Serpentine alteration as source of high dissolved silicon and elevated δ30Si values to the marine Si cycle

Serpentine alteration is recognized as an important process for element cycling, however, related silicon fluxes are unknown. Pore fluids from serpentinite seamounts sampled in the Mariana forearc region during IODP Expedition 366 were investigated for their Si, B, and Sr isotope signatures (δ30Si, δ11B, and 87Sr/86Sr, respectively) to study serpentinization in the mantle wedge and shallow serpentine alteration to authigenic clays by seawater. While serpentinization in the mantle wedge caused no significant Si isotope fractionation, implying closed system conditions, serpentine alteration by seawater led to the formation of authigenic phyllosilicates, causing the highest natural fluid δ30Si values measured to date (up to +5.2 ± 0.2‰). Here we show that seafloor alteration of serpentinites is a source of Si to the ocean with extremely high fluid δ30Si values, which can explain anomalies in the marine Si budget like in the Cascadia Basin and which has to be considered in future investigations of the global marine Si cycle.

If the study site is unique as a serpentinization site (L57), what does that mean for the representivity of the data and conclusions? In general, the manuscript lacks an assessment and discussion of how the fractionations derived are governed and applicable elsewhere (even assuming they are correct).
Despite the fact that the Mariana seamounts represent a unique serpentinization site, the controlling mineral reactions are comparable/ identical to mineral reactions occurring at the entire seafloor. The process of mafic and ultramafic mineral dissolution (basalt or serpentine alteration) and the subsequent re-precipitation of Si and other cations as authigenic phyllosilicates is reported from numerous seafloor sites (e.g. Iberian Abyssal Plain (Milliken et al., 1996), Juan de Fuca plate (Wheat and Mottl, 2000). Especially the authigenic mineral saponite is reported as the most common secondary mineral in altered oceanic crust  and as saponite also forms in the investigated Mariana seamounts (Morrow et al., 2019), we strongly argue for a transferability of the Mariana seamount results to general alteration reactions of the oceanic crust. We included the discussion of saponite formation in section 1.3 and clarified the transferability of the results from the Mariana seamounts to the oceanic crust in general in lines 258-275.
Temperature as a possible controlling factor of isotope fractionation is not considered. This seems particularly relevant given experimental and field evidence has demonstrated a relatively high degree of sensitivity to temperature changes in the ranges relevant here.
We agree that temperature changes can significantly affect isotope fractionation. However, taking the thermal gradients into account (Yinazao ~20-30°C/km, Fantangisña summit ~11.7°C/km; Fryer et al., 2018) the temperature change with sampling depth is only minor. At Yinazao, the deepest sample is located at 21 m below seafloor, which translates into a maximum temperature increase of 0.6°C (from 4°C seafloor to 4.6°C at 21 m depth using the highest measured thermal gradient of 30°C/km). The deepest sample at Fantangisña located 175 m below seafloor shows a temperature increase of 2°C (from 4 to 6°C). Temperature variations of ≤2°C are to date not yet detectable in isotope fractionations and generally within measurement uncertainty (e.g. . Therefore, we exclude significant impact of temperature variations with sampling depth on the Si and B isotope variability in studied pore fluids. We added a short sentence noting this in lines 78-79. I believe the numerical modeling can be improved. I question if a Rayleigh model is the right approach, since this implies that a package of Si can only be progressively depleted, while it seems more likely that Si is both added and removed by different reactions simultaneously (indeed, this is even mentioned on L147). This invalidates the assumptions inherent in the Rayleigh model for isotope ratio evolution. In general, no motivation is given for choosing a Rayleigh model. Also, does it make sense to choose a 'closed system' model (S-L27)? In general, I don't find the modeling approach to calculate a fractionation (in the supplement) very convincing. How is the 'best fit scenario' derived and quantified? It seems like there should be many parameter assemblages that work well, and there is no need for each sample to be explained by the same model (or set of model parameters).
We have substantially revised the model section of the manuscript in lines 151-228 of the revised manuscript (see also comment to remark #1). We agree that a Rayleigh model does not account for the complexity and duality of processes occurring and was mainly included to illustrate a simplified process of authigenic mineral precipitation. The Rayleigh model and the closed-system numerical model were removed from the revised manuscript and instead a numerical reactive transport-reaction model (steady state conditions) was included. Sensitivity tests concerning the magnitude of Si isotope fractionation are included in the model to constrain the associated δ 30 Si values. The δ 30 Si values can only be reproduced in the model with a Si isotope fractionation of -3‰ (see Fig. 6c; lines 184-187).
Given that the interpretation of data for samples from 1498B is different ('mantle sourced fluids' rather than low-T alteration processes) it might be worthwhile to also differentiate them in the figures.
The symbols of the mantle-sourced fluid were changed to better differentiate them from the seawater-altered fluids.
If the secondary phases that are inferred to form are phyllosilicates, It's not clear to me how this process of 'serpentine alteration' is qualitatively different from authigenic clay formation observed or interpreted elsewhere in different marine settings. I wonder if this body of literature, both Si specific and not, might help with some of the arguments made here.
The alteration of mafic and ultramafic rocks and the formation of authigenic phyllosilicates is occurring in the Mariana seamounts and the results are transferrable to the entire oceanic crust as we have argued above (comment to reviewer remark #3 'Especially the authigenic mineral saponite is reported as the most common secondary mineral in altered oceanic crust  and as saponite also forms in the investigated Mariana seamounts (Morrow et al., 2019), we strongly argue for a transferability of the Mariana seamount results to general alteration reactions of the oceanic crust.'). We agree that the transferability might not have been stated clearly in the original manuscript and we have revised the section in lines 258-275. Additionally, we added a short paragraph in which we compare the mineral reactions occurring in the Mariana seamounts to authigenic mineral formation in continental margin settings (lines 194-209).
Minor comments L16: 'seawater alteration on the seafloor' is odd formulation, rephrase.
The sentence was rephrased. L45: Strange phrasing, "the dependence of Si isotope fractionation on…" or similar might be better.
The sentence was rephrased accordingly.
L48/L51: Are these two statements contradictory? i.e. B is both incorporated into serpentine but is also incompatible? This is not a contradiction -although B prefers the liquid phase (D(s/f)=0.25), a lot of B can also be stored in the crystal lattice of the serpentine compared to other fluid mobile elements (Li, Cs, Sr), meaning that B-rich fluids precipitating serpentine provide B-rich serpentine. The large amount of B capable of being incorporated into serpentine is relative to other minerals, but B still remains incompatible (just less incompatible with serpentine than with other minerals). Ie its Kd between water and mineral is less than 1 (~0.25, Tenthorey and Hermann, 2004), but higher than many other incompatible elements. When fluids have very high B concentrations there is a high capability for serpentine to incorporate B into its structure. We included a short comment in lines 57-59.
L73/ Table 1S: It seems the 'increasing Si concentrations with sediment depth' is a bit of an oversimplification.
We agree that the description was an over simplification of the profiles and modified the text accordingly (lines 86-87).
L74: But some of the values are substantially below overlying seawater δ30Si and concentration, which is unexpected and not explained. (i.e. the lowest δ30Si value is also low Si concentration, which implies Si removal rather than addition as inferred). Conversely, some of the The statement that most of the δ 30 Si values are above seawater is valid (10 out of 13 samples have higher δ 30 Si values compared to seawater). The δ 30 Si values lower than seawater δ 30 Si are explained now at the end of section 1.2.2 (lines 232-244) and are likely caused by different mineral solubilities related to the ambient fluid. In order to cause these low pore fluid δ 30 Si values, mineral phases with low isotopic values need to dissolve. We speculate that the most likely mineral phases are earlier formed authigenic clays. The authigenic clays likely formed during higher ascent rates of the deep mantle-derived fluids in the past causing oversaturation of these phases and precipitation. At present, the ascent rates likely decreased shifting the dominant fluid phase to seawater, which causes redissolution of these phases. Significant variabilities between ascent rates were observed between the different seamounts in the Mariana forearc region (between 2 mm yr -1 to 4 cm yr -1 ; Mottl et al., 2004) and thus changes over time are not unlikely. Further investigations are necessary to constrain thoroughly the fluid ascent rates and possible changes over time. However, the complex mixing of fluids with different Si concentrations (close to cero in the deep mantle and ~140µM in seawater) and simultaneously occurring Si mineral dissolution and re-precipitation likely explains the Si concentration ranges and δ 30 Si variability in the studied samples.
L182: This value seems high, Gunnarsson and Arnorsson seem to report lower values at 8 C. Gunnarsson and Arnórsson (2000) report an amorphous silica solubility of ~80ppm (2.8mM, reported as 3mM in the main text) at 8°C. However, this part was removed from the revised manuscript and replaced with the results and interpretation from the reactive transport model (lines 151-228).
L207: These electronegativity arguments and the Meheut et al. paper cited refer to equilibrium isotope fractionation, which to my knowledge do not approach near the magnitude of -4.5‰, which requires a kinetic isotope effect (and no relation with Mg has yet been described.) We agree that the Méheut and Schauble (2014) refers to equilibrium fractionation and the reference was removed from the manuscript.
The percentage of Si derived from low-temperature weathering relative to the riverine flux is taken from Wheat et al. (2017), who based their calculation on riverine fluxes by Elderfield and Schultz (1996) and additionally took the global heat loss and thermal anomalies into account. However, we have revised this section given the large range of Si fluxes related to low-temperature oceanic crust weathering reported in literature (e.g. Wheat and McManus, 2005). In order to fully access and model the Si flux from oceanic crust alteration, better constraints on alteration rates and related Si fluxes from various marine settings is required.
L240b: Why would this (empirical) estimate of ridge flank circulation, which is certainly crude, not already include serpentine alteration?
The reported ridge flank Si fluxes originate from a fluid discharged from altered basaltic crust (e.g. Juan de Fuca Plate (Wheat and Mottl, 2000)). To the best of our knowledge, no Si fluxes related to serpentine alteration are reported. It is likely, that the general fluid flux has similar magnitudes as derived from seafloor basalt alteration studies (in the range of 0.02 to 0.8 Tmol Si yr -1 ; Wheat and McManus, 2005), but the associated Si concentration remains unknown. We rearranged this part of the discussion in lines 258-281. Table S1: Some samples have Si concentrations below detection: how is this possible within the framework advanced here?
The reported Si concentrations were removed from Table S1.

Reviewer #2 (Remarks to the Author):
This manuscript tries to use Si isotopes of pore fluids from serpentine seamounts to study serpentinization in the mantle wedge and during seawater alteration on the seafloor, and explain anomalies in the marine Si budget. However, because the discussion is significantly flawed, this submission does not match the requirement for being published in Nature Communication.
First, I am not sure why it is necessary to report B isotopes in this manuscript. These data did not help explain any process in serpentinization or get the conclusion of this study. I am also curious why the authors try to use Si and B isotopes of pore fluids to decipher the formation of authigenic secondary phases in subducted slab?
The combination of Si and B isotopes helped to reveal the processes during seawater alteration of serpentine and secondary mineral precipitation. A large number of publications have shown that Si isotopes are a good tracer for alteration reactions, however, Si isotopes have never been measured in relation to serpentine, in contrast to B isotopes. B isotopes are well studied to trace primary serpentinization reactions as well as secondary alteration reactions. Thus, the combination of a new tracer (Si isotopes) with a well-established tracer (B isotopes) provides the possibility to establish Si isotopes as a new proxy to trace serpentine alteration reactions. In this manuscript, we show for the first time the very good correlation of both isotope systems during alteration reactions in that the light isotope is incorporated in the secondary mineral for both isotope systems. We have revised former section 1.2.3 (revised section 1.2.2) to clarify this coupling by incorporating a transport-reaction model using both isotope systems together with further geochemical data to identify and disentangle mineral reactions (lines 151-228 in the revised manuscript). We can show that processes within the seamount are controlled by serpentine dissolution and precipitation of authigenic minerals in contact with seawater and related changes in mineral solubility. The combination of Si and B isotopes has thus the potential to identify different alteration reactions during seafloor weathering. Authigenic phases in the subducted slab, as mentioned by the reviewer, are not discussed in this manuscript. Si and B isotopes from deep mantle-derived fluids, discussed in section 1.2.1, relate to primary serpentinization process.
Secondly, the manuscript did not discuss why the δ30Si values of pore fluids increased with depth increasing. Higher temperature? Or different mineral phases? Is it possible that there is other mechanism to produce this heavy Si isotope signature in pore fluids? A deeper source, which can mix with pore fluids?
We agree that the variations in pore fluid δ 30 Si and δ 11 B values were not sufficiently discussed in the manuscript. We have revised former section 1.2.3 (new section 1.2.2) by including a transport-reaction model which takes the ascent of a deep mantle-derived fluid into account that mixes with seawater in the upper 30m of the seamounts (lines 151-228 in the revised manuscript). The geochemical and isotopic data are well reproduced by the model and alteration processes can be related to changing mineral solubilities and associated dissolution and precipitation reactions. Concentration data of Mg 2+ , Sr, and Clas well as 87 Sr/ 86 Sr and porosity were used to further constrain the model results (revised figures 3 and 4). See also answer to comment #1 of reviewer #1. Temperature should only have a minor effect on the isotope data of the seawater-derived fluids in that thermal gradients cause only a temperature difference of ≤2°C between the shallowest and deepest sample (Fryer et al., 2018). We included a comment in the main text in lines 78-79.
Also, a mass balance model is needed if trying to constrain the silicon fluxes of the global marine Si cycle.
We refrained from calculating a mass balance model for the global marine cycle, because we provide the first Si, B (isotope) data related to serpentine alteration. We strongly argue that the mineralogical changes involved in serpentine alteration are transferable to basalt alteration (see revised section 1.3, lines 258-275), however, this needs to be proven in further studies. In order to set-up a realistic seafloor alteration model further isotopic constraints are required to test this hypothesis. Additionally, reported Si fluxes related to seafloor basalt alteration show a large range (e.g. 0.02 to 0.8 Tmol Si yr -1 ;  and are thus not well enough constrained to justify a mass balance model. Nevertheless, we provide the first Si and B isotope data, which indicates that seafloor alteration reactions have the potential to influence oceanic Si budgets and hopefully inspire further studies to study this process.
The English of this manuscript should be improved.
The English of the manuscript was completely revised by an English native speaker.
Line 126-129: Not really understand what the authors try to discuss here.
This section discusses the low isotopic values detected in the top 5 m of the seamounts. The low B and Si isotope values are related to the dissolution of a mineral phase with low δ 30 Si values, most likely earlier formed authigenic clays. This section was revised taking into account the model results and is now part of section 1.2.2, lines 232-244. We speculate that the authigenic clays likely formed during interaction with the deep mantle-derived fluids in the past caused by higher fluid ascent rates. With respect to the deep fluids, these phases are oversaturated in the fluids and thus precipitate. At present, lower ascent rates of the deep fluid likely cause seawater entrainment and the earlier formed authigenic phases dissolve again due to their undersaturation in seawater (see also the revised Fig. 5 and 6). Ascent rates can vary significantly between seamounts in the Mariana forearc region (between 2 mm yr -1 to 4 cm yr -1 ; Mottl et al., 2004) and thus changes over time are not unlikely as well. Further investigations are necessary to constrain thoroughly the fluid ascent rates and possible changes over time.
Line 145-147: Why not? No evidence has been discussed here to prove that "pore fluid δ30Si values cannot originate from serpentine dissolution alone".
We conclude that the pore fluid δ 30 Si values cannot be generated by serpentine dissolution alone, because the pore fluids should then have the Si isotope signature from the serpentine, that is to say -0.3‰. Therefore, we conclude that another process must occur, shifting pore fluid δ 30 Si to the observed high values. The most obvious process is Si reprecipitation and with that the precipitation of the light 28 Si, enriching the fluid in 30 Si. In order to clarify this process, we added a short paragraph in lines 173-176.
Line 155-156: The thin sections of drill samples will give direct evidence for the formation of authigenic secondary phases rather than both Si and B isotopes. It is not necessary to use these two isotope systems to decipher mineral phases in these drill core samples (pore fluids).
We agree that investigations of thin sections will give direct evidence on the formation of authigenic phases and that the sentence might have been misleading. The main intention of studying both isotope systems was the possibility to trace mineral reactions and establish Si as a new tracer for alteration reactions at the seafloor. Often, effusing fluids from basement rocks are easier to access compared to basement rocks themselves. We show that information on mineral reactions can be drawn from Si and B isotope variations in these fluids and that these isotope systems can be used to trace reactions at depth. We revised section 1.3 accordingly.
Line 165-166: "B isotopes stay relatively stable at δ11B values of +38‰ (Fig. 3a)". I cannot find the stable δ11B values of +38‰ in Fig. 3a. There is only one spot of Yinazao sample has δ11B value of +38.2‰. There are five samples with δ11B values around +38‰ (+37.5 to +38.6‰), and their δ30Si values varied from +0.4 to +2.7‰. However, the samples with high δ11B values (15-R-1 and 3-F-2), their δ30Si values are also in this range. It cannot be explained as "Instead, B is potentially trapped without species preferences in nanotubes within the crystal structure and δ11B remains unfractionated". It is not appropriate to select some data to give an explanation, and leave the other data behind.
Former section 1.2.3 (revised section 1.2.2) was completely revised to clarify the mineral reactions during serpentine alteration. We removed the Rayleigh and closed-system model and instead included a transport-reaction model, which is briefly explained in the main text and a full description can be found in the Method section and the Supplement. We can now show that the alteration reactions are controlled by mineral solubilities in that serpentine dissolves in the upper 30m of the seamount and secondary minerals like talc reach oversaturation and precipitate, thereby incorporating the light isotopes of both investigated systems and shifting fluid δ 30 Si and δ 11 B to the observed high values. We added new figures to illustrate the model results together with the measured geochemical data (revised Fig. 2 -6) and discuss the model results in lines 151-228.
Review #3 of Serpentine alteration and the impact on the marine Si cycle Authors Sonja Geilert, Patricia Grasse, Klaus Wallmann, Volker Liebetrau, Catriona D. Menzies

General Comments
I think this is a good paper worth publishing in Nature communications after some revisions and moderate changes. The Si isotope data is a novel approach to understand the role of serpentinites in the Si global cycle and along B provide good insight in the formation and nature of serpentine mud volcanoes and its potential to alter Si signatures in the ocean.
My main criticism to the paper is that it tries to expand the findings from the Mariana's trench serpentinite mud volcanoes to most of oceanic crust alteration without providing evidence for this link. While their findings are robust for serpentinite alteration, they do not have supporting evidence for basaltic crust. Serpentinization reactions in peridotites are very different from basalt alteration reactions. I think it is fine for the authors to hypothesize that Si isotopes during basalt alteration might behave similarly and further study is required as their data do not directly apply to basalt alteration. I think this needs to be modified.
We agree that we can only speculate about the Si isotope signature of fluids related to basalt alteration. Here, we want to show that serpentine alteration (not formation!) is comparable to basalt alteration given that similar reaction products are formed. Reaction products for Si phases are mainly phyllosilicates like saponite, a Mg-rich smectite. We hypothesize that similar mineralogical reactions during alteration will result in similar isotopic fractionations and that is why we think, that the results from serpentine alteration are transferable to basalt alteration. We rearranged the discussion concerning this part to clarify the similarity of authigenic minerals (section 1.3, lines 258-271) and defined the meaning of serpentine alteration in line 34-38, to avoid confusion with preceding serpentine formation.

Specific comments
Line 26 It is unclear who or why serpentinites are expected to play a fundamental role in Si exchange, is there a reference for this or is this the authors interpretation?
We included the reference Frost & Beard (2007) who discuss in detail the role of Si and silica activity in a serpentinizing environment (lines 27-29).
Line 40-44 Authors do not mention the variability in d30Si in seawater. Review by Poitrasson (2017) show seawater variability in d30Si from +0.5 to +4.0‰. Part of this variability is associated geographic location and some with depth see Holtzer and Brzezinski (2015).This moves to line 75 where authors compare pore fluids with NW Pacific seawater. The variability needs to be mentioned in the text and point that at depth is relatively constant in the Pacific.
We added a sentence of the δ 30 Si variability in in the Pacific in lines 46-49 and emphasize the homogeneity of Pacific Ocean deep waters. We refer to Reynolds et al. (226), Grasse et al. (2013) and the review by Sutton et al. (2018), that provides the latest data compilation. We have focused the discussion in the main text on Pacific δ 30 Si values, given that the study area is located within the NW Pacific. We show that surface waters from the photic zone can have high δ 30 Si values, up to 4.4‰, but that the deep waters are relatively homogeneous with an average δ 30 Si of +1.2±0.2‰ (Grasse et al., 2013). Line 81-83 If it is within analytical uncertainty, I recommend this sentence to be removed from the manuscript.
The reported range in δ 30 Si for the serpentinite muds is outside the analytical uncertainty, which is reported in the method section (long-term uncertainty ≤0.13‰; sample uncertainty between 0.1 and 0.4‰). Therefore, we will leave the reported range in δ 30 Si values in the manuscript.
Line 103 Add B concentrations as done for d11B.
The B concentrations were added in line 131.
Line 103-109 I suggest these lines are reorganized. The fact that fluids and clasts overlap is solid and points to no fractionation. I suggest that this is linked and then the authors explain their interpretation for no fractionation (ie B coordination).
The lines concerning deep-mantle serpentinization and the impact on δ 11 B were reorganized accordingly (lines 126-135).
Line 110 Value of d30Si in pore fluids is close to seawater which is opposite to what is seen with Sr. Serpentinization fluids would have Si concentration an order of magnitude lower than seawater at 300C (see Klein et al. 2009) so I am worried that this is reflecting minute amounts of SW. Though lost city seawater has lower Si so this might no be relevant We agree that the δ 30 Si values are likely affected by shallow overprinting and do not represent deep serpentinizing fluids as stated in the original manuscript. The silica concentration is about a magnitude lower compared to seawater (<40µM (below detection limit) versus 145µM Si, Table S1), however, the δ 30 Si value is higher than seawater (1.6‰ versus 1.05‰; Reynolds et al., 2006). This indicates that it is a deep, Si-depleted fluid as expected from Sr isotopic values, but Si precipitation processes shifted the δ 30 Si values to higher values. We can only speculate that the precipitation process occurred close to the surface when temperature and pressure drop (see lines 135 to 142 in the revised manuscript). So far, due to the extremely low Si concentrations in serpentinizing fluids, no comparable δ 30 Si values exist providing evidence for this hypothesis.
Line 119 If closed system dissociation would not matter much would it?
Also in a closed system, when Si dissociates at pH > 8.5, the precipitating phase would be enriched in 30 Si, shifting fluid δ 30 Si to low values. Thus, if a pH effect on Si fractionation would occur than fluid δ 30 Si values should be below its initial δ 30 Si values, which is that of olivine (~-0.3‰). Given that δ 30 Si values are higher than that of olivine, we exclude a pHinduced fractionation and assume a kinetically-controlled fractionation in dependence of decreasing T and P after ascent and deposition of the serpentinite muds (see lines 135 to 142 in the revised manuscript).
Line 130 B increase in the pore fluid?
Adsorbed B is released into the pore fluid during interaction with seawater, thus increasing B concentrations. This section was revised and is now part of section 1.2.2, lines 232-244.
Line 130-133 I recommend adding a reference for the borate fractionation. Apart from being the uppermost part of the seamounts are pelagic clays and silts related to the serpentinite muds?
A reference for boric acid-borate fractionation was added (lines 211-216). The pelagic clays and silts form the uppermost layer of the mud volcanoes and are not directly related to the serpentinite muds and the process of serpentinization in general. For Yinazao seamount, pelagic muds are only reported in the upper about 5 m of the seamount (Fryer et al., 2018). At Fantangisña seamount, pelagic clays dominate at the top and at the base of the seamount with only few layers intersecting the serpentinite muds (about ≤ 10%; Fryer et al., 2018). Therefore, pelagic clays play only a subordinate role in the investigated core sections and will not control the Si isotope signatures at depths > 5m.
Line 143 Need clarification for serpentinitites. Are the authors referring to clasts brought up by the fluids or only serpentinite muds.
We refer to all serpentine minerals either in clasts as pervasive serpentine or as serpentine veins or in the serpentinized mud itself. When seawater reacts with serpentinite, serpentine minerals dissolves through hydrolysis following Milliken et al. (1996): We included a short explanation as well as the reaction in lines 166-172. Line 242 I agree with the associated signature from serpentinites but not from basalts. I think it is fine for the authors to hypothesize that basalt alteration might be similar and further study is required but their data do not directly apply to basalt alteration and this needs to be modified.
We have revised the discussion in section 1.3 to clarify why serpentine alteration is comparable to basalt alteration (similar mineralogy of authigenic products). See also comment to 'Main criticism'.    (Table S2, Fig. S1). A small shift to higher δ 30 Si values with depth is present, however, the 101 detected range is within measurement uncertainty and therefore to be interpreted with 102 caution. 103 The combination of pore fluid δ 11 B and δ 30 Si values shows a very good correlation of these 104 two isotope systems (Fig. 2). For both systems, the isotope values are below and above the 105 seawater signature, indicating dissolution of presumably serpentine and precipitation of 106 authigenic mineral phases, respectively. In order to decipher the processes controlling 107 mineral dissolution and precipitation, fluid sources and compositional changes in relation to 108 depth need to be examined.  During desorption in reaction with seawater, the 10 B is thus re-released into the pore fluid 174 shifting δ 11 B to lower values compared to seawater. In general, organic matter 175 decomposition might also contribute to this isotopic shift as also the 10 B isotope was found 176 to beis preferentially released during this process 35,36 . However, as alkalinity, phosphate, and 177 ammonia are not reported todo not increase (Table T6 in Table S1).  With apologies that this review is slightly overdue.

Strontium sample preparation and isotope analyses by TIMS
Geilert and colleagues present a manuscript that details silicon, boron, and strontium isotope geochemistry (δ30Si, δ11B, 87Sr/86Sr) of fluids and muds from seamounts experiencing serpentization and subsequent serpentine alteration. They argue for no significant Si isotope fractionation during serpentinization reactions, but substantial fractionation during the subsequent dissolution of serpentine and (re)precipitation of authigenic mineral phases (that they represent by talc in their model). This creates fluids with δ30Si locally >5‰, the highest fluids measured to date, and a potentially important source in the ocean Si budget. In this revised version, the manuscript benefits from a more quantitative model and clarification of several points. I think the central conclusion -that high δ30Si fluids are generated by the alteration (and reprecipation) of serpentine minerals/muds -is robust, although the complexities of natural systems will always hinder any attempts to numerically model the processes. While I have some quibbles about the RTM, I think the novelty of the data and the general interpretative framework are sufficient to warrant publication.
In my original review of the manuscript, I raised several (overlapping) points that are now largely dealt with in the revised version. The largest change is the incorporation of a reactive transport model, and this helps to address many of the issues. Specifically: -the descriptive nature of the results: the new model goes some way to addressing this. As and aside, I would also encourage the authors to make their model code accessible. Perhaps I miss it in the online portal, but simply displaying an advection-diffusion-reaction equation (L671/SL 17) is not the same as making the underlying scripts available.
-the generalizability of the interpretations to other regions of the ocean: the authors have made a good attempt at justifying the extension of their results to elsewhere (e.g. revised ms section 1.3), while also emphasising that this would need to be demonstrated in other field studies. I find this reasonably convincing.
-the role of temperature as a potential control on fractionation: I had misinterpreted the potential range of temperature experienced by their fluids today (i.e. only with a few deg C, L78 in revised ms). However, the response does not account for the thermal history of the circulating seawater, but I interpret from e.g. Fig 4d that the penetration of seawater into the seamounts is not deep -and therefore not at elevated T, as otherwise it would have distinct Sr isotope composition etc.? Perhaps mention this around L78 for the sake of clarity, if so.
-issues with the numerical model previously used (e.g. assumptions associated with Rayleigh distillation model, etc.). These are largely resolved with the RTM model now employed, though other issues now appear. For example, model sensitivity to the prescribed parameters (Tables S3,4,5) is not investigated. Perhaps this is not necessary: my understanding is that the true value of model lies not in its ability to pinpoint one precise set of parameters, but rather to show that the dissolution of serpentine and subsequent re-precipitation of secondary phases (whether talc, or something else) can produce fluid δ30Si in the same range as the observations. One thing that I missed is a quantification of the fraction of Si being re-sequested into the talc. Presumably there is a trade-off between fluid d30Si and fluid [Si] -as d30Si increases, [Si] decreases. Does this have implications for the importance of the flux at a regional-to global-scale?
The issues raised by other reviewers included the utility of δ11B, and the ability of extrapolating insights or fluxes from serpentine alteration reactions (as done here) to the broader suite of basalt alteration reactions. I think the authors make a reasonable case in the revised version as to the benefits of a dual-isotope approach, and the arguments based on secondary phase mineralogy are helpful in assessing the extension of the conclusions to other regions of the ocean. However, it seems to me that δ11B is not being used as a well-understood tracer to 'calibrate' the δ30Si data, as the text/rebuttal seems to imply, but rather as a system that is thought to largely be sensitive to the same processes (i.e. secondary phase formation), so that correlation between the two supports the interpretations, and aids in their model parameterisation. I think this is OK, but if this is not the case then the text should be clarified to highlight where some knowledge of the system is inferred on the basis of d11B and/or [B] alone, and then the silicon isotope interpretation made on the basis of this knowledge.
Some minor comments on the revised version of the manuscript: L151: "Motivated". L162: "the rapid dissolution of serpentine, which is undersaturated with respect to serpentine"something is off here, please rephrase.

General comments
All my comments from my previous review have been addressed in a satisfactory way. They provide an updated manuscript. It retains their Si and B isotope data and provide a novel approach to understand the role of serpentinites and its alteration in the Si global cycle. They added a new section using a onedimensional transport-reaction model to explain their data and estimate fractionation factors during serpentinite formation and alteration. Dealing with open systems is always problematic and the authors have done a good work in addressing the problematics and come with a model to interpret their data that is solid. Overall their model requires significant serpentine dissolution and loss of some Si from the serpentine muds to seawater. This is shown using pore fluid d30Si and I think is a solid result.
However, before I recommend this paper for publication it requires some minor changes in the text as well as some important clarifications with data treatment and figure 2.

Specific comments
Line 30 I suggest that the authors change seawater to hydrous fluids and modify the following sentence accordingly. I agree that serpentinization is usually thought as a seawater-rock interaction. However, serpentinization can occur with any hydrous fluid and in settings different than the seafloor. On the context of this manuscript the serpentine being brought up is a result of water/rock interactions in the mantle wedge and the fluid is likely different in composition than seawater.
Line 104 Authors mention a "very good correlation" between the two isotope systems. I think this is correct for the pore fluid data with seawater fingerprint. However, their figure is shown in a way that appears to indicate that the correlation continues through their serpentine mud data. This is an artifact of the break in the y-axis (figure from table S1 data below). They also left out a single data point for mantle derived fluid that I think should be added to the figure.
Authors also need to clarify how they derived the serpentine mud data point. Their Table S1 does not have pore fluid data for d30Si 1498B and Table S2 does not have serpentine mud data for d11B. My guess is that the authors are pairing "mantle-like" fluid d11B from Fantangisña 1498B with d30Si measured in the serpentine muds from 1498B but it is not clear from the figure or the text, if this is the case it needs to be stated in the text or the figure caption as the authors are combining different measurements to get to that data point (d30Si measured in the serpentine muds and d11B measured in pore fluids with mantle derived signature) while leaving out their only data point with both measurements for mantle-derived signature.
Line 160 I think that there is a typo "slap" should be slab. Please find enclosed our revised manuscript 'Serpentine alteration and the impact on the marine Si cycle'. We are very pleased that our revised manuscript was appreciated by the reviewers and that their responses were mainly positive. We have incorporated the suggestions from reviewer #1 and #3 regarding Fig. 2 by 1) including the mantle-derived fluid data point, 2) calculating and including a mixing curve between seawater and the mantle-derived fluid (see also supplement for calculations), and 3) removing the break on the yaxis. Further, the model code was uploaded in the journal online system and the minor comments by the reviewers were assessed and corrected in the manuscript.
We hope that you agree that the revisions have improved the manuscript and that you will consider the revised version for publication.
With best regards on-behalf of all co-authors Sonja Geilert

Reviewer #1 (Remarks to the Author):
Re-review of manuscript NCOMMS-19-35155 submitted to Nature Communications by Sonja Geilert and colleagues: Serpentine alteration and the impact on the marine Si cycle With apologies that this review is slightly overdue.
Geilert and colleagues present a manuscript that details silicon, boron, and strontium isotope geochemistry (δ30Si, δ11B, 87Sr/86Sr) of fluids and muds from seamounts experiencing serpentization and subsequent serpentine alteration. They argue for no significant Si isotope fractionation during serpentinization reactions, but substantial fractionation during the subsequent dissolution of serpentine and (re)precipitation of authigenic mineral phases (that they represent by talc in their model). This creates fluids with δ30Si locally >5‰, the highest fluids measured to date, and a potentially important source in the ocean Si budget. In this revised version, the manuscript benefits from a more quantitative model and clarification of several points. I think the central conclusion -that high δ30Si fluids are generated by the alteration (and reprecipation) of serpentine minerals/muds -is robust, although the complexities of natural systems will always hinder any attempts to numerically model the processes. While I have some quibbles about the RTM, I think the novelty of the data and the general interpretative framework are sufficient to warrant publication.
In my original review of the manuscript, I raised several (overlapping) points that are now largely dealt with in the revised version. The largest change is the incorporation of a reactive transport model, and this helps to address many of the issues. Specifically: -the descriptive nature of the results: the new model goes some way to addressing this. As and aside, I would also encourage the authors to make their model code accessible. Perhaps I miss it in the online portal, but simply displaying an advection-diffusion-reaction equation (L671/SL 17) is not the same as making the underlying scripts available. The MATHEMATICA model code was uploaded in the online portal of the journal.
-the generalizability of the interpretations to other regions of the ocean: the authors have made a good attempt at justifying the extension of their results to elsewhere (e.g. revised ms section 1.3), while also emphasising that this would need to be demonstrated in other field studies. I find this reasonably convincing.
-the role of temperature as a potential control on fractionation: I had misinterpreted the potential range of temperature experienced by their fluids today (i.e. only with a few deg C, L78 in revised ms). However, the response does not account for the thermal history of the circulating seawater, but I interpret from e.g. Fig 4d that the penetration of seawater into the seamounts is not deep -and therefore not at elevated T, as otherwise it would have distinct Sr isotope composition etc.? Perhaps mention this around L78 for the sake of clarity, if so. A note was added in lines 83-86 regarding the penetration depth of seawater.
-issues with the numerical model previously used (e.g. assumptions associated with Rayleigh distillation model, etc.). These are largely resolved with the RTM model now employed, though other issues now appear. For example, model sensitivity to the prescribed parameters (Tables S3,4,5) is not investigated. Perhaps this is not necessary: my understanding is that the true value of model lies not in its ability to pinpoint one precise set of parameters, but rather to show that the dissolution of serpentine and subsequent re-precipitation of secondary phases (whether talc, or something else) can produce fluid δ30Si in the same range as the observations. One thing that I missed is a quantification of the fraction of Si being re-sequested into the talc. Presumably there is a trade-off between fluid d30Si and fluid [Si] -as d30Si increases, [Si] decreases. Does this have implications for the importance of the flux at a regional-to global-scale? We have investigated sensitivity tests for the Si isotope fractionation with different Δ 30 Si shown in Fig. 6c, which is one of the least constrained variables in the model. The reviewer is correct, that further sensitivity tests were not carried out. The model was set up to show general pathways of processes during serpentine alteration. We are not able to derive reliable values for transport velocities, mixing rates, and reaction rates because our simple 1-D steady state model does not resolve the 3-dimensional spatial structure of the two different seamounts and the temporal variability of transport processes. Given these model limitations -that are acknowledged in the main manuscript-we decided to not conduct further sensitivity tests. The depth-integrated rates of serpentine dissolution and authigenic mineral precipitation calculated in the model nearly balance each other out (773 versus 768 µM Si cm -2 yr -1 , respectively) so that about 99% Si is removed from pore fluids during authigenic mineral formation (now stated in lines 199-200 in the revised manuscript). This would result in a flux of 5 µM Si cm -2 yr -1 . However, given that the model cannot sufficiently resolve the 3D nature of the system (as also stated in the manuscript) we refrain from transferring the flux results to a regional-and global-scale.
The issues raised by other reviewers included the utility of δ11B, and the ability of extrapolating insights or fluxes from serpentine alteration reactions (as done here) to the broader suite of basalt alteration reactions. I think the authors make a reasonable case in the revised version as to the benefits of a dual-isotope approach, and the arguments based on secondary phase mineralogy are helpful in assessing the extension of the conclusions to other regions of the ocean. However, it seems to me that δ11B is not being used as a well-understood tracer to 'calibrate' the δ30Si data, as the text/rebuttal seems to imply, but rather as a system that is thought to largely be sensitive to the same processes (i.e. secondary phase formation), so that correlation between the two supports the interpretations, and aids in their model parameterisation. I think this is OK, but if this is not the case then the text should be clarified to highlight where some knowledge of the system is inferred on the basis of d11B and/or [B] alone, and then the silicon isotope interpretation made on the basis of this knowledge. We agree with the reviewers comment that both isotope systems show a similar response to fluidrock reactions and both isotope systems were used to identify the involved processes. We followed the suggestion of the reviewer and did not change the phrasing in the main text.
Some minor comments on the revised version of the manuscript: L151: "Motivated". The word was corrected.
L162: "the rapid dissolution of serpentine, which is undersaturated with respect to serpentine" -something is off here, please rephrase. The sentence was revised. Are all figures necessary in the main text? It seems like 3 and 4 could potentially be combined, while 5 might be better suited in the supplement. On figure 2, is it possible to display a) mixing hyperbolas between seawater and mantle fluid, and b) model predicted vectors of fluid evolution in δ30Si-δ11B space? This would allow the reader to assess the extent to which the general assumptions implicit in the RTM can replicate the range of data observed. Given that the current number of figures is still below the maximum number allowed by the journal, we have decided to leave all figures in the main text. Regarding Fig. 2, we have included a mixing curve between seawater and the deep mantle fluid. We decided not to include the model data in Fig. 2, because we did not want to shift the focus of Fig. 2 from the measured data to the model data. Further, the current logic and structure of the manuscript would be affected and the incorporation of the model data would require re-writing and re-organization of the manuscript.
Reviewer #2 (Remarks to the Author): Dear editor, this revised manuscript has addressed most of the comments from the reviewers. I am happy with their revision. i would like to suggest acceptance if the authors can address a few more things.

A few words need further clarification.
line 139: change Si fractionation to Si isotope fractionation. The word 'isotope' was included in the sentence.
line 149: what is "isotope values"? The expression was changed to 'isotope value of both elements'. line 241: change "More data is needed" to "More data are needed". The word was changed to plural.
2. line 57-58: "B is an incompatible element". "incompatible element" is not referred to the partitioning behaviour between fluid and solid, but melt and solid.
We have rephrased the sentence in lines 61-63: ' All my comments from my previous review have been addressed in a satisfactory way. They provide an updated manuscript. It retains their Si and B isotope data and provide a novel approach to understand the role of serpentinites and its alteration in the Si global cycle. They added a new section using a one-dimensional transport-reaction model to explain their data and estimate fractionation factors during serpentinite formation and alteration. Dealing with open systems is always problematic and the authors have done a good work in addressing the problematics and come with a model to interpret their data that is solid. Overall their model requires significant serpentine dissolution and loss of some Si from the serpentine muds to seawater. This is shown using pore fluid d30Si and I think is a solid result. However, before I recommend this paper for publication it requires some minor changes in the text as well as some important clarifications with data treatment and figure 2.

Specific comments
Line 30 I suggest that the authors change seawater to hydrous fluids and modify the following sentence accordingly. I agree that serpentinization is usually thought as a seawater-rock interaction. However, serpentinization can occur with any hydrous fluid and in settings different than the seafloor. On the context of this manuscript the serpentine being brought up is a result of water/rock interactions in the mantle wedge and the fluid is likely different in composition than seawater. The word seawater was replaced by hydrous fluids accordingly.
Line 104 Authors mention a "very good correlation" between the two isotope systems. I think this is correct for the pore fluid data with seawater fingerprint. However, their figure is shown in a way that appears to indicate that the correlation continues through their serpentine mud data. This is an artifact of the break in the y-axis (figure from table S1 data below). They also left out a single data point for mantle derived fluid that I think should be added to the figure. We agree that the figure implied a wrong impression of the data and we changed it accordingly (removal of the y-axis break). Additionally, we added the data point from the mantle-derived fluid and emphasized in the caption that the δ 30 Si value may be affected by precipitation during fluid ascent.
Authors also need to clarify how they derived the serpentine mud data point. Their Table S1 does not have pore fluid data for d30Si 1498B and Table S2 does not have serpentine mud data for d11B. My guess is that the authors are pairing "mantle-like" fluid d11B from Fantangisña 1498B with d30Si measured in the serpentine muds from 1498B but it is not clear from the figure or the text, if this is the case it needs to be stated in the text or the figure caption as the authors are combining different measurements to get to that data point (d30Si measured in the serpentine muds and d11B measured in pore fluids with mantle derived signature) while leaving out their only data point with both measurements for mantle-derived signature. We added an explanation for the serpentine mud data point in the caption and included the data point from the mantle-derived fluid.
Line 138 Typo "serpentinized" missing an i. The typo was corrected.
Line 160 I think that there is a typo "slap" should be slab. The typo was corrected.
Line 161-163 This sentence needs some rewriting to improve clarity on what is undersaturated with respect to serpentine. The sentence was revised. Two seamounts were investigated (Yinazao and Fantangisña), which are located at 55 km 78 and 62 km from the trench axis, respectively (Fig. 1a, b). Three drill cores were recovered 79 from Yinazao (1491B and C, 1492B) and three from Fantangisña seamount (1498A and B, 80 1497B; Fig. 1a, b; Table S1). Drill cores were taken from the flanks of the seamounts (except 81 coring location 1497B, which was located close to the seamount summit), in order to study   (Fig. 1c). In contrast 99 to the large isotope variability observed in the fluids, serpentinite muds show only a small 100 range in δ 30 Si between -0.6‰ and +0.1‰, independent of seamount and sampling site 101 (Table S2, Fig. S1). The fluids are further characterized by high B concentrations (from 245 to 102 758 µM B) with a wide range in δ 11 B (+16.1 to +43.5‰), which encompasses δ 11 B of 103 seawater (seawater B concentration: 432.6 µM 29 ; δ 11 B SW : +39.6‰ 30 ) (Fig. 1c). 104 The combination of pore fluid δ 11 B and δ 30 Si values shows a very good correlation of these 105 two isotope systems (Fig. 2). For both systems, the isotope values are below and above the 106 seawater signature, indicating dissolution of presumably serpentine and precipitation of 107 authigenic mineral phases, respectively. In order to decipher the processes controlling 108 mineral dissolution and precipitation, fluid sources and compositional changes in relation to 109 depth need to be examined.  (Fig. 3, 4). The strong increase in Si and B concentrations in the surface layer is induced by 170 rapid dissolution of serpentine, which is undersaturated with respect to serpentine due to 171 the entrainment of pH-neutral seawater (Fig. 5). The persistent mixing in the surface layer 172 with seawater supports high dissolution rates (~1.2 µmol cm -3 yr -1 ) whereas the rates are low 173 in the deeper layers (<10 -5 µmol cm -3 yr -1 ) that are not affected by mixing with ambient 174 seawater (Fig. 5). Serpentine, which formed in the mantle wedge, becomes unstable after 175 deposition on the seafloor either as serpentinized clasts or as serpentinite mud and begins 176 to dissolve during interaction with seawater via hydrolysis 35 , following: 177 178 3 ( 2 5 )( ) 4 + 5 2 → 3 2+ + 2 4 4 + 6 − , 179 180 thereby releasing Si and B into the pore fluid (see also upper red circle in Fig. 1b; Fig. 6a,b). The coupling of Si and B isotopes showed that the light isotope is preferentially incorporated 239 in authigenic minerals and that the combination of both isotope systems can trace 240 serpentine alteration reactions (Fig. 2, 6). Pore fluids with δ 30 Si and δ 11 B values lower than 241 seawater (δ 30 Si from +0.4‰ to +0.8‰; δ 11 B from +37.5‰ to +38.2‰; Fig. 1 Table S1). We further hypothesize, that the Si isotope signature of  (Table S2)  Mg 2+ concentrations (Fig. 3) and modeled silica concentrations (Fig. 6). c) Rate of serpentine