In this study, an iron blast furnace slag (BFS) and a basic oxygen furnace slag (BOFS) were obtained from Liberty Steelworks in Whyalla, Australia, and an electric arc furnace slag (EAFS) was obtained from BlueScope in Wollongong, Australia. Each slag was crushed using a Retsch BB50 jaw crusher and ground using a Pulverizer Laboratory ring mill, resulting in grain sizes averaging between 50 and 200 μm (Supplementary material). The elemental composition of each slag was determined using an inductively coupled plasma mass spectrometer (ICP-MS). The slag dissolution experiments were conducted using MilliQ (18.2 MΩ) and natural seawater (NSW). NSW was collected and treated as described in Moras et al. (2023).

A series of dissolution experiments were conducted in MilliQ to estimate the potential for alkalinity formation per mg of slag dissolving. Each slag was added in concentrations of 10, 25 and 50 mg kg⁻¹ of MilliQ in high-quality borosilicate 3.3 1 L Schott bottles. The nine bottles had initially little headspace to minimize CO₂ gas exchange and hence being able to attribute measured changes in DIC upon slag addition to carbonate dissolution or precipitation. They were kept at room temperature (air conditioning set at 21°C) and placed on a roller table at about 6 rotations per minute to keep the added slag particles in suspension. The pH was recorded daily to monitor the dissolution kinetics (calibrated using Metrohm buffers). Once a stable value was reached, TA samples were collected. Another set of TA samples was taken two weeks later to confirm that a steady state had been reached. The released TA was considered stable when the changes over two weeks was less than 5 μmol kg⁻¹.

Using the data from the MilliQ experiments, another set of experiments was conducted to assess the dissolution of slag in NSW. Five slag additions were performed, aiming to increase TA by 50, 100, 200, 350, and 500 μmol kg⁻¹. This corresponded to 5.04, 8.53, 24.84, 122.73, and 606.61 mg kg⁻¹ for BFS, 3.87, 7.80, 15.73, 27.58, and 38.44 mg kg⁻¹ for BOFS, and 4.48, 8.98, 17.95, 31.26, and 44.73 mg kg⁻¹ for EAFS. These additions were selected to cover a wide range of TA increase with a maximum total alkalinity target of +500 μmol kg⁻¹, identified as the limit beyond which secondary CaCO₃ precipitation could occur (Moras et al., 2022). The high-quality borosilicate 3.3 2 L Schott bottles were constantly rotated on a roller table for four weeks at about 6 rotations per minute and at room temperature (air conditioning set at 21°C). TA and DIC samples were collected before slag addition, and on days 1, 4, 7, 14, 21, and 28 days after slag addition. Similarly, samples for ICP-MS analysis were taken before slag addition for elemental analysis and 7, 14, 21, and 28 days after (see below for sampling procedures).

Samples for TA and DIC were collected and analyzed as per Moras et al. (2022). TA was analyzed in duplicate and DIC in triplicate. The TA and DIC values were corrected against secondary in-house reference materials since certified ones from Prof. Andrew G. Dickson’s laboratory could not be supplied during the COVID-19 pandemic (Moras et al., 2023). The TA and DIC measurement uncertainties were calculated by error propagation. The DIC measurement uncertainty was estimated at ±1.1 μmol kg⁻¹, and the TA measurement uncertainty at ±1.8 μmol kg⁻¹. Remaining carbonate chemistry parameters such as CaCO₃ saturation state of aragonite (Ω_A) or pH on the total scale were estimated from TA, DIC, salinity and temperature data (see Moras et al., 2022 for details) using the CO2SYS script for MATLAB (Sharp et al., 2021).

Elemental composition of slag was determined on an Agilent 7,700 ICP-MS, coupled to a laser ablation unit (NWR213, Electro Scientific Industries, Inc.). Powdered slags were embedded in resin for solid ICP-MS analysis. The samples were then polished and measured, before being corrected against standard reference materials, obtained from the National Institute of Standards and Technology (NIST, batch #610 and #612). For liquid elemental ICP-MS analysis, 12 mL of water were taken with a 60 mL luer lock sterile syringe and filtered through a Nylon Captiva EconoFilter (0.45 μm) into 15 mL Falcon tubes. Samples were diluted 10 times with 3%v/v nitric acid. Measurements were carried out on the same ICP-MS as for the solid analyses and measurements were corrected against a seawater reference material from the National Research Council of Canada NASS-6.

Additionally, to estimate the carbon content in powdered slags, ~100 mg of each slag was weighed and wrapped in tin capsules in triplicates, before being pressed into small balls of about 5 mm diameter. The samples were then analyzed using a Thermo Fisher Flash Elemental Analyzer, coupled to a Delta V Plus isotope ratio mass spectrometer.

The grain size distribution for individual slags was very similar, ranging mostly between 0.5 and 350 μm in diameter, with peaks around 50–200 μm (Appendix Figure A1). Concerning elemental composition, all three slags were rich in calcium, varying from 227 ± 51 mg g⁻¹ for BFS, to 387 ± 144 and 338 ± 95 mg g⁻¹ for BOFS and EAFS, respectively. Silicon and magnesium were also present in all slags, varying between ~64 and ~ 115 mg g⁻¹ for silicon, and ~ 43 and ~ 81 mg g⁻¹ for magnesium. With 61 ± 13 mg g⁻¹ of aluminium, BFS was nearly four times richer than BOFS and EAFS. In contrast, BFS contained hardly any iron, while concentrations in BOFS and EAFS were 134 ± 58 and 310 ± 35 mg g⁻¹, respectively. Some potentially toxic elements were also found, with 17–61 mg g⁻¹ of aluminium in all slags, up to 4.3 mg g⁻¹ of chromium in EAFS as well as 52.5 mg g⁻¹ of manganese. Finally, BFS contained ~1.5 ± 0.3 mg of carbon per gram of slag, BOFS ~7.6 ± 0.4 mg g⁻¹ and EAFS ~4.2 ± 0.1 mg g⁻¹ (Appendix Table A1).

The alkalinity generation potential of the three slags was tested in MilliQ water, at concentrations of 10, 25 and 50 mg of slag per kg of MilliQ. For BFS additions, stable conditions were reached after three weeks, while for BOFS and EAFS additions stable conditions were already observed after five days of experiment, meaning that complete alkalinity generation was about 4 times faster than for BFS. The alkalinity generation for BFS was not linear (Figure 1). In fact, with increasing amounts of BFS added, less alkalinity was generated. In contrast, over the range tested, alkalinity generation in BOFS and EAFS was linear. Also, about 10% more alkalinity was leached per mg of BOFS compared to EAFS.

Using the TA generation potential data for MilliQ, five additions per slag were conducted in NSW, with ΔTA targets of +50, +100, +200, +350 and + 500 μmol kg⁻¹. The maximum TA measured for the BFS experiments was much lower than the expected one, i.e., about half (Figure 2). While the TA remained stable in most treatments once reaching its maximum concentrations, the TA in the highest BFS addition decreased slightly after it had reached its maximum, i.e., ~230 μmol kg⁻¹, between day 7 and 14 (Figure 2). DIC increased overtime, varying between +20.2 and + 32.9 μmol kg⁻¹ over the 28 days of experiments, except for the highest BFS addition, where DIC increased by 56.3 μmol kg⁻¹. Ω_A followed the TA trends, with similar patterns for each addition except for the highest one. In the higher BFS experiment, Ω_A increased over the first 7 days to up to 4.8, before decreasing to 3.8 and then remaining stable around 4.0. In the BOFS and EAFS experiments, similar DIC, TA and Ω_A trends were observed (Figure 2). For both steel slags, TA and Ω_A quickly increased during the first day, reaching a steady plateau about a week later. In contrast to TA, DIC increased steadily over time, ranging between 20 and 40 μmol kg⁻¹, with lower concentrations for low and higher for high slag additions.

During the dissolution of all slags, [Si], [Al], and [Mn] increased gradually over the four weeks of experiments with more pronounced increases when more slag had been added (Figure 3; Appendix Figures A3–A5). This pattern was particularly clear in the changes in [Al] and [Mn] from BOFS and EAFS. However, in some instances, stable concentrations were observed over 4 weeks, in particularly for [P] or [Ni], and some elements concentrations decreased over time, e.g., [Fe], [Cr], or [Sn] (Figure 3; Appendix Figures A3–A5).

This study tested the dissolution of iron and steel slags in seawater, and their potential to generate alkalinity and leach trace elements in lab-scale experiments. Among the three slags sourced, one iron slag and two steel slags were used, with elemental compositions similar to those reported in the literature (Reddy et al., 2019). A first very interesting observation was that BFS dissolved much slower than BOFS and EAFS (Figures 1, 2). A logarithmic trend between BFS addition and TA generated was observed in MilliQ, indicative of reaching saturation, i.e., the more BFS is added, the less TA is generated per mass of feedstock. Furthermore, in all NSW treatments, BFS solubility was 46–48% lower than in MilliQ, further reducing its OAE suitability. The reason for the saturation of TA generation and lower solubility of BFS in seawater than in MilliQ is unclear at this stage. Based on the calcium and magnesium measurements in each slag, and assuming that all calcium and magnesium are present as CaO and MgO, respectively, all the TA generated by BOFS and EAFS (~12.72 mmol g⁻¹ and ~ 11.18 mmol g⁻¹, respectively) can be attributed to fully dissolving these compounds (Reddy et al., 2019). This is only the case for the lowest BFS addition in MilliQ, so it appears that calcium is present in a less soluble form than CaO. Which one is unknown, but it is not calcite or aragonite (insoluble in seawater) due to the low carbon content measured in all slags.

Nevertheless, an estimate of the carbon capture potential using slag can be derived. Ignoring BFS, due to the saturation behavior and further reduced solubility in seawater, on average 12.0 ± 1.1 mmol of alkalinity was generated per g of steel slag. With an average of ~0.8 mole of CO₂ captured per mole of TA generated in oceanic surface waters (Moras et al., 2022; He and Tyka, 2023; Schulz et al., 2023), ~9.6 mmol of CO₂ could be captured, corresponding to about 420 mg of CO₂ per gram of steel slag. At a global annual steel production of nearly 1.9 gigatonnes (1 Gt = 10⁹ tonnes) in 2021 (Worldsteel, 2022), and assuming that for every tonne of crude steel produced, ~0.125 tonnes of steel slag are generated (Tuck, 2023), using all available steel slag for OAE purposes could lead to about 100 Mt. of CO₂ captured per year at the present day. Similarly, modeling global steel production into the future for various shared socioeconomic pathways (Renforth, 2019), up to 22 Gt of cumulative CO₂ capture could be realized until 2,100 (Figure 4). Hence, the CO₂ sequestration potential of steel slags is significant and may be economically viable as no extra investments are required for production (Renforth and Henderson, 2017; Renforth, 2019). The numbers presented here are at the upper boundary for steel slag as they assume full CO₂ equilibration with the atmosphere after the alkalinity increase. However, they do not account for the CO₂ emissions during steel production. Furthermore, the capture potential is highly dependent on the location of OAE implementation. For example, while model simulations suggest that a ratio of TA generated to CO₂ captured close to 0.8 can be realized within years to decades, there are some places in which surface waters with increased TA will not stay in contact with the atmosphere for long enough to allow for full equilibration (He and Tyka, 2023).

For all three slags in all treatments, an immediate increase in DIC was measured, which cannot be explained by CO₂ ingassing as there was no headspace initially. Although, for BOFS and EAFS, the increase was relatively small and a measured decrease at the second and third sampling point, respectively, suggests that there might have been sampling/measurement issues, which however has no implications for any conclusions drawn here. For BOFS and EAFS, there was a general trend of a steady DIC increase over time, consistent with CO₂ ingassing, as expected. Altogether, this suggests that BOFS and EAFS do not release significant amounts of DIC during dissolution, which is important as it would reduce the CO₂ uptake potential. One can also come to the same conclusion by assuming that all the carbon contained in the slags would be carbonates that would dissolve in seawater (either soluble calcium or magnesium carbonates). At 7.6 and 4.3 mg of CO₃²⁻ per gram of BOFS and EAFS, respectively, only a ~ 5 and ~ 3 μmol kg⁻¹ increase in DIC at the highest TA addition would be expected.

In contrast, for BFS, DIC could have increased via the dissolution of carbonates by up to 15 μmol kg⁻¹ in the highest TA treatment. However, the measured DIC increase was much higher, and at this stage it is not clear what could explain this observation. In any case, this does not affect any of the conclusions drawn previously, and BFS has already been flagged as a suboptimal OAE feedstock. Finally, the apparent decrease in TA between the first and second week in the highest TA treatments with BFS is most likely a measurement artifact from imperfect sample filtration. CaCO₃ precipitation is unlikely as it should have continued in a runaway fashion as described previously, and a drop in DIC should also have been observed (Moras et al., 2022). Instead, it could result from the dissolution of not yet dissolved slag during TA titrations, even though no visual observations of particles were made (note that BFS dissolution kinetics were much slower than for BOFS and EAFS).

An important finding was the general stability of TA observed after the increase at the start of the experiments. Recent studies have highlighted the issue of potential secondary CaCO₃ precipitation following an increase in TA for OAE purposes (Fuhr et al., 2022; Moras et al., 2022; Hartmann et al., 2023). For instance, Moras et al. (2022) reported that at an Ω_A ~ 7.0, so called “runaway CaCO₃ precipitation” can occur, ultimately removing more TA than had been added. Surprisingly in our experiments, despite Ω_A levels above 7.0 in two of the steel slag experiments, no drop in TA was observed for almost a month. In contrast, in experiments with Mg(OH)₂, Ca(OH)₂ and CaO, CaCO₃ precipitation was observed within a few days at these conditions (Moras et al., 2022, Hartmann et al., 2023). The mechanisms behind CaCO₃ precipitation are complex but several factors are known to enhance or inhibit precipitation. The presence of mineral particles in suspension have been found as a main trigger for runaway CaCO₃ precipitation (Morse et al., 2003; Lioliou et al., 2007; Bustos-Serrano et al., 2009; Tang et al., 2020). Observed both in nature and in laboratory settings, these particles act as precipitation nuclei for CaCO₃. The reason for not observing any CaCO₃ precipitation here could be that steel slag has a lower lattice compatibility for CaCO₃ than pure CaO, Ca(OH)₂ and Mg(OH)₂, although how that would work is not clear, as it is CaO in those slags that is mainly dissolving and generating TA. Also, rather-unlikely is that an increase in dissolved Mg concentrations during slag dissolution could act as a natural inhibitor of CaCO₃ precipitation (Berner, 1975; Rushdi et al., 1992; Morse et al., 1997; Pan et al., 2021). Even if all Mg present in slags would be MgO, with Mg content varying between 4 and 8%, a maximum increase of about 90 μmol kg⁻¹ could be explained at the highest TA treatment. This is relatively small, considering a natural Mg background of ~52.8 mmol kg⁻¹ in natural seawater at a salinity of 35 (Dickson et al., 2007; Riebesell et al., 2011), and rather unlikely to significantly inhibit CaCO₃ precipitation.

The most striking difference between steel slags and the calcium and magnesium minerals investigated previously are their dissolution kinetics. While CaO and MgO fully dissolved within hours, steel slag dissolution was on the order of days. A lower dissolution rate would suggest a lower pH in the diffusive boundary layer around particles suspended in the water column, given all other factors being equal (particles size, mixing, etc.). Despite similar bulk seawater Ω_A, a lower pH at the particle surface could then indeed explain why no CaCO₃ precipitation was found here. It has to be kept in mind, though, that dissolution rates measured here were for 21°C, meaning that minerals contained in slag with retrograde solubility such as Ca(OH)₂ would dissolve faster at lower temperatures, which could lead to higher pH at particle surfaces. If that would lower the Ω_A threshold at which secondary CaCO₃ precipitation is actually observed, however, is not clear as at the same time, precipitation kinetics are slowed down at lower temperatures (Zhong and Mucci, 1989). Hence, slag dissolution experiments at different temperature are required.

If alkaline materials such as slags with various metal impurities would be used for OAE, an open question is the potential release of bioactive elements during dissolution. Using the slag composition data and the known amounts added to NSW, the maximum increase in potentially bioactive element concentrations can be calculated. For example, about 100 μmol kg⁻¹ of Ti could leach in the highest BFS addition and about 3.7 μmol kg⁻¹ of Cr in the highest EAFS addition. When comparing the potential changes in NSW elemental composition from solid ICP-MS analysis to the actual measured changes in NSW elemental composition, in most cases, the measured concentration was much lower than the potential ones, i.e., the plotted data being positioned to the right of the 1:1 line (Figure 3). In some instances, such as for Li or Ni, the measured concentration was close to the potential one, at least for the lower addition. A few data points were above the 1:1 line, most likely the result of measurement uncertainties. In some instances, such as for Si, Al and Mg, seawater concentrations clearly increased toward higher slag additions. Finally, for elements such as Li, Ni and Sn, seawater concentrations hardly changed indicating that some elements do not dissolve in seawater, at least not within the experiment timeframe. These elements may be locked up in minerals that have a low solubility in seawater, such as for instance Fe(III).

Using the Australian and New Zealand guidelines for fresh and marine water quality, nine potentially toxic elements for which harmful concentrations are reported were identified in the slags (Commonwealth of Australia, 2023). These were aluminium, chromium, cobalt, copper, lead, manganese, nickel and zinc. In all of our experiments, potentially toxic element concentrations were all lower than the safe levels reported in the Australian and New Zealand guidelines for fresh and marine water quality (Table 1). For BFS, Al would be the first element to reach toxic levels, followed by Mn. The same would be the case for EAFS, while for BOFS, the element order would be reversed. This means that up to 509.9 mg kg⁻¹ of BFS, 149.4 mg kg⁻¹ of BOFS and 332.7 mg kg⁻¹ of EAFS can be dissolved before potentially reaching toxic levels for an element. Such additions would correspond to an alkalinity increase of 483.7, 1899.8 and 3718.9 μmol kg⁻¹, respectively. For the two steel slags, this is most likely much higher than one would aim for, as of potentially negative effects of increased pH_T on marine biota (between 9.5 and 10.6 at 21°C, assuming open ocean conditions, i.e., initial TA = 2,300 μmol kg⁻¹, DIC = 2,100 μmol kg⁻¹ and salinity of 35), as well as pseudo-homogeneous CaCO₃ precipitation onto naturally occurring colloids and organic matter at Ω_A values above 12.3 (Ω_A would increase between 24.6 and 31.8, respectively). In essence, dissolution of trace elements, particularly in steel slags, appears unlikely to exceed potentially toxic levels, at least for the type of slags tested here. Similarly, the quick dilution naturally occurring following slag dissolution, especially in an open-ocean setting, would further reduce the concentration of potentially toxic elements. Such dilution would also allow for more slag to be dissolved.