Hydrothermal oxidation of the ferrous iron contained in the steel slag (Equation 1) by liquid water leads to the formation of hydrogen and magnetite (Malvoisin et al., 2013). This reaction was investigated in a 500 ml autoclave equipped with gas and solution sampling capillaries (see Crouzet et al., 2017). From 2 to 20 g of steel slag are inserted in 200 ml of either de-ionized water or 1 M acetic acid solution (starting pH of 2.5). The autoclave is heated by two external collars. Temperatures in the 250–350°C range were investigated. An initial pressure of 50–100 bar of argon (Ar) was set at ambient temperature to reach 180–230 bar at the temperature set-point. The slag suspension is stirred at a rotation speed of 500 rpm. H₂ production will be expressed hereafter as the mass of H₂ divided by the equivalent FeO mass content of steel slag. Potentially, up to 9.28 g H₂ per kg FeO can be produced according to the following reaction:

The sampled gas is injected into a gas chromatograph (Clarus 500 GC, Perkin Elmer) run under a constant argon gas flow of 30 mL/min; the gas components (H₂, CO₂, N₂, O₂, CO, CH₄) are separated with a polymer filled column (Restek ShinCarbon®). Chemical identification and quantification were performed by comparing retention times with a reference gas mixture using a thermal conductivity detector (TCD). The temperature of the detector, the injection system and the oven were respectively set to 250, 100, and 80°C. The determination of the produced H₂ mass which involves sampling and GC measurement is believed to be accurate to ±10%.

Aliquots of 2 ml aqueous solution were sampled with a plunging capillary and collected in a 20 ml syringe after filtering through a 0.45 μm porous nitrocellulose membrane. Iron content was then determined by UV-spectroscopy after complexation by orthophenantroline. This method allowed a fast quantification with a detection limit of 0.1 ppm and an estimated error of 10%. When lower iron concentrations were expected, total iron content was quantified by ICP-OES. The pH of the recovered aliquot was measured at ambient temperature with pH paper indicators.

The effect of temperature and initial acid acetic concentration on the hydrothermal oxidation of steel slag was first investigated by conducting a set of experiments in 2-cm long gold-capsules. Around 80 mg of steel slag and 80 mg of de-ionized water are loaded in a gold tube (outer diameter of 4.4 mm). The two ends of the tube are welded shut. The encapsulated sample is placed in a cold-seal pressure vessel, itself introduced in a horizontal furnace. Three different initial concentrations of acetic acid, 1, 2, and 4 moles per liter, were investigated at temperatures of 150 and 300°C for a run duration of 3 days. Argon pressure was fixed to 300 bar. The gas entrapped in the capsule was recovered following the method described by Malvoisin et al. (2013) and injected with a 250 μl syringe into a gas chromatograph for analysis.

Three experiments were first conducted with steelmaking slag and de-ionized water in a sampling autoclave at 250, 300, and 350°C (Figure 1A; APW-250C, APW-300C & APW-350C in Table 3). Whatever the investigated temperature, about half of the H₂ is produced during the first 24 h. After 24 h and up to the end of the experiment (i.e., 80 h), the H₂ production rate slows down and remains roughly constant. Hydrogen production kinetics were all successfully fitted to a square root function of time (Equation 2). A significant dependence in temperature was observed. After 72 h, 0.87, 2.13, and 8.07 g of H₂ / kg equivalent FeO were respectively produced at 250, 300, and 350°C. The A_n constant (Equation 2) was estimated to 0.11, 0.28, and 0.87 g H₂/h^1/2/kg FeO at 250, 300, and 350°C, respectively (Table 3). Assuming an Arrhenius behavior, an activation energy of 28 ± 4 kJ/mol is retrieved (Figure 1B).

The steel slag oxidized at 300°C was further characterized (APW-300C). Time-resolved composition of the aqueous solution was monitored by ICP-OES (Figure 2). By order of abundance, Ca, Si, Mg, Fe, Ni, and Mn were found to be the main dissolved components. Their concentrations remained roughly stable as reaction proceeded. Average calcium concentration reached 10⁻³ mol/L whereas iron concentration was close to 10⁻⁷M and so could not be quantified by the complexation–UV spectroscopy method. An estimated pH of 11 is obtained from the measured Ca concentration (assumed to occur as Ca²⁺,_aq) and neglecting possible CO₂ aqueous species.

XRPD was conducted on dried solid sample (Figure 3). Mineral modes obtained from semi-quantitative Rietveld refinement are presented in Table 2. Although Ca-species dominate in the HP-HT solution (Figure 2), CaO remains the main oxide component in the solid residue after hydrothermal oxidation. Members of the hibschite-katoite series Ca₃Al₂(SiO₄)_3−x(OH)_4x were found to represent the dominant mineral phases. Jaffeite, an hydrated calcium silicate Ca₆Si₂O₇(OH)₆, was also detected. Calcium is also found in calcite, portlandite and hydroxylapatite. Iron was only detected as wüstite and magnetite. Minor brucite, Mg(OH)₂, is observed.

In addition to hydrogen production, steel slag hydrothermal oxidation led to the formation of magnetite as shown by XRPD (Figure 3). In order to better characterize the magnetite component of the experimental product, the magnetite fraction was separated using a two-step magnetic separation procedure, respectively in water and in a 2% hydrochloric solution. The result of the first step of separation in water can be evaluated on Figure 4 where magnetite represents now, by far, the dominant mineral constituent. However, many other calcium-bearing phases initially present in oxidized steel slag remained. Consequently, a second magnetic separation step was performed in dilute hydrochloric acid in order to dissolve and remove these unwanted phases. A solid fraction composed of only magnetic oxides was successfully recovered (Figure 5a). From Rietveld semi-quantitative analysis, magnetite was found to represent 80 wt.% of the sample whereas the remaining 20 wt.% are assigned to wüstite.

Images of the separated iron oxide particles are displayed on Figures 5b,c. Particles were found to fall in three size ranges: 10–30 nm, 100–300 nm, and 1–10 μm. Cubic shaped particles are dominant within the 10–30 nm range. Flat and needle shaped particles are predominant in the 100–300 nm range. SAED pattern was obtained from a single nanoparticle and indexed as magnetite (Figure 5d). However, note that the SAED pattern is also close to that of maghemite.

Steel slag was reacted with 500 ml solution of 4 M acetic acid at ambient temperature for 60 min in order to dissolve portlandite, calcite and larnite. Chemical analyses performed by μXRF before and after the treatment (Table 1) confirmed the selective dissolution of Ca and Si; Ca and Si contents were divided by 2.5 and 1.5, respectively. Every other oxide proportions were subsequently increased by a factor of about 2. Mineralogical changes were inspected using XRPD (Table 2). After acetic acid attack, only iron-bearing oxides were found as residual crystalline phases; wüstite and brownmillerite are the dominant mineral phases which coexist with minor magnetite whereas no more crystalline silicates were detected.

Hydrothermal oxidation of the decalcified steel-slag sample was performed at 150 and 250°C in a sampling autoclave with a 1 M acid acetic solution (APAc-1M-250C sample). Under such acidic conditions, the hastelloy^TM autoclave may itself react with the solution and produce H₂. A blank run with the same amount of 1 M solution was performed at 150 and 250°C. Whereas no H₂ was produced by the metal of the autoclave at 150°C, a total of about 1 mmol H₂ was produced after 51 h at 250°C (Figure 7). It should however be noted that this H₂ yield produced by the autoclave oxidation with pure 1 M acetic solution is an upper bound to the effective autoclave contribution in the presence of a slag sample since the partial dissolution of the latter will increase the pH of the reacting solution.

Around 2 g of decalcified slag was first oxidized for 24 h at 150°C. No H₂ was detected at this temperature (Figure 7). Ferrous iron dissolution occurred to a level of 8.10⁻⁴M (Figure 7) corresponding to less than 2 mol.% of the bulk iron content of the slag.

In the absence of significant reaction, temperature was then raised to 250°C. Pressure was decreased before heating in order to reach 180 bar at 250°C. Immediately after reaching 250°C, hydrogen production started at a roughly constant rate until a plateau of H₂ concentration was attained after an overall run duration of 72 h (i.e., after 48 h at 250°C). A yield of 6.93 g H₂/kg FeO equivalent is achieved after 75 h (i.e., 51 h at 250°C, Table 3) which, again, includes the H₂ arising from the autoclave corrosion under acidic conditions.

In parallel, ferrous concentration increased by one order of magnitude to reach 8.10⁻³M at 250°C. Ferrous iron concentration remained roughly constant until the end of experiment. A constant pH of 4 was measured at ambient in aqueous aliquots sampled at 150 and 250°C.

XRPD analysis was performed on the decalcified sample after its oxidation at 250°C in a 1 M acetic acid solution (Figure 8). Only three iron-bearing phases were identified: hematite, magnetite and a chlorite-group mineral with its characteristic 14 Å basal reflection. Phase abundances are presented in Table 2 but have to be considered with caution due to the unknown composition of the observed chlorite.

FE-SEM investigations were performed on the recovered particles without preliminary magnetic separation (Figure 9). Clusters of 5–50 nm magnetite particles with octahedral morphologies were easily found. Larger particles ranging from 100 to 300 nm with sheet-like morphologies were also observed which are interpreted as chlorite.

Chemical analysis performed by μXRF on the retrieved solid sample showed results closed to the decalcified steel slag before oxidation. Differences were mainly found in the unexpected presence of Ni (autoclave contamination) and the decrease in Mg and Ca.

Hydrogen production kinetics associated with the hydrothermal oxidation of a relatively fresh BOF steel slag sample (a few weeks of storage in open air) in de-ionized water has been studied at 250, 300, and 350°C (Figure 1). A similar overall kinetic behavior was observed at the three temperatures with the highest H₂ production rate encountered during the first 24 h, which progressively decreased for longer run duration. A very similar kinetic behavior was described by Crouzet et al. (2017) for reagent grade FeO reacting with pure water at 300°C. Even the H₂ yield normalized to bulk FeO content is very close at 300°C; after about 72 h, 2.13 g H₂/kg FeO equivalent were produced from steel slag oxidation whereas about 1.9 g H₂/kg FeO were found from FeO reagent grade oxidation. Two different models however were used to fit the FeO and steel slag oxidation kinetics. Steel-slag oxidation kinetics could not be successfully fitted to a first-order kinetic model, at least for the investigated run duration. Instead, H₂ production kinetics for steel slag were fitted to a simple square root function of time. After 72 h of reaction, 0.87, 2.13, and 8.07 g H₂/kg equivalent FeO were respectively produced at 250, 300, and 350°C (Figure 1A). Increasing the temperature led to higher H₂ production. By fitting steel slag oxidation kinetics with a square root model and assuming an Arrhenius behavior, an activation energy of 28 ± 4 kJ/mol was retrieved (Figure 1B). This value is in line with range of activation energies reported by Malvoisin et al. (2013), of 15 ± 2 and 41 ± 12 kJ/mol for uncarbonated and carbonated BOF slag, respectively.

On the only basis of two temperatures, the temperature dependence of steel slag oxidation kinetics as measured in the sampling autoclave is confirmed in experiments with acetic acid performed in gold capsule. Retrieved H₂ production yields were respectively multiplied by 23 and 25 at 1 and 2 M by raising the temperature from 150 to 300°C, corresponding to an approximate activation energy in the order of 20 kJ/mol. Moreover, Crouzet et al. (2017) found an activation energy of 27 kJ/mol for the oxidation of FeO in the presence of 0.05 M acetic acid in gold capsule. This activation energy does not allow to extrapolate oxidation kinetics obtained below 300°C toward higher temperature due to the likely thermal decomposition of acetic acid. However, it can be tentatively concluded that the clustering of activation energies around 30 kJ/mol for the production reaction of H₂ in steel slag and pure FeO, with or without acid acetic addition, is an indication of the same rate-limiting process, i.e., the ferrous iron dissolution from the wüstite component.

Crouzet et al. (2017) reported that iron dissolution is the rate limiting factor for the hydrothermal H₂ production from FeO in similar P-T conditions as those investigated here. They showed that the presence of mild acetic acid solution significantly enhances iron dissolution kinetics owing to a ligand-promoted dissolution process (Stumm, 1997; Hamer et al., 2003).

However, using the same acid acetic concentration will definitely not yield the same pH conditions in the case of the steel slag due to its high alkalinity (Huijgen and Comans, 2005; Malvoisin et al., 2013; Costa et al., 2016; Crouzet et al., in press). In order to achieve acidic conditions during steel slag oxidation with an in-situ pH < 5–6, two routes were tested through (1) the use of concentrated acetic acid to ensure the dissolution of all the Ca of the steel slag and (2) calcium removal at room temperature followed by the oxidation of the solid residue in a 1 M acetic acid solution.

For both routes, a clear kinetics enhancement has been noticed at 250 and 300°C. At 150°C, however, on the contrary to what has been found for reagent grade FeO (Crouzet et al., 2017), no significant H₂ production has been detected.

H₂ production and Fe(II) dissolution in 1 M acetic acid solution from Ca depleted steel slag sample have both been monitored in the sampling autoclave at a temperature initially fixed to 150°C and then raised to 250°C due to the absence of H₂ production (Figure 7). Aqueous iron saturation with respect to FeO and Fe₃O₄ in 1 M acetic acid has been calculated at 150 and 250°C and compared to the experimental data. Furthermore, at both temperatures, assuming that the slag is only composed of FeO and Fe₃O₄ and neglecting the effect of residual Ca, an in-situ pH in the 3.9–5.2 range is calculated which consistent with the pH of ca. 4 determined on aqueous aliquot after cooling. At 150°C, no significant H₂ is produced and the Fe²⁺ aqueous concentration remains at the 8.10⁻⁴M level. Aqueous Fe concentration at 150°C is by one to two orders of magnitude below Fe₃O₄ and FeO saturation, respectively. Either aqueous Fe is controlled by the solubility of a less soluble iron oxide/hydroxide phase(s) or the aqueous Fe content is limited by the slow FeO dissolution at this temperature. The fact that magnetite supersaturation is not attained is consistent with the absence of H₂ production. The situation becomes different once T is increased to 250°C. H₂ production starts to proceed and Fe²⁺ aqueous concentration increases by one order of magnitude to reach 8.10⁻³M which corresponds to magnetite saturation. FeO dissolution and Fe₃O₄ precipitation processes are thus triggered. Contrary to what has been observed by Crouzet et al. (2017) with respect to the hydrothermal oxidation rate of pure FeO at 150°C, FeO dissolution is not fast enough to produce significant magnetite supersaturation. These results on decalcified steel slag confirmed, however, the positive effect of acidic pH and acetic acid on the hydrothermal oxidation kinetics of FeO.

The positive effect of acetic acid on H₂ production was also evidenced from the oxidation experiments performed in gold capsules on native steel slag samples in the presence of concentrated acetic acid solutions. At 300°C, H₂ yield at 1, 2, and 4 M in gold capsules was found to be about four times higher than in the experiment performed with distilled water in the sampling autoclave. The effect of concentrated acid acetic might even be larger considering the fact that capsule experiments have a lower H₂ yield compared to sampling autoclave experiments (Crouzet et al., 2017). Finally, H₂ yields were also found to increase with initial acetic acid concentrations at 150 and 300°C.

By combining GC analysis of H₂ and saturation magnetization (Js, see Malvoisin et al., 2012 for the method), Malvoisin et al. (2013) showed a one to one molar relationship between H₂ and magnetite production in a BOF slag treated hydrothermally (pure water) in gold capsule using externally heated pressure vessels. However, Malvoisin et al. (2013) focused on H₂ as valuable product but did not consider the potential value of magnetite and they therefore overlooked grains with the smallest sizes. The identification, here, of abundant magnetite with sizes below 1 μm opens the way to a new valorization of steel slags owing to the multiple applications of magnetite in the industry (see Introduction section).

By applying a simple wet magnetic separation with a permanent magnet, we succeeded in extracting from the hydrothermally oxidized slag (pure water) an iron oxide-rich product containing a high magnetite proportion (Figure 4). FE-SEM and TEM imaging revealed that most of the newly formed magnetite crystals occur as nanoparticle agglomerates (Figure 5) in both original steel slag and decalcified samples, i.e., at basic and acidic pH values. Three main granulometric classes were identified: 10–30 nm, 100–300 nm, and 1–10 μm. Most of the particles display an octahedral morphology as it is usually encountered for grains of that size (Sun and Zeng, 2002; Daou et al., 2006; Horak et al., 2007; Iida et al., 2007). Needle-shaped grains are also observed whereas such morphology was neither observed in iron oxides produced hydrothermally from FeO (Crouzet et al., 2017) nor in steel slags oxidized with a CO₂ partial pressure (Crouzet et al., in press).

Due to its applications in medicine as contrast agent, a considerable number of studies were dedicated to the synthesis of magnetite nanoparticles (Daou et al., 2006; Gnanaprakash et al., 2007; Iida et al., 2007; Ge et al., 2009). Most of the syntheses are based on the co-precipitation method. This method consists in the dissolution of a stoichiometric mixture of ferrous and ferric salts in aqueous solution. The pH is then increased through the addition of a base (NH₄OH, N(CH₃)₄OH, H₂N(CH₂)₆NH₂…) to instantaneously decrease the iron solubility and promote the precipitation of magnetite nanoparticles. The particle size distribution was found to be dependent upon the initial proportion of ferrous ions in the starting solution (Iida et al., 2007). Higher ferrous initial proportion leads to the formation of smaller particles. To get larger particles and reduce their dispersity, Daou et al. (2006) proposed to perform an additional hydrothermal step at a temperature of 250°C. Particles centered on 39 nm were then obtained. Ge et al. (2009) proposed to use only ferrous salt as starting material. After the addition of the base under vigorous stirring in air and the formation of a magnetite suspension, hydrothermal conditions (134°C) are applied. The authors also confirmed the effect of the initial ferrous salt concentration on the size distribution.

As the co-precipitation method, the production of magnetite from hydrothermal steel slag oxidation occurred through a dissolution—precipitation process for which dissolution was identified as the limiting factor in pure water. Iron speciation and concentration (supersaturation level) are however drastically different when nanomagnetite particles are synthesized from hydrothermal oxidation of steelmaking slags and from co-precipitation methods. Due to the relatively high pH imposed by native steel slag, aqueous iron concentration remains very low, even at high temperature, e.g., 10⁻⁷ M of ferrous iron was encountered at 300°C. Contrary to the co-precipitation method, magnetite particles were not all produced simultaneously but over the whole oxidation reaction as shown with the time-resolved monitoring of hydrogen production. This particularity likely contributes to the formation of the observed particle size distribution. The largest micrometric particles may be related to the magnetite particles initially present in steel slag and which possibly grew during the whole hydrothermal treatment possibly by Ostwald ripening.