Olivine Dissolution in Seawater: Implications for CO2 Sequestration through Enhanced Weathering in Coastal Environments

Enhanced weathering of (ultra)basic silicate rocks such as olivine-rich dunite has been proposed as a large-scale climate engineering approach. When implemented in coastal environments, olivine weathering is expected to increase seawater alkalinity, thus resulting in additional CO2 uptake from the atmosphere. However, the mechanisms of marine olivine weathering and its effect on seawater–carbonate chemistry remain poorly understood. Here, we present results from batch reaction experiments, in which forsteritic olivine was subjected to rotational agitation in different seawater media for periods of days to months. Olivine dissolution caused a significant increase in alkalinity of the seawater with a consequent DIC increase due to CO2 invasion, thus confirming viability of the basic concept of enhanced silicate weathering. However, our experiments also identified several important challenges with respect to the detailed quantification of the CO2 sequestration efficiency under field conditions, which include nonstoichiometric dissolution, potential pore water saturation in the seabed, and the potential occurrence of secondary reactions. Before enhanced weathering of olivine in coastal environments can be considered an option for realizing negative CO2 emissions for climate mitigation purposes, these aspects need further experimental assessment.


■ INTRODUCTION
Climate engineering approaches that aim to deliberately and actively remove greenhouse gases from the atmosphere are categorized as carbon dioxide removal (CDR) or negative emission technologies (NETs). 1 CDR or NETs are seen as a possible future complement to current climate policies, which are presently only focused on the reduction of CO 2 emissions. Model studies emphasize that the large-scale implementation of NETs will be needed to limit global warming to within a 2°C increase with respect to preindustrial conditions. 2,3 The worldwide commitment to attain this target, thus avoiding a "dangerous" level of climate change, has been strengthened by the recent COP21 Paris Agreement. 4 Enhanced silicate weathering (ESW) is a NET approach in which the natural process of (silicate) rock weathering is artificially stimulated. 5,6 The technique has been recognized as a potentially promising strategy for CO 2 removal from the atmosphere while at the same time counteracting ocean acidification. 1,6−9 The implementation of ESW requires suitable source rock to be mined, ground to small grain sizes, and subsequently spread over suitable areas. 5 The mineral grains dissolve (i.e., chemical weathering), through which CO 2 is eventually captured from the atmosphere. 6,10 Olivine (Mg 2−x Fe x SiO 4 ) is an abundant and fast-weathering ultramafic silicate mineral and has been advanced as a prime candidate mineral for ESW application. 5,11 The dissolution of olivine in an aqueous environment consumes protons or equally increases alkalinity, 6,12,13 and so increases CO 2 uptake by the aqueous medium (Supporting Information section 1).
In theory, ESW can be applied in terrestrial soils, 5,6 in the surface mixed layer of the open ocean, 13 or by spreading minerals onto sediments of the coastal zone and continental shelf. 14 The largest application domain for ESW would be the open ocean, but model analysis suggests that the olivine particles need to be ground to very small sizes to facilitate dissolution in the surface ocean. 13 As such, the high grinding costs and CO 2 emissions during production potentially limit this approach. An alternative scheme is the application of ESW to coastal and shelf environments, where it could be integrated into existing coastal zone management practices, such as dredging operations, land reclamation, and beach nourishment. A theoretical examination of the concept of coastal ESW 14 has indicated advantages as well as challenges. One important knowledge gap is that detailed experimental investigations of olivine dissolution under natural conditions (i.e., realistic for coastal ESW) are lacking. A better understanding of the rate and mechanism of olivine dissolution in natural marine environments is needed to better evaluate the feasibility and potential of coastal ESW as a NET. Previous work has largely focused on olivine dissolution under laboratory conditions using artificial seawater solutions. 15−18 Such idealized approaches potentially exclude important geochemical and environmental influences that could be relevant under field conditions.
Here, we specifically address a number of questions related to the application of enhanced silicate weathering in natural coastal environments: (1) What is the rate of olivine dissolution in natural seawater and how does this differ from artificial seawater? (2) Does olivine dissolve stoichiometrically in natural seawater? (3) What dissolution products can be used to efficiently monitor the dissolution rate of olivine in coastal sediments, i.e., quantify the efficiency of enhanced silicate weathering? (4) To what extent does secondary mineral formation diminish the CO 2 sequestration efficiency of olivine dissolution in seawater?
We present results from dissolution experiments with simulated grain−grain collisions, in which olivine was dissolved in natural filtered seawater as well as in artificial seawater media with modified cation composition. Potential proxies for quantifying the dissolution rate of olivine are analyzed and compared. Based on these results, we discuss a number of challenges for ESW in coastal environments.
Different reactive seawater media were used as supernatant. Filtered seawater (FSW) was collected as natural seawater from the Oosterschelde tidal basin (The Netherlands) and filtered over a Mahle amaGuard FP 0.2 [μm] woven cotton filter (Mahle Benelux, The Netherlands). In addition, three types of artificial seawater were prepared according to the ASTM Standard Practice D 1141−98 19 (Table S3): (1) plain artificial seawater (ASW), (2) artificial seawater with Ca 2+ replaced by Na + (hereafter named ASW-Ca), and (3) artificial seawater with both Ca 2+ and Mg 2+ replaced by Na + (hereafter named ASW-CaMg).
Experiments. Specific amounts of olivine and quartz grains were added to a specific volume of seawater in 500 mL borosilicate glass bottles, which were then subjected to continuous rotating movements on a CH-4103 rotating shaking platform (INFORS AG, Switzerland) set at 155 rpm. The bottles were closed with membrane screw caps, equipped with a 0.5 mm thick silicone septum that prevented evaporation but allowed gas exchange. To prevent photosynthesis, the experiment took place in the dark, while the bottles were wrapped in aluminum foil (leaving the top of the membrane cap uncovered).
A total of three types of agitation experiments were conducted: A1, A2, and A3 (Supporting Information section 3 and Table S4). In both the A1 and A2 experiments, 0.1 mol of either olivine (OLI) or quartz (QUA) were added to 300 mL of FSW, with one control treatment (i.e., only seawater; SW). A total of three replicates were conducted per treatment (n = 3). Experiment A1 was conducted under ambient conditions, i.e. without regulation of temperature (range: 13.5−20°C) and pCO 2 (range: 445−525 ppmv), and lasted for 88 days. The second agitation experiment, A2, had essentially the same setup as A1 except for a few modifications. First, natural seawater was bubbled with air prior to the experiment to ensure CO 2 equilibration with the surrounding atmosphere. Second, the experiment had a much shorter duration (20 days), and it was conducted under stable temperature and pCO 2 conditions. The third experiment, A3, was designed to specifically investigate the effect of the composition of seawater on the dissolution rate of olivine (quartz was not investigated). Agitation experiment A3 was designed to examine the effect of the composition of seawater on the dissolution rate. A3 was also conducted under stable temperature and pCO 2 conditions and used atmosphereequilibrated reactive fluids, bubbled with air. The dissolution of olivine (OLI) was monitored in four reactive fluids (FSW, ASW, ASW-Ca, and ASW-CaMg; see the Supporting Information section 2) and compared to control treatments (respective solution media without olivine). Instead of 0.10 mol olivine used in A1 and A2, and 0.03 mol was used in A3 (Table  S3).
To test the impact of agitation, a layer of olivine sand was placed also in a nonmoving cylindrical container with FSW (n = 1). In this nonagitated treatment, the olivine itself was not agitated. Rather, the overlying water was stirred, and only samples for solid-phase analysis were collected.
Water and Solid-Phase Analysis. The overlying water of the slurry batch reaction experiments was sampled at regular time intervals and analyzed for temperature, salinity, pH, total alkalinity (TA), dissolved inorganic carbon (DIC), dissolved silicate (Si), dissolved nickel (Ni), and dissolved magnesium (Mg) using standard analytical procedures 20,21 (see section 3 of the Supporting Information). All solute concentrations are reported as μmol per kg of seawater (μmol kg −1 ).
Upon completion of the A3 experiment, olivine grains were recovered from the agitated and nonagitated treatments, inspected for dissolution features, and analyzed for carbonate precipitates, according to Nieuwenhuize et al.,20 to yield the mass percentage of inorganic carbon (mass% C inorg ). The elemental composition of the olivine particle surfaces were investigated using scanning electron microscope energydispersive X-ray spectroscopy (SEM−EDX). Additional details on both water and solid-phase analyses can be found in section 3 of the Supporting Information.

Article
Olivine Dissolution Rate Calculations and Simulations. The accumulation over time of the reaction products in the reactor vessels is reported as excess concentration values, ΔC i (t) = C treatment (t) − C control (t). In this, the control refers to the treatment without the addition of any solid minerals. A total of three empirical mathematical models were implemented to describe ΔC i (t) as a function of the incubation time, and from these model fits, the accumulation rate R i [μmol kg −1 day −1 ] of compound i and the associated area-specific dissolution rate constant k i [μmol m −2 day −1 ] were derived (section 4 of the Supporting Information). Note that in the case of stoichiometric dissolution, the rate constants k i should be, at least in theory, identical for all olivine dissolution products.
The accumulation of weathering products and the change of solution chemistry during the batch dissolution experiments was also mechanistically simulated using the geochemical software package PHREEQC v2 22 · In these simulations, the solution chemistry (and, hence, the saturation states) were free to evolve with time as a consequence of mineral dissolution (kinetic rate equations specified in section 4 of the Supporting Information). Solubility constants were taken from the MINTEQ.dat and LLNL.dat databases to calculate the saturation states of solid phases in the solution. The measured initial composition of the solution, which was specific for each treatment, was used as the starting conditions for the PHREEQC simulations.
All mathematical analyses, apart from the PHREEQC analyses, and plotting were done using the open source R framework for statistical computing. 23 ■ RESULTS

Olivine and Quartz Dissolution in Natural Filtered
Seawater. In the A1 and A2 experiments, we investigated the dissolution of olivine and quartz in natural filtered seawater. In both A1 and A2, there was a clear ΔSi signal in the quartz treatment (QUA), most likely caused by dissolution of very fine quartz particles ( Figure 1). ΔSi increased until ∼18 μmol kg −1 within the first week of the experiments, after which it remained constant. There was no discernible Ni release in the A1 and A2 quartz treatment (Figure 1), and hardly any response from the carbonate system. The ΔpH increased by 0.05 within the first 2 weeks but then decreased again to its initial value by the end of the experiment. While ΔTA remained constant with time, ΔDIC decreased with 22 μmol kg −1 during the first 15 days, likely caused by CO 2 outgassing, as the initial solution in A1 may not have been in equilibrium with the atmosphere (sections 2 and 3 of the Supporting Information). In the A2 experiment, the experimental procedure was improved, and the FSW medium was bubbled with ambient air at the start of the experiment. As a result, the carbonate system variables ΔDIC, ΔTA, or ΔpH did not change significantly over time (linear regression, p = 0.35, p = 0.28, and p = 0.696, respectively).

Article
In the olivine treatment (OLI) of experiment A1, increases with time of both ΔSi and ΔNi were observed, suggesting olivine dissolution ( Figure 1). While ΔNi leveled off at 3.2 μmol kg −1 , suggesting that an equilibrium was reached, ΔSi increased almost linearly over the 88 day incubation period, displaying a much-stronger dissolution than in the QUA treatment. Experiment A2 showed a similar temporal evolution of ΔSi and ΔNi ( Figure 1). ΔTA showed a pronounced increase over the first 5 days (Figure 1), leveling off thereafter at ΔTA = 103 μmol kg −1 . ΔDIC increased in a similar way, albeit over a slightly longer period of ∼15 days, leveling off at ΔDIC = 93 μmol kg −1 . The pH in the olivine treatment increased rapidly over the first 5 days by ∼0.1, after which it decreased again and leveled off at ΔpH = 0.02 ( Figure 1). The carbonate system in A2 showed a similar pattern, with ΔTA and ΔDIC leveling off at 104 and 74 μmol kg −1 , respectively ( Figure 1). Similar to experiment A1, the ΔpH in A2 increased strongly within the first 6 days by ∼0.15, after which it decreased again to ΔpH = 0.06 ( Figure 1).
Olivine Dissolution in Artificial Seawater Media with Different Cation Composition. In experiment A3, large differences in the release of dissolution products were observed between the different seawater media. All four media displayed a quasi-linear ΔSi response with time ( Figure 2). The ΔSi attained at the end of the experiment was lowest in the natural seawater (FSW: 68 μmol Si kg −1 ) and artificial seawater (ASW: 82 μmol Si kg −1 ) and increased markedly when Ca 2+ and Mg 2+ were replaced by Na + in the medium (ASW-Ca: 122 μmol Si kg −1 ; ASW-CaMg: 162 μmol Si kg −1 ).
The Ni release showed a saturation-type response, which varied between media. In the FSW, ASW, and ASW-Ca treatments (Figure 2), the ΔNi concentration showed a comparable accumulation (plateau concentrations between 1.19 and 1.68 μmol Ni kg −1 ). In contrast, the ASW-CaMg treatment showed hardly any Ni accumulation, apart from a small initial release, which was taken up again by the end of the experiment.
The carbonate system (TA, pH, and DIC) responded very differently in the Ca-and Mg-free seawater compared to the three other treatments. Although the shape of the response curves were similar, the overall accumulation of alkalinity (ΔTA = 340 ± 14 μmol kg −1 ) and dissolved inorganic carbon (ΔDIC = 317 ± 11 μmol kg −1 ) was substantially higher in the ASW-CaMg compared to the other treatments (range ΔTA = 41−69
The ΔpH showed an "overshoot" response, with a sharp initial increase in the first 5 days, reaching maximum between 4 and 7 days, after which the ΔpH gradually decreased and tended toward an asymptotic equilibrium value at the end of the experiment (Figure 2). Consistent with the stronger alkalinity accumulation in the Ca-and Mg-free seawater, the long-term pH (pH t137 minus pH t0 ) increase was much higher in the Ca-and Mg-free seawater (ΔpH = 0.06), compared to the other three treatments (range in ΔpH = −0.01 to 0.035 at 137 days).
The accumulation rates R were determined by the best model fit 23 (see Figure S6 for representative examples) to the response curves of the olivine dissolution products ΔSi, ΔNi, ΔMg, and ΔTA and the ensuing CO 2 sequestration ΔDIC ( Table 1). As emphasized above, different dissolution products tended to have different response curves within the same treatment, thus indicating nonstoichiometric dissolution (Table  1). In the case of ΔSi, the accumulation response was generally best described by a linear model (model 1, Table S5), while for ΔTA, ΔDIC, ΔMg, and ΔNi, the profiles were typically best fitted with a saturation model (model 2, Table S5) or a combination of short-term saturation with a long-term linear increase (model 3, Table S5). Only in the ASW-CaMg treatment (experiment A3) could all variables be described best by model 3 (Table S5) To further illustrate the absence of stoichiometric dissolution in either natural or artificial seawater in the presence of magnesium, Figure S8 plots the accumulation of response variables in experiment A3 normalized for stoichiometry (Table  S1: ΔSi/1, ΔTA/4, ΔNi/0.0075, and ΔMg/1.87). Only in the ASW-CaMg treatment did olivine dissolution tend to become stoichiometric, as shown by the similar responses for ΔMg and ΔTA ( Figure S8).
Because of nonstoichiometric dissolution, the olivine dissolution rate constant k i showed a dependence on the response variable (ΔSi, ΔNi, ΔMg, ΔTA, and ΔDIC; Table 2).
The rate constant based on ΔNi (k ΔNi ) is the highest of all response variables and is similar across all treatments (31−74 μmol of olivine m −2 day −1 ). In the ASW-CaMg treatment, k ΔMg (63 μmol of olivine m −2 day −1 ) was consistent with k ΔNi values, while k ΔSi were an order of magnitude lower than k ΔNi values in the reactive fluid media containing Mg 2+ (Table 1 and 2). The exclusion of Mg 2+ in the Mg-free reactive fluid (ASW-CaMg) increased k ΔSi by 1 order of magnitude. The values of both k ΔTA and k ΔDIC show substantial variation between treatments, and are highest in the ASW-CaMg treatment. The temperaturenormalized 24 (to 25°C) mean values for k i (where i = ΔSi, ΔNi, ΔTA, or ΔDIC), for the FSW and ASW cases are shown in Figure 3 (the ASW-Ca and ASW-CaMg treatments are considered unrealistic for ESW and are thus excluded).
SEM−EDX. SEM−EDX analyses of mineral grains from fresh, unreacted olivine were generally angular, with sharp edges ( Figure 4A). In contrast, olivine grains that had been rotating during the entire experiment (137 days) were generally subrounded ( Figure 4B), suggesting abrasion due to grain− grain collisions. The Mg-to-Si atomic ratios (Mg/Si) at the surface of the unreacted particles were significantly higher (mean ± SEM Mg/Si = 2.11 ± 0.02, n grains = 6; Figure S9) than for grains that were agitated in solution (mean ± SEM Mg/Si    Environmental Science & Technology Article 1.7 ± 0.04−2 ± 0.03, n grains = 3−10; Figure S9). This suggests preferential mobilization of Mg during dissolution, consistent with the higher dissolution rates obtained for Mg and Ni compared to Si. The preferential leaching of Mg 2+ (lowest Mgto-Si ratio) was most prominent in the ASW-CaMg treatment ( Figure S10), where areas with Mg/Si ≤ 1 and lower were observed. No carbonate minerals were observed on any of the analyzed olivine grains.
Potential for Carbonate Precipitation. The inorganic carbon content (C inorg ) in the solid mineral phase recovered from experiment A3 was very low (mean C inorg < 0.005%) and was not significantly different between the four treatments (one-way ANOVA, p = 0.112; Figure S9). Small changes in the solid-phase carbonate content (which are difficult to measure) could nevertheless be associated with substantial changes in the alkalinity of the supernatant. Although not significantly different, the difference in C inorg content between FSW and ASW-CaMg was 0.003 mass percent. If this difference would be real and caused by carbonate precipitation, this would imply that the FSW contained 37 μmol kg −1 of CaCO 3 in excess to the ASW-CaMg, when expressed per unit volume of fluid. The absence of this precipitation would hence cause the alkalinity to be 74 μmol kg −1 higher in ASW-CaMg. However, the measured ΔTA difference between the ASW-CaMg treatment and the FSW and ASW was much higher, amounting to ca. 300 μmol TA kg −1 at the end of experiment A3 (Figure 2). Accordingly, carbonate precipitation cannot explain the difference in alkalinity between the FSW and ASW-CaMg treatments, and so it was likely that more olivine dissolution took place in the ASW-CaMg treatment.

■ DISCUSSION
The dissolution experiments here demonstrate several features regarding olivine weathering in seawater and its potential applications for ESW in coastal settings. First, the basic principle underlying ESW in seawater appears to work. Olivine dissolution in natural seawater under nonsterile laboratory conditions consistently causes alkalinization, followed by CO 2 invasion from the atmosphere into the seawater, at rates in agreement with those estimated by previous studies. 14,24,25 Second, apparent nonstoichiometric dissolution complicates the experimental determination of the rate and extent of olivine dissolution within the seabed, making it more challenging to assess of the efficiency of ESW. Therefore, the quantification of the actual olivine dissolution rate under realistic in situ conditions will require a multiparameter approach, combining flux measurements of dissolved silicate, dissolved metals (nickel and iron), and alkalinity, with appropriate experimental controls. Third, the rate of olivine dissolution within the seabed can be limited by saturation effects, which could decrease the efficiency of ESW applications. Fourthly, the extent to which secondary reactions impact the CO 2 sequestration efficiency of olivine dissolution under in situ conditions remains unresolved and is an important issue to address in further studies on coastal ESW. We will now discuss each of these aspects in more detail.
Quantification of the Olivine Dissolution Rate. To be implemented as a negative emission technology for climate change mitigation, 7 the carbon sequestering potential of marine olivine dissolution needs to be quantified. In other words: How much olivine dissolution occurs within the seabed? What is the time frame in which olivine particles react? How much CO 2 is eventually taken up by the seawater as a consequence of ESW?
The overall CO 2 sequestration rate (R CO 2 ; mol CO 2 per m 2 of seabed per unit of time) can be expressed as To determine the effectiveness of coastal ESW, both the factors γ CO 2 and R OLI need to be accurately constrained. The CO 2 sequestration efficiency γ CO 2 specifies the net amount of CO 2 that is taken up from the atmosphere during the dissolution of 1 kg of olivine within the seafloor (this parameter will be further discussed below). For a given amount of finely ground olivine distributed onto the seafloor (C OLI ; mol of olivine m −2 of seabed), the olivine dissolution rate (R OLI ; mol of olivine per

Environmental Science & Technology
Article m 2 of seabed per unit of time) determines over what time frame the ESW application will be effective (dissolution period τ = C OLI /R OLI ). The olivine dissolution rate R OLI further depends on the specific surface area of the mineral grains (A surface ; m 2 g −1 ) and the intrinsic dissolution rate constant (k OLI ; mol of olivine per m 2 of grain surface area per unit of time). In practical ESW applications, the olivine dissolution rate (R OLI ) within the seabed can be determined experimentally by monitoring the release of olivine dissolution products from the seabed. This poses the question as to which dissolution product (e.g., Mg 2+ , Si, and TA) should be monitored as a reliable proxy for the olivine dissolution rate in field-type experiments. The use of both dissolved silicate and alkalinity is nontrivial as these are generated in sediments by other processes than olivine dissolution. 26 Any observed sediment efflux of dissolved silicate and alkalinity can thus not be exclusively attributed to olivine dissolution. Furthermore, Mg 2+ cannot be used as a dissolution proxy due to the high background concentration in seawater (∼50 mmol of Mg 2+ kg −1 of seawater), and thus, its accumulation in the overlying water cannot be reliably measured.
Our experiments suggest that Ni 2+ could be a suitable dissolution proxy, generating a dissolutive accumulation, which substantially supersedes the ambient seawater concentration 27,28 (0.002−0.16 μmol Ni kg −1 ). This way, pore water accumulation and sediment fluxes of Ni can be accurately measured using standard analytical techniques for trace metals (e.g., inductively coupled plasma−mass spectrometry). However, to qualify as a good proxy for sedimentary dissolution of olivine, two important conditions need to be fulfilled. First, the efflux of the olivine proxy (Ni) from the sediment should also match the release rate of the weathering products in the pore solution. In this regard, Ni 2+ seems an advantageous proxy. The natural cycling of Ni 2+ in coastal sediments is restricted, and so the observed Ni 2+ efflux from the sediment in olivine addition experiments can be fully attributed to olivine dissolution. Still, in future studies, it should be verified whether the Ni 2+ release is modulated by diagenetic effects within the sediment (e.g., sorption onto minerals).
A second important condition is that stoichiometric dissolution of olivine occurs, so that the Ni 2+ release can be properly rescaled to the overall olivine dissolution rate R OLI by means of the Ni content of the olivine source rock that is used 25 · However, our experiments suggest that this not the case (Tables 1 and 2). The experiments A1−A3 show that, when assessed over short-time scales, the dissolution of olivine in seawater is nonstoichiometric, implying incongruent dissolution under Earth surface conditions 29 · In the absence of secondary precipitation reactions, nonstoichiometric dissolution cannot continue indefinitely. Experiments of longer time scales should therefore clarify to what extent the Ni 2+ release from sediments can be a valid proxy for in situ olivine dissolution.
In our dissolution experiments, the k i value for Si (determined at the initial time t 0 ) was ca. 30 times lower compared to that of Ni (or Mg). This suggests a preferential release of divalent cations, respective to silicate. The preferential release of metal cations (Ni 2+ and Mg 2+ ) compared to Si, and the observed quasi-linear increase in ΔSi are typical for solid-state diffusion in silicate minerals, 29 which facilitates the formation of a "surface (altered) layer". 30 The crystal ionic radius of nickel (83 pm) is only slightly smaller than that of magnesium 31 (86 pm), implying that both metal ions will have a similar rate of solid-state diffusion. Both Pokrovsky and Schott 25 and Palandri and Kharaka 22 already suggested that for slightly alkaline solutions (e.g., seawater), forsterite dissolution at steady-state is controlled by the decomposition of a protonated surface complex, which is silica-rich and magnesium-deficient. Maher et al. 32 postulate that olivine dissolution occurs as a series of boundary layer processes, in which primary dissolution of cations is followed by dissolution of silicic acid ions, which may subsequently repolymerize at the surface. This implies that the measured dissolved silica release rate is a net value, which may not serve as the sole proxy for olivine dissolution. Although in some SEM−EDX images -particularly those from the ASW-CaMg treatment (Figure S10), the particle surface did look as if flakes of surface material had been detaching, and the examined olivine grains did not show any evidence of secondary silicate precipitates. 17,33−35 A significant finding here is that SEM−EDX analyses show decreasing Mg-to-Si atomic ratios of the forsterite surface between initial substrate and reacted material ( Figure S9). Rather than a buildup of thick silica formations, these decreasing Mg-to-Si ratios corroborate the mechanism of a cation-leached, surface altered layer formation by preferential dissolution and subsequent repolymerization processes sensu Hellmann et al. 30 and Maher et al. 32 The time scale on which the weathering takes place in this study is much longer than in high-temperature and high-pressure studies or studies in which an elevated pCO 2 is employed. 29 This so-called "unstrained dissolution", 36 combined with physical disturbances, such as grain abrasion, does not allow for the buildup of a conspicuous passivating layer or thick silicate precipitates.
The nonstoichiometric dissolution as observed in the experiments here emphasizes that the proper quantification of olivine dissolution in field-type ESW experiments requires a careful experimental design. Overall, the nonstoichiometric dissolution of olivine makes the experimental assessment of ESW more challenging. One cannot simply measure one dissolution proxy (e.g., Ni 2+ ) and estimate the release of other reaction products by application of reaction stoichiometry. Moreover, both dissolved silicate and alkalinity are generated in sediments by other processes than olivine dissolution. 26 From a biogeochemical perspective, it is crucial to know how olivine dissolution stimulates the efflux of dissolved silicate and alkalinity from the seabed because the efflux of alkalinity is the ultimate driver of CO 2 uptake, 37 while silicate could stimulate primary productivity by marine diatoms. Hence, a multiparameter assessment, combining flux measurements of Ni 2+ , dissolved silicate, and alkalinity with appropriate experimental controls, seems to provide the best strategy to confidently determine the olivine dissolution rate under in situ conditions. Impact of Saturation. The values for the dissolution rate constant k of olivine in seawater obtained in this study are consistent with literature values. For the temperature ranges used in FSW and ASW, the dissolution rate constant varied between 1.9 ± 0.8 μmol of olivine m −2 day −1 for k Si and 56 ± 18 μmol of olivine m −2 day −1 for k Ni (mean ± SD values). Normalized for temperature differences, these k value ranges corresponded well with the mean value of 14 μmol of olivine m −2 day −1 , as compiled by Palandri and Kharaka 22 and Hangx and Spiers, 14 of which the latter had an order of magnitude of variation around the mean (Figure 3).
The nonstoichiometric dissolution in the seawater media FSW and ASW, together with the saturation behavior observed in the TA, Mg 2+ , and Ni 2+ results (Figure 1), suggest that the

Environmental Science & Technology
Article olivine dissolution approached thermodynamic equilibrium, thus slowing down the reaction. Only by using a lower solubility product for forsterite (log K = 26.448) than those found in the PHREEQC databases did model simulations indeed show a slowing of the dissolution reaction by saturation, mirroring our experimental observations. Furthermore, the DIC accumulation followed that of alkalinity perfectly in all cases, albeit with a time lag. This lag is due to the relatively slow process of CO 2 invasion 38 and is also observed in the pH response, which first increases to reach a maximum and then subsequently decreases again. This pH response reflects the initial removal of protons through olivine dissolution, followed by a replenishment of the proton pool by lagged CO 2 transfer.
The observed time response of the reaction products in our experiments provide a first idea about the possible influence of saturation effects under in situ conditions. Our experiments show that saturation occurs within a time frame of about 20 days (Figures 1 and 2), for an experimental setup with 15 g of olivine in 300 mL of seawater (i.e., 20 mL of solution g −1 of olivine). Assuming the same dissolution rate occurs under in situ pore water conditions, the ratio of pore solution to olivine will be lower. For example, if 10−20% of the solid sediment consists of olivine (mixing a 1−2 cm olivine layer into the top 10 cm of sediment) and assuming a porosity of 0.8 and an olivine particle density of 3.3 g mL −1 , we obtain a ratio of 6−12 mL of solution g −1 of olivine. Based on our results, we determined that such a pore solution will be saturated within 4.5 to 9 days, after which olivine dissolution will slow down and cease. However, the pore water of coastal sediments is also regularly refreshed through physical, advective pore water flow induced by waves and currents 39 and biological irrigation by burrowing macrofauna. 40,41 Coastal sediments subject to moderate and high bioirrigation show flushing rates in the range of 10−100 L m −2 day −1 , 39 implying that the pore solution of the first 10 cm would be refreshed on a time scale of 0.5−8 days (assuming a porosity range from 0.5 to 0.8). In these sediments, irrigation appears sufficient to counteract the saturation of olivine dissolution in the pore water. However, in more cohesive (muddy) coastal sediments with flushing rates <10 L m −2 day −1 , saturation effects can be expected, which could decrease the efficiency of enhanced olivine weathering applications. Therefore, a judicious choice of the application location seems warranted. 42 CO 2 Sequestration Efficiency. The experiments in this study were performed in a setup that allowed free gas exchange with the atmosphere, while internal biological processes affecting the DIC pool were excluded (i.e., primary production and microbial degradation of organic matter). 26,37 Accordingly, the observed DIC increase in the experiments can be entirely attributed to CO 2 invasion induced by olivine dissolution, illustrating the proof-of-principle that ESW enhanced silicate weathering works as a NET.
The CO 2 sequestration efficiency expresses the amount of CO 2 transferred across the air−sea interface per unit mass of silicate rock that dissolves within the seabed and can be written as This formulation reflects the two consecutive steps in the process of CO 2 sequestration. In a first step, olivine dissolution takes place (rate R OLI ), which increases alkalinity in the pore solution (rate R TA ). This alkalinity increase will then shift the acid−base equilibrium from dissolved CO 2 to bicarbonate and carbonate, thus stimulating a CO 2 uptake from the atmosphere across the air−sea interface (rate R CO 2 ). 26,37 The CO 2 sensitivity(∂ΣCO 2 /∂TA) pCO 2 specifies how much CO 2 is taken up from the atmosphere for each mole of alkalinity that is released from the seabed. This thermodynamic factor is evaluated at a given partial pressure of CO 2 in the atmosphere and is dependent on the local salinity, temperature, and chemical composition of the coastal seawater. 43 Calculating the CO 2 sensitivity over the entire experimental period, and for all the experiments that had full ionic strength of seawater (A1, A2, A3: FSW and A3: ASW), we obtain a CO 2 sensitivity of 0.84 ± 0.1 (mol of DIC mol −1 of TA), which is in close agreement with the theoretical value 0.854 for seawater at the experimental conditions employed (T = 17°C, S= 33, TA = 2400 μmol L −1 , and pCO 2 = 400 ppmv). 44 Accordingly, the CO 2 uptake in our experiments appears entirely congruent with the standard acid−base thermodynamics of the carbonate system in seawater. Nevertheless, the alkalinity increase during olivine dissolution, R TA /R OLI , was less than expected. Traditionally, olivine dissolution is described by the reaction equation: Because 4 moles of protons are consumed per mole of olivine dissolved, and hence 4 mol of alkalinity are produced, a ratio (ΔTA/ΔSi = 4 (R TA /R OLI = 4)) is expected. Only in the ASW-CaMg treatment, the ΔTA/ΔSi approached the expected value of 4, while it was substantially less in the FSW, ASW, and ASW-Ca treatments. These observations suggests that the Equation 3 does not provide a complete description of the overall olivine dissolution process, but that secondary reactions could be active. Overall, CO 2 sequestration efficiency can be formulated as Here, 4 denotes the theoretical stoichiometry between olivine dissolution and CO 2 , 5,14 and x denotes a reduction in the CO 2 sequestration efficiency due to secondary reactions. Because our slurry experiments were conducted with oxygenated seawater, one such possible reaction is the aerobic oxidation of ferrous iron: This reoxidation process produces free protons, thus consuming again the alkalinity generated during dissolution of the Fe-component of olivine. The olivine employed here contains 6% of Fe (Table S1), which would reduce the alkalinity release by an equal percentage. However, this reduction is not enough to explain the observed ΔTA/ΔSi values. Another possibility to reduce the ΔTA-to-ΔSi ratio is calcium carbonate precipitation: Although the supernatant in the FSW and ASW was saturated with respect to calcite and aragonite, magnesium is known to act as an inhibitor for CaCO 3 nucleation in seawater, limiting sequestration efficiency of enhanced olivine weathering in seawater. However, no sepiolite was found in the XRD analyses. The extent to which secondary reactions impact the CO 2 sequestration efficiency of olivine dissolution under in situ conditions within the seabed remains an important issue to address in further studies on coastal ESW.
Due to the exclusion of Mg 2+ and Ca 2+ , the ionic strength of the ASW-CaMg medium was lower than that of the ASW (Table S3). The ionic strength of the ASW was 0.72 mol kg −1 , while that of ASW-Ca was 0.015 mol kg −1 (or 2%) lower than ASW. Equally, the ionic strength of ASW-CaMg was 0.0975 mol kg −1 (or 13%) lower than that of ASW. Ionic strength impacts the activity coefficients of aqueous species and has been found to impact dissolution kinetics, particularly at lower pH. 16 Still, other factors (i.e., pCO 2 , pH, and saturation state) exhibit a much larger influence on dissolution kinetics 48,49 · Given the relatively high pH in the reactive fluids (pH 7.9−8.2) and the fact that all solutions were highly undersaturated with respect to fosterite, the impact of the lower ionic strength of the ASW-CaMg was likely to be very small.
The rate at which CO 2 is sequestered due to olivine dissolution in seawater can thus be formulated as the following relation: Here, 4 denotes the theoretical stoichiometry between olivine dissolution and CO 2 , 5,14 R OLI is the olivine dissolution rate, γ CO 2 is the reaction efficiency of the CO 2 sequestration in seawater, and x is the molar fraction of Fe in the olivine source material.
Olivine Application in a Coastal Geo-Engineering Framework. To place coastal ESW in a broader perspective, a real-world example illustrates its carbon-capturing potential. The Netherlands is a densely populated, industrialized country, with a GDP of ca. 850 billion USD (2013) and ca. 50% of its surface area below sea level. 50 To protect the coastal region of the country where ca. 60% of the GDP is produced, 51,52 continuous large-scale sand nourishments are needed. Between 2000 and 2010, ca. 12 million m 3 (Mm 3 ) sand per year have been deployed along The Netherlands' coast, which is expected to increase due to predicted climate change-induced sea level rise 51,52 (https://www.noordzeeloket.nl/en/functions-and-use/ surface-mining-and-quarrying/).
In a thought experiment, the sand used in these coastal nourishments is replaced by finely ground olivine, as used in the experiments described here. In a hypothetical one-time application of 12 Mm 3 (≈ 26 Mt) of olivine sand, parameter values for k ΔTA obtained in our experiments (Table 1) were implemented in the Olsen 53 shrinking core model for olivine carbonation (assuming the measured olivine particle size distribution; see section 2 of the Supporting Information). This model has been previously implemented in ten Berge et al., 54 describing total mass of olivine weathered and consequential CO 2 captured (section 8 of the Supporting Information). Our simulations showed a cumulative weathering of 4% of the olivine after the first year, 12% after 5 years, 35% after 25 years, 57% after 50 years, and 84% after 100 years ( Figure 5A). After 200 years, 98% of the initially applied 12 Mm 3 olivine will be dissolved. These values are in accordance with those presented by Hangx and Spiers,14 in which 100 μm (median diameter: D 50 ) olivine grains would take >100 years to dissolve.
Making use of the earlier derived relationship, eq 6, with γ CO 2 = 0.84 and x = 0.06 as discussed above, the amount of carbon dioxide taken up can be estimated. With annual 12 Mm 3 applications, for periods of 1, 5, 10, and 25 years, the CO 2 -

Environmental Science & Technology
Article capturing rate would increase from ca. 2.5 Mton CO 2 year −1 to a peak value of ca. 9 Mton CO 2 year −1 after 25 years of coastal olivine application ( Figure 5B). This would be the equivalent of 5% of The Netherlands' yearly 170 Mton CO 2 emissions (2013 value; http://data.worldbank.org/indicator/). Once the application stops, the remaining olivine will dissolve in about 250 years, with decreasing yearly CO 2 uptake rates ( Figure 5B). The long time scale over which ESW is effective has two important implications. First, the process of issuing and validating carbon credits for ESW will need to take into account that CO 2 sequestration is not immediately realized at once but stretched out over a century-scale time window. Second, given the long-lasting effects, any potential ecosystem impacts need to be properly assessed and evaluated upfront in small-scale field trials before large-scale ESW application can start.
Environmental Implications. From an ecological perspective, the potential secondary effects of (large-scale) olivine dissolution are a critical issue. Although dilution processes in marine coastal environments will likely prevent accumulation to toxic levels of dissolution products, it is important (and obligatory in e.g. the European Union) to perform upscaling calculations of dissolution product concentrations and their conceivable effects on the marine ecosystem. The main consequences of forsteritic olivine dissolution are increases in Mg 2+ , Si, TA, DIC, Fe 2+ , and Ni 2+ , and their ecosystem effects should be thoroughly assessed. In addition, the geophysical consequences of olivine distribution in coastal ecosystems should be assessed, such as the increase in suspended particulate matter, sediment pore space clogging and smothering effects due to the higher specific density of olivine. While increases in alkalinity and DIC are a desired effect for climate engineering purposes, the increase in Mg 2+ is not expected to pose a significant threat because of the high background concentration in seawater. Increases in dissolved Si and dissolved Fe can stimulate primary production and thus lead to additional CO 2 sequestration, as recently assessed by model analysis 55 · However, the ultimate impacts on coastal foodwebs of fertilizing by olivine dissolution are uncertain and need further investigation.
The impact of increased nickel flux on marine ecosystems is a matter of potential concern, and has only been scarcely touched upon. Nickel leaches from the olivine mineral matrix in its ionic Ni(II) form. Dissolved nickel occurs in trace concentrations in seawater (0.03−0.16 μmol kg −1 ; 27 ), as low as 0.002−0.006 μmol kg −1 in the central southern North Sea and up to 0.04 μmol kg −1 in the Rhine delta area. 28 In comparison, background nickel concentrations in the control treatments ranged between 0.14 μmol kg −1 in the FSW in experiment A3 and 0.45 μmol kg −1 in experiment A1, while background Ni concentrations in the artificial seawater media in experiment A3 (ASW, ASW-Ca, and ASW-CaMg) were an order of magnitude lower, between 0.017 and 0.032 μmol kg −1 .
The ecotoxicology of nickel in marine organisms and ecosystems is summarized on the Web site of the UK Marine Special Areas of Conservation (http://www.ukmarinesac.org. uk/) and established for the UK at a chronic concentration of 0.25 μmol L −1 . Nickel toxicity has been reported in a number of cases: 56−58 negative effects on spawning in mysiid shrimps at 2.4 μmol L −1 , DNA damage with associated physiological and cytotoxic effects in the blue mussel Mytilus edulis at 0.3 μmol L −1 , disrupting ionoregulatory functions in the green crab Carcinus maenas between 8.5 and 51 μmol L −1 in very low-salinity seawater (0.006 PSU), and organ oxidative stress in the killifish Fundulus heteroclitus, also mainly in freshwater. However, one of the conclusions of Blewett et al. 57 and Blewett and Wood 58 is that higher, seawater-like salinities (e.g., 30−38) seem to be negatively correlated with Ni-induced effects. In general, higher salinities are inversely correlated with Ni 2+ seawater concentrations. 59 Although bioaccumulation of nickel in individual organisms occurs, there seems to be little evidence of biomagnification throughout (marine) foodwebs, 27 although Kumblad et al. 60 present results that suggest the contrary. The potential toxicity of nickel, combined with rather large uncertainties about the magnitude and direction of its response effects, make it paramount to further investigate its ecotoxicological effects within the framework of large-scale application of olivine in coastal environments.
Containment is not an issue for ESW. Before any field-scale application, there should be proper field trials in quasicontained conditions, such as mesocosm setups, which can be upscaled in, e.g., tidal harbor basins. In the case that a mesoscale field trial (∼100 m 2 ) would be undertaken, common dredging equipment would be used to apply the olivine into the (coastal) environment. The same equipment and expertise can be used to remove the olivine sand, should any acute unforeseen situation develop.
The CO 2 sequestration induced by ESW is governed by the acid−base thermodynamics of seawater, which are wellunderstood, 26,37,38 therefore rendering the containment of CO 2 in the ocean highly predictable. The central premise of ESW is that it increases the ocean's alkalinity, enabling more CO 2 to be dissolved into seawater at any given pCO 2 compared to the situation in which no alkalinity is added to the ocean. This CO 2 will be stay dissolved in the ocean (or contained) as long as no other process changes the alkalinity of the ocean. In the ESW, the CO 2 storage reservoir (the ocean) is an open system, as CO 2 can be freely exchanged between atmosphere and ocean across the air−sea interface. Because leakage cannot occur in an open system, storage of CO 2 in the ocean is therefore leakage-proof. The evidence that CO 2 will be contained for long periods of time is given by observations on the long-term (>1000 year) carbon cycle and the impact of natural silicate weathering: the long-term fate of fossil CO 2 is to be absorbed in the ocean. 61 If ESW is applied to coastal systems in a geo-engineering framework, it will be crucial to determine in situ olivine dissolution rates to determine the efficiency of the method. 42 Once in the natural sediment, the olivine will be subject to very different biogeochemical and geophysical conditions. Microbial mineralization processes could greatly increase the CO 2 concentration in the sediment's pore waters, 62 while benthic macrofauna process vast quantities of sediment for their sustenance and mobility. 63,64 These processes are likely to speed up the dissolution process within marine sediments. Large-scale sediment transport and wave action are expected to cause constant particle abrasion and faster mechanical weathering, in turn facilitating faster chemical weathering. If ESW is ever to be applied in a geo-engineering framework, it is of paramount importance to investigate the effects of all of these natural processes on the dissolution of olivine in coastal environments.

Environmental Science & Technology
Article ■ ASSOCIATED CONTENT

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.est.6b05942.
Experimental details of olivine solution, composition of minerals and media, overview of experimental conditions, calculation of rate constants, model simulations, accumulation of reaction products, solid-phase analysis, and calculation of total mass of olivine weathered and CO 2 captured. (PDF)