostly <0.01 ppb.

Whole-Rock Cr Isotope Analysis
Whole-rock Cr isotope compositions were measured at the CAS Key Laboratory of Crust-Mantle Materials and Environments at the University of Science and Technology of China (USTC), Hefei.The procedures of sample dissolution and column chemistry have been reported by Shen et al. (2018) and Q. Zhang et al. (2019).In brief, approximately 5 mg of rock powder was dissolved in a mixture of twice-distilled HF-HCl-HNO 3 in capped beakers on a hot plate at 130°C overnight.After the complete dissolution of the sample, the solution was evaporated and then completely redissolved in 1 mL 6 N HCl.The Cr concentrations of the sample solutions were determined by ICP-MS to ensure that the mass of Cr in the aliquot was 1 μg.Then, the solutions were mixed with an appropriate amount of the 50 Cr to 54 Cr double spike.The molar ratio of 50 Cr to 54 Cr and weight proportion between the double spike and sample solution followed Shen et al. (2018).Cr separation was achieved by two-step cation chromatography using Bio-Rad AG 50 W-X8 resin (200-400 mesh).The total procedure blank was <3 ng, which was negligible compared to the separated Cr from the sample (1 μg).
Purified Cr samples were analyzed by a Thermo Scientific Neptune Plus multicollector-ICP-MS instrument at USTC.The detailed analytical procedure followed Shen et al. (2018) and Q. Zhang et al. (2019).All the sample solutions were diluted with 2% HNO 3 to 200 ppb and introduced with an Aridus II™ desolvating nebulizer.The carrier gas was high purity argon without the addition of N 2 to minimize potential polyatomic interferences.Analyses were performed in medium-to high-resolution modes (5,500 < M/ΔM < 11,000).The ion intensities of Cr isotopes and relevant isobaric interferences were measured in static mode on Faraday collectors, and all isobaric interferences were carefully corrected (Q.Zhang et al., 2019).Prior to each analytical session, the spiked NIST SRM 3112a standard was analyzed.After every 4-5 sample measurements, the spiked internal standard (SCP) was analyzed to monitor data quality.During the whole analytical period, NIST 3112a and SCP yielded δ 53 Cr values of −0.09 ± 0.04‰ (n = 11) and −0.03 ± 0.04‰ (n = 46), respectively, which are consistent with the reported long-term mean values of −0.07 ± 0.06‰ and −0.02 ± 0.04‰, respectively (Shen et al., 2021).The peridotite reference material JP-1 shows a δ 53 Cr of −0.10 ± 0.03‰ (n = 5), consistent with previous results (Bonnand et al., 2016(Bonnand et al., , 2020)).

Results
The δ 53 Cr values are listed in Table 1, and a summary of δ 53 Cr for meta-serpentinites and related reservoirs is shown in Table 2.The newly analyzed trace element results for MNS- 2, MNS-3, MNS-4, Al08-07, and Al08-38 Note.Mineral abbreviations are after Whitney and Evans (2010).The major and trace elements of the samples are listed in Table S1 in Supporting Information S1.

Table 1 Continued
and detailed major and trace elements for other samples are summarized in Table S1 in Supporting Information S1.We also compiled the reported low-grade serpentinite in Table S2 in Supporting Information S1.The compositional data in Table 1, Tables S1 and S2 in Supporting Information S1 are available in Xiong et al. (2023).
The δ 53 Cr values of oceanic serpentinites range from −0.17‰ to 0.61‰, in agreement with the literature results (Figure 2a).The olivine-rich vein from ET yields a δ 53 Cr value of −0.03 ± 0.05‰ (n = 2, 2SD), higher than its corresponding wall-rock Atg-serpentinite with a δ 53 Cr of −0.39 ± 0.01‰ (n = 2, 2SD).The chlorite-rich vein in the mylonite zone yields a δ 53 Cr value of −0.19 ± 0.05‰ (n = 3, 2SD), and its corresponding wall rocks show δ 53 Cr values of −0.16‰ and −0.04‰.The Almirez Atg-serpentinites have variable δ 53 Cr values from −0.27‰ to 0.20‰.Among the subgroups, Atg-serpentinite, Cpx-bearing serpentinite and Opx-bearing serpentinite yield δ 53 Cr values of −0.24‰ to −0.15‰ (n = 4), −0.17‰ to −0.08‰ (n = 4), and −0.27‰ to 0.20‰ (n = 2), respectively (Table 1).In the Chl-harzburgite group, the δ 53 Cr values of the granofelsic, spinifex-like and recrystallized   To compare the data from the meta-serpentinites, we performed a statistical analysis of the samples.According to Shapiro normality tests, some rock types are not normally distributed, and therefore we used the non-parametric Mann-Whitney test to assess the potential differences among the δ 53 Cr of various rock types.An Opx-bearing Atg-serpentinite with remarkably higher SiO 2 than others is considered an outlier and is excluded from further discussion (Figure S1 in Supporting Information S1).Because low-grade serpentinite with lower Cr contents can display fractionated δ 53 Cr values (Figure 2a), samples with higher Cr contents of 1,800-4,000 ppm are also selected for comparison.The results are listed in Table S3 in Supporting Information S1.

Discussion
The presented sample suite spans a large δ 53 Cr range of ∼1‰.In addition, there are two outstanding observations among the (meta-)serpentinites: (a) low-grade serpentinites with Cr <∼1,800 ppm have significantly higher δ 53 Cr values than those samples with Cr >1,800 ppm (Figure 2a), whereas there is no clear trend between Cr concentrations and δ 53 Cr values of mantle xenoliths (Jerram et al., 2022;Ping et al., 2022); (b) The studied Atg-serpentinites have lower average δ 53 Cr value by ∼0.08‰ than both low-grade serpentinites and Chl-harzburgites, regardless of whether the Cr contents are restricted to 1,800-4,000 ppm or not (Table 2; Figures 2b and 3).Notably, some meta-serpentinites yield low δ 53 Cr that are rarely found in low-grade serpentinite (Figure 2), and the obviously variable δ 53 Cr values of fluid-induced veins from −0.32 to −0.03‰ imply that prograde serpentinite devolatilization could produce resolvable Cr isotope fractionation.Although these observations suggest the significant role of serpentinization and devolatilization in the Cr isotope systematics of the (meta-)serpentinites, other processes must be assessed first.These processes are (a) melt extraction/ percolation of mantle protoliths prior to serpentinization; (b) external fluid influx in the subduction zone; and (c) retrogression and rehydration during exhumation.

Effect of Mantle Melting and Melt-Peridotite Reaction
The chemical composition of peridotites prior to serpentinization is mainly the combined result of melt extraction and late refertilization (e.g., Bodinier & Godard, 2014;Niu, 2004).In the Al 2 O 3 /SiO 2 versus MgO/SiO 2 plot, the serpentinites follow a magmatic depletion trend and fall within the range of abyssal peridotites (Figure S2 in Supporting Information S1).The good correlations between Al 2 O 3 , Sc, and V indicate that these elements mostly preserve the magmatic signature of the protolith (Figures 4a and 4b, see also Marchesi et al., 2013).There are no systematic differences among these elements for each group of meta-serpentinite.Similar to both peridotite xenoliths and abyssal peridotites worldwide, the Cr contents of the samples show no correlations with Al 2 O 3 (Figure 4c).The vast Cr content heterogeneity is likely caused by different elemental behaviors during melt extraction under variable P-T-fO 2 conditions and local effects of melt-rock reactions (Bodinier & Godard, 2014;Liang & Elthon, 1990;Roeder & Reynolds, 1991).
Previous studies have shown that although melts have slightly lower δ 53 Cr than peridotite (Δ 53 Cr residue-melt < 0.1‰, Bonnand et al., 2020;Jerram et al., 2020;Ma et al., 2022;Shen et al., 2020), melt extraction has a rather limited effect on the Cr isotope composition of residual peridotites (<0.01‰,Jerram et al., 2022), as shown by the lack of correlations between Al 2 O 3 /SiO 2 and δ 53 Cr (Figure 4d).In contrast, δ 53 Cr variations in mantle rocks can be produced by later melt refertilization (Jerram et al., 2022;Xia et al., 2017).Under the Cr concentration gradient between ultramafic peridotites and percolated melts, kinetic exchange would lead to the preferential loss of isotopically lighter Cr into the melts, leaving residues with higher δ 53 Cr values (Jerram et al., 2022;Xia et al., 2017).On the other hand, the direct contribution of low-δ 53 Cr melts into the percolated peridotites may result in mantle rocks with lower δ 53 Cr (Ping et al., 2022).A melt percolation process should display a correlation between δ 53 Cr and incompatible elements (Jerram et al., 2022).Considering that the original REEs (except for La and Eu in some cases) of protoliths are commonly preserved during (de)serpentinization (Deschamps et al., 2013;Niu, 2004), REE ratios can be used as tracers for melt percolation (Pettke & Bretscher, 2022).The presented (meta-)serpentinites are generally depleted, and their δ 53 Cr values show no correlations with (Ce/Yb) N .Therefore, melt percolation cannot explain the systematic trend of the sample suite (Figure 4e).

Effect of External Fluid Influx in the Subduction Zone
During subduction, serpentinites could have experienced cryptic metasomatism by external fluids from crustal rocks, leading to the enrichment of FMEs and related isotopes (Cannaò et al., 2015;Debret et al., 2021;Garrido et al., 2005;López Sánchez-Vizcaíno et al., 2005).However, no systematic enrichment of FMEs among each group of samples is observed, and there are no correlations between δ 53 Cr and FMEs of the meta-serpentinites (Figure 5).In fact, Cr abundances in oceanic crust (∼320 ppm, White & Klein, 2014) and global subducting sediments (∼70 ppm, Plank, 2014) are much lower than those of mantle peridotites and serpentinites (normally ∼2,000-4,000 ppm, Figure 4c).Hence, from a mass balance perspective, the effect of interaction with crustal fluids on Cr isotope variations in subducted serpentinites is negligible.

Effect of Retrogression and Rehydration During Exhumation
Previous studies have emphasized the geochemical effect of retrogression during the exhumation of meta-serpentinites (e.g., Pettke & Bretscher, 2022).Among the sample suites, retrogression has been observed in transitional lithologies and Chl-harzburgites from CdA (Bretscher et al., 2018;Debret et al., 2015;Padrón-Navarta et al., 2011) and a Grt-peridotite from CdG (Scambelluri et al., 2014).The main hydrous mineral in Atg-serpentinite is antigorite, while that in Chl-harzburgite is chlorite.Meanwhile, antigorite and chlorite are the main Al-bearing minerals in both rock types, dominating their bulk Al 2 O 3 contents.As such, the theoretical H 2 O stored by peak assemblages can be assessed by the bulk rock Al 2 O 3 contents (Bretscher et al., 2018;Pettke & Bretscher, 2022).For Atg-serpentinite, the general match of measured loss on ignition (LOI) and predicted range suggests the minor effect of retrogression (Figure 6a), in agreement with previous observations (Bretscher et al., 2018;Padrón-Navarta et al., 2011).In contrast, the LOI of Chl-harzburgites is significantly higher than the predicted range (Figure 6b), manifesting the presence of later-formed serpentine and talc (Bretscher et al., 2018;Pettke & Bretscher, 2022;Vieira Duarte et al., 2021).
If retrogression upon exhumation played a role in the Cr isotope composition of the meta-serpentinites, a systematic difference in δ 53 Cr between each group with various extents of rehydration would be expected, which is, however, not observed (Figure 6).Therefore, late retrograde reactions have an insignificant effect on the Cr isotope systematics, in agreement with previous observations (Shen et al., 2015).
In summary, magmatic processes, external fluid metasomatism and retrogression cannot explain the systematic trends of (meta-)serpentinites.Alternatively, the observed variation in δ 53 Cr of (meta-)serpentinite is most likely associated with Cr isotope fractionation during serpentinization and prograde devolatilization.

Oceanic Serpentinization Stage (Stage 0)
The studied oceanic serpentinites display a large δ 53 Cr variation from BSE-like to a much higher value of 0.61‰, consistent with previous results.In addition, the compiled low-grade serpentinites show elevated δ 53 Cr at low Cr concentrations (Figure 2a).Because the extent of serpentinization and late weathering have negligible effects on the Cr isotope composition of serpentinite (Novak et al., 2022), such a trend was previously explained either by redox-related serpentinization or by the loss of isotopically lighter Cr 3+ during serpentinization (non-redox-related) (Farkaš et al., 2013;Kraemer et al., 2021;Wang et al., 2016).The redox-related interpretation involves the oxidation of immobile Cr 3+ into soluble Cr 6+ that is lost during serpentinization.Although the oxidation of Cr 3+ to Cr 6+ during serpentinization can be feasible in some cases (Oze et al., 2016), the accompanying Cr isotope behavior, which is crucial for the isotope composition of the residues, is still poorly constrained.Previous experimental studies documented highly variable Cr isotope fractionation factors (even opposite fractionation directions) during the oxidation of Cr 3+ to Cr 6+ (Bain & Bullen, 2005;Zink et al., 2010), which are significantly different from the calculated equilibrium fractionation factors (Schauble et al., 2004).Hence, the behavior of Cr isotopes in the redox process may be very complex and needs further constraints (Wang et al., 2015).Alternatively, more recent studies found that significant Cr isotope fractionation can be induced by the removal of Cr 3+ in the presence of complexing ligands (Babechuk et al., 2017(Babechuk et al., , 2018;;McClain & Maher, 2016;Saad et al., 2017).In this regard, the observed trend in Figure 2a can be due to the removal of isotopically light Cr during serpentinization (Babechuk et al., 2017(Babechuk et al., , 2018;;Wang et al., 2016).The Rayleigh fractionation modeling shows that a fractionation factor between fluid and serpentinite (Δ 53 Cr fluid-rock ) of −0.05 to −0.6‰ is needed to account for the range of low-grade serpentinites (Figure 2a).
Nevertheless, higher δ 53 Cr values in low-grade oceanic serpentinites are only observed in samples with lower Cr contents (Figure 2a).When focusing on samples with higher Cr contents of 1,800-4,000 ppm, their δ 53 Cr values are consistent with those of peridotite xenoliths (Figure 3b).This suggests that resolvable Cr isotope fractionation during serpentinization requires significant loss of Cr.

From Low-Grade Serpentinite to Atg-Serpentinite (Stage 0-1)
The shallow and low-temperature origin of the ET and Almirez serpentinites has been confirmed by previous field and geochemical studies (Garrido et al., 2005;Kendrick et al., 2011;Marchesi et al., 2013;Scambelluri et al., 2001).The higher grade Atg-serpentinite formed through a phase transition from chrysotile/lizardite to antigorite.The absence of brucite and the presence of olivine may further indicate the occurrence of the bruciteout reaction at higher P-T conditions.These processes lead to the release of approximately 3 wt.%fluids (Figure 7).Atg-serpentinite has an average δ 53 Cr of −0.17 ± 0.19‰ (n = 12, 2SD), which is statistically lower than that of low-grade serpentinite of −0.05 ± 0.30‰ (n = 80, 2SD) (Figure 3).For samples with Cr contents of 1,800-4,000 ppm, such a difference is still significant (Table S3 in Supporting Information S1).The resolvable lighter Cr isotope compositions of Atg-serpentinite cannot be explained by Cr loss during serpentinization, which would otherwise result in enrichment of heavier Cr in the residues (Figure 2a).In fact, for both Atg-serpentinite and Chl-harzburgite, the mean δ 53 Cr of samples with restricted Cr contents (1,800-4,000 ppm) is comparable to that of samples with unrestricted Cr contents (Table 2), indicating that Cr loss during initial serpentinization had an insignificant effect on the Cr isotope compositions of the present higher grade rock types (Figure 2b).Instead, the lower δ 53 Cr values of Atg-serpentinite than those of the low-grade serpentinites suggest isotopic fractionation during prograde metamorphism.
In Atg-serpentinite, almost all Cr is hosted by antigorite and magnetite (Padrón-Navarta et al., 2011;Vieira Duarte et al., 2021).Cr is partitioned together with other trivalent cations, such as Al 3+ in antigorite (Padrón-Navarta et al., 2013), while it substitutes octahedral Fe 3+ in magnetite.The transformation of lizardite into antigorite, despite associated water loss, presumably would not remarkably fractionate Cr isotopes as they belong to the same serpentine group.Because antigorite contains much less Fe 3+ than lizardite (Evans et al., 2012), prograde serpentine phase transformation would facilitate magnetite crystallization for excess Fe 3+ (Vieira Duarte et al., 2021).As a result of more Fe 3+ competing octahedral positions in magnetite, the corresponding Cr contents would be much lower.Indeed, Vieira Duarte et al. (2021) found that the Cr contents of the magnetite formed at this stage (Cr 2 O 3 < 6 wt.%) are much lower than those of relict magnetite (Cr 2 O 3 up to 10 wt.%).The ionic model calculation predicted that spinel is isotopically heavier in Cr than other mantle minerals (Cpx, Opx and Ol) (Shen et al., 2018).Magnetite with spinel-like crystallography should also be enriched in heavier Cr.Novak et al. (2017) showed that the Cr-magnetite-rich part in serpentinite displays a much higher δ 53 Cr of +0.4‰ than the residual fraction of −0.29‰.Importantly, serpentinite phase transformation is accompanied by the release of high-salinity fluids (<50 wt.% NaCl equiv., Scambelluri et al., 1997Scambelluri et al., , 2004;;Kendrick et al., 2011), which effectively enhances Cr mobility by up to several orders of magnitude (Huang et al., 2019;Klein-BenDavid et al., 2011;Watenphul et al., 2014).
Therefore, we infer that during the formation of Atg-serpentinite, Cr was redistributed in both antigorite and magnetite recrystallization.The acquisition of much more Fe 3+ in magnetite facilitates an intense redistribution and partial loss of heavier Cr, as Cr is more readily mobile in the presence of high-salinity fluids (Crossley et al., 2017;Huang et al., 2019;Watenphul et al., 2014).This hypothesis is strengthened by the remarkably higher δ 53 Cr of a fluid-induced vein (ET42) than its serpentinite wall rock (ET42A) from ET.The other vein (ETF10) in the mylonite zone contains abundant chlorite that is not found in the undeformed veins (Scambelluri et al., 2001).Considering that both its trace element composition and δ 18 O value are different from those of other HP veins (Früh-Green et al., 2004;Scambelluri et al., 2001), it seems that such a vein sample in the high-strain domain records a much more complex process.
For a better constraint, we performed batch and Rayleigh distillation models to estimate Cr isotope variation during serpentinite dehydration at this stage, following the methods of Pons et al. (2016) (Figure 8a).As there is no available Cr isotope fractionation factor between fluid and rock for the meta-serpentinite system, we used paired vein-wall rock samples to estimate the fractionation factor (Figure 7a).The initial rock value is the mean δ 53 Cr of low-grade serpentinite with Cr contents of 1,800-4,000 ppm.The results show that with a fluid-rock fractionation factor of ∼0.3‰, a loss of ∼15% Cr from the original serpentinite is needed for the mean δ 53 Cr of Atg-serpentinite.The average anhydrous Cr concentration of low-grade serpentinite is ∼2,600 ppm (assuming ∼13 wt.% water), while that of the Atg-serpentinites is ∼2,400 ppm.In addition to their loss of ∼2-6 wt.% water (m f ), the remaining Cr fraction at this stage can be estimated by C final /C initial × (1 − m f ) ≈ 0.87 -0.90 (Pons et al., 2016).Considering the local variability of both Cr content and δ 53 Cr, the preferential loss of heavier Cr reasonably accounts for the lower average δ 53 Cr of Atg-serpentinite than that of low-grade serpentinite.

From Atg-Serpentinite to Chl-Harzburgite (Stage 1-2)
The contact boundary between Atg-serpentinite and Chl-harzburgite in Almirez shows no signs of tectonic discontinuity (Padrón-Navarta et al., 2011).Transitional lithologies at the contact steadily show the disappearance of antigorite and the growth of orthopyroxene and olivine.The contact is commonly interpreted as representing a reaction front formed by subduction HP dehydration of Atg-serpentinite to produce Chl-harzburgite (Garrido et al., 2005;Marchesi et al., 2013;Padrón-Navarta et al., 2011;Trommsdorff et al., 1998).It is also proposed that this Atg-out front corresponds to an oxidation serpentinization front, separating the serpentinite protoliths of Atg-serpentinite and Chl-harzburgite with variable compositions that were caused by different types of melt refertilization and extents of seafloor serpentinization (Bretscher et al., 2018;Pettke & Bretscher, 2022;Piccolli et al., 2019).As mentioned above, both rock types show overlapping Al 2 O 3 , V, and Sc, indicating a similar magmatic history (Figure 4), which is also supported by the lack of systematic differences in Fe isotopes among both rock types (Debret et al., 2021).In addition to textural variations pointing to metamorphic growth of Chl-harzburgite after a serpentinite protolith (Dilissen et al., 2018(Dilissen et al., , 2021;;Padrón-Navarta et al., 2015), thermodynamic modeling shows that open-system serpentinite dehydration can account for the oxidation differences between the two types of rocks (Padrón-Navarta et al., 2023).Therefore, the observed higher average δ 53 Cr of the Chl-harzburgites than that of Atg-serpentinites by ∼0.08‰ cannot be dominated by the heterogeneity of the protolith.Instead, in view of the large amounts of released aqueous fluids (∼8 wt.%, Figure 8), a more feasible scenario would be that Cr was also mobile during metamorphic dehydration of Atg-serpentinite to form Chl-harzburgite, inducing Cr isotope fractionation at this stage.Resolvable fluid-rock Cr isotope fractionation is evidenced by the obviously low δ 53 Cr in an HP syn-metamorphic vein (Alm-13/3v).
In Almirez, the antigorite breakdown reaction leads to the formation of Cr-rich magnetite, and mass balance calculations reveal that magnetite in Chl-harzburgite contains more bulk Cr than that in Atg-serpentinite (Vieira Duarte et al., 2021).This means that the Cr excess from the breakdown of antigorite would be accommodated in magnetite (Vieira Duarte et al., 2021).Consequently, the stabilization of magnetite prevents the loss of isotopically heavy Cr during intense Cr redistribution.We model the possible δ 53 Cr variation at this stage, assuming that the initial material value is the average δ 53 Cr of Atg-serpentinite (Figure 8b).The fractionation factor is inferred from isotope compositions of the vein Alm-13/3v and adjacent Chl-harzburgite samples (Alm-1, Alm-6)  (Halama et al., 2012).The results show that with a fluid-rock fractionation factor of −0.25‰, a loss of ∼20% Cr from the initial rock can match the observed average δ 53 Cr of Chl-harzburgite.As the average Cr content of Chl-harzburgite is ∼2,200 ppm, and in addition to their loss of ∼8 wt.% water at this stage (Figure 7), the remaining Cr fraction is approximately C final /C initial × 0.92 ≈ 0.84.In view of the variability in both Cr content and δ 53 Cr, it is reasonable to infer that the lower δ 53 Cr of Atg-serpentinite than that of Chl-harzburgite is caused by the preferential loss of isotopically lighter Cr during Cr redistribution in the presence of Cl-bearing fluids.Such a process cannot explain Chl-harzburgite Al06-05b with a very low δ 53 Cr value of <−0.4‰ (Figure 8b).Given that this sample shows no evidence of melt percolation (Figure 4e), its low δ 53 Cr may be related to interaction with fluids with isotopically lighter Cr at this stage or to other unknown processes.

From Chl-Harzburgite to Grt-Peridotite (Stage 2-3)
Grt-peridotites from CdG have Cr isotope compositions similar to those of Chl-harzburgite.During stage 2-3, the transition from Chl-harzburgite to Grt-peridotite releases a limited amount of fluid from the breakdown of chlorite compared to that from antigorite breakdown (Figure 7).The study on Gagnone meta-peridotite documented a much lower Cl concentration in fluids released from chlorite breakdown (Scambelluri et al., 2015).This is further reinforced by the similar bulk Cl concentrations in Chl-harzburgite (36 ppm) and Grt-peridotite (∼40 ppm) from Gagnone (Kendrick et al., 2018).The lack of resolvable Cr isotope fractionation from Chl-harzburgite to Grt-peridotite is likely ascribed to the limited Cr mobility due to the lack of aqueous Cl complexes in fluids.

Implications
The scarcity of Cr isotope data in subduction-related metamorphic rocks has hindered our understanding of Cr isotope behavior during metamorphic devolatilization reactions in subduction zones (Shen et al., 2015(Shen et al., , 2021;;Wang et al., 2016).Based on a unique sample suite, this study aims to fill the gap in the δ 53 Cr values of serpentinites metamorphosed under diverse P-T conditions in subduction zones.The fact that higher δ 53 Cr of low-grade serpentinite always correlates with lower Cr contents while samples with higher Cr contents (>∼1,800 ppm) typically have BSE-like compositions indicates that the positive shift in δ 53 Cr cannot be the result of isotope exchange with isotopically heavier seawater (Wang et al., 2016).The most straightforward interpretation is that isotopically lighter Cr is preferentially removed during serpentinization.This scenario implies that the fluids involved in serpentinization are expected to exhibit a notable enrichment in light Cr, which is significantly different from the surface water with very high δ 53 Cr values (Qin & Wang, 2017).
Dehydration of serpentinite releases aqueous fluids rich in FMEs and Cr.Previous studies have shown that the interaction with serpentinite-derived fluids led to remarkable Cr enrichment in eclogite from the Western Alps (Angiboust et al., 2014;Spandler et al., 2011).The obviously variable δ 53 Cr values of veins imply resolvable Cr isotope fractionation during prograde serpentinite dehydration.These serpentinite-derived fluids with variable δ 53 Cr values can interact with subducted slab rocks and modify their Cr isotope composition.On the other hand, as serpentinite is the main host of Cr in subducted slabs, the Cr isotope signature of deserpentinization fluids in arc volcanism should be a more faithful tracer than that of FME (e.g., Li, Sr, and Ba) and their isotopic compositions.However, rare Cr isotope investigations have been conducted on metasomatized mantle wedge peridotites and associated arc lavas.Mantle wedge peridotites that are greatly affected by serpentinite-derived fluids, such as those in the source of lavas from the South Sandwich Arc (Tonarini et al., 2011) and the Lesser Antilles Arc (Cooper et al., 2020), may have different Cr isotope compositions from others that lack serpentinite imprints.This hypothesis needs to be tested in the future based on the Cr isotope systematics of supra-subduction peridotites and worldwide arc lavas.

Conclusions
We present here for the first time high-precision Cr isotope compositions for a unique suite of meta-serpentinites that experienced oceanic hydration and subduction-zone dehydration under increasing P-T conditions.The whole sample set has a large δ 53 Cr range from −0.45‰ to 0.61‰.The initial mantle-related magmatic processes, as well as the external influx during prograde metamorphism and rock rehydration during exhumation rehydration, had no significant effects on the Cr isotope composition of the subducted serpentinites.Oceanic serpentinization with remarkable Cr loss can lead to residues with higher δ 53 Cr values, while samples with Cr >1,800 ppm typically preserve their BSE-like Cr isotope composition.The subducted Atg-serpentinites show lower δ 53 Cr values than both low-grade serpentinite and higher-grade Chl-harzburgite.This is explained by Cr isotope fractionation during prograde dehydration.During the transformation of low-grade serpentinite to Atg-serpentinite, the recrystallization of new magnetite containing less Cr induces the loss of heavy Cr into Cl-rich fluids.In contrast, the antigorite-breakdown reaction would stabilize magnetite that fixed heavy Cr, resulting in the preferential loss of isotopically lighter Cr.No resolvable Cr isotope fractionation occurs during the transition from Chl-harzburgite to Grt-peridotite, which is ascribed to the limited Cr mobility in Cl-poor aqueous fluids.Our study shows that resolvable Cr isotope fractionation could have occurred during serpentinite dehydration in subduction zones.Fluids released from serpentinite may affect the Cr isotope compositions of other metamorphic rocks within subduction zones, as well as of the overlying mantle wedge peridotites and their derived arc magmas.

Figure 1 .
Figure1.Computed P-T phase relations in subducted serpentinite (modified afterMenzel et al., 2019Menzel et al., , 2020) ) showing the peak P-T conditions of samples from the different localities in this work.The numbers on each curve refer to the mineral reactions listed in the text.Yellow, green and blue bands and lighter bands with * are the P-T trajectories of hot, intermediate and cold slabs at the slab-mantle interface and at the Moho depth within the slab after the thermal models ofSyracuse et al. (2010), respectively.The blue dashed arrow shows the average metamorphic field gradient of subducted lithologies from exhumed metamorphic terranes(Penniston-Dorland et al., 2015).

Figure 3 .
Figure 3. Boxplot of δ 53 Cr for relevant reservoirs and samples in this study (a).Also shown is the boxplot of (meta-)serpentinites with Cr contents between 1,800 and 4,000 ppm (b).Data for bulk silicate Earth and peridotite xenoliths are from Jerram et al. (2022) and Ping et al. (2022).

Figure 4 .
Figure 4. Plots of element and Cr isotope variations for the studied samples.Panels (a-c) also show the coefficient of determination (R 2 ) of the linear fitting for the studied samples.Values for primitive mantle in (c-e) are from McDonough and Sun (1995).Data sources: abyssal peridotites, Niu (2004); mantle peridotite xenoliths, Jerram et al. (2022) and Ping et al. (2022).

Figure 6 .
Figure 6.Plot of loss on ignition (LOI) versus Al 2 O 3 for Atg-serpentinite (a) and Chl-harzburgite (b).The pink area in (a) and yellow area in (b) are constrained by the compositions of antigorite and chlorite, respectively (see text for detail).Rehydration of olivine and/or pyroxene to serpentine and talc would increase the bulk LOI.Compositional data for antigorite and chlorite are from Padrón-Navarta et al. (2011).

Figure 7 .
Figure 7. δ 53 Cr versus LOI c for the studied samples with Cr contents of 1,800-4,000 ppm.To remove the effect of rehydration, the shown loss on ignition (LOI) values are calculated based on peak assemblages (LOI c ); see text for details.Stars represent the average values of each rock type.The initial low-grade serpentinite is assumed to contain 13 wt.%H 2 O.

Figure 8 .
Figure 8. Batch and Rayleigh distillation models for the evolution of δ 53 Cr during serpentinite dehydration from stage 0-1 (a) to stage 1-2 (b).In the batch model, the δ 53 Cr of the final rock is calculated by δ 53 Cr initial = δ 53 Cr final × F + δ 53 Cr fluid × (1 − F) and δ 53 Cr fluid − δ 53 Cr final = 1,000 × lnα, while the δ 53 Cr value of the final rock in the Rayleigh model is calculated by δ 53 Cr final − δ 53 Cr initial = (1,000 + δ 53 Cr initial ) × (F (α−1) − 1), where F is the fraction of Cr in the residue.The fractionation factor α between fluid and meta-serpentinite for each stage is estimated based on the δ 53 Cr of the vein and its wall rock.The initial rock values are the mean δ 53 Cr of low-grade serpentinite with Cr contents of 1,800-4,000 ppm in panel (a) and the mean δ 53 Cr of Atg-serpentinite in panel (b).

Table 2
Cr Isotope Results for Different Reservoirs and the Studied (Meta-) Serpentinite