**Synthesis of Poly-Alumino-Ferric Sulphate Coagulant from Acid Mine Drainage by Precipitation**

#### **Brian Mwewa 1,2,\*, Sre´cko Stopi´c 3, Sehliselo Ndlovu 1,2, Geo**ff**rey S. Simate 1, Buhle Xakalashe <sup>3</sup> and Bernd Friedrich <sup>3</sup>**


Received: 3 October 2019; Accepted: 24 October 2019; Published: 29 October 2019

**Abstract:** The wastes generated from both operational and abandoned coal and metal mining are an environmental concern. These wastes, including acid mine drainage (AMD), are treated to abate the devastating effects they have on the environment before disposal. However, AMD contains valuable resources that can be recovered to subsidize treatment costs. Two of the major constituents of coal AMD are iron and aluminium, which can be recovered and engineered to function as coagulants. This work examines the potential of producing a poly-alumino-ferric sulphate (AMD-PAFS) coagulant from coal acidic drainage solutions. The co-precipitation of iron and aluminium is conducted at pH values of 5.0, 6.0 and 7.0 using sodium hydroxide in order to evaluate the recovery of iron and aluminium as hydroxide precipitates while minimizing the co-precipitation of the other heavy metals. The precipitation at pH 5.0 yields iron and aluminium recovery of 99.9 and 94.7%, respectively. An increase in the pH from 5.0 to 7.0 increases the recovery of aluminium to 99.1%, while the recovery of iron remains the same. The precipitate formed at pH 5.0 is used to produce a coagulant consisting of 89.5% and 10.0% iron and aluminium, respectively. The production of the coagulant is carried out by dissolving the precipitate in 5.0% (w/w) sulphuric acid. Subsequently, the treatment of the brewery wastewater shows that the AMD-PAFS coagulant is as efficient as the conventional poly ferric sulphate (PFS) coagulant. The turbidity removal is 91.9 and 87.8%, while the chemical oxygen demand (COD) removal is 56.0 and 64.0% for AMD-PAFS and PFS coagulants, respectively. The developed process, which can easily be incorporated into existing AMD treatment plants, not only reduces the sludge disposal problems but also creates revenue from waste.

**Keywords:** acid mine drainage; precipitation; iron; aluminium; coagulation; water treatment

#### **1. Introduction**

Acid mine drainage (AMD) is one of the largest environmental threats facing the world today. It is rated second only to global warming and stratospheric ozone depletion in terms of its ecological effects [1]. Environmentalists have termed AMD the single most significant threat to South Africa's environment. AMD is caused by the oxidation of sulfur, present in the mineral pyrite (Fe2S). When exposed to water and air, either during mining operations, once the mine has been abandoned or as a result of natural weathering, the pyrite is oxidized, which leads to the generation of high acidity and ferrous iron-impacted waters [2,3]. There are a series of reactions and side reactions involved during the formation of AMD, with the overall reaction given by Equation (1). The presence

of AMD has the potential to devastate streams, rivers and aquatic life [4–8]. Such devastating scenarios necessitate the treatment of AMD to abate the effects it has on the environment.

$$4\text{FeS}\_2 + 14\text{H}\_2\text{O} + 15\text{O}\_2 \to 4\text{Fe(OH)}\_3 + 16\text{H}^+ + 8\text{SO}\_4^{2-} \tag{1}$$

For many decades, the most widely applied method for the treatment of AMD is an active treatment process involving chemical-neutralization reagents [9,10]. This technology entails the addition of lime to acidic waters to raise the pH and precipitate the dissolving of metals. This process produces a hydroxide sludge, which typically contains 2–5% solids [11]. The voluminous sludge is difficult to dispose of because of the scarcity of land. In addition, the process produces metastable phases whose long-term stability has not been established. Therefore, post-precipitate stabilization before final solids disposal is required. However, the metals in AMD can be recovered with the objective of obtaining valuable products while meeting the effluent discharge limitation [12]. This is one of the potential ways to extend the use of natural resources. This paradigm shift has led to a number of studies being conducted to investigate the recovery of valuable products from AMD, including iron oxide pigments for production of paint [13–15]; ferric oxide nanoparticles [16]; inorganic pigments [17]; metals like Fe, Al, Zn, and Cu [18–21]; and acid and water [22–25], and the use of AMD neutralization sludge in brick and cement production, and as an artificial soil additive [26–28].

There is also huge potential to recover alternative coagulants from AMD for water treatment. The coagulants that are widely used to remove a broad range of impurities from effluent, including colloidal particles and dissolved organic substances, are metal salts such as aluminium sulphate Al2(SO4)3.5H2O, aluminium chloride AlCl3, polyaluminium chloride AlCl3, ferric sulphate Fe2(SO4)3.5H2O and ferric chloride FeCl3 [29]. The actual coagulant species involved in the coagulation process are formed after the coagulant chemicals are added to water. The addition of these cationic species to water results in colloidal destabilization as they specifically interact with and neutralize the negatively charged colloidal particles. For example, when aluminium sulphate/chloride is dissolved in water, the Al ion Al3<sup>+</sup>, immediately coordinates with six water molecules, Al(H2O) 3+ <sup>6</sup> [29,30]. The hydrolysis reactions (e.g., Equation (2)) proceed with the formation of numerous mononuclear species, e.g., Al(OH) <sup>2</sup>+, Al(OH) + <sup>2</sup> , Al(OH)<sup>3</sup> (molecule) and Al(OH) − <sup>4</sup> , followed by the formation of three polynuclear species including but not limited to Al2(OH) 4+ <sup>2</sup> , Al3(OH) 5+ <sup>4</sup> and Al13O4(OH) 7+ <sup>24</sup> , as well as a solid precipitate [Al(OH)3]. The hydrolysis of Fe is very similar in many respects to that of Al. Flynn Jr. [31] studied the hydrolysis of ferric iron and reported five mononuclear species Fe3+, Fe(OH) <sup>2</sup>+, Fe(OH) + <sup>2</sup> , Fe(OH)<sup>3</sup> molecule and Fe(OH) − <sup>4</sup> , and dimeric species Fe2(OH) 4+ <sup>2</sup> and Fe3(OH) 5+ <sup>4</sup> .

$$2\text{Al}^{3+} + 2\text{H}\_2\text{O} \Leftrightarrow \text{Al}\_2(\text{OH})\_2^{4+} + 2\text{H}^+ \tag{2}$$

The high concentration of Fe and Al in AMD, as high as 5000 mg/L for Fe and 500 mg/L for Al, has led to studies that have focused on developing an understanding of its potential reuse as a coagulant in wastewater treatment. For example, a novel application of AMD for coagulation/flocculation of microalga biomass was developed by Salama et al. [32]. A coagulation efficiency of 89% and 93% was obtained for *S. obliquus* and *C. vulgaris*, respectively, with a 10% dose of AMD as a coagulant. Lopes et al. [33] used mine water directly as a coagulant for the treatment of sewage wastewater. AMD was effective in the removal of suspended solids, organic matter, phosphorus and bacteria of the coliform group. Another process for the direct use of AMD as a coagulant in municipal wastewater treatment was tested by Rao et al. [34] and compared with FeCl3. The AMD was found to be as effective as FeCl3. However, the treated water contained high residual heavy metals from AMD. This precluded its general use in water treatment without pretreatment to remove heavy metals. This led to other studies being conducted to recover ferric sulphate coagulant by reacting the ferric hydroxide precipitate formed from AMD at pH 3.5–3.6 with sulphuric acid [34,35]. The use of dodecylamine surfactant to avoid co-precipitation of other metals, thereby improving the purity of the precipitate, was also tested. The recovered coagulant was effective in municipal wastewater treatment and compared favourably with conventional coagulants.

This study is motivated by the work done by Jiang and Graham [36], who produced a poly-aluminoiron sulphate (PAFS) coagulant using chemical grade aluminium sulphate Al2(SO4)3.5H2O and ferric sulphate Fe2(SO4)3.5H2O salts as the two primary raw materials. The coagulant was evaluated for the removal of colour and dissolved organic carbon from drinking water and showed similar or better performance to conventional coagulants. In addition, the PAFS achieved the lowest residual metal-ion (Fe and Al) concentration when compared to ferric sulphate and aluminium sulphate. The high Fe and Al concentration in AMD means similar coagulants can be recovered from such mine-impacted waters. The specific objective of this study is to evaluate the recovery of an AMD-derived poly-alumino-ferric sulphate (AMD-PAFS) coagulant from coal AMD using chemical precipitation between pH 5.0 and 7.0. The efficiency of the AMD-PAFS is compared with conventional PFS coagulant in the treatment of brewery wastewater to remove turbidity, COD total dissolved solids (TDS). The effect of the coagulants on the electric conductivity (EC) of the wastewater is also evaluated.

#### **2. Materials and Methods**

#### *2.1. Materials*

The AMD sample was collected from Mpumalanga, South Africa. The sample was stored in a sealed polyethylene container. Before an experimental run, the solid debris and all the precipitated iron were removed by filtration using a grade 4 Whatman filter paper. Analytical-grade sodium hydroxide and sulphuric acid were used to prepare solutions for pH adjustment. All the solutions were prepared using deionised water. The conventional coagulant poly ferric sulphate (PFS) used for comparative tests was supplied by Merck, South Africa. The AMD-PAFS coagulant and PFC coagulant were tested on the brewery wastewater obtained from South African (SA) Breweries.

#### *2.2. Experimental Procedure*

#### 2.2.1. Iron and Aluminium Co-Precipitation

All precipitation experiments were conducted in a2Lreactor, shown in Figure 1. The agitator was fitted with a two-radial-blade impeller, and a speed of 300 rpm was used for all the experimental runs. In order to maximize mixing, the reactor was fitted with four equally spaced baffles. The reactor closure had ports for electrodes to measure pH and temperature. The experimental procedure involved oxidization of Fe (II) to Fe (III) by aeration for a period of 24 h. The oxidation of Fe (II) to Fe (III) is essential to the precipitation of Fe at low pH. Fe (III) precipitates at the pH range of 3–4, while Fe (II) does not precipitate at a pH < 6 [37]. The Fe (II) concentration was monitored by wet chemistry using potassium dichromate titration method [38]. Table 1 presents the summary of the experimental conditions. All experiments were performed in triplicates. The experimental procedure involved maintaining the temperature of the reactor contents at ambient temperature (25 ◦C) using the infrared heater. The agitation was then increased to the required speed, and the pH was adjusted by automatically injecting either 4.0 M sodium hydroxide or 0.1 M sulphuric acid using a Glass Chem reactor system, which has an automatic titrator. The accuracy of the pH control was 0.1 pH units. After attaining the required pH, the experiment was allowed to proceed for a period of one hour. The precipitate was separated from the effluent by vacuum filtration, followed by washing with deionised water to remove the entrained effluent solution. The precipitate was then left in the oven for 24 h at 80 ◦C to dry. 6 g of the dried precipitate was dissolved in 50 mL of 5.0% (w/w) sulphuric acid to obtain a clear solution, which was then used as a coagulant.

**Figure 1.** Picture of the experimental setup for the acid mine drainage (AMD) precipitation experiments using NaOH and H2SO4 at 25 ◦C.


#### 2.2.2. Metal Analysis

All solution and precipitate samples were analyzed for Fe, Al, Ca, Mn, Mg, Cu, Zn, Ni and Co using inductive coupled plasma mass spectroscopy (ICP-MS 7700X), from Agilent Chemetrix. The concentration of the sulphate was determined using ion chromatography. Metal recovery (*R*) was calculated according to Equation (3), as follows:

$$\mathcal{R} = \frac{\mathcal{C}\_0 - \mathcal{C}\_1}{\mathcal{C}\_0} \tag{3}$$

where*C0* is the concentration of a particular metal species in raw AMD (mg/L) and*C1* is the concentration of a metal species in the effluent (mg/L) after precipitation. Tabak et al. [39] defined the precipitate purity as the ratio of a desired precipitated metal species to the sum of all the metal species that have been precipitated. Based on this definition, the precipitate purity (*P*) was calculated according to Equation (4), as follows:

$$p = \frac{\mathcal{C}\_i}{\sum\_{i}^{n} \mathcal{C}\_j} \times 100\% \tag{4}$$

where *Ci* is the concentration of the individual or sum of the species of interest (%), n is the total number of metal species and *Cj* is the concentration of all the metal species precipitated (%). In this case, *Ci* was regarded as the total concentration of iron and aluminium in the precipitate.

#### *2.3. Water Treatment by Coagulation*

A six-beaker jar tester apparatus was used with each beaker containing 500 mL of brewery wastewater samples. The same concentration of the AMD-PAFS and conventional PFS was added to the water and pH adjusted to 7.0. The water samples were agitated for 3 min at a paddle speed of 200 rpm, followed by 10 min of slow mixing at a speed of 20 rpm and sedimentation of 30 min. Supernatant samples were withdrawn at 5 cm below the surface of the water samples. The performance evaluation was based on pH, EC, turbidity, COD and TDS measurement. The pH, EC and TDS were measured using the Hanna HI 9812-5 pH/EC/TDS/temperature portable meter (Hanna Instruments, Johannesburg, South Africa). The meter was calibrated with standard solutions of pH 4.0 and 7.0 before use. The supernatant was measured for turbidity and COD using a Merck Pharo 300 spectroquant, (Merck, Johannesburg, South Africa). The unit of measurement for turbidity was the Formazin attenuation units (FAU). The analysis methods followed the "Standard Method for Examination of Water and Wastewater" [40].

#### **3. Results and Discussion**

#### *3.1. Precipitation*

The general characteristic of the raw AMD is presented in Table 2. The characteristics of AMD are typical of the South African coal AMD solutions [41,42]. As can be seen from the table, this included high concentration of Fe, Al, Ca, Mg and Mn with minor concentrations of Ni, Zn, Cu and Co. The total Fe composition in the raw AMD was 80% as Fe (II) and 20% as Fe (III). The table also shows the SA standard for wastewater discharge into a water resource as well as the characteristics of the effluents obtained in this study at different precipitating pH values. When the pH was raised to 5.0, the Fe and Al concentrations were 2.7 and 14.0 mg/L in the effluent, respectively. This translated to 99.9 and 96.5% Fe and Al removal, respectively, calculated using Equation (3). These recoveries, which are averages of the triplicate results, are depicted in Figure 2 with the error bars related to the standard deviation. An increase in pH to 7.0 resulted in Al concentration of 3.7 mg/L in the effluent and the recovery being 99.1%, but the iron recovery remained at 99.9%. Other workers have also found similar results [16,19,21]. For example, during the synthesis of magnetic nanoparticles from AMD, Wei et al. [16] reduced Fe from 169 mg/L at pH 2.6 to 0.09 mg/L at pH 6.7 and Al from 71 mg/L to 0.2 mg/L under the same pH conditions. This represented 99.9% and 99.7% Fe and Al recovery, respectively. Figure 3 presents the effect of pH on the solubility of the other major heavy metals. The results show that the precipitation of Ca is almost negligible in the tested pH range. However, Mn and Mg effluent concentration were reduced from 93.9 mg/L to 83.6 mg/L and 474.0 mg/L to 457 mg/L, respectively. This represented 10.9% and 3.6% co-precipitation of Mn and Mg, respectively. Other minor elements, including Zn, Cu and Co, did not precipitate.The precipitate purity was calculated using Equation (4) and gave 99.0, 99.0 and 98.0% for pH 5.0, 6.0 and 7.0, respectively. The results obtained in the study are comparable with results obtained by other researchers. Michalková et al. [18] obtained less than 0.05% of Zn, Co, Cu and Ni in the AMD precipitated using sodium hydroxide at pH 6.9. In a study by Wei et al. [16], the precipitation of Ca, Mg, Mn and Ni during neutralization at pH 6.7 was 6.02, 5.57, 16.67 and 37.27%, respectively.


**Table 2.** Summary of the chemical composition of raw AMD and effluents after precipitation.

<sup>a</sup> data from [43] (NA = not applicable).

**Figure 2.** Effect of pH on Fe and Al recovery through precipitation of the AMD using NaOH and H2SO4 25 ◦C.

**Figure 3.** Solubility of the major heavy metals as a function of pH in the raw and treated AMD solutions at 25 ◦C.

#### *3.2. Coagulation*

#### 3.2.1. Coagulant Production and Testing

The precipitate obtained at pH 5.0 was dissolved in 5.0% (w/w) sulphuric acid to produce a coagulant. Table 3 summarizes the characteristics of the AMD-PAFS as well as the commercial PFS and polyaluminium sulphate (PAS) coagulants. The coagulant was composed of 89.5% Fe and 10.0% Al with trace amounts of Mn, Mg and Ca. The total mass concentration of Fe and Al in the AMD-PAFS coagulant was 96,644 mg/L, which compares well with the commercial PFS with the Fe mass concentrations 115,000 mg/L. The AMD-PAFS was compared with the commercial FPS in the treatment of brewery wastewater. Table 4 shows the characteristics of the brewery wastewater used in the study.


**Table 3.** Chemical composition of the poly-alumino-ferric sulphate coagulant produced by precipitation at pH and conventional poly ferric sulphate and poly aluminium sulphate.

<sup>b</sup> data from [11] (ND = not detected).

**Table 4.** Characteristics of the brewery wastewater that was treated with the AMD-PAFS and PFS coagulant.


3.2.2. Effect of the Coagulant on Turbidity and Chemical Oxygen Demand Removal

Turbidity, which is the cloudiness of the water, has long been the targeted substance during the coagulation and flocculation processes and is largely used as an indicator for the efficiency of the coagulation process [44]. It is the principal physical characteristic of water and expresses the optical property that causes light to be scattered and absorbed by particles and molecules rather than transmitted in a straight line through the water sample. The turbidity removal efficiency was determined by adding different doses of the coagulants from 10 mg/L to 150 mg/L. As shown in Figure 4, the percentage removal of the turbidity of the brewery water samples increased from 18.1% at 10 mg/L to 91.92% at 150 mg/L AMD-PAFS. This compared favourable with results obtained from the use of PFS, where 22.3% and 87.8% turbidity removal at 10 and 150 mg/L PFS were obtained, respectively. The increment in the removal of turbidity was due to the increment of the activity site

of the coagulants. The AMD-PAFS coagulant not only compared favourably with PFS coagulants but also with other synthetic coagulants. For example, a poly-aluminium-silicate-chloride coagulant (PSiFAC) was synthesized and tested in the treatment of simulated surface water [45]. A 99% turbidity removal was obtained at a PSiFAC concentration of 100 mg/L. The relatively high performance of the PSiFAC can be attributed to the presence of the silicate species. The silicate species increases the bridge effect and thereby slows down the formation of Fe(OH)3 precipitate, which results in enhanced coagulation [46].

(**b**)

**Figure 4.** Performance evaluation of (**a**) poly-alumino-ferric sulphate (AMD-PAFS) coagulant and (**b**) poly ferric sulphate coagulant (initial pH 7.0, temperature = 24.6 ◦C) during the treatment of brewery wastewater.

The COD is the amount of oxygen required to break down an inorganic pollutant in water or wastewater. Contrary to turbidity removal, the COD removal at 150 mg/L PFS was 64%, which was higher than the COD removal of 56% obtained at the same concentration of AMD-PAFS. This result is consistent with the previous studies such as the study by Xing and Sun [47], who obtained 72.4% COD removal from antibiotic fermentation wastewater, which had an initial COD concentration of 3279 mg/L by using 200 mg/L PFS coagulant. However, one important observation from this study was the formation of the emulsion at 150 mg/L AMD-PAFS, which could be an indication of excess coagulant.

#### 3.2.3. Effect of the Coagulants on Electric Conductivity

The EC is the measure of the dissolved ionic components in water and hence the electric characteristics. The EC gives an indication of the amount of total dissolved substitution in water [48]. As shown in Figure 4, the electric conductivity of the brewery wastewater increased as the dose of the coagulants increased. The conductivity of the original water sample was 3510 μS/cm, but it was increased to 4010 and 4110 μS/cm for AMD-PAFS and PFS coagulants, respectively. The sporadic rise in EC observed in all the samples tested could be due to the presence of the dissolved ions of the wastewater coupled with the dissolved ions of the coagulants and the pH regulator (NaOH). Similar observations have been made by other researchers [49,50].

#### 3.2.4. Effect of the Coagulant Dose on Total Dissolved Solids

TDS is one of the key parameters that can be used for water quality analysis. It is related to the quantity of material in water that can pass a filter size of 2 μm. The TDS increases the conductivities of water due to the presence of dissolved impurities [51]. The TDS in water influence the quality of drinking water such as taste, alkalinity, hardness and corrosion properties. As shown in Figure 5, the TDS of the untreated brewery wastewater was 1810 mg/L. The TDS increases only slightly with an increase in coagulant dose for both the AMD-PAFS and PFS coagulants. In general, the increase in TDS is due to an increase in the number of solute particles or ions as a result of coagulant addition. The principal anions contributing to the TDS value include the carbonate, bicarbonate, chloride, sulphate and nitrates, and cations such as calcium, magnesium, potassium and sodium [52]. The sulphate components of the tested coagulants contributed to the increase in the TDS of the treated water when the coagulant dose was increased.

**Figure 5.** Total dissolved solids as a function of coagulant dose during treatment of brewery wastewater.

#### **4. Conclusion**

The recovery of Fe and Al from coal generated AMD at pH 5.0 was 99.9% Fe and 94.7% Al. With an increased pH of up to 7.0, the overall Al recovery increased to 99.1%. Although Al precipitation was 99.1% at pH 7.0, the precipitate formed at pH 5.0 was chosen for coagulant production due to the reduced chances of co-precipitation of other impurities should they exist in substantially higher concentrations. Dissolution of precipitate in 5.0% (w/w) sulphuric acid produced a coagulant containing 89.5% Fe and 10.0% Al. The coagulant produced had comparable characteristics to the PFS commercial coagulant. The subsequent brewery wastewater treatment tests showed that the AMD-derived coagulant was as effective as the conventional coagulants in the removal of COD and turbidity. This process can be easily integrated in existing AMD treatment plants, which would provide revenue and thereby subsidize the treatment costs. Furthermore, the issues associated with disposal of the voluminous sludge could be avoided, as the coagulant recovery would reduce the sludge volume by 95.0%.

**Author Contributions:** Conceptualization, B.M.; Methodology, B.M., S.N., G.S.S., B.X.; Validation, B.M.; Formal analysis, B.M., G.S.S., S.S., S.N., B.X.; Investigation, B.M., and B.X.; Resources, S.N., B.F.; Data curation, B.M.; Writing—Original draft preparation, B.M.; Writing—Review and editing, B.M., B.X., G.S.S., S.S., and S.N.; Visualization, B.M., B.X., and G.S.S.; Supervision, S.N., G.S.S., S.S., and B.F.; Project administration, S.N., and B.F.; Funding acquisition, S.N., and B.F.

**Funding:** This research was funded by the National Research Foundation (NRF) and the Department of Science and Technology (DST) of South Africa through the Germany-South Africa collaborative Project "AddWater" ref# 105,879 and the SARChI chair in Hydrometallurgy and Sustainable Development (SARCI150223114415 Grant# 98350) and the international office of the BMBF in Germany under the AddWater Project (No. 01DG17024).

**Acknowledgments:** The National Research Foundation (NRF) and the Department of Science and Technology (DST) of South Africa are gratefully acknowledged for their financial contribution to this work through the Germany/South Africa Collaborative Project "AddWater" ref# 105,879 and the SARChI chair in Hydrometallurgy and Sustainable Development (SARCI150223114415 Grant# 98350). The international office of the BMBF in Germany for the financial support under the AddWater Project (No. 01DG17024) is also acknowledged.

**Conflicts of Interest:** The authors declare no conflict of interest. The sponsors had no role in the design of the study; in the collection, analysis or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **Chemical Stability of Zirconolite for Proliferation Resistance under Conditions Typically Required for the Leaching of Highly Refractory Uranium Minerals**

#### **Aleksandar N. Nikoloski 1,\*, Rorie Gilligan 1, Jonathan Squire <sup>2</sup> and Ewan R. Maddrell <sup>3</sup>**


Received: 11 July 2019; Accepted: 12 September 2019; Published: 1 October 2019

**Abstract:** In this study, synthetic zirconolite samples with a target composition Ca0.75Ce0.25ZrTi2O7, prepared using two different methods, were used to study the stability of zirconolite for nuclear waste immobilisation. Particular focus was on plutonium, with cerium used as a substitute. The testing of destabilisation was conducted under conditions previously applied to other highly refractory uranium minerals that have been considered for safe storage of nuclear waste, brannerite and betafite. Acid (HCl, H2SO4) leaching for up to 5 h and alkaline (NaHCO3, Na2CO3) leaching for up to 24 h was done to enable comparison with brannerite leached under the same conditions. Ferric ion was added as an oxidant. Under these conditions, the synthetic zirconolite dissolved much slower than brannerite and betafite. While the most intense conditions were observed previously to result in near complete dissolution of brannerite in under 5 h, zirconolite was not observed to undergo significant attack over this timescale. Fine zirconolite dissolved faster than the coarse material, indicating that dissolution rate is related to surface area. This data and the long term stability of zirconolite indicate that it is a good material for long-term sequestration of radioisotopes. Besides its long term durability in the disposal environment, a wasteform for fissile material immobilisation must demonstrate proliferation resistance such that the fissile elements cannot be retrieved by leaching of the wasteform. This study, in conjunction with the previous studies on brannerite and betafite leaching, strongly indicates that the addition of depleted uranium to the wasteform, to avert long term criticality events, is detrimental to proliferation resistance. Given the demonstrated durability of zirconolite, long term criticality risks in the disposal environment seem a remote possibility, which supports its selection, above brannerite or betafite, as the optimal wasteform for the disposition of nuclear waste, including of surplus plutonium.

**Keywords:** uranium; zirconolite; brannerite; betafite; leaching; kinetics

#### **1. Introduction**

Zirconolite, CaZrTi2O7 is one of several titanate phases present in synthetic titanate ceramics developed for the immobilisation of actinides and fission products in spent nuclear fuel. Other phases include pyrochlore, brannerite and zircon [1].

These minerals frequently contain uranium and/or thorium. In zirconolite, uranium undergoes extensive substitution onto the calcium site [2]. In brannerite, uranium is an essential element while in pyrochlore and zirconolite, actinides and light rare earth elements (REEs) can also take the place of calcium. Synthetic forms of these minerals can substitute other actinides as well, such as plutonium, americium and curium, in the calcium site, being likely too large for the other sites. The high chemical durability of these materials suggests that they could be ideal for the sequestration of surplus plutonium

and other actinides present in spent nuclear fuel. There are reports of zirconolites that have been observed to show evidences of post-crystallization corrosion [3].

Zirconolite has been identified in weathered gravel in Sri Lanka [4], in Western Australian dolerite intrusions [5], in lunar granite [6] and other varied geological settings and/or minero-genetic conditions [7]. Zirconolites up to 650 million years old have been identified in which 206Pb/ 238U, 207Pb/ 235U and 206Pb/ 207Pb isotope ratios give consistent ages, indicating that no uranium has been lost from the zirconolite despite the host rock having undergone extensive weathering over the 650 million years since formation [2]. It is also worth noting the existence of 2 billion year old zirconolite from Phalaborwa in South Africa as referred in some of the references of available papers [2]. This makes zirconolite potentially ideal for sequestering the radioactive elements present in spent nuclear fuel over the millennia required for them to decay into less harmful substances.

Cerium is often used as a substitute for plutonium in studies of nuclear waste ceramics. Cerium and plutonium have very close ionic radii (Ce4<sup>+</sup> = 97 pm and Pu4<sup>+</sup> = 96 pm) when coordinated by eight other atoms [8], as in the Ca site in zirconolite [1]. However, cerium is far safer to work with than plutonium.

Lumpkin [1] compared several mineral phases for waste immobilisation. The advantages of zirconolite over others include its high aqueous durability and chemical flexibility, though it is less tolerant to radiation dose than some other phases. The relative aqueous stability of several phases from pH 2–12 is as follows: zirconolite > pyrochlore > brannerite >> perovskite [1].

Brannerite is known to dissolve quickly in sulphuric and hydrochloric acids under oxidising conditions [9,10], while betafite (pyrochlore) will dissolve at a lower rate under similar conditions [11]. By comparing the leaching of zirconolite with the leaching of brannerite and betafite under these conditions, the stability of zirconolite as a host for actinides can be evaluated and demonstrated.

#### **2. Materials and Methods**

#### *2.1. Sample Preparation*

Synthetic zirconolite samples were prepared using two different methods. The zirconolite target composition was Ca0.75Ce0.25ZrTi2O7, with Ce as a substitute for Pu. Assuming that the feed mixtures are homogeneous, the composition of the feed mixtures for both methods should be CaO 11.5 wt. %, Ce2O3 11.2 wt. %, ZrO2 33.7 wt. %, and TiO2 43.6 wt. %. The oxidation state of Ti was expected to be mixed 4/3+ to maintain charge balance. The preparation methods are outlined below.

i. Alkoxide route

Required quantities of zirconium n-propoxide and titanium isopropoxide were hydrolysed with a solution containing the necessary amounts of calcium and cerium nitrate. The slurry was then stir-dried in a stainless steel beaker on a hot plate. Once dried the product was calcined at 750 ◦C in air for 8 h.

ii. Oxide route

ZrO2 and TiO2 as approximately 1 μm particle size powders, CeO2 as a 5 μm particle size powder and calcium nitrate were combined to form a slurry. This was stir dried and calcined as for the alkoxide route.

The powders produced by both routes were then planetary milled as a slurry for 20 min with propan-2-ol as a carrier fluid, dried and sieved. The powders were then blended with 2.2 wt% Ti metal in a Turbula mixer and packed into stainless steel hot isostatic pressing (HIP) cans 3.5 cm diameter by 5 cm high. The Ti metal acted as an in-can reducing agent to convert Ce4<sup>+</sup> to Ce3<sup>+</sup> and Ti4<sup>+</sup> to Ti3<sup>+</sup> to ensure correct charge balance in the zirconolite. The HIP cans were then sealed and evacuated, and then hot isostatically pressed at 1320 ◦C and 100 MPa for 2 h.

#### *2.2. Sample Characterisation*

Mineralogical analysis by X-ray diffraction (XRD) was performed with a GBC Enhanced Multi-material Analyser (EMMA) (GBC Scientific Equipment, Braeside, Victoria, Australia) at Murdoch University. Samples were placed directly onto X-ray absorbing silicon discs within circular metal sample holders. Samples were introduced under a drop of ethanol and the ethanol was allowed to evaporate prior to the analysis.

The X-ray tube was operated at a voltage of 35.0 kV and current of 28.0 mA. Diffraction patterns were collected over a range of 20◦ ≤ 2θ ≤ 70◦ using a 1◦ diverging slit, a 0.2◦ receiving slit and a 1◦ scattering slit. A step size of 0.02◦ was used, with a speed of 1◦/min (1.2 s per step) with five passes. Cu Kα X-rays were used. A Kα<sup>2</sup> strip was performed on the diffraction patterns, with a Kα2/Kα<sup>1</sup> ratio of 0.51. Initial scans showed no peaks of interest below 20◦.

Scanning electron microscopy (SEM) observations were performed with a JEOL JCM-6000 bench top SEM with an energy dispersive X-ray spectroscopy (EDX) analyser (JEOL Ltd., Tokyo, Japan). An accelerating voltage of 15 kV was used to produce the SEM images of the samples. Both secondary electron (SE) and backscattered electron (BSE) modes were utilised. Particles were mounted on carbon discs. The cross-sections of the particles were prepared by embedding in epoxy resin and subsequent polishing with silicon carbide. A 15 kV accelerating voltage was used for the semi-quantitative EDX analyses, the highest possible with the instrument used in this study. All EDX analyses were run for 60 s. All images associated with EDX analyses were taken in BSE. For line-scan analyses, the counting time was set to 15 s per step. X-ray elemental maps were produced with a resolution of 384 × 512 pixels and a counting time of 10 × 0.2 ms per pixel. The standard colour scheme for the element maps adhered to throughout this report is red for calcium, green for zirconium and blue for titanium. Cerium was not included on the element maps due to the overlap of the Ce Lα peak at 4.83 keV with the Ti Kβ peak at 4.93 keV.

All aqueous samples were analysed for calcium, titanium, zirconium and cerium with a Thermo-Fisher iCAP-Q ICP-MS instrument (Thermo-Fisher Scientific, Bremen, Germany) at Murdoch University. The purity of zirconolite samples was verified by digestions and ICP-MS analysis performed at a commercial minerals laboratory. The coarse zirconolite was assayed twice.

#### *2.3. Leaching Study*

Similar conditions for the leaching study were used to those previously reported for brannerite leaching [9,10,12]. Acid (HCl, H2SO4) leaching experiments were run for five hours and alkaline leaching tests were run for 24 h to enable comparison with brannerite leached under the same conditions. Coarse zirconolite was used in the majority of the experiments. The highest temperature experiment for each lixiviant was repeated with fine zirconolite. The conditions used in the leaching experiments are listed in Table 1.


**Table 1.** Leaching conditions used in this study.

As with the brannerite leaching experiments, Fe3<sup>+</sup> was added as an oxidant. Iron was added as 0.05 mol/L FeCl3 in the chloride leaching experiments, 0.05 mol/L Fe(SO4)1.5 in the sulphate leaching experiments and 0.025 mol/L K3Fe(CN)6 in the carbonate leaching experiments.

#### **3. Results and Discussion**

#### *3.1. Feed Characterisation*

The two feed samples produced using the different methods had different size distributions—63–125 μm for the alkoxide route sample and 125–250 μm for the oxide route sample. Wet screening was used to narrow down the size range of each sample. These are labelled 'fine' sample and 'coarse' sample, respectively.

#### 3.1.1. Feed Assays

Chemical analyses of the synthetic zirconolite by ICP-MS presented in Table 2; Table 3 show that the synthetic zirconolite from both methods was of high purity. Hafnium was the main non-formula element identified. Hafnium is often found with zirconium, and separating the two presents a significant technical challenge.

**Table 2.** Major elements (>0.1 wt%) in the zirconolite feed samples.


**Table 3.** Minor elements (>100 ppm) in the zirconolite feed samples in ppm.


These zirconolite specimens had average formulas of Ca2<sup>+</sup>0.71Ce<sup>3</sup><sup>+</sup>0.21(Zr1.03Hf0.01)Ti1.95O7 for the coarse zirconolite and Ca2<sup>+</sup>0.72Ce3<sup>+</sup>0.21(Zr1.02Hf0.01)Ti1.95O7 for the fine zirconolite. Si has been excluded based on EDX results, showing that it was present in a separate minor SiO2 phase. Both were slightly Ti deficient compared to the ideal zirconolite composition, but had a higher amount of Ti than typical natural samples. Titanium is commonly replaced by Fe<sup>3</sup>+/Nb5<sup>+</sup> in natural zirconolite. Tantalum may also be present in this site in small amounts [13] Cerium, REEs and actinides replace calcium in the zirconolite crystal structure [2,4,14–16].

#### 3.1.2. XRD

XRD also showed feed samples produced by both methods to be effectively pure zirconolite, a solid solution (Ca0.75Ce0.25)ZrTi2O7. The XRD data showed the presence of zirconolite and perovskite

as major phases. Zirconolite exists in three polytypes [17]; and reference diffraction patterns for them have been superimposed on the measured diffraction pattern (Figure 1). Titanium dioxide (rutile) was detected in small amounts. The other polymorphs of titanium dioxide, anatase and brookite, were not detected. Perovskite was detected as a minor phase.

**Figure 1.** X-ray diffraction (XRD) pattern of the zirconolite material with relevant PDF references.

#### 3.1.3. SEM, EDX

The examination by SEM showed the feed samples contained very small inclusions of a second phase, and possibly some unreacted ZrO2. Backscattered electron images (Figure 2) showed that the inclusions have a lower average atomic mass. EDX analyses (Figures 3 and 4) indicated that this material was titanium dioxide, though it is not possible to tell from the EDX analyses which polymorph of titanium dioxide was present. The XRD results indicate that it was most likely rutile, possibly a relic of the Ti metal added for redox control during the HIP process. Neither silicon nor hafnium were detected in EDX analyses of zirconolite. When silicon was detected, it occurred as a separate phase (SiO2), while hafnium at 0.6% of the mass was below the detection limit for EDX analyses.

EDX analyses of the zirconium-free regions showed that they contained cerium along with calcium and titanium (Figures 3 and 4). This material was probably the same perovskite phase identified by XRD. The brightness of this phase in the BSE images suggests that was not pure Ca perovskite however. Pure Ca perovskite has a low average atomic number (*Z*avg) (16.5) close to that of rutile (16.4) [18] while the calculated *Z*avg of zirconolite exceeds 22. While there were some subtle variations in the BSE brightness of the zirconolite/perovskite regions, these variations did not clearly correlate with variations in composition as determined by EDX spectra or elemental maps.

**Figure 2.** Rutile inclusions within zirconolite grains surrounded by perovskite. **Left**: backscattered electron (BSE) images, **right**: element map.

**Figure 3.** Spectra of spots analysed in the top half of Figure 2.

**Figure 4.** Energy dispersive X-ray spectroscopy (EDX) spectra of spots analysed in the bottom half of Figure 2.

It is difficult to resolve the cerium Lα peak from the titanium Kβ peak. For this reason, cerium was not included on any of the element maps. Closer examination shows the Ce Lβ peaks in some spectra allowing the presence of cerium to be confirmed. Of all the spots analysed on the feed sample, cerium was most prominent in the spectrum of spot 35 in Figure 4.

Eight oxygen atoms coordinate the calcium site in zirconolite, but the calcium site in perovskite is larger and coordinated by 12 oxygen atoms [19]. Hence, calcium within perovskite undergoes extensive isomorphous substitution with uranium, thorium and REEs [20]. Perovskite is commonly formed as a side-product in the synthesis of zirconolite and other titanate ceramics [14,21] and is an intentional phase in Synroc C, to host Sr-90. Studies on polyphase heterogeneous actinide titanate ceramics show that large lanthanide ions like Ce3+/Nd3<sup>+</sup> and trivalent actinides (Pu3<sup>+</sup>, Am3<sup>+</sup>, Cm3<sup>+</sup>) favour the Ca site of perovskite over zirconolite. This explains the presence of cerium in the perovskite phase in this sample. The partition coefficient between zirconolite and perovskite was lower for larger cations [21].

Zirconolite is significantly more stable than perovskite [1,8]; thus, the formation of perovskite in synthetic samples intended for uranium sequestration should be minimised as much as possible. Pöml et al. [14] succeeded in synthesising a cerium doped zirconolite without detectable levels of perovskite by adding a stoichiometric excess of ZrO2 during synthesis.

Along with the three major separate phases, rutile, Ce-perovskite and Ce-zirconolite, one of the spectra indicates the presence of a fourth minor Zr oxide phase (spot 25 in Figure 3). The boundaries between phases in the coarse zirconolite sample are clear and distinct unlike those observed in brannerite [22]. The boundaries between phases in this sample are clear and distinct unlike those observed in natural brannerite [22], as is apparent from EDX line analyses across a rutile inclusion. Rutile inclusions were typically surrounded by smaller perovskite inclusions though not all perovskite inclusions were associated with rutile.

#### *3.2. Leaching Kinetics*

Under similar leaching conditions, the synthetic zirconolite dissolved much more slowly than natural brannerite [9,12,23] and betafite [11,24,25]. Cerium extraction from zirconolite followed linear kinetics in sulphuric acid (Figure 5). After five hours of leaching, cerium extraction had yet to plateau. Titanium dissolved at a slower rate than cerium but faster than zirconium. This suggests that zirconolite is not dissolving in significant amounts. Based on the observed leaching kinetics, perovskite is more susceptible to leaching than zirconolite. Calcium extraction kinetics were not included due to the significant levels of analytical error in measuring the calcium concentrations in solution. Apart from Ca, these results are consistent with earlier work that showed the typical order of elemental dissolution rates from zirconolite in acidic solutions is Ca > Ce > Ti > Zr [14].

**Figure 5.** Leaching kinetics under various conditions in sulphuric acid media.

Similar trends were apparent during leaching in chloride media (Figure 6) to those observed for sulphate media. Cerium dissolved faster than titanium, which in turn dissolved faster than zirconium. While extraction rates were lower in chloride media compared to sulphate media at the same temperature and acid concentration, variations in acid concentration had a larger effect on the rate of dissolution in chloride media. Both of these behaviours have been observed when leaching brannerite in chloride and sulphate media over a wide range of temperatures and acid concentrations [10]. The order of uranium and titanium extraction from brannerite was approximately 0.5 with respect to H2SO4 while the order was approximately 1 with respect to HCl [9,10].

**Figure 6.** Leaching kinetics under various conditions in hydrochloric acid media.

After five hours of leaching, the extent of cerium dissolution was 3–6 times higher than that of titanium in hydrochloric acid and around 100 times higher than that of zirconium in 0.25 M HCl. In 1.00 M HCl, the Ce/Zr ratio decreased to approximately 20, with acid concentration having a significant effect on the dissolution rate of zirconium in chloride media. These trends match those observed in long term leaching studies on synthetic actinide waste forms such as zirconolite and pyrochlore at pH 2 in 0.01 M HNO3 solution [26], and are in agreement with the relative solubility of the simple oxides of these elements (Figure 7; Reactions 1–5).

$$\text{Ca} \quad \text{CaO} + 2\text{ H}^+ \rightarrow \text{Ca}^{2+} + \text{H}\_2\text{O} \tag{\text{Reaction 1}}$$

$$\text{Ce} \quad \text{(III)} \ 0.5\,\text{Ce}\_2\text{O}\_3 + 3\,\text{H}^+ \rightarrow \text{Ce}^{3+} + 1.5\,\text{H}\_2\text{O} \tag{Reaction 2}$$

$$\text{Ce} \quad \text{(IV)}\\\text{CeO}\_2 + 4\text{ H}^+ \rightarrow \text{Ce}^{4+} + 2\text{ H}\_2\text{O} \tag{\text{Reaction 3}}$$

	- Zr ZrO2 <sup>+</sup> 2 H<sup>+</sup> <sup>→</sup> ZrO2<sup>+</sup> <sup>+</sup> H2O (Reaction 5)

The relative rates of leaching were different in an alkaline environment (Figure 8). As with the acid leaching experiments, rates of dissolution in alkaline media were significantly slower than those observed for brannerite. Titanium dissolved faster than cerium, which dissolved much faster than zirconium. If the zirconolite was allowed to react for longer, it is expected that titanium would re-precipitate as titanium dioxide as observed with brannerite and Ti rich uranium ore [12,28]. Titanium is somewhat amphoteric and may dissolve as Ti(OH)5 - at high pH [29,30].

**Figure 7.** Solubilities of simple oxides of Ca, Ce, Ti and Zr from 0–100 ◦C as calculated using software [27]. Log K values for Reactions 1–5 were used.

While Ce2O3 will readily dissolve in acidic or neutral conditions, CeO2 is far less soluble (Figure 7). Oxidising conditions may have caused cerium as Ce3<sup>+</sup> or CeOOH to be oxidised to insoluble CeO2. Calculations [26] have indicated that this process is favourable (see Reactions 6 and 7). Increasing the pH will make Reaction 7 even more favourable.

$$\begin{array}{l} \text{0.5 } \text{Ce}\_2\text{O}\_3 + 0.5 \text{ H}\_2\text{O} \rightarrow \text{Ce}\text{OOH}\_{\text{(aq)}}\\ \Lambda\_{\text{rxn}}\text{G}^{70^\circ \text{C}} = -25.0 \text{ kJ/mol} \end{array} \tag{\text{Reaction 6}}$$

$$\begin{array}{c} \text{CeOOH}\_{\text{(aq)}} + \text{Fe(CN)}\_{6}^{3-} + \text{OH}^{-} \rightarrow \text{CeO}\_{2} + \text{Fe(CN)}\_{6}^{4-} + \text{H}\_{2}\text{O} \\ \qquad \Delta\_{\text{rxn}}\text{G}^{70^{\circ}\text{C}} = -129.4 \text{ kJ/mol} \end{array} \tag{\text{Reaction }7}$$

While the unreliability of the calcium assay data makes it impossible to determine its behaviour in solution, secondary calcite phases have been observed when leaching brannerite (~2% Ca) under similar conditions [12]. The leaching of calcium from perovskite in carbonate solutions forming anatase and calcite has been observed in natural titanium deposits [31]. The process may take place according to the following reactions:

$$\text{CaTiO}\_3 + 2\text{H}\_2\text{CO}\_3 \rightarrow \text{TiO}\_{2(\text{anatase})} + \text{Ca(HCO}\_3\text{)}\_{2(\text{aq})} + 2\text{H}\_2\text{O} \tag{\text{Reaction 8}}$$

$$\text{Ca(HCO}\_3\text{)}\_{2(aq)} \to \text{CaCO}\_3 + \text{CO}\_2 + \text{H}\_2\text{O} \tag{\text{Reaction 9}}$$

The overall process is described in Reaction 10:

$$\text{CaTiO}\_3 + \text{CO}\_2(\text{aq}) \rightarrow \text{TiO}\_{2(\text{amatase})} + \text{CaCO}\_3 \cdot \text{Arx} \\ \text{G}^{70^\circ \text{C}} = -48.4 \text{ kJ/mol} \qquad \qquad \text{(Reaction 10)}$$

Clearly, this process is favourable under these conditions.

**Figure 8.** Extraction of titanium, zirconium and cerium from zirconolite in alkaline media at 70 ◦C.

#### *3.3. Activation Energy*

The rate of cerium dissolution showed a strong dependence on temperature. The average rate of dissolution between 1 and 5 h residence time was used to calculate the activation energy in the temperature range 30 to 85 ◦C, shown in Table 4. This can be considered to be the initial rate of extraction given the long periods over which zirconolite is known to dissolve [14,26]. Arrhenius plots for the sulphate leaching experiments showing data from tests conducted at 30, 50, 70 and 85 ◦C are shown in (Figure 9).

**Table 4.** Activation energy (kJ/mol) for the dissolution of Ce, Ti and Zr based on extraction rates from 1–5 h.


Activation energies (Table 4) were also calculated for the hydrochloric acid leaching tests, though the results are less certain as chloride leaching was only done at two temperatures, 50 and 85 ◦C. Arrhenius plots may have multiple regions corresponding to different rate determining steps [32], thus, such plots derived from only two temperature points may be unreliable.

**Figure 9.** Arrhenius plot for Ce, Ti and Zr leaching in 0.25M H2SO4, based on the average rate of leaching from 1–5 h.

Omitting the outlying 85 ◦C point from the titanium calculation gives an activation energy of 21.4 kJ/mol in sulphate media, very close to the calculated activation energy for zirconium dissolution in the same media, possibly indicative of a similar dissolution mechanism.

Longer term leaching experiments over 14 days in 1 M HCl at 100–200 ◦C gave an activation energy of approximately 20 kJ/mol for the dissolution of cerium and titanium from synthetic zirconolite [14]. Leaching experiments with similar synthetic samples by Zhang et al. [33] between 25 and 75 ◦C and over a pH range of 2–12, showed that the activation energy for uranium release from zirconolite varied with pH when these calculations were repeated with data presented by Zhang et al. [33]. The activation energy for uranium release was typically 15–20 kJ/mol with the one outlier being the pH 4.1 tests which gave a calculated activation energy value of 37 kJ/mol.

Comparisons with other studies [9,10] showed that zirconolite underwent slower dissolution than brannerite or even betafite when leached under similar conditions (Figure 10).

**Figure 10.** Arrhenius plots for the extraction of various elements from brannerite, betafite and zirconolite in sulphuric acid.

#### *3.4. Leached Residue Characterisation*

Unlike the brannerite studied previously, there were few apparent signs of corrosion after the leaching of zirconolite. Images, element maps and spectra were taken of zirconolite particles leached at the highest temperature in each lixiviant.

**Figure 11.** Grain boundary in zirconolite after leaching in 0.25 M HCl at 85 ◦C. **Left**: BSE image, **right**: element map.

Figure 11 shows a rutile inclusion surrounded by perovskite following chloride leaching. It is possible that the perovskite regions may have undergone some corrosion. There was no sign of pitting on the zirconolite (light green in all element maps) visible at the resolution of these images, though line analyses indicated that the outermost 2–5 μm were enriched in titanium and zirconium and depleted of calcium and cerium relative to the core of the particles.

These apparent changes in the distribution of elements in the solid phase are corroborated by the leaching kinetics data, which showed that the extraction of cerium was consistently higher than that of titanium and significantly higher than that of zirconium (Figures 5 and 6).

There were less visible signs of corrosion in the sulphate leaching system (Figure 12), although line EDX analyses indicate some selective leaching of calcium and cerium in the outermost 2–5 μm layer of zirconolite. Line A intersects a rutile inclusion and a perovskite grain, while line B runs across a protrusion of zirconolite corroded on both sides, both ends showing decreased Ca/Ce relative to Ti/Zr. There were no visible signs of corrosion on the rutile inclusions. Past experience with ilmenite [34] suggests minimal corrosion occurs in 0.25–1.00 M H2SO4 at 95 ◦C during 5 h of contact.

**Figure 12.** A large zirconolite particle with rutile inclusions after leaching in 0.25 M H2SO4 at 85 ◦C for 5 h Left: BSE image, right: element map.

There were minimal signs of corrosion in carbonate leaching media (Figures 13 and 14). Once again, this is consistent with the leaching kinetics (Figure 8) and past experience with brannerite [12,28].

**Figure 13.** Backscattered electron SEM image (**left**) and Ca–Zr-Ti map (**right**) of zirconolite after 24 h of leaching in sodium carbonate at 70 ◦C with the locations of EDX analyses. Spectra are shown in Figure 14.

**Figure 14.** EDX spectra of rutile and perovskite inclusions in a zirconolite particle after leaching in sodium carbonate.

The alteration of zirconolite to secondary phases such as anatase and baddeleyite has been observed in earlier work [14,33]; however, these leaching tests typically ran over longer time periods of at least two weeks.

#### *3.5. Reaction Mechanisms*

Particle size had a clear effect on the rate of cerium dissolution in sulphate and chloride media. The rate at which various elements dissolve from zirconolite has been observed to be proportional to the surface area in contact with the lixiviant until saturation is reached and secondary phases begin to form [14]. The extraction rates of titanium and zirconium were less affected by particle size (Figure 5; Figure 6), as might be expected if they were forming secondary phases.

Calcium, aluminium and cerium dissolution were proportional to surface area, while zirconium and titanium were observed to plateau due to the formation of secondary solid phases on reaching saturation [14]. The reaction for zirconolite dissolution given by Pöml et al. [14] is:

$$\begin{aligned} \text{Tr}(\text{Ca}\_{1-x}\text{Cr}\_{x})Zr(\text{Ti}\_{2-y}Al\_{y})\text{O}\_{7} + (6+2y)H^{+} \\ \rightarrow \text{ZrO}\_{2} + (2-y)TiO\_{2} + (1-x)Ca^{2+} + xCe^{4+} + yAl^{3+} + (3+y)H\_{2}O \end{aligned} \tag{\text{Reaction 11}}$$

Titanium and zirconium precipitation does not occur until ZrO2 and TiO2 exceed saturation at the zirconolite-solution interface [14]. These experiments were not run for long enough for the zirconium and titanium concentrations in the bulk solution to plateau, though the outer 5 μm of zirconolite leached in sulphate media at 85 ◦C was enriched in titanium and zirconium indicating that saturation may have been reached at the solid-aqueous interface. The relative rates of extraction observed in this study matched those identified in longer term leaching studies [14,26].

It is under oxidising conditions that the similarities between cerium and plutonium break down. Cerium is oxidised to insoluble CeO2 (Reaction 7), while insoluble PuO2 can be oxidised further and remobilised as PuO2 <sup>+</sup> and PuO2 <sup>2</sup><sup>+</sup> complexes [35], similar to what is typically seen with uranium.

#### *3.6. Crystallinity and Leachability*

There are two reasons for the much lower extent of dissolution observed in the zirconolite leaching compared with earlier work with brannerite. Zirconolite is known to be highly chemically stable. This sample is also highly crystalline as is apparent from the XRD results (Figure 1). Crystalline phases are more refractory than metamict materials [2,23]. Even within the same mineral sample, heavily altered metamict zones are more susceptible to corrosion than less altered zones [22].

The process of metamictisation, by which a radioactive crystalline material gradually becomes amorphous from internal irradiation [36] decreases its chemical and physical stability [14]. However, if a metamict material is recrystallised by heating, the material becomes less soluble. This has been documented for brannerite [23], betafite [11] and synthetic zirconolite [26].

Synthetic titanates have been synthesised with plutonium-238 (half-life = 87.7 years) to study the rate and effects of radiation damage over the course of five years. Strachan et al. [26] showed that the degree of radiation damage has little to no effect on the chemical stability of zirconolite and pyrochlore. Similarly, studies of natural zirconolites have shown them to be highly chemically durable having survived 600 Ma or more in nature [2]. This means that measurements of the chemical stability of zirconolite based on recently prepared non-metamict samples are likely to be applicable to aged and metamict samples as well. Furthermore, this indicates that zirconolite is a good material to use for the long-term sequestration of radioisotopes.

The minor perovskite phase seemed to have undergone more corrosion than the surrounding zirconolite phase. This could be related to the formation of Ce3<sup>+</sup>, which is easy to form, as observed in the fabrication, and is what stabilises perovskite. Perovskite should be avoided when synthesising zirconolite, as this may increase the proportion of soluble and mobile plutonium. Pöml et al. [14] achieved this by adding a stoichiometric excess of ZrO2. However, this is likely to be less of a concern for the immobilisation of tetravalent actinide ions.

In addition to long term durability in the disposal environment, a wasteform for fissile material immobilisation must demonstrate proliferation resistance such that the fissile material cannot be retrieved by dissolution of the wasteform. When the United States Department of Energy was developing wasteform options for the disposition of surplus weapons grade plutonium, zirconolite, the initial choice of wasteform, was replaced by betafite (therein referred to as pyrochlore) on the grounds that the U–238 in the wasteform would guard against certain long term criticality events in the disposal environment [37]. This study, in conjunction with our previous work on betafite leaching, strongly indicates that the addition of depleted uranium to the wasteform is detrimental to proliferation resistance. Given the demonstrated durability of zirconolite, from both natural and synthetic samples, long term criticality risks in the disposal environment seem a remote possibility, and this supports the selection of zirconolite, above betafite, as the wasteform for disposition of surplus plutonium.

#### **4. Conclusions**

Synthetic zirconolite is significantly more stable than natural brannerite or betafite. The most intense conditions used in this study did not cause synthetic zirconolite to undergo significant leaching or visible corrosion despite the same conditions being sufficient for near complete dissolution of natural brannerite in under five hours [9,12,22].

There was some evidence for incongruent dissolution, as the outer 5 μm of some leached zirconolite particles were enriched in titanium and zirconium, indicating that these elements had exceeded saturation at the aqueous-solid interface.

Fine zirconolite dissolved faster than the coarse material, indicating that the rate of dissolution is related to surface area. In practice, the rate of dissolution could therefore be further minimised by forming the zirconolite waste ceramics into larger solid masses.

**Author Contributions:** Conceptualisation, A.N.N., E.R.M.; methodology, A.N.N., R.G., E.R.M. and J.S.; validation, A.N.N.; formal analysis, A.N.N., R.G.; investigation, A.N.N., R.G., E.R.M., J.S.; resources, A.N.N.; data curation, A.N.N., R.G.; writing—original draft preparation, R.G., A.N.N.; writing—review and editing, A.N.N., R.G; supervision, A.N.N.; project administration, A.N.N.; funding acquisition, A.N.N.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors acknowledge Murdoch University for access to equipment.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Synthesis of Nanosilica via Olivine Mineral Carbonation under High Pressure in an Autoclave**

#### **Srecko Stopic 1,\*, Christian Dertmann 1, Ichiro Koiwa 2, Dario Kremer 3, Hermann Wotruba 3, Simon Etzold 4, Rainer Telle 4, Pol Knops <sup>5</sup> and Bernd Friedrich <sup>1</sup>**


Received: 24 May 2019; Accepted: 19 June 2019; Published: 24 June 2019

**Abstract:** Silicon dioxide nanoparticles, also known as silica nanoparticles or nanosilica, are the basis for a great deal of biomedical and catalytic research due to their stability, low toxicity and ability to be functionalized with a range of molecules and polymers. A novel synthesis route is based on CO2 absorption/sequestration in an autoclave by forsterite (Mg2SiO4), which is part of the mineral group of olivines. Therefore, it is a feasible and safe method to bind carbon dioxide in carbonate compounds such as magnesite forming at the same time as the spherical particles of silica. Indifference to traditional methods of synthesis of nanosilica such as sol gel, ultrasonic spray pyrolysis method and hydrothermal synthesis using some acids and alkaline solutions, this synthesis method takes place in water solution at 175 ◦C and above 100 bar. Our first experiments have studied the influence of some additives such as sodium bicarbonate, oxalic acid and ascorbic acid, solid/liquid ratio and particle size on the carbonation efficiency, without any consideration of formed silica. This paper focuses on a carbonation mechanism for synthesis of nanosilica under high pressure and high temperature in an autoclave, its morphological characteristics and important parameters for silica precipitation such as pH-value and rotating speed.

**Keywords:** silica; synthesis; olivine carbonation; autoclave; precipitation

#### **1. Introduction**

As a result of the high nickel production costs associated with traditional pyrometallurgical techniques and the depletion of high-grade sulfide ores, renewed interest has developed concern on the production of nickel and cobalt by high pressure acid leaching (PAL) of nickel laterites. More than one third of the world's nickel is nowadays produced from laterite ores [1,2]. Laterites account for two thirds of the world's nickel resources. It is therefore likely that increasing amounts of nickel will be produced from laterites. Since laterite type ores naturally occur close to the surface, economical open pit mining techniques are employed to recover the ore after removal of the overburden [3]. The laterite ore consists of fresh saprolite such as K0.4(Si3.0Al1.0)4.0(Al2.0Mg0.3)2.33O10(OH)2 and nontronite such as Na0,3(Fe<sup>3</sup><sup>+</sup>)2(Si,Al)4O10(OH)2·nH2O. These silicate ores represent the various layers in the laterite bedrock. The limonite consists mainly of goethite. This continues to a nontronite rich zone. Saprolite is the next layer, which is distinguished from its rich magnesium silicate content. The lateritic ore mostly has a low level of Ni (1–3%), Co (max. 0.1%), Fe (20–30%) and high level of SiO2 (more than 50%). The treatment of silicate based ores with different acids under an atmospheric pressure leads to formation of silica gel and breaking of leaching process. Silica gel represents an amorphous and porous form of silicon dioxide consisting of an irregular tridimensional framework of alternating silicon and oxygen atoms with nanometer-scale pores [4].

Similarly, high Si content in red mud and its slags produced by pyrometallurgical treatment for the Fe removal makes these secondary resources untreatable with conventional acid leaching routes due to the formation of silica gel. Alkan et al. [5] studied red mud and slags synthesized by electric arc furnace smelting, which contain rather moderate and extensive SiO2. In the next step, the formed slag was exposed separately by red mud to dry digestion with sulfuric acid at room temperature aiming at selective Sc recovery without Ti and Si dissolution. An empirical dry digestion-leaching model was proposed for each starting material in a comparative manner in order to prevent the formation of silica gel using sulfuric acid.

The Eudialyte concentrate is a potential rare earth elements (REE) primary resource due to its good solubility in acid, low radioactivity and relatively high REE content (about 2%), but also contains more than 50% of silica. The treatment of the Eudialyte concentrate can produce silica gel during a treatment with some acids [6]. The main challenge is avoiding the formation of silica gel, which is non-filterable when using acid to extract REE. Ma et al. [7] have studied neural network modeling for the optimization of the extraction of rare earth elements from the Eudialyte concentrate by dry digestion and a subsequent leaching avoiding the formation of silica gel in the presence of the hydrochloric acid.

Development of ceramic nanoparticles such as silica, alumina and titania with improved properties has been studied with much success in several areas such as synthesis and surface science [8,9]. Advancement in nanotechnology has led to the production of nanosized silica, which has been widely used as filler in catalysis and glass industry. The silica particles extracted from natural resources contains metal impurities and are not favorable for advanced scientific and industrial applications.

The sol-gel process is widely applied to produce silica, glass, and ceramic materials due to its ability to form pure and homogenous products at mild conditions. The process involves hydrolysis and condensation of metal alkoxides (Si(OR)4) such as tetraethylorthosilicate (TEOS, Si(OC2H5)4) or inorganic salts such as sodium silicate (Na2SiO3) in the presence of mineral acid (e.g., HCl) or base (e.g., NH3) as catalyst [10]. The synthesis of spherical hollow silica particles from sodium silicate solution with boric acid or urea as an additive was carried out by the ultrasonic spray pyrolysis method. This work dealt with the effect of four parameters (the concentration of the boric acid and urea, feed rate of reactant, reaction temperature and time) on particle size and standard deviation. As a result, the mean particle size and standard deviation decreased with increasing of all parameters except urea [11]. Ratanathavorn et al. [12] have studied silica nanoparticles synthesis by ultrasonic spray pyrolysis (USP) technique using tetraethylorthosilicate (TEOS) as a precursor in order to produce a fixative material for cream perfume fomulation. The results showed that the synthesis temperature of 500 ◦C provided the smallest size of silica nanoparticle, about 106 nm. The particle size decreased from 347 nm to 106 nm when the synthesis temperature increased from 300 ◦C to 500 ◦C.

The ultra-small hollow silica nanoparticles were synthesized using the prepared amorphous calcium carbonate (ACC) particles as a template. The ACC particles were firstly prepared by the carbonation method, which the procedure was conducted in the methanol solvent to form the Ca(OCH3)2 layers on the ACC particles. An effect of methanol concentration on the morphology of ACC particles was also investigated [13]. ACC particles were prepared by a carbonation method via bubbling CO2 gas into calcium ions dispersing in methanol solution. An effect of methanol concentration on the CaCO3 formation was investigated. The pH of the ACC preparation was studied in a range of 9.4 and 10. After that, ultra-small HSNPs were synthesized using the prepared ACC particles in the one-pot process. The results suggested that the synthesis of HSNPs using the ultra-small ACC particles via the one-pot process is one of the most effective methods to produce ultra-small HSNP regarding to save energy and cost.

In a mineral carbonation process, silicate minerals can also be used as feedstock to form carbonates and H4SiO4, that are chemically stable in a geological timeframe. Silicate minerals usually are richer in alkaline earth metal content such as magnesium, sodium, and calcium. Common silicate minerals suitable for carbonation are forsterite (Mg2SiO4), antigorite (Mg3Si2O5(OH)4) and wollastonite (CaSiO3) and their overall reaction rates are given in Equations (1)–(3):

$$\mathrm{Mg\_2SiO\_4(s) + 2CO\_2(g) + 2H\_2O(l) = 2MgCO\_3(s) + H\_4SiO\_4(aq) + 89 \text{ kJ/mol}}\tag{1}$$

$$\mathrm{Mg\_3Si\_2O\_5(OH)\_4(s) + 3CO\_2(g) + 2H\_2O(l) = 3MgCO\_3(s) + 2H\_4SiO\_4(aq) + 64 \text{ kJ/mol}} \tag{2}$$

$$\text{CaSiO}\_3(\text{s}) + \text{CO}\_2(\text{g}) + 2\text{H}\_2\text{O}(\text{l}) = \text{CaCO}\_3(\text{s}) + \text{H}\_4\text{SiO}\_4(\text{aq}) + 90 \text{ kJ/mol} \tag{3}$$

Stopic et al. [14] have shown the reaction path of direct forsterite carbonation in the aqueous solution without any deeper consideration of the formed silica particles as shown Equations (1)–(3) and Figure 1.

**Figure 1.** Reaction path of direct forsterite carbonation in the aqueous solution.

Although olivine is one mixed crystalline material (Mg, Fe)2SiO4, for simplicity, olivine consists only of Mg2SiO4, namely forsterite. First, gaseous carbon dioxide dissolves in the aqueous solution. Simultaneously, forsterite is dissolved in the aqueous solution (Equation (4)) forming aqueous silicic acid, then precipitates as amorphous silica (Equation (5)), which is a by-product, and lastly magnesium ions and carbonate form magnesite as shown with Equation (6):

$$\mathrm{Mg\_2SiO\_4(s)} + 4\,\mathrm{H^+(aq)} \stackrel{\mathrm{Na\_{\mathbb{R}}SiO\_4}}{\rightarrow} 2\,\mathrm{Mg^{2+}(aq)} + \mathrm{H\_4SiO\_4(aq)}\tag{4}$$

$$2\text{ H}\_{4}\text{SiO}\_{4}(\text{aq}) \stackrel{r\_{\text{SiO}\_{2}}}{\rightarrow} \text{SiO}\_{2}(\text{s}) + 2\text{ H}\_{2}\text{O}(\text{l})\tag{5}$$

$$\text{Mg}^{2+}\text{(aq)} + \text{CO}\_3^{2-}\text{(aq)} \stackrel{\text{DMgCO}\_3}{\rightarrow} \text{MgCO}\_3\text{(s)}\tag{6}$$

The determination of process parameters such as temperature, pressure and pH for maximum overall conversion rates is elementary. Direct CO2 sequestration at high pressure with olivine as a feedstock has already been performed in numerous studies at different temperatures and pressures with or without the use of additives such as carboxylic acid, and sodium hydroxide. It is reported that optimal reaction conditions are in the temperature range of 150–185 ◦C and in the pressure range of 135–150 bar [15]. Additives are reported to have a positive influence on carbonation rate, but without a study in detail. Optimal addition of additives are reported by Bearat et al. [16] in studies about the mechanism that limits aqueous olivine carbonation reactivity under the optimum sequestration reaction conditions observed as follows: 1 M NaCl + 0.64 M NaHCO3, at 185 ◦C and P

(CO2) about 135 bar. A reaction limiting silica-rich passivating layer forms on the feedstock grains, slowing down carbonate formation and raising process costs. Eikeland [17] reported that NaCl does not have significant influence on carbonation rate. The presented results show a conversion rate of more than 90% using a NaHCO3 concentration of 0.5 M, without adding NaCl. Ideally, the solid phases exist as pure phases without growing together. In reality, different observations are made on the behavior of solid phases. Daval et al. [18] reported about the high influence of amorphous silica layer formation on the dissolution rate of olivine at 90 ◦C and elevated pressure of carbon dioxide. This passivating layer may be either built up from non-stoichiometric dissolution, precipitation of amorphous silica on forsterite particles or a combination of both. These previously mentioned results suggest that the formation of amorphous silica layers plays an important role in controlling the rate of olivine dissolution by passivating the surface of olivine, an effect that has yet to be quantified and incorporated into standard reactive-transport codes. In contrast to that, Oelkers et al. [19] and Hänchen [20] observed stoichiometric dissolution and no build-up of a passivating layer except during start-up of experiments. Furthermore, magnesite may precipitate on undissolved forsterite particles leading to a surface area reduction and therefore a reduction on forsterite dissolution rate, which was reported by Turri et al. [21]. In addition to this undesired intermixing of solids, they observed pure particles of magnesite to be predominant in the intermediate particle class, amorphous silica particles to be mainly present in the smallest particle class and unreacted olivine particles to be predominant in the largest particle class. This knowledge may be of value for subsequent separation of products such as magnesium carbonate and silica.

CO2 sequestration with olivine as a feedstock was performed in a rocking batch autoclave at 175 ◦C and 100 bars in an aqueous solution and a CO2-rich gas phase from 0.5 to 12 h. Turri et al. [21] showed maintainable recovery of separate fractions of silica, carbonates and unreacted olivine. Characterization of the recovered solids revealed that carbonates predominate in particle size range below 40 μm. The larger, residue fraction of the final product after carbonation consisted mainly of unreacted olivine, while silica is more present in the form of very fine spherical particles. An addition of sodium hydrogen carbonate at 0.64 M, oxalic acid at 0.5 M and ascorbic acid at 0.01 M was successfully applied in order to obtain maximal carbonation, what leads also to a complete formation of silica.

Our paper deals with the formation of magnesium carbonate and especially nanosilica using an olivine from Norway (40.1 MgO, and 48.7 wt % SiO2) and with special attention on morphological characteristics of the obtained product and water solution after filtration, which was determined by structure and composition analysis (XRD, SEM; EDS; TEM and STEM).

#### **2. Experimental Section**

#### *2.1. Materials*

The samples used represent Steinsvik olivine from Norway as analyzed by a PW2404 XRF device (Malvern Panalytical B.V., Eindhoven, The Netherlands) and as shown in Table 1.

**Table 1.** Chemical composition of the investigated olivine from Norway (fraction between 20 and 63 μm) as analyzed by X-ray fluorescence (XRF) in wt %.


#### *2.2. Procedures*

The treatment of olivine was performed using the operations such as milling, sieving, carbonation in an autoclave, filtration and chemical analysis of solid and liquid sample shown at Figures 2 and 3. According to Reference 21 (Turri et al.) and Reference 14 (Stopic et al.) the carbonation tests have been carried out in the 1500 mL autoclave from Büchi Kiloclave Type 3E, Switzerland (as shown at

Figure 4) at 175 ◦C with 117 bar pure grade CO2 in the presence and the absence of the additives such as sodium bicarbonate, oxalic acid and ascorbic acid in duration of 2–4 h. An amount ranging from 100 to 300 g sample has been added to 1000 mL solution with mixing rate 600 revolution per min in different experiments. After reaction, the liquid had very low contents of metal cations and was analyzed via the induced coupled plasma optical emission spectrometry ICP OES analysis (SPECTRO ARCOS, SPECTRO Analytical Instruments GmbH, Kleve, Germany). Characterization of the solid products was restricted to the X-ray powder diffraction XRD (Bruker AXS, Karlsruhe, Germany) and X-ray fluorescence XRF analyses using Device PW2404 (Malvern Panalytical B.V., Eindhoven, The Netherlands).

**Figure 2.** Procedure for experimental work and characterization of samples.

**Figure 3.** Carbonation process of olivine and sampling.

After milling the particle size fraction 20–63 μm was tested in an autoclave. The change of temperature and pressure during a heating from room temperature to 175 ◦C and a subsequent carbonation of 100 g olivine in 1000 mL water was followed in time. Carbonation was performed by injection of carbon dioxide from a bottle at the fixed temperature within 2 h. After this reaction time, the pressure was decreased to the atmospheric values and the solution was cooled to room temperature. After opening the cover of the autoclave the solution was filtrated as shown by Figures 2 and 3. Subsequent to drying of the solid residue, XRD analysis was performed for the product and an initial sample of olivine. X-ray powder diffraction patterns were collected by a Bruker-AXS D8 Advance diffractometer in Bragg–Brentano geometry, equipped with a copper tube coupled with a primary nickel filter providing Cu Kα1,2 radiation and LynxEye detector. The microstructure of the solid samples was examined using a scanning electron microscope (SEM)–JEOL6380 LV (JEOL Ltd., Tokyo, Japan). Energy dispersive X-Ray spectroscopy (EDS) was utilized by JSM-6000 (JEOL Ltd., Tokyo, Japan) to reveal elemental composition of the samples analyzed by SEM. The characterization of solid and liquid products was performed using the ICP-OES analysis (SPECTRO ARCOS, SPECTRO Analytical Instruments GmbH, Kleve, Germany).

**Figure 4.** Sketch and picture of the autoclave: 1. Pressure pipe (Stahlflex); 2. needle valve; 3. tube for gas exhaust; 4. reactor shell with heating and cooling; 5. temperature sample head; 6. propeller mixer; 7. outlet valve; 8. analog manometer; 9. motor for the magnetic coupled stirrer; 10. cable for the measurement cutting site; 11. gas inlet; 12. testing rode for pressure; 13. working volume.

The values of temperature and pressure during carbonation are presented at Figure 5.

**Figure 5.** Pressure, temperature and time curves during work in autoclave (R-reaction; A-autoclave).

#### **3. Results and Discussion**

#### *3.1. Product Characterization–Analysis via XRD and SEM of Solid Product after Carbonation*

To evaluate the overall capability of the carbonation process, an experiment was performed using Norwegian olivine (20–63 μm) at 175 ◦C, 120 bar, 120 min, 600 rpm, in the presence of additives of sodium carbonate, oxalic and ascorbic acid considering the present mineralogical phases detected via XRD before and after the carbonation (Figure 6). A measurement range from 5–90◦ 2θ in 0.02◦ steps at 2 s per step are the chosen XRD parameters. XRD analysis was combined with a subsequent semi-quantitative evaluation of the mineral phase fractions. Table 2 provides information about the main mineral phase fractions as detected in the olivine samples within the given accuracy range. Forsterite, enstatite, lizardite and talc exhibited the most considerable mineralogical phase amounts. A decreased content of forsterite confirmed that the carbonation was successfully performed, what is indicated by 20–25% of magnesite in structure. The content of enstatite, lizardite and talc was not significantly changed, pointing out that these mineralogical phases show less reactivity within the performed process than forsterite.

**Figure 6.** XRD analysis of Norwegian olivine samples; initial state (top) and carbonation product (bottom).


**Table 2.** The semi-quantitative XRD analysis before and after carbonation.

The presence of magnesite and silica was confirmed using the SEM analysis of the solid product, as shown in Figure 7.

**Figure 7.** SEM analysis of initial olivine sample after carbonation.

As illustrated by Figure 7, SEM-analysis has confirmed that very small particles of SiO2 and magnesite are formed as rhombohedrons or hexagonal prisms at the surface of partially carbonated magnesium silicate. The challenge of future work is a separation of the formed nanosilica particles from the product.

#### *3.2. Product Characterization–Analysis of Precipitate from a Water*

After filtration of the carbonated products, the white and yellow precipitate appeared during staying after 7–10 days, as shown in Figure 8. Then, this precipitate was separated from water solution and dried at 110 ◦C after the night. The obtained dried products were analyzed by XRD, SEM, EDS, and scanning transmission electron microscopy (STEM) by JEM-2100F (JEOL Ltd., Tokyo, Japan).

**Figure 8.** Precipitated solid residues from water solution after carbonation at 175 ◦C, 120 bar, s/L 1:10 (V1-600 rpm, 120 min; V2-600 rpm, 240 min; V3-1800 rpm, 120 min; V3F-1800 rpm, 40 min; V3F2-1800 rpm, 2 min).

As shown in Figure 9, an increase of stirring speed from 600 rpm to 1800 rpm in an autoclave leads to increased formation of products changing color from a white to yellow one. At the same way an increased reaction time leads to an increased production of solid residue. We suppose that an increased stirring speed has a positive influence for the separation of a formed silica rich layer at a non-reacted magnesium silicate. At the other side the pressure in an autoclave was increased from 120 to 170 bar with an increasing stirring speed from 600 to 1800 rpm, what is an additional support for the silica separation and precipitation from solution.

**Figure 9.** Relationship between operational pressure in an autoclave and stirring speed.

As shown in Table 3 and with Equation (7), the consumption of hydrogen ions, pH-Value was increased from the starting value 7.2 to maximal value of 8.57, forming the precipitate of SiO2.


**Table 3.** The change of pH-Value of solution and pressure during a heating and carbonation.

$$\mathrm{Mg\_{1.8}Fe\_{0.2}SiO\_4 + H^+ \to 1.8 Mg^{2+} + 0.2 \, Fe^{2+} + SiO\_2 + H\_2O} \tag{7}$$

The maximal pressure amounted 162.2–170.1 bar at 1800 rpm, what leads to an increased separation of nanosilica. Oelkers et al. [22] studied the products after carbonation of olivine via the reaction (8). They found that quartz is not stable at partial CO2 pressures between 21.4 and 223.8 bar at temperatures from 120 to 200 ◦C. The spherical particles are containing high amounts of silicon and oxide atoms, which is confirmed by the EDS analysis (Figure 9).

$$0.5\,\mathrm{Mg}\_2\mathrm{SiO}\_4 + \mathrm{CO}\_2 \to \mathrm{MgCO}\_3 + 0.5\,\mathrm{SiO}\_2\tag{8}$$

J. Götze and M. Göbbels [23] have mentioned that an increase of pH-value above six increases the solubility of silica. EDS Mapping was used for analysis of this dried final product drawn from water solution, as shown in Figure 10.

**Figure 10.** *Cont*.

**Figure 10.** X-ray spectroscopy (EDS) mapping of final product obtained in the experiment V3F, (**a**) BF-SEM Image, (**b**) C K-carbon mapping, (**c**) Mg K-magnesium mapping, (**d**) O K-oxygen mapping, and (**e**) Si K-silicon mapping.

As shown in Figure 11, the TEM analysis of silicon oxide confirms that the formed silica particles have a spherical shape, with diameters of approximately 400–500 nm and the particles are amorphous.

**Figure 11.** The TEM and STEM analysis of the obtained products.

Since an extensive share of the investigated material in Figure 10 shows low carbon content, there is a high possibility that this area is Mg2SiO4, what is previously shown in Figure 6. A new model for formation of nanosilica from magnesium silicate (forsterite) in used olivine was shown in Figure 12.

**Figure 12.** The proposed schematic model for the formation of SiO2 from Mg2SiO4 in an autoclave.

We assumed that high stirring speed and high pressure in autoclave lead to the formation of ideally spherical silica particles together with larger fractions of magnesium carbonate. Weng et al. [24] found during a study of the kinetics and mechanism of mineral carbonation of olivine for CO2 sequestration that the addition of sodium bicarbonate can dramatically increase the ionic strength and aid the dissolution of Si to temporarily aqueous H4SiO4 followed by decomposition to amorphous silica and consequently the removal of Si-rich layer. The aqueous silicon was not stable and can be decomposed into amorphous silica, which was extensively observed in the aqueous solution after carbonation and settled down for more than one month at room temperature.

#### *3.3. Conclusions*

Synthesis of nanosilica was studied via carbonation of olivine using size fraction between 20 and 63 μm with solid/liquid ratios of 1:10 at 175 ◦C and partial pressure of CO2 more than 100 bar in an autoclave in the presence of additives such as sodium bicarbonate, oxalic and ascorbic acid. Under the above-mentioned conditions the ideally spherical particles of silica below 500 nm with amorphous grains were produced during carbonation. In comparison to ultrasonic spray pyrolysis, sol-gel and carbonation method via bubbling CO2 gas into calcium ions dispersing in methanol solution under an atmospheric pressure, this synthesis was performed in a water solution in a closed reactor (autoclave) under higher pressure conditions above 100 bar avoiding the formation of silica gel, what blocks the metal extraction. An increase of stirring speed from 600 rpm to 1800 rpm raises the pressure from 120 to 170 bar and leads to an increase of silica production because of a removal of passivated silica formed layer at forsterite particles. The precipitated silicate particles were separated at pH-values between 8.32 and 8.57. In order to validate the first results in a 0.25 L-and 1.5 L autoclave, new scale up experiments will be performed in 10 L and 1000 L-autoclaves. Due to a good filterability of the carbonated product, separation of nanosilica from magnesite and magnesium carbonate shall be considered in future work in detail.

**Author Contributions:** S.S. conceptualized, managed the research, and co-wrote the paper. D.K. performed the preparation of the olivine materials (grinding, sieving) and co-wrote the paper. H.W. co-wrote the paper. C.D. participated in our experimental part, analyzed the data and co-wrote the paper. S.E. supervised the XRFand XRD-analyses and co-wrote the paper with R.T. and B.F. supervised the personnel, provided funding and co-wrote the paper. I.K. performed the XRD, SEM, REM and STEM analysis of nanosilica. P.K. conceptualized the research and helped in the discussion of the morphological characteristics of obtained magnesium carbonate and nanosilica.

**Funding:** This research was funded by BMBF (Federal Ministry of Education and Research) in Berlin, grant number 033RCO14B (CO2MIN Project in period from 01.06.2017 to 31.05.2020).

**Acknowledgments:** For a continuous support and preparation of Figure 1 previously published in Metals in 2018, we would like to thank Andreas Bremen, AVT, RWTH Aachen University.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Synthesis of Tribological WS2 Powder from WO3 Prepared by Ultrasonic Spray Pyrolysis (USP)**

#### **Nataša Gaji´c 1,\*, Željko Kamberovi´c 2, Zoran Anđi´c 3, Jarmila Trpˇcevská 4, Beatrice Plešingerova <sup>4</sup> and Marija Kora´c <sup>2</sup>**


Received: 30 December 2018; Accepted: 22 February 2019; Published: 28 February 2019

**Abstract:** This paper describes the synthesis of tungsten disulfide (WS2) powder by the sulfurization of tungsten trioxide (WO3) particles in the presence of additive potassium carbonate (K2CO3) in nitrogen (N2) atmosphere, first at lower temperature (200 ◦C) and followed by reduction at higher temperature (900 ◦C). In addition, the ultrasonic spray pyrolysis of ammonium meta-tungstate hydrate (AMT) was used for the production of WO3 particles at 650 ◦C in air. The HSC Chemistry® software package 9.0 was used for the analysis of chemistry and thermodynamic parameters of the processes for WS2 powder synthesis. The crystalline structure and phase composition of all synthesized powders were analyzed by X-ray diffraction (XRD) measurements. The morphology and chemical composition of these samples were examined by scanning electron microscopy (SEM) combined with energy dispersive X-ray analysis (EDX).

**Keywords:** tribology materials; tungsten disulfide; ultrasonic spray pyrolysis; tungsten trioxide

#### **1. Introduction**

According to current studies, nearly one-quarter of the world's total energy consumption originates from tribological contacts [1]. By improving the performance of current tribological materials, energy losses due to friction and wear could potentially be reduced by 40% in the long term (15 years) and by 18% in the short term (8 years) [1]. Hence, in the face of increasing global requirements for saving energy, there is no doubt that the search for efficient tribological materials with improved performances will continue in the coming years because the application conditions of future tribomechanical systems will undoubtedly be much more demanding than the current ones [2]. Today's market of tribological materials has unique requirements for the release of harmful substances (Cd, Pb, Cr, etc.) and the prevention of their migration into the environment during their use [3]. It is generally considered that these elements and their compounds, which can be introduced into food or water, are difficult to eliminate from the body and according to current studies can have carcinogenic effects [4]. For all of the above reasons, despite their excellent performances, tribological materials must also be acceptable from economic and ecological points of view. Among the various tribological materials, WS2 has attracted a large amount of attention from researchers for its specific properties and extensive promising applications [5]. Even in small quantities, WS2 contributes to the high performance of tribological materials and their specific properties (chemical stability in a wide temperature range, the possibility of revitalizing the surface that they protect, they are corrosion-resistant, inert, non-toxic, non-magnetic, and have lamellar structure, low shear strength, high oxidation and thermal degradation resistance, and so forth), which are of particular importance for the functioning of modern tribomechanical systems. WS2 is applicable in various types of industries, such as automotive, aerospace, military, medical, and so forth.

WS2 crystals have a hexagonal structure composed of a layer of tungsten atoms packed in between two layers of sulfur atoms, as shown in Figure 1a. The bonding between W-S layers is very strong and covalent, but layers of S atoms are loosely bound through weak van der Waals forces. This structure is responsible for the interlayer mechanical weakness with low shear strength, which results in a macroscopic lubricating effect. Figure 1b shows the lubrication mechanism of WS2 single sheets.

**Figure 1.** (**a**) Hexagonal structure of WS2 (top view); (**b**) The relative sliding between two single layers of WS2 (slide view).

So far, various methods of producing WS2 of different size and morphology using WO3 as precursor have been established, such as: gas–solid phase reaction [6–8], chemical vapor deposition (CVD) [9–11], hydrothermal method [12–14], mechanochemical activation method [15], and solid–solid phase reaction [16–18].

Gas–solid reactions present a very simple approach to generating WS2. In most cases, WO3 is reacted with a sulfur-containing compound for extended periods of time at high temperatures. For example, WS2 has been synthesized in a tubular furnace through gas phase reactions between WO3 particles and H2/H2S gases at a temperature of 840 ◦C for 30 min in Ar gas flow [6], by sulfidation of hexagonal WO3 with H2S/H2 (15% H2S) at different temperatures: 400, 500, and 800 ◦C for 4 h [7], and by sulfurization of WO3 nanostructured thin film in a mixture of H2S and Ar gas (10:90) at different partial pressure values [8]. However, this method involves exposure to extremely toxic and harmful H2S at elevated temperatures.

WS2 flakes have been synthesized successfully on SiO2/Si substrate by the sulfurization of WO3 powder at high temperatures by the CVD method [9–11]. Furthermore, there was no poisonous H2S gas released during these experiments and WS2 of a high purity was obtained. Although this method has many advantages, it is very demanding to coordinate the complicated relations among many process parameters.

Using the hydrothermal method [12–14], WS2 has been synthesized by autoclaving a mixture of WO3 and sulfur precursor, followed by washing and drying of the resulting product. Although various inexpensive precursors can be used for this method, the productivity of this process is low and additional thermal treatment is necessary because the obtained WS2 has an amorphous structure.

Wu Z. et al. [15] have synthesized WS2 nanosheets by novel mechanochemical activation methods in which a ball-milled mixture of WO3 and S powder was annealed at 600 ◦C for 2 h in an atmosphere of Ar gas. This method seems to be environmentally advantageous and may be an alternative to the traditional route of synthesis, but it is a very complex process and a robust method for the production of particles with small dimension. Another difficulty arises from the fact that there is still a lack of clear interpretation of the exact reaction and activation mechanism.

Regarding solid–solid phase reaction, the synthesis of WS2 powder was carried out by sulfurization of the WO3 powder with thiourea in a N2 atmosphere at 850 ◦C for 1 h in a horizontal tube furnace [16–18].

In this investigation the sulfurization of the WO3 particles with S powder in the presence of additive K2CO3 in a nitrogen atmosphere was studied. In order to obtain an adequate precursor for synthesis, this research involved the production of WO3 particles using an ultrasonic spray pyrolysis method. It is a simple and low-cost method, which in continuous operation can generate spherical, non-agglomerated submicron particles by using commercially available (inexpensive) precursors [19]. The as-prepared WO3 particles were used for WS2 synthesis without any post processing because they were free of impurities.

#### **2. Experimental**

#### *2.1. Materials and Methods*

The raw materials used for the experimental test performed in this work were: WO3 obtained by ultrasonic spray pyrolysis (USP), commercial WO3 (Chemapol, Prague, Czech Republic), sulfur (Solvay & CPC Barium Strontium GmbH & Co, Hannover, Germany, powder with characteristic size <45 μm, purity 99.95%), and K2CO3 (Zorka, Šabac, Serbia). Commercially available WS2 powder (SpeedUP INTERNATIONAL, Belgrade, Serbia, with a minimum 99.4% WS2 and WO3) was used for comparative analysis. The production of precursor, WO3 powder, as well as the synthesis of WS2 powders were carried out in a rotary tilting tube furnace (ST-1200RGV).

The HSC Chemistry® software package 9.0 was used for the analysis of the chemistry and thermodynamic parameters of the processes for the synthesis of WS2 powder [20]. The determination of appropriate conditions for WS2 synthesis was crucial for this analysis. Therefore, the Gibbs free energy of theoretically suspected chemical reactions during the WS2 synthesis process, the phase stability diagram for the W–O–S system, and the equilibrium composition of the species in the WO3–S–K2CO3 system were calculated using HSC Chemistry software.

All obtained powders, as well as the commercial WS2 powder, were subjected to analysis on a scanning electron microscope (SEM) equipped with energy dispersive X-ray analysis (EDX), a MIRA FE-SEM, from TESCAN Inc. In SEM images of the WO3 powder, well-dispersed powder without any agglomerates was clearly seen and the determination of particle size distribution for the WO3 powder was done using the Image Pro Plus Software. However, the particle size of the WS2 powder was not identifiable using image analysis due to the poor dispersion of samples. Size measurements of the obtained WS2 powders were thus performed using a laser particle size analyzer (Malvern Instruments, Malvern, UK). In addition, all samples were subjected to X-ray diffraction analyses using a Philips PW 1710, X-Pert Pro diffractometer (Co Kα radiation, generated at 40 kV and 30 mA). Measurements were carried out at an angle interval 10◦ < 2θ < 119◦ with step 0.017◦. XRD results were analyzed by employing the Rietveld method with the help of PowderCell Software and the RIFRANE® programme.

#### *2.2. Synthesis of WO3 Powder*

The apparatus for WO3 synthesis consisted of an aerosol generator ("Profi Sonic", Prizma, Kragujevac), a horizontal reactor with a quartz glass tube (0.5 m diameter and 1.2 m length), a vacuum pump (VP125), and powder collectors (Figure 2).

**Figure 2.** Schematic illustration of experimental setup for the ultrasonic spray pyrolysis (USP) synthesis of WO3.

Firstly, the ammonium meta-tungstate hydrate (AMT—H26N6O41W12·aq) diluted in distilled water was put into the particle nebulizer to generate an aerosol. The concentration of AMT solution was 10 mmol/L. For this ultrasonic nebulizer system, the resonance frequency was set to 1.7 MHz. A vacuum pump and air with a flow rate of 5 L/min was used to introduce the generated aerosol droplets into the tubular reactor. Prior to the introduction, the temperature of the reactor was raised to 650 ◦C. The pressure of the system was adjusted using the reactor pressure controller. The calculated retention time of droplets in the reaction zone was estimated to be about one second. After thermal decomposition of the transported aerosol in the furnace, the formed WO3 was partially collected in the bottles with water and alcohol. During the spray pyrolysis process, the evaporation of water from aerosol droplets increases the concentration of AMT in the droplets. Finally, AMT is thermally decomposed according to Equation (1):

$$2\text{ H}\_2\text{N}\_6\text{O}\_{41}\text{W}\_{12} \to 24\text{ WO}\_3 + 12\text{ NH}\_3\uparrow + 8\text{ H}\_2\text{O}\uparrow + \text{O}\_2\uparrow. \tag{1}$$

As shown in Figure 3, the mechanism of WO3 particle formation proposed by Arutanti et al. [21] is comprised of different steps starting from an initial solution of AMT.

**Figure 3.** The mechanism of the formation of WO3 particles in the spray pyrolysis method.

#### *2.3. Synthesis of WS2 Powder*

A schematic illustration of the WS2 synthesis process is shown in Figure 4.

**Figure 4.** Schematic illustration of the experimental setup for the synthesis of WS2 powder.

K2CO3 powder (5 wt.%) was added to the WO3 and S powder mixture with a weight ratio of 60:40. The powders were mixed and ground for 15 min in the mixer. Then, the as-prepared mixture was transferred to a covered ceramic boat. The boat was placed in the center of the quartz tube and into the furnace. High-purity nitrogen gas was introduced through one side of the furnace, whilst the other side of the quartz tube was connected to a cooling system and outlet gas washing system. The total flow rate of the N2 gas was fixed at 200 cm3·min−<sup>1</sup> for all experimental conditions. Prior to heating the furnace, nitrogen gas was flushed constantly for about 30 minutes to remove residual air in the furnace. First, the furnace with the sample was heated at a lower temperature (200 ◦C) at a rate of 10 ◦C/min and maintained under these conditions for 2 h. Then, it was further heated at a rate of 5 ◦C/min followed by reduction at a higher temperature (900 ◦C). After 2 h, the furnace was turned off and allowed to cool down to room temperature. Nitrogen gas flow was stopped and the WS2 powder was collected from the boat.

#### **3. Result and Discussion**

#### *3.1. Thermodynamic Analysis*

Results of the thermodynamic analysis of the process of WS2 powder synthesis are discussed below. Using the assumption of raw materials for the composition of the synthesis process, the following chemical reactions (Equations (2)–(4)) were considered:

$$\rm{H\_3WO\_3 + K\_2CO\_3 + TS} = K\_2O \cdot NO\_3 + 2WS\_2 + CO\_2(g) + 3SO\_2(g), \tag{2}$$

3WO3 + K2CO3 + 5S = K2O·WO3 + CO2(g) + 2SO2(g) + WO2 + WS3, (3)

$$\rm{BWO\_3 + K\_2CO\_3 + 4S = K\_2O \cdot WO\_3 + CO\_2(g) + 2SO\_2(g) + WO\_2 + WS\_2} \tag{4}$$

Figure 5 shows the calculated results of the Gibbs energy of reactions versus temperature from the reaction of Equations (2)–(4). The temperature range that was considered was up to 1000 ◦C.

**Figure 5.** The change in Gibbs free energy (ΔG) versus temperature for the different possible reactions that take place during the WS2 synthesis.

As shown in Figure 5, it can be clearly seen that the ΔG of all reactions has a negative value at temperatures higher than 300 ◦C, which means that all of the reactions are theoretically possible from the thermodynamic point of view. However, among them, the changes of the Gibbs energy of reaction presented with Equation (2) has a more negative value relative to Equations (3) and (4) (i.e., Equation (2) is dominant).

The equilibrium composition of the WO3–K2CO3–S system at different temperatures was calculated using the HSC computer program based on the Gibbs energy minimization method, and the results are shown in Figure 6.

**Figure 6.** Equilibrium amount of WO3–S–K2CO3 reactant system in N2 gas atmosphere.

The highest stability of WS2 was achieved at approximately 800 ◦C when the stability of the reactants were decreasing. K2O·WO3 and WO2 phases also became stable at the same temperature. The main gaseous products of this reaction were gaseous SO2 and CO2.

In the analyzed system, the main products of the reactions at a temperature of 900 ◦C were WS2, K2O·WO3, WO2, CO2, and SO2. These results were only qualitative.

By using thermodynamic software, a phase stability diagram for the W–O–S system for constant partial pressure of oxygen was constructed (Figure 7).

**Figure 7.** Temperature–partial pressures (Tpp) phase stability diagram for the W–O–S system at constant oxygen partial pressure.

Considering the logarithmic sulfur vapor pressure of 1 atm [22] and experimental temperature of 900 ◦C, it was found that the predominant stability area was of WS2. At these temperatures, the vapor pressure of sulfur is such that it led to the saturation of the atmosphere with the sulfur vapor in the furnace. This meant that the amount of sulfur gas phase necessary to react with the WO3 was sufficient to establish contact between the mentioned phases. Further, the diffusion of sulfur into WO3 was enabled to form WS2 powder.

The synthesis of WS2 was performed in two stages: (i) initiation of synthesis at a low temperature (200 ◦C) followed by (ii) high-temperature (900 ◦C) reduction. The temperature of the first stage was selected to prolong the contact time of sulfur and WO3 in the starting powder mixture, before sulfur self-ignition (232 ◦C) which promotes the transformation of sulfur to a gas phase and the evaporation of a large amount of S and/or SO2. This loss of sulfur obstructs its diffusion into WO3 and consequently obstructs the synthesis of WS2. Thermodynamic analysis of the second stage indicated that the optimal temperature of the synthesis was above 800 ◦C. However, at lower temperatures oxides were formed and the chemical composition of the mixture was changed, and a lower level of crystallization occurred.

For that reason, the second stage of synthesis was carried out at 900 ◦C to prevent oxide formation and to increase the degree of crystallinity in the final WS2 product.

#### *3.2. Characterization of Synthesized WO3 Powder*

Figure 8a,b shows the SEM images of the synthesized WO3 powder by USP method under different magnifications. The result of EDX analyses from the presented scanning surface is given in Figure 8c, which reveals that the sample consisted of the elements tungsten and oxygen, and no other elements were observed.

**Figure 8.** (**a**,**b**) SEM images and (**c**) EDX spectrum of WO3 powder synthesized by USP.

The XRD pattern of the prepared WO3 particles is shown in Figure 9. It is evident from the pattern that no diffraction peaks from other elements of compounds were found in the samples. Therefore, it is obvious that the as-prepared samples were composed of WO3. The XRD pattern suggested that the prepared particles had two types of crystal structures: hexagonal and monoclinic.

**Figure 9.** XRD pattern of WO3 powder prepared by USP method.

The particle size distribution determination for the WO3 powder prepared by USP is presented in Figure 10.

**Figure 10.** Particle size distribution for WO3 powder prepared by USP method.

#### *3.3. Characterization of WS2 Powder*

Figure 11a demonstrates an SEM image of powder obtained using the WO3 precursor powder prepared by the USP method. The sample was composed of a large number of ultrathin WS2 flakes. Furthermore, WO2, WO3, and K2O·WO3 phases with larger dimensions of particles were noticeable. The structure of the ultrathin flakes from the marked locations is presented more clearly in Figure 11b. The results demonstrate the presence of flower-shaped WS2 particles composed of nanoflakes of 200–500 nm in length and 50 nm in mean thickness. A further EDX analysis of the sample presented in Figure 11c reveals that the product was composed of W, S, O, and K, which suggests that the powder was composed of mainly WS2 (Spectrum 1) and WO2, WO3, and K2O·WO3 (Spectrum 2) phases, in accordance with the results presented in Figure 6.

**Figure 11.** (**a**,**b**) SEM images and (**c**) EDX spectrum for WS2 synthesized using WO3 prepared by USP as precursor.

The results of the XRD analysis employing the Rietveld method are shown in Figure 12. These results revealed that WS2 phase was predominant (present in two crystalized forms: hexagonal and rhombohedral) and was accompanied by the presence of WO2 (9.80%), WO3 (3.04%), and K2O·WO3 (1.56%) phases.

The obtained WS2 particles were tested for their size and size distribution and the mean particle size was found to be around 950 nm (Figure 13).

WS2 powder synthesized using WO3 prepared by USP as precursor was selected as a lubrication additive in SF SAE 15W-40 motor oil. WS2 powder was ultrasonically dispersed into the base oil for 10 min. In addition, a base oil without any additive was also prepared. Wear tests were conducted using a ball-on-disc configuration on a Bruker UMT-3 tribometer (Bruker, Billerica, MA, USA). This test method involves a ball-shaped upper specimen that slides against a rotating disk as a lower specimen under a prescribed set of conditions. Both the ball and the steel disc were cleaned with acetone and dried with a normal stream of air before the test. A normal load of 50 N and a linear sliding speed of 0.1 m/s were used for the experiments, for a sliding distance of 500 m. The results of base oil with additive showed that the average value of the friction coefficient was *μ* ~ 0.1. It was concluded that the friction coefficient of the base oil (*μ* ~ 0.16) was improved by adding WS2. The mass ratio of WS2/base oil in samples was 1.0 wt.%.

The generated WS2 powder had plate-like particles, which were oriented differently when the starting material was commercial WO3 powder (Figure 14a). It can also be clearly seen from Figure 14b that WS2 particles were clustered together and exhibited evident agglomeration of up to 1 μm in size. The thickness of the WS2 plate-like particles was approximately 100 nm and their lengths varied from 500 nm to 1 μm. Results of EDX analyses from marked locations showed the presence of W and S elements and a certain content of O and K elements which indicated the presence of WS2 (Spectrum 2) and WO2, WO3, and K2O·WO3 (Spectrum 1) phases (Figure 14c).

**Figure 12.** XRD pattern of WS2 synthesized using WO3 prepared by USP as precursor.

**Figure 13.** Particle size distribution by intensity of WS2 particles synthesized using WO3 prepared by USP as precursor.

**Figure 14.** (**a**,**b**) SEM images and (**c**) EDX spectrum for WS2 synthesized using commercial WO3 as precursor.

The sample of WS2 powder synthesized using commercial WO3 as precursor was analyzed by XRD (Figure 15). Results showed the predominance of the WS2 phase, present in two crystalized forms: hexagonal and rhombohedral. WO3, WO2, K2O·WO3, and S phases were also identified, of which WO3 was hexagonal and monoclinic. The strong and sharp diffraction peaks in the pattern indicated that the product was very highly crystallized.

**Figure 15.** XRD pattern of WS2 synthesized using commercial WO3 as precursor.

The obtained WS2 particles were tested for their size and size distribution and the mean particle size was found to be around 500 nm (Figure 16).

**Figure 16.** Particle size distribution by intensity of WS2 particles synthesized using commercial WO3 powder as precursor.

Commercial WS2 powder as evidenced by SEM images consisted of nanoparticles with the presence of large agglomerates (Figure 17a). The uniform shape of particles can be seen in Figure 17b. On the basis of the EDX analysis as well as the high content of W and S, the presence of O was also confirmed (Figure 17c).

**Figure 17.** (**a**,**b**) SEM images and (**c**) EDX spectrum for commercial WS2.

The XRD analysis of the commercial WS2 powder indicated that there were major WS2 and WO3 phases and a minor WO3·0.33H2O phase, with low levels of crystallinity (Figure 18c). The presence of the amorphous phase was dominant, although the crystalline phase was also present.

**Figure 18.** XRD pattern of commercial WS2 powder.

Comparative analysis of the obtained powders with commercial WS2 powder showed differences in particle size and agglomerates. The commercial WS2 powder had the smallest particle size, but had more agglomerates compared to synthesized powders. Agglomerated powders are not good for tribological properties [23].

Particles of the synthesized WS2 powders ranged from submicrometers to micrometers in size. The mixture of these two powders should provide better friction and wear performance between contacting surfaces, according to N. Wu et al [24].

Apart from the presence of sulfide as the main product, commercial WS2 powders had a WO3 (8.70%) phase, whereas synthesized WS2 powders had K2O·WO3 (1.52–1.56%), WO2 (9.80–10.34%), and WO3 (2.04–3.04%) phases. In order to obtain WS2 powders that will provide low and stable friction coefficients, the presence of WO3 should be avoided [25]. In synthesized powders, the WO3 phase was less abundant relative to the levels found in commercial WS2 powders. Both tungsten dioxides and tungsten disulfides exhibited similar lubrication performances [26]. The materials with low oxygen content were more resistant to wear [27]. In this respect, WO2 in synthesized WS2 powders led to a lower friction coefficient. In addition, the K2O·WO3 phase in synthesized powders improved the thermal stability of obtained WS2 powders [28].

#### **4. Conclusions**

In response to the requirements for improving performances of currently available tribological materials, this work demonstrated the synthesis of WS2 powder using ultrafine WO3 powder prepared by USP method. In addition, we synthesized WS2 powder using commercial WO3 powder as precursor.

The synthesis of WS2 powder with the addition of K2CO3 as a fluxing agent reduced the WO3 precursor and protected the sample against oxidation, which is one of the advantages of the applied method.

Further, the results of the XRD analysis of synthesized powders were in accordance with thermodynamic predictions. In the synthesized WS2 powders, K2O·WO3 and WO2 phases were present in addition to the WS2 phase. These phases are useful from the aspect of thermal stability and friction coefficient control of the powders.

As a result of this investigation, two sizes of WS2 particles ranging from submicrometers to micrometers were obtained. Hence, future efforts are planned to investigate mixed micro/submicron lubrication systems composed of the synthesized WS2 powders.

In summary, tribological WS2 powder was successfully prepared by a facile and environmentfriendly method.

**Author Contributions:** Conceptualization, Ž.K.; Formal analysis, J.T.; Investigation, N.G.; Methodology, Ž.K. and Z.A.; Resources, M.K.; Software, N.G. and B.P.; Supervision, Z.A.; Validation, J.T. and B.P.; Writing—original draft, Ž.K.; Writing—review & editing, M.K.

**Funding:** This research was funded by the Ministry of Education, Science and Technological Development of the Republic of Serbia, project No. 34033.

**Acknowledgments:** This paper was done with the financial support of the Ministry of Education, Science and Technological Development of the Republic of Serbia and it is a result of project No. 34033.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **Influence of the Shape of Copper Powder Particles on the Crystal Structure and Some Decisive Characteristics of the Metal Powders**

#### **Ljiljana Avramovi´c 1, Vesna M. Maksimovi´c 2, Zvezdana Bašˇcarevi´c 3, Nenad Ignjatovi´c 4, Mile Bugarin 1, Radmila Markovi´c <sup>1</sup> and Nebojša D. Nikoli´c 5,\***


Received: 13 November 2018; Accepted: 31 December 2018; Published: 9 January 2019

**Abstract:** Three different forms of Cu powder particles obtained by either galvanostatic electrolysis or a non-electrolytic method were analyzed by a scanning electron microscope (SEM), X-ray diffraction (XRD) and particle size distribution (PSD). Electrolytic procedures were performed under different hydrogen evolution conditions, leading to the formation of either 3D branched dendrites or disperse cauliflower-like particles. The third type of particles were compact agglomerates of the Cu grains, whose structural characteristics indicated that they were formed by a non-electrolytic method. Unlike the sharp tips that characterize the usual form of Cu dendrites, the ends of both the trunk and branches were globules in the formed dendrites, indicating that a novel type of Cu dendrites was formed in this investigation. Although the macro structures of the particles were extremely varied, they had very similar micro structures because they were constructed by spherical grains. The Cu crystallites were randomly oriented in the dendrites and compact agglomerates of the Cu grains, while the disperse cauliflower-like particles showed (220) and (311) preferred orientation. This indicates that the applied current density affects not only the morphology of the particles, but also their crystal structure. The best performance, defined by the largest specific surface area and the smallest particle size, was by the galvanostatically produced powder consisting of disperse cauliflower-like particles.

**Keywords:** copper; powder; electrolysis; hydrogen; SEM; XRD; PSD

#### **1. Introduction**

Copper powders have found a wide industrial application for a very long time [1]. Cu in a powder form is often used in the electrical and electronic industries due to its excellent electrical and thermal characteristics. The self-lubricating bearing is probably the most common application of Cu powder and about 70% of the total Cu powder production in a granular form is used for that purpose. This application takes advantage of the ability to produce a component with controlled interconnected and surface-connected porosity. Copper powders are also used in such nonstructural applications as

brazing, cold soldering, and mechanical plating, as well as for medals and medallions, metal-plastic decorative products and a variety of chemical and medical purposes.

The Cu powder production methods can be divided into electrolytic and non-electrolytic (or chemical) methods. Processes such as ultrasonic spray pyrolysis [2], solvothermal synthesis [3], cementation [4], chemical reduction methods [5–8], the high-energy electrical explosion method [9], atomization [10], pyrolysis [11], polyol processes [12–14], hydrometallurgy [15], etc. belong to the group of non-electrochemical or chemical processes of synthesis. The chemical reduction method is often referred to as electroless deposition [15,16]. Various reducing agents, such as ascorbic acid [5,8,16], hydrazine hydrate [7], sodium borohydride [6], and formaldehyde [15], are used in the chemical reduction processes. Copper, in the form of nanoparticles, can be synthesized in ethylene glycol (EG) using copper sulphate as a precursor, and vanadium sulfate as an atypical reductant [17]. Metal nanoparticles can also be obtained by biosynthesis processes using the microorganisms, which has clear advantages compared to chemical synthesis methods [18]. The biosynthesis processes are environmentally friendly, no toxic chemicals or reagents are needed for these processes, and it is possible to synthesize particles that cannot be obtained using chemical synthesis methods.

Aside from the above-mentioned methods of synthesis, electrolysis is often used for Cu powder synthesis. The advantage of this method of synthesis can primarily be attributed to the fact that the shape and size of particles can be easily regulated by the choice of electrolysis regime and parameters [19]. The regimes, both constant (potentiostatic [20] and galvanostatic [21–23]) and periodically changing [19,24–26], are used for the Cu synthesis in powder form. The electrolysis parameters that affect the final shape of the particles, are: the type and composition of electrolytes, presence of additives, temperature, type of cathode, electrolysis time, etc. [19,21,25,27–30]. Special attention has been dedicated to the effect of the hydrogen evolution reaction as the parallel reaction to Cu electrolysis in the production range of Cu powder [20,24].

The shape of Cu powder particles is related to their synthesis method. The four various forms of Cu particles can be identified as: almost ideal micro spheres, irregular rough particles, disperse cauliflower-like particles and dendrites. Dendrites are the most commonly observed shape of particles and they can be obtained by both electrolysis and some chemical methods such as the galvanic replacement reaction method [31]. The very disperse cauliflower-like particles are formed by the electrolysis in conditions of strong hydrogen evolution as the parallel reaction [19,20]. Irregular particles are obtained by water atomization [32] while spherical particles are obtained by the gas-atomizing process [32]. Aggregates of non-uniform irregular particles are obtained by the polyol process [12]. Almost ideal Cu spheres can be also obtained by the polyol process [12]. Polyhedral, non-agglomerated monodispersed particles are obtained using ascorbic acid as the reducing agent [4]. The non-agglomerated almost spherical particles are obtained by ultrasonic spray pyrolysis [2].

Recently, a strong correlation between the morphologies of silver powder particles and their crystal structure was found [33,34]. Here, this type of investigation has been continued with the aim of establishing the existence of the same correlation for Cu powders. For this reason, the three completely various morphologies of Cu powder particles were analyzed and correlated with their crystal structure. Two of them are obtained by electrolysis under completely different conditions, without and with vigorous hydrogen evolution. The third type was commercially supplied, and on the basis of morphological characteristics, it is clear that it was obtained by some non-electrochemical method. In order to examine the effect of the shape of particles on the decisive characteristics that define the behavior of the powders as a collection of particles, the specific surface area (SSA) and particle size distribution (PSD) were also analyzed.

#### **2. Materials and Methods**

Electrolyte, formed by dissolution of 0.10 mol·dm−<sup>3</sup> CuSO4 in 0.50 mol·dm−<sup>3</sup> H2SO4, was used for the electrolytic production of Cu powders. Cu powder was galvanostatically produced at the current densities of 14.4 and 384 mA·cm−2. Hereinafter, these powders are denoted as Cu(14.4) for powder produced at 14.4 mA·cm−<sup>2</sup> and Cu(384) for powder produced at 384 mA·cm−2. Electrolysis was carried out at a temperature of 21.0 ± 0.5 ◦C in an open cell of cylindrical shape. The cylindrical Cu wire of the overall surface area 0.50 cm2 was used as the working electrode. The counter electrode was Cu foil. Figure 1 shows a configuration of electrodes in the cell. The ultra-pure water and p.a. reagents (CuSO4 × 5H2O, H2SO4) were used for electrolyte preparation for the Cu powder synthesis. Powder was produced with the electricity amount of 10 mA·h·cm<sup>−</sup>2, and removed from the electrode surface after this amount of electricity was reached.

**Figure 1.** (**a**) Scheme of the cell used for production of the Cu powders by electrolysis. WE—working electrode; CE—counter electrode, (**b**) top view, (**c**) side view.

The electrolytically produced powder was compared to the commercially supplied powder, Sigma-Aldrich company (Saint Louis, MO, USA), product No. 326453 (powder (spheroidal), 14–25 μm, 99%). Hereinafter, this powder is denoted as Cu(CHEM).

A scanning electron microscope (SEM)—TESCAN Digital Microscopy company (model VEGA3, Brno, Czech Republic) was used for the morphological characterization of the produced particles.

The Rigaku Ultima IV diffractometer (Rigaku Co. Ltd., Tokyo, Japan) with CuKα radiation was used to study the crystal structure of the produced powders. Four reflections in the 2*θ* range between 30◦ and 95◦ were recorded in order to determine the preferred orientation of powder particles. For this purpose, calculation of the Texture Coefficient, *TC*(*hkl*) and Relative Texture Coefficient, *RTC*(*hkl*), based on an analysis of data obtained by the X-ray diffraction (XRD) method, was made. It was made in the following way: using an intensity of each reflection (*hkl*) plane, the following ratios (in percentage) were calculated by applying Equation (1) [35]:

$$R(hkl) = \frac{I(lkl)}{\sum\_{i}^{4} I(l\_{i}k\_{i}l\_{i})} \times 100\tag{1}$$

where *<sup>I</sup>*(*hkl*) is the intensity of each (*hkl*) reflection plane, while <sup>4</sup> ∑ *i I*(*hikili*) represents the sum of intensities of all recorded reflection planes. Note: the values of intensities are given in cps.

Then, using the determined *R*(*hkl*) coefficients, the values of the Texture Coefficient, *TC*(*hkl*), were calculated by applying Equation (2):

$$TC(hkl) = \frac{R\left(hkl\right)}{R\_{\text{s}}\left(hkl\right)}\tag{2}$$

*Metals* **2019**, *9*, 56

where *R*s(*hkl*) is determined in the same way as presented in Equation (1), but taking into consideration the Cu standard (04-0836).

In this way, accurate quantitative information about the absolute intensity of each of these reflections were obtained determining the *TC*(*hkl*) coefficients. Also, the intensity of each reflection plane in relation to the other reflection planes represents the relevant information, and is defined as the Relative Texture Coefficient, *RTC*(*hkl*) according to Equation (3):

$$\text{RTC}(hkl) = \frac{\text{TC}(hkl)}{\sum\_{i}^{4} \text{TC}(h\_{i}k\_{i}l\_{i})} \times 100\tag{3}$$

The *RTC*(*hkl*) coefficient defines the intensity of the considered (*hkl*) orientation in relation to the standard which is included in the *TC* values.

A MALVERN Instruments MASTERSIZER 2000 (MALVERN Instruments Ltd., Malvern, Worcestershire, UK) device was used to determine the particle size distribution (PSD) and specific surface area (SSA) of the powders. Malvern Software (Version 5.60, MALVERN Instruments Ltd., Malvern, Worcestershire, UK) was used in order to obtain the SSA values.

#### *Determination of the Average Current Efficiency of Hydrogen Evolution (ηI,av(H2))*

In order to quantify the amount of hydrogen, generated in the galvanostatic electrolysis, the average current efficiency of hydrogen evolution, ηI,av(H2) was determined. For that purpose, the working and counter electrodes of Cu were placed in a burette which was positioned so that the overall volume of evolved hydrogen in the galvanostatic electrolysis remained in it. The surface area of the Cu working electrode, placed in a burette, was 0.63 cm2. The current efficiency of hydrogen evolution in time *t*i, ηI,i(H2), in %, is given by Equation (4):

$$
\eta\_{\rm I,i}(\rm H\_2) = \frac{V(\rm H\_2)}{\mu(\rm H\_2)S\_0it\_i} 100 \tag{4}
$$

where

$$
\mu(\text{H}\_2) = \frac{V}{nF} = \frac{24,120 \text{ cm}^3}{2 \times 26.8 \text{ Ah}} = 450 \frac{\text{cm}^3}{\text{Ah}} \tag{5}
$$

and *V*(H2) is the volume of evolved hydrogen in time *t*i, *nF* is the number of Faradays per mole of spent ions, *V* is the molar volume of gas at temperature of 21.0 ◦C (i.e., 24,120 cm3), *S*<sup>0</sup> is the surface of working electrode, and *j* is the current density of electrolysis. The average values of current efficiency for the hydrogen evolution reaction, ηI,av(H2) are determined as ηI,av(H2)=(1/*t*) *t* 0 ηI,i(H2)d*t*, where *t* is time of electrolysis.

#### **3. Results**

#### *3.1. Morphological Analysis of Cu Powders Produced by the Electrolytic and Non-Electrolytic Methods*

The polarization curve for the copper electrodeposition from 0.10 mol dm−<sup>3</sup> CuSO4 in 0.50 mol dm−<sup>3</sup> H2SO4 is shown in Figure 2. The plateau of the limiting diffusion current density was in the range of overpotentials between 300 and 750 mV, with a limiting diffusion current density value (*j*L) of 9.6 mA cm<sup>−</sup>2. In the galvanostatic regime of electrolysis, Cu in the powder form is formed at current densities larger than the limiting diffusion current density [19].

Morphologies of the Cu particles, obtained at current densities corresponding to 1.5 (*<sup>j</sup>* = 14.4 mA·cm<sup>−</sup>2) and 40 (*<sup>j</sup>* = 384 mA·cm<sup>−</sup>2) times larger values than the limiting diffusion current density, are shown in Figures 3 and 4, respectively. Considering the fact that the hydrogen evolution as the second reaction in the Cu electrolysis, commences inside the limiting diffusion current density plateau, a strong difference in the amount of generated hydrogen at these current densities, with strong

consequences for the morphology, structure and decisive characteristics of the obtained powders, was expected. Accordingly, inspection of the presented microphotographs shows the strong effect of the applied current densities on the morphology of Cu powder particles.

**Figure 2.** Polarization curve for the Cu electrodeposition from 0.10 mol·dm−<sup>3</sup> CuSO4 in 0.50 mol·dm−<sup>3</sup> H2SO4.

Figure 3a shows the Cu deposit obtained immediately after finishing the process of electrolysis at 14.4 mA·cm−2. The three morphological forms can be identified on this microphotograph: very branched 3D (three dimensional) dendritic forms, holes originating from detached hydrogen bubbles and cauliflower-like agglomerates of Cu grains. It can be seen that the shape of the dendritic forms (Figure 3a,b) was different from all of the so far observed forms of Cu dendrites. In previously observed forms of Cu dendrites, the tips of both the trunk and branches were sharp, while in this form they end with globules (Figure 3c,d). This form is observed for the first time in this investigation, and for this reason, it can be said that this shape represents a completely novel type of Cu dendrites. The size of the globules was from 3–5 μm for those in the branches, to about 10 μm at the tops of both branches and trunk. The size of the holes formed from the detached hydrogen bubbles, approached 100 μm. The typical cauliflower-like agglomerate of Cu grains, formed between the dendrites and holes is shown in Figure 3e. It consists of small agglomerates of Cu grains, surrounded by irregular channels. The size of the individual grains in these agglomerates was considerably smaller than the size of globules, and it was about 1 μm. The particles, obtained after removing the deposit from the electrode surface, are shown in Figure 3f,g. From a macromorphological point of view, no difference is observed between those on the electrode surface after the finished process of electrolysis and those after their removal. It is necessary to note the existence of channel structure inside the cauliflower-like particles (Figure 3f), which is a result of hydrogen evolution reaction.

**Figure 3.** *Cont*.

**Figure 3.** Morphologies of the powdered copper obtained at a current density of 14.4 mA·cm−2: (**a**) appearance of electrode surface after the process of electrolysis, (**b**) dendrite, (**c**) tops of branches of dendrite, (**d**) top of trunk of dendrite, (**e**) cauliflower-like agglomerates of Cu grains formed among dendrites and holes, (**f**,**g**) particles obtained after removal from the electrode surface.

Figure 4a shows the SEM microphotograph of copper electrode, obtained at a current density of 384 mA·cm−2. A typical honeycomb-like structure, constructed from holes formed from the detached hydrogen bubbles (Figure 4b) surrounded by the disperse cauliflower-like agglomerates of Cu grains (Figure 4c,d) was obtained. Increasing the current density from 14.4 to 384 mA·cm−<sup>2</sup> led to an intensification of the hydrogen evolution reaction, which is manifest by the increase in the number of holes and by a decrease in their size. In this case, a hole size was about 70 μm. As a result of intensification of the hydrogen evolution, the cauliflower-like agglomerates of Cu grains were more disperse than those obtained at 14.4 mA·cm<sup>−</sup>2. The size of the grains in these agglomerates was about 200 nm, that is, approximately five times smaller than those formed at 14.4 mA·cm<sup>−</sup>2. The typical forms of Cu particles, obtained by removal of deposit from the electrode surface, are shown in Figure 4e,f. Inhibition of dendritic growth and the cauliflower-like character of the formed particles as a result of vigorous hydrogen evolution, are clearly visible in Figure 4e,f.

**Figure 4.** Morphologies of the powdered copper obtained at a current density of 384 mA·cm−2: (**a**) appearance of the electrode surface after the process of electrolysis (honeycomb-like structure), (**b**) hole formed from the detached hydrogen bubble, (**c**) disperse cauliflower-like agglomerates of Cu grains formed around holes, (**d**) typical agglomerate constructing disperse cauliflower-like agglomerates, (**e**,**f**) particles obtained after removal from the electrode surface.

Figure 5 shows the morphology of commercially supplied Cu powder. It can be seen in Figure 5 that this powder consists of agglomerates of approximately spherical (spheroidal) grains of various size. It can be seen that these agglomerates have a relatively compact structure. The size of these agglomerates was larger than 10 μm, and they were constructed from grain sized in the range of 1–10 μm. Also, some of them had very irregular shapes. Considering the fact that these forms have never been observed among electrolytically synthesized particles, it is clear that they are obtained in some non-electrochemical way. The irregular, relatively compact forms constructed from approximately spherical grains in a wide range of sizes confirmed this assumption.

**Figure 5.** Typical copper particles obtained by chemical processes: (**a**) ×500, (**b**) ×4000, (**c**) ×4500, (**d**) ×3500.

#### *3.2. Structural Analysis of Cu Powders Produced by the Electrolytic and Non-Electrolytic Methods*

The XRD patterns of particles, obtained by removal the deposits obtained at current densities of 14.4 and 384 mA·cm−2, and Cu standard (04-0836), are shown in Figure 6. The peaks at 2*<sup>θ</sup>* angles of 43.3◦, 50.4◦, 74.1◦ and 89.9◦ correspond to (111), (200), (220) and (311) crystal planes. The appearance of peaks at these angles confirms a face centered cubic (FCC) crystal lattice of Cu. A larger ratio of Cu crystallites oriented in the (111) plane observed in the both diffractograms, can be ascribed to the lower surface energy of this plane in relation to the other planes, because the surface energy (γ) values follow a trend: γ<sup>111</sup> < γ<sup>100</sup> < γ<sup>110</sup> [36,37]. The preferred orientation of the obtained particles was investigated by an analysis of the peak intensity ratios (111)/(200), (111)/(220) and (111)/(311), and by determination the *TC*(*hkl*) and *RTC*(*hkl*) coefficients.

**Figure 6.** XRD patterns of Cu particles obtained at current densities of 14.4 and 384 mA·cm<sup>−</sup>2, and Cu standard (04-0836).

The values of peak intensity ratios for the Cu particles produced at current densities of 14.4 mA·cm−<sup>2</sup> (Cu(14.4)) and 384 mA·cm−<sup>2</sup> (Cu(384)), and the same ratios for Cu standard are given in Table 1. It can be seen that (111)/(220) and (111)/(311) for the Cu(384) powder are considerably smaller than those for the Cu standard. On the other hand, for the Cu(14.4) powder, these ratios are still relatively close to the values for the Cu standard, indicating the random orientation of Cu crystallites in the particles shown in Figure 3. It is known [33] that the values of ratios considerably larger than those of the standard indicate the presence of (111) preferred orientation. However, there is no data about what happens with the crystal orientation when the peak intensity ratios are smaller than those of the standard. Therefore, the additional analysis of the crystal orientation of particles was carried out by determining the Texture Coefficient, *TC*(*hkl*), and Relative Texture Coefficient, *RTC*(*hkl*).

**Table 1.** Ratios of intensities of the diffraction peaks for the analyzed powders and Cu standard.


The values of *TC*(*hkl*) and *RTC*(*hkl*) coefficients obtained for the electrolytically produced particles at current densities of 14.4 and 384 mA·cm−<sup>2</sup> (Cu(14.4) and Cu(384), respectively) are given in Table 2. The values of *TC*(*hkl*) coefficients larger than 1 indicate the existence of preferred orientation [33–35]. On the other hand, since four main reflections were analyzed, the values of *RTC* coefficients larger than 25% indicate the existence of a preferred orientation.

**Table 2.** Texture calculations for Cu powders obtained by the galvanostatic regime at current densities (*j*) of 14.4 and 384 mA·cm−<sup>2</sup> and for commercially available powder (*<sup>j</sup>* = 14.4 mA·cm−2—14.4, *<sup>j</sup>* = 384 mA·cm<sup>−</sup>2—384, commercially available powder—CHEM, s—Cu standard).


The data analysis, presented in Table 2, shows that the values of the *RTC* coefficients for Cu(14.4) powder are about 25%, confirming the randomly orientated crystallites in the particles obtained at a current density of 14.4 mA·cm−2. However, a different situation is observed when the particles produced at 384 mA·cm−<sup>2</sup> were analyzed. The *RTC* coefficients for (220) and (311) planes are larger than those for the Cu standard, indicating the existence of a preferred orientation in these planes. Thus, it is clear that peak intensity ratios smaller than the values for the Cu standard (Table 1) indicate the preferred orientation in that plane.

Figure 7 shows the XRD pattern of chemically synthesized Cu powder (Cu(CHEM)) and Cu standard. The appearance of peaks at the same angles, as in the case of electrolytically produced Cu powders, confirms the face centered cubic (FCC) crystal lattice of Cu. Also, similar to the electrolytically produced powders, the Cu crystallites were predominately oriented in the (111) plane. The peak intensity ratios (111)/(200), (111)/(220) and (111)/(311) for this powder are included in Table 1. It can be seen in Table 1 that the values are very close to those for the Cu standard, confirming that the powder consists of randomly distributed spherical grains.

**Figure 7.** XRD pattern of chemically synthesized Cu powder, and Cu standard (04-0836).

The values of *TC*(*hkl*) and *RTC*(*hkl*) coefficients for this powder are shown in Table 2. The values of *TC* coefficients very close to 1, and *RTC* coefficients about 25%, clearly point to the random orientation of the Cu crystallites in these particles.

#### *3.3. Analysis of the Specific Surface Area (SSA) and Particle Size Distribution (PSD) of the Considered Powders*

To compare the powders obtained by the different methods of synthesis and under different conditions of electrolysis, the specific surface area (SSA) and particle size distribution (PSD), as two very important characteristics that describe the behavior of powders as a collection of the particles were analyzed. The values of the SSA for the analyzed powders are summarized in Table 3. It can be seen that the largest SSA value is obtained for the powder that consisted of disperse cauliflower-like particles obtained by electrolysis at 384 mA·cm<sup>−</sup>2. The SSA of powder consisting of particles produced at this current density, was almost two time larger than the value for chemically synthesized powder. The smallest SSA value was obtained for the powder including the ordered dendritic structure galvanostatically produced at 14.4 mA·cm<sup>−</sup>2.

**Table 3.** Values of specific surface area (SSA), particle size with the maximum volume ratio (PSMVR), and the average current efficiency for hydrogen evolution reaction (ηI,av(H2)) for powders obtained by electrolysis at current densities of 14.4 mA·cm−<sup>2</sup> (Cu(14.4)) and 384 mA·cm−<sup>2</sup> (Cu(384)), and chemically synthesized powder (Cu(CHEM)).


The PSD curves obtained for all three types of particles are shown in Figure 8. A uniform distribution of particles was obtained in all cases. The maximum volume ratios shift towards the smaller particle size in the same way as the increase of SSA. The maximum volume ratio for the powder produced at 384 mA·cm−<sup>2</sup> corresponds to a particle size of 10.9 <sup>μ</sup>m (this value as well as those obtained for the other powders are included in Table 3), and it is more than 50% smaller than the maximum volume ratio obtained at a current density of 14.4 mA·cm<sup>−</sup>2, corresponding to a particle size of 23.2 <sup>μ</sup>m. The presence of larger particles in powder produced at 384 mA·cm−<sup>2</sup> (Figure 4e) is also identified in the PSD curve for this powder. For the chemically synthesized powder, the maximum volume ratio corresponds to a particle size of 14.6 μm, and it is situated between those for the electrolytically produced powders.

**Figure 8.** Particle size distribution (PSD) curves obtained for the particles produced at current densities of 14.4 and 384 mA·cm<sup>−</sup>2, and chemically synthesized powder.

#### **4. Discussion**

Three different morphologies of Cu particles were analyzed in this study: the 3D dendrites, disperse cauliflower-like agglomerates of Cu grains and compact agglomerates of Cu grains. Two of these were obtained by the electrolysis processes: 3D dendrites and disperse cauliflower-like agglomerates of Cu grains. Hydrogen evolution, as a parallel reaction to the Cu electrolysis at high current densities and overpotentials, had a crucial role in creating the final morphology of these particles. The amount of produced hydrogen is quantified by determination of the average current efficiency for hydrogen evolution, as presented in Figure 9 for the Cu electrolysis at current densities of 14.4 and 384 mA·cm<sup>−</sup>2.

Figure 9 shows the dependencies of the evolved hydrogen volume and current efficiency for the hydrogen evolution reaction in a time *t*<sup>i</sup> on the electrolysis time obtained at current densities of 14.4 (Figure 9a) and 384 mA·cm−<sup>2</sup> (Figure 9b). It is clear that increasing the current density leads to a strong intensification in hydrogen evolution. The values of the average current efficiency for the hydrogen evolution reaction, obtained at the current densities of 14.4 and 384 mA·cm<sup>−</sup>2, are given in Table 3.

**Figure 9.** The dependencies of the evolved hydrogen volume and current efficiency for the hydrogen evolution reaction on the electrolysis time obtained at current densities of: (**a**) 14.4 mA·cm−<sup>2</sup> and (**b**) 384 mA·cm<sup>−</sup>2.

The amount of hydrogen that evolved at a current density of 14.4 mA·cm−<sup>2</sup> was not sufficient to have any effect on the hydrodynamic conditions in the near-electrode layer. The proof for this is the rare holes formed from detached hydrogen bubbles and a somewhat non-uniform electrode surface determined by very branched 3D dendrites and cauliflower-like agglomerates of Cu grains (Figure 3a,b). As already mentioned, the 3D dendrites shown here represent a novel type of dendrite observed for the first time in this investigation. Formation of this type of dendrites can be explained from an electrochemical point of view in the following way: decrease of potential occurs with the electrolysis time in the galvanostatic regime of electrolysis. In this case, the potential at the end of electrolysis processes, corresponding to an amount passed electricity of 10 mA·h·cm−2, was about 275 mV. This potential corresponds to the mixed activation-diffusion control at which formation of globules like those shown in Figure 3 occurs [19,38].

On the other hand, a very uniform honeycomb-like structure was formed at a current density of 384 mA·cm−<sup>2</sup> (Figure 4). The amount of evolved hydrogen at this current density was vigorous enough to have a strong effect on the hydrodynamic conditions in the near-electrode layer. Namely, hydrogen generated during electrolysis causes a strong stirring of electrolyte in the near-electrode layer, leading to a decrease in diffusion layer thickness, an increase in the limiting diffusion current density, and a decrease in the degree of diffusion control in the electrodeposition process. The mechanism for the formation of the very disperse cauliflower-like particles, shown in Figure 4, was completely different to the one responsible for the formation of the dendritic particles (Figure 3). The concept

of "effective overpotential" can be applied to the formation of this particle type [19,39]. According to this concept, when hydrogen evolution is vigorous enough, the electrodeposition process occurs at an overpotential that is effectively lower than the specified one, and this overpotential is denoted as "effective" in the deposition process. From the morphological point of view, this means that the morphologies of metal deposits become similar to those obtained at some lower overpotentials at which the hydrogen evolution does not occur or is very slow. Formation of the cauliflower-like particles instead of dendrites confirms that there is a lower degree of diffusion control at this current density than at 14.4 mA·cm<sup>−</sup>2.

The third type of particles are irregular compact agglomerates of Cu grains (Figure 5). On the basis of the macromorphology, which is determined by the compact irregular shapes, it is clear that the particles were formed by some non-electrochemical method of synthesis.

The XRD analysis showed that the chemically obtained particles and those obtained by electrolysis at 14.4 mA·cm−<sup>2</sup> were randomly oriented. Unlike these, the electrolytically obtained particles at 384 mA·cm−<sup>2</sup> showed (220) and (311) preferred orientation. The different preferred orientation of the particles produced by the galvanostatic regime of electrolysis can be explained from the electrochemical point of view using the basic laws related to the metal electrocrystallization processes [19]. Namely, the (111) plane is a "slow-growing" one, while the other planes ((200), (220) and (311)) are the "fast-growing" planes [33,40]. In the growth process, the "fast-growing" planes disappear, while the "slow-growing" (111) plane survives. This explains the predominant orientation of Cu crystallites in the (111) plane in particles obtained at both current densities. On the other hand, a nucleation rate depends on the potential according to Equation (6) [19]:

$$J = K\_1 \exp\left(-\frac{K\_2}{E^2}\right) \tag{6}$$

where *J* is the nucleation rate, *K*<sup>1</sup> and *K*<sup>2</sup> are the constants independent of potential, and *E* is the potential. According to Equation (6), the nucleation rate increases with increasing the potential of electrolysis. At current density of 14.4 mA·cm<sup>−</sup>2, a potential response was in the (275–900) mV range, while at 384 mA·cm<sup>−</sup>2, this response was in the (1100–1250) mV range. Therefore, the nucleation rate was considerably larger at 384 mA·cm−<sup>2</sup> than at 14.4 mA·cm−2, causing the formation of a larger number of nuclei in the initial stage of electrolysis at 384 mA·cm−<sup>2</sup> than at 14.4 mA·cm−2. With the increased number of initially formed nuclei, there is an increased probability that the larger number of the "fast-growing" planes survive the growth process. In this way, the larger number of Cu crystallites oriented in (220) and (311) planes in the particles produced at 384 mA·cm−<sup>2</sup> than at 14.4 mA·cm−<sup>2</sup> can be explained.

The correlation between the morphology of the particles and crystal structure is also observed for some types of chemically synthesized particles. Namely, Cu nanowires synthesized under hydrothermal conditions using oleyl amine (OLA) and glucose as a reducing agent showed an XRD pattern with a higher intensity (200) peak than (111) peak [41]. The larger ratio of Cu crystallites oriented in the (200) plane than in (111) plane can be ascribed to the existence of a nanocubic structure [42], where a relatively large number of nanocubes coexist with the nanowires synthesized using OLA.

It is clear from the above consideration that electrolysis processes have certain advantages in the metal powder production over other methods. The advantages of this method include the easy control of the shape and size of particles by the choice of parameters of electrolysis. On the other hand, the particles obtained by chemical processes show a tendency towards aggregation [43], as observed here, which is avoided in the electrolysis processes by a vigorous hydrogen evolution reaction that can prevent this process. It is clear that the "current of hydrogen" formed through the interior of cauliflower-like particles (Figure 4e,f) prevents this aggregation.

#### **5. Conclusions**

Three types of particles with completely different shapes were analyzed by SEM, XRD and PSD. Two of them were obtained by electrolysis: the 3D dendrites and very disperse cauliflower-like agglomerates of the Cu grains. The third type of particle is a commercially available powder consisting of relatively compact agglomerates of the Cu grains. On the basis of the shape of particles, it was concluded that these particles were obtained by some non-electrochemical method.

Copper powders were synthesized by the galvanostatic regime of electrolysis at current densities of 14.4 and 384 mA·cm−<sup>2</sup> at a temperature of 21.0 ± 0.5 ◦C using 0.10 mol·dm−<sup>3</sup> CuSO4 in 0.50 mol·dm−<sup>3</sup> H2SO4. These current densities were selected to enable formation of either dendritic or cauliflower-like particles.

The dendrites obtained in this investigation ended in globules, which is different to the usual shape of Cu dendrites which have sharp tips at the top of the trunk and branches. The formation of globules at the tops of both the trunk and branches indicated the formation of a novel type of the Cu dendrites.

The Cu crystallites in the particles were randomly oriented (in both the electrolytic-produced powder at the current density of 14.4 mA·cm−<sup>2</sup> and chemically synthesized powder) or showed (220) and (311) random orientation (in the galvanostatically produced powder at 384 mA·cm<sup>−</sup>2). The change in preferred orientation for the galvanostatically produced powder was discussed on the basis of the theory of metal electrocrystallization.

The specific surface area of the Cu powders increased in the following order: SSA(14.4) < SSA(CHEM) < SSA(384).The particle size corresponding to the maximum volume ratio decreased similarly.

On the basis of comprehensive analysis of the three various forms of Cu powder particles, it is concluded that the powder obtained by electrolysis at the current density of 384 mA·cm−<sup>2</sup> is more advantageous owing to it having the largest SSA and smallest particle size.

**Author Contributions:** L.A. performed the formation of copper powders by an electrolytic procedure; V.M.M. performed the XRD analysis and discussion of the corresponding data; Z.B. performed the SEM characterization of copper powders; N.I. performed the PSD analysis of copper powders; R.M. performed the preparation of powders for the SEM, XRD and PSD analysis; M.B. provided the reagents and equipment for the electrochemical experiments; and N.D.N. conceived and wrote the paper.

**Funding:** This research received no external funding.

**Acknowledgments:** This work was supported by the Ministry of Education, Science, and Technological Development of the Republic of Serbia under the research Project: "Electrochemical Synthesis and Characterization of Nanostructured Functional Materials for Application in the New Technologies" (Project No. 172046).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **Preparation of Vanadium Oxides from a Vanadium (IV) Strip Liquor Extracted from Vanadium-Bearing Shale Using an Eco-Friendly Method**

#### **Yiqian Ma 1,\*, Xuewen Wang 2, Srecko Stopic 1, Mingyu Wang 2, Dario Kremer 3, Hermann Wotruba <sup>3</sup> and Bernd Friedrich <sup>1</sup>**


Received: 5 November 2018; Accepted: 23 November 2018; Published: 27 November 2018

**Abstract:** In the traditional vanadium precipitation process, the use of ammonium salts can produce serious pollution problems from the ammonia waste-water and the ammonia gas generated during the processing. In this reported study, an eco-friendly technology was investigated to prepare vanadium oxides from a typical vanadium (IV) strip liquor, obtained after the hydrometallurgical treatment of a vanadium-bearing shale. Thermodynamic analysis demonstrated that VO(OH)2 could be prepared as a precursor over a suitable solution pH range. Experimental results showed that by adjusting the pH to around 5.6, at room temperature, 98.6% of the vanadium in the strip liquor was formed into hydroxide, in 5 min. After obtaining the VO(OH)2, it was washed with dilute acid to minimize the level of impurities. VO2 and V2O5 were then produced by reacting the VO(OH)2 with air or argon, in a tube furnace. The XRD analyses of the products showed that VO2 had been produced in air and V2O5 had been produced in argon. The purity of the VO2 was 98.82% after calcining for 2 h at 550 ◦C, in argon flow, at a rate of 50 mL/min. It was found that the purity of the V2O5 was 98.70%, using the same reaction conditions in air. Compared to the traditional precipitation method that uses ammonium salt, followed by calcination, this proposed method is eco-friendly and employs less quantities of reagents and energy, and two types of products can be produced. Consequently, this process could promote the sustainable development of the vanadium chemical industry.

**Keywords:** vanadium precipitation; vanadium oxides; vanadium-bearing shale; vanadium strip liquor

#### **1. Introduction**

Vanadium is used in the preparation of alloys, catalysts, medicines, redox batteries, and ceramics, in the chemical industry, which makes it an essential metal in modern industry [1–6]. Although it enjoys a variety of applications, most vanadium is used in large quantities as an additive in the production of steel, primarily in the form of V2O5. As such, it can significantly improve the performance of steel [7]. The low-valence vanadium oxides (VO2 and V2O3) are suitable for the manufacture of high vanadium-containing ferrovanadium and vanadium nitride [8,9]. In addition, VO2 has a metal-semiconductor phase change property, which makes it suitable for use in photoelectric and magnetic thermally-sensitive materials, for a variety of applications, including smart window layers, memory layers, thermal sensors, and electrical and infrared light switching devices [10–12].

Currently, the common process used for producing V2O5 from various raw material sources, including vanadium-titanomagnetite, spent catalyst, fly ash from oil industry, and vanadium-bearing

shale, can be summarized as pretreatment → leaching → purification → precipitation → calcination [13–17]. After obtaining a purified and enriched vanadium solution, ammonium precipitation with ammonium salts ensures a high precipitation efficiency and high-purity of the final product (V2O5). The precipitated ammonium polyvanadate is then decomposed into V2O5, at ~400 ◦C [18–20]. To produce VO2, the common method is to further reduce the V2O5, at high temperature [21,22]. However, the use of ammonium salts causes serious pollution problems when ammonia waste-water and ammonia gas are produced during the processing [6]. The ammonia nitrogen wastewater has posed a great threat to the purity of water resources and treatment of the waste-water, to meet the strict discharge requirements, is costly [23,24]. Some new technologies, such as microbial reduction and membrane distillation, have been developed to treat the ammonia nitrogen waste-water [25,26], but simply avoiding the use of ammonium salts, altogether, in the process would be more reasonable. For example, recently, Liu et al. [27] studied a vanadium oxide electrowinnng process, from an alkaline solution. Zhang et al. [28] reported an eco-friendly technology of hydrothermal hydrogen reduction, to prepare a pure vanadium oxide product from a vanadium enriched solution. Both of these methods produced vanadium oxide directly from a vanadium solution, based on the characterization of the solutions. Autoclave and electro-decomposition units were utilized in their studies.

Vanadium-bearing shale, which is also called stone coal, accounts for 87% of the resource reserves of vanadium in China, and is widely distributed over many provinces [6]. This shale is characterized by its low vanadium-content and complex, polymetallic ores. Extraction of vanadium from this resource has been studied, extensively, over the past decade, especially, for the leaching and purification process [15,29,30]. Acid leaching, followed by solvent extraction, is the most widely applied process for treating this material and this approach has been successfully industrialized [31,32]. There are two mature solvent extraction methods for the treatment of acid leach solution—extraction of V(IV) using an acidic organophosphorus extractant, such as D2EHPA [31,33], and extraction of V(V) using an amine extractant, such as N235 [32]. After stripping, a purified vanadium liquor is obtained. In the case of the V(IV) strip liquor, it is first oxidized, if the ammonium precipitation method is to be used, and the precipitation process requires a temperature of ~90 ◦C. V2O5 is produced after calcination of the precipitate (ammonium polyvanadate).

The description of the conventional processes lends credence to the significance of developing an eco-friendly method for the vanadium precipitation process for the vanadium extraction industry. In this reported work, a novel technology for preparing vanadium oxides from a V(IV) strip liquor was studied to surmount the shortcomings of the conventional methods for vanadium oxide production. The experimental approach proposed here would be simpler than others described in the literature [27,28], and the direct preparation of a low-valence vanadium oxide was also attempted. Based on the solution's chemistry, the precursor was initially prepared and two types of products were obtained, using different atmospheres.

#### **2. Thermodynamic Analysis**

The potential-pH diagram of the V-H2O system at 298 K is shown in Figure 1. These thermodynamic data were obtained from the literature [34]. As can be seen from this diagram, various species, in different valence states, exist in a solution of vanadium, based on its pH. Furthermore, Figure 2 shows the V-H2O system, at low pH and the electrode potentials of *E*<sup>θ</sup>(Fe3+/Fe2+) = 0.77 V, *<sup>E</sup>*<sup>θ</sup>(Fe2+/Fe) = <sup>−</sup>0.46 V [35]. It is known that Fe(II) cannot co-exist with V(V) at low pH, because there is no overlap in their respective predominant-areas of the diagram. In the acid leach solution of the vanadium-bearing shale, Fe(III) is the major impurity that can adversely affect the V(IV) extraction by acidic organophosphorus extractants, like D2EHPA. Therefore, prior to solvent extraction of V(IV), addition of Fe powder to the solution would reduce the Fe(III) to Fe(II), thereby, avoiding the adverse effects of Fe(III) and ensuring that vanadium is extracted from the solution in the form of V(IV). Correspondingly, the vanadium in the strip liquor is at the +4 state.

**Figure 1.** *E*-pH diagram of the system V-H2O at 298.15 K (activity of vanadium is 1.0).

**Figure 2.** *E*-pH diagram of the V-H2O and Fe-H2O systems at 298.15 K and low pH (Activities: Vanadium 1.0, iron 1.0).

The pH value of the solution directly affects the nature of the chemical species of the vanadium in the solution. Table 1 lists the chemical reactions and Δr*G*<sup>θ</sup> in the V(IV)-S-H2O system at 278.15 K. After calculating, the reaction equilibrium constants and the relationship between the pH and logV(IV) of the different species, were obtained. Using these data, the activity-pH diagrams for V(IV)-water-sulfur systems was plotted (Figure 3). As can be seen from these data, the activity of V(IV), in solution, would be the lowest when the pH of the solution was around 5.0. Therefore, VO(OH)2 can be obtained through hydrolysis, by adjusting the pH of the strip liquor.

**Table 1.** Chemical reactions and Δr*G*<sup>θ</sup> in the V(IV)-S-H2O system at 278.15 K (Δr*G*<sup>θ</sup> data were obtained from the literature [35]).


**Figure 3.** Activity-pH diagram for V(IV)-S-H2O system at 298.15 K (activity of sulfur 1.0).

#### **3. Materials and Methods**

#### *3.1. Materials and Analysis*

The strip V(IV) liquor was obtained from a demo plant test, which was part of a project concerned with the extraction of vanadium from a decarburized vanadium-bearing shale. The vanadium-bearing shale containing 0.65 wt.% V2O5, 6.51 wt.% Al2O3, and 2.57 wt.% Fe2O3 was supplied by Datang Huayin Electric Power Co., Ltd. (Changsha, China). The H2SO4 and NaOH used in the study were of analytical grade, and all aqueous solutions were prepared using distilled water. The air and Argon gas used in this work were 99.9% pure and were supplied by Changsha Zhanyuan Gas Co., Ltd. (Changsha, China).

Vanadium was titrated with ammonium ferrous sulphate. The contents of other elements were measured by inductively-coupled plasma emission spectroscopy (ICP) with a PS-6 PLASMA SPECTROVAC, BAIRD (Waltham, MA, USA). Thermal gravimetric analysis (TGA), (Pyris 1 TGA, Perkin Elmer, Waltham, MA, USA) was used to identify the chemical reactions that occurred in the samples during the calcination processes. The products were characterized by X-ray diffraction (XRPD, PANalytical X'PERT-PRO diffractometer, Malvern Panalytical, Eindhoven, The Netherland).

#### *3.2. Procedures*

#### 3.2.1. Source of the V(IV) Strip Liquor

The process flowchart for preparing the strip V(IV) liquor from the vanadium-bearing shale is shown in Figure 4. As shown, the sulfuric acid roasting, after water leaching, achieved a high leaching efficiency of vanadium. This technology and the detailed parameters of the process have been reported by our group [36]. The main composition of the acid leach solution is listed in Table 2, where it can be seen that it was a typical acid vanadium leach solution, with impurities like iron and aluminum. After removing the Al using alum, iron powder was used as the reductant to reduce the iron in solution from Fe(III) to Fe(II), avoiding the interference of Fe(III) in the V(IV) extraction. This principle was detailed in the previous thermodynamic analysis. After this initial treatment, Na2CO3 is added to the mixture to adjust the pH of the leach solution to about 2.5. Four stages of solvent extraction was then conducted, under the following conditions—the kerosene solution with 20% (*v*/*v*) D2EHPA and 5% (*v*/*v*) TBP, *A*/*O* phase ratio of 1:1, and an equilibrium time of 5 min. At last, the loaded organic phase was stripped by 1.0 mol/L H2SO4, with four stages, at an *O*/*A* phase ratio of 5:1 and equilibrium time of 5 min. The extraction and stripping process can be expressed by Equations (1) and (2):

$$\text{Extraction: }\text{VO}^{2+}\text{(aq)} + 2\text{(HA)}\_{2(\text{o})} \rightleftharpoons \text{VOA} \cdot 2\text{HA}\_{\text{(o)}} + 2\text{H}^{+}\text{(aq)}\tag{1}$$

$$\text{Stirping: VOA}\_2\text{-}2\text{HA}\_{\text{(o)}} + 2\text{H}^+\text{(aq)} \rightleftharpoons \text{VO}^{2+}\text{(aq)} + 2\text{(HA)}\_{2\text{(o)}}\tag{2}$$

**Figure 4.** Proposed flowchart for preparation of the V(IV) strip liquor from vanadium-bearing shale.

**Table 2.** Chemical composition of the acid leach solution and the V(IV) strip liquor.


A blue band of V(IV) liquor, with a pH of about 0, containing 22.1 g/L vanadium was obtained. This band of V(IV) liquor was the raw material used to prepare the vanadium oxides and its composition is shown in Table 2. The results in Table 2 show that impurities had been removed in the process of enriching the vanadium, during the solvent extraction process.

#### 3.2.2. Preparation of VO2 and V2O5

Figure 5 shows the proposed flowchart for the preparation of vanadium oxides from the V(IV) strip liquor. Before calcination, the VO(OH)2 precursor was prepared. Hydrolysis precipitation was used to precipitate the vanadium from the strip liquor. The precipitation percentage was affected by the final pH value and the temperature. Another issue that arose in this process was the control of the impurities in the hydrolysis products. There were few impurity ions in the strip liquor (Table 2), but the few Fe(III) could co-precipitate with VO(OH)2. It was due to the oxidation of Fe(II) by air and then the co-extraction with V(IV). Fortunately, the vanadium oxide products can have a small amount of iron, if they are applied in the steel industry. On the other hand, the hydrolysis product contained some inclusion of Na+, as NaOH was the neutralizer. Na+ should be eliminated by washing the precipitate. After obtaining a relatively pure VO(OH)2, a tube furnace (OTL 1200, Nanjing Nanda Instrument Co., Ltd. Nanjing, China) was used to calcine the vanadium hydroxide. The maximum working temperature of the tube furnace was 1100 ◦C and the maximum heating rate was 10 ◦C/min. The Schematic drawing of the experimental apparatus is shown in Figure 6. The reactions for the overall process can be expressed as follows:

$$2\text{VO}^{2+} + 2\text{OH}^- \rightarrow \text{VO(OH)}\_2 \downarrow \tag{3}$$

$$4\text{VO(OH)}\_{2} + \text{O}\_{2} \xrightarrow[550^{\circ}\text{C}]{\text{Air}} 2\text{V}\_{2}\text{O}\_{5} + 4\text{H}\_{2}\text{O}\tag{4}$$

$$\text{VO(OH)}\_{2} \xrightarrow[550 \text{ } ^\circ \text{C}]{Air} \text{VO}\_{2} + \text{H}\_{2}\text{O} \tag{5}$$

**Figure 5.** Proposed flowchart for preparation of vanadium oxides from the V(IV) strip liquor.

**Figure 6.** Schematic drawing of experimental apparatus.

#### **4. Results and Discussion**

#### *4.1. Effect of pH on Vanadium Precipitation*

As shown in Figure 3, The V(IV) in the strip liquor can be hydrolyzed and precipitated by neutralization. The effect of the final solution pH on the V(IV) precipitation was determined by varying the pH of the solution from 4.2 to 10.7 and the results are shown in Figure 7.

**Figure 7.** Effect of final pH on the vanadium precipitation percentage.

These experiments were conducted at room temperature for 30 min. As can be seen, the V(IV) precipitation yield increased with an increase in pH from 4.2 to 5.6, and it attained 99.5% at a pH of 5.6, then decreased. This effect was consistent with the data depicted in the activity-pH diagram for the V(IV)-water-sulfur system. The V(IV) was initially precipitated as VO(OH)2, but this would dissolve slowly when the pH was increased further, converting the VO(OH)2 to HV2O5 −. Therefore, the optimum solution pH for V(IV) precipitation was around 5.6.

#### *4.2. Effect of Time on Vanadium Precipitation*

The effect of precipitation time on the V(IV) precipitation was evaluated at room temperature, after adjusting the pH to the optimum value of 5.6. Some samples were taken at various times, during the reaction, and the vanadium content of the solution after precipitation was determined. As shown in Figure 8, the precipitation of the V(IV) product was quite rapid. The precipitation yield reached 98.6% in 5 min, and no further increase was noted, with a prolonged reaction time. Therefore, 5 min of reaction was considered to be sufficient for the full precipitation of the desired product, after adjusting the solution pH to ~5.6.

**Figure 8.** Effect of reaction time on the concentration of vanadium precipitate.

#### *4.3. Effect of Temperature on Vanadium Precipitation*

The effect of temperature on the vanadium precipitation is shown in Figure 9. As can been seen from these results, the V(IV) precipitation yield was comparable at temperatures below 40 ◦C, but higher temperatures resulted in a decrease of the precipitation yield. This effect could be due to the re-dissolution of the precipitation at these higher temperatures. Therefore, it was concluded that room temperature was the optimum temperature for this process.

**Figure 9.** Effect of temperature on the vanadium precipitation percentage.

#### *4.4. Removal of Na<sup>+</sup> by Washing*

The resulting precipitate always contained some Na+, which needed to be removed before the calcining process could be conducted. It was found that washing the precipitate, with acidic water (pH ~2.0 with H2SO4), was effective in removing the unwanted Na+, and this washing needed to be conducted with stirring. Figure 10 shows the effect of washing on the liquid/solid ratio of the precipitate.

**Figure 10.** Effect of liquid to solid ratio on the washing of Na+.

As can be seen from the results in Figure 10, the Na<sup>+</sup> content in the VO(OH)2 decreased with the increase in the liquid/solid ratio, but the loss of vanadium increased, correspondingly. To meet the composition requirements of the products, an acidic water/precipitate weight ratio of 6/1 (mL/g) was used to wash out the Na+, and then a relative pure precursor, VO(OH)2 was obtained.

#### *4.5. Preparation of VO2 and V2O5 by Calcination*

Thermal (TG-DTA) analysis was utilized to study the behavior of VO(OH)2, during calcination in inert (argon) atmosphere and oxygen (air) atmosphere. Figure 11 shows the TG-DTA curves of the precursor in argon atmosphere.

**Figure 11.** Thermal (TG-DTA) analysis of the VO(OH)2 in argon atmosphere.

These results show that there was a small weight loss below 200 ◦C that was due to the loss of physically-adsorbed water, on the precipitate. A large weight loss occurred from 215 ◦C to 325 ◦C, which was accompanied by an evident endothermic peak (DTA), centered at about 238 ◦C. This result was attributed to the decomposition of the VO(OH)2. No additional weight change and endothermic peaks could be observed in the thermogram, after the temperature was greater than 325 ◦C. We can infer from these results that VO2 was produced and it gradually crystallized. As a comparative result, Figure 12 shows the thermal (TG-DTA) curves of the VO(OH)2 calcined in an air atmosphere. Below 270 ◦C, the precipitate behaved similarly to the precipitate calcined in argon. The decomposition of the VO(OH)2 also produced an endothermic peak (DTA) centered at about 238 ◦C, accompanied by a fast weight loss in the sample (TG). However, a small weight increase occurred from 275 ◦C to 410 ◦C, which was accompanied by a small exothermic peak (DTA). This was attributed to the oxidizing reaction, where the product transformed from VO2 to V2O5. In addition, the DTA curve, in air, resulted in an obvious endothermic peak, centered at about 690 ◦C. This was considered to be attributable to the melting point of V2O5 being 690 ◦C.

**Figure 12.** Thermal (TG-DTA) analysis of the VO(OH)2 in air atmosphere.

Based on the results of the thermal (TG-DTA) analysis, the VO(OH)2 calcined in the tube furnace in argon and air atmosphere resulted in VO2, in air, and V2O5, in Ar. In the case of the VO2, the calcining was conducted with a temperature heating rate of 10 ◦C/min and an argon flow rate 50 mL/min. The VO2 powder product was obtained after 2 h of calcination at 550 ◦C. In the same manner, the V2O5

powder was produced, using the same conditions, with argon substituted for air. The two products were then characterized by XRD and the results are shown in Figures 13 and 14.

**Figure 13.** XRD patterns of the product obtained after calcination in argon.

**Figure 14.** XRD patterns of the product obtained after calcination in air.

By comparing the XRD patterns of the experimental products with the standard patterns for VO2 and V2O5, it was found that the XRD patterns of the products were identical to those of the pure VO2 and V2O5. This verified that the desired products had been successfully prepared by calcining the VO(OH)2 in argon and air. The compositions of the products (Table 3) were evaluated using ICP analysis, after dissolving the material in a suitable solvent. The results indicated that relatively pure products had been prepared using the proposed process flowchart.

**Table 3.** Chemical composition of VO2 and V2O5 products.


#### **5. Conclusions**

In this reported study, a new method for the preparation of VO2 and V2O5 from a typical V(IV) strip liquor, which avoids using ammonium salts, was investigated. Thermodynamic analysis of the resulting experimental products showed that a vanadium precursor, VO(OH)2 could be obtained from

the liquor, over a suitable pH range. It was found that adjusting the pH of the liquor to ~5.6, at room temperature, caused the vanadium precipitation yield to quickly reach its maximum. After obtaining the VO(OH)2 product, washing it with dilute acid ensured a low impurity-content. VO2 and V2O5 were then prepared from the VO(OH)2, using a tube furnace, with sequential atmospheres of air and argon. Characterization of products confirmed their structure and purity, which demonstrated the feasibility and performance of this process. According to the product standards [37], the products prepared by this method are suitable for applications in metallurgy, like steel additive. Further purification is necessary for a high-purity product. Nevertheless, compared to the traditional precipitation method, this proposed eco-friendly method employs simple equipment, together with a low reagent and energy consumption. Therefore, it is a feasible process for promoting cleaner production of vanadium oxide, in the vanadium chemical industry.

**Author Contributions:** Y.M performed the experiments and wrote the paper; X.W, M.W. and S.S. contributed the reagents/materials/analysis tools; Y.M., D.K., H.W. and B.F. analyzed the data.

**Funding:** This research received no external funding.

**Acknowledgments:** One of the authors (Y.M.) is grateful to the Chinese Government for providing a scholarship.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Synthesis of Magnesium Carbonate via Carbonation under High Pressure in an Autoclave**

#### **Srecko Stopic 1,\*, Christian Dertmann 1, Giuseppe Modolo 2, Philip Kegler 2, Stefan Neumeier 2, Dario Kremer 3, Hermann Wotruba 3, Simon Etzold 4, Rainer Telle 4, Diego Rosani 5, Pol Knops <sup>6</sup> and Bernd Friedrich <sup>1</sup>**


Received: 5 November 2018; Accepted: 23 November 2018; Published: 27 November 2018

**Abstract:** Magnesium carbonate powders are essential in the manufacture of basic refractories capable of withstanding extremely high temperatures and for special types of cement and powders used in the paper, rubber, and pharmaceutical industries. A novel synthesis route is based on CO2 absorption/sequestration by minerals. This combines the global challenge of climate change with materials development. Carbon dioxide has the fourth highest composition in earth's atmosphere next to nitrogen, oxygen and argon and plays a big role in global warming due to the greenhouse effect. Because of the significant increase of CO2 emissions, mineral carbonation is a promising process in which carbon oxide reacts with materials with high metal oxide composition to form chemically stable and insoluble metal carbonate. The formed carbonate has long-term stability and does not influence the earth's atmosphere. Therefore, it is a feasible and safe method to bind carbon dioxide in carbonate compounds such as magnesite. The subject of this work is the carbonation of an olivine (Mg2SiO4) and synthetic magnesia sample (>97 wt% MgO) under high pressure and temperature in an autoclave. Early experiments have studied the influence of some additives such as sodium bicarbonate, oxalic acid and ascorbic acid, solid/liquid ratio, and particle size on the carbonation efficiency. The obtained results for carbonation of olivine have confirmed the formation of magnesium carbonate in the presence of additives and complete carbonation of the MgO sample in the absence of additives.

**Keywords:** MgCO3-powder; synthesis; CO2- absorption; olivine carbonation; autoclave; thermal decomposition; CO2 utilization

#### **1. Introduction**

The significance of the results coming from greenhouse gas (GHG) emissions to both the atmosphere and our lives has already been urged and is nowadays well-known. Because of the continuous increase of CO2 concentration in the atmosphere since the industrial revolution, various techniques are proposed. Carbon capture and utilization (CCU) is considered as the most promising technique in order to use the product in cement, transforming it into insoluble carbonate (mainly calcite and magnesite), that is able to remain stable in a geological timeframe [1].

In order to accelerate mineral carbonation, some pretreatment processes are required (microwave heating, grinding, sieving, separation, thermal decomposition, and chemical treatment). The main goal of pre-treatment processes is to increase the carbonation rate and improve the process kinetics. Typical pre-treatment methods are particle size reduction, magnetic separation and thermal treatment. Particle size reduction incorporates various grinding methods for an increase of specific surface area. In magnetic separation, undesired ferrous particles are separated from the rest of the feedstock. Thermal treatment is necessary for hydrated minerals, such as serpentine that contains H2O molecules in the mineral structure. Pre-treatment is usually required in direct carbonation processes [2]. However, one must consider the balance between increase of reaction rate and additional energy costs, possible CO2-production related to energy supply and the influence on the beneficial utilization of the final products. The major problem of pre-treatment is its high energy input, i.e., thermal treatment should be avoided due to the high energy demand and CO2-emissions (depending on the energy source). Overall, the most potentially economical pre-treatment proved to be size reduction [3]. Although there are large resources, it is not a feasible feedstock material because of its crystallographic stability and thus the necessary step of thermal pre-treatment.

Industrially produced by-products containing alkaline metals are also feasible for mineral carbonation, such as numerous types of slags, scraps, red gypsum, combustion residues, fly ashes and other forms of metal oxide materials such as red mud [4]. Unlike natural feasible materials, industrial by-products usually do not require mining and pretreatment processes for utilization because they already have high alkaline metal contents which are sufficient for mineral carbonation [5]. Because of high availability in minerals and secondary materials among all of the possible materials selections, calcium oxide and magnesium oxide are the most favorable options, as shown in Equations (1) and (2) [6,7].

$$\text{CaO}\_{(s)} + \text{CO}\_{2(g)} = \text{CaCO}\_{3(S)} + 179 \text{kJ/mol} \tag{1}$$

$$\text{MgO}\_{\text{(s)}} + \text{CO}\_{2(g)} = \text{MgCO}\_{3(s)} + 118 \text{kJ/mol} \tag{2}$$

Although the carbonation process is an exothermic reaction, it requires an additional heat for better dissolution of carbon dioxide in water in order to form magnesium carbonate. The benefits of exothermic mineral carbonation may result in a positive net energy balance, which improves the net efficiency of a combined cycle power plant [8]. Calculating energy balances of the process is essential for the determination of the process' profitability which might be influenced by high energy costs. Furthermore, the overall reduction of carbon dioxide emissions has to be considered due to further emissions in the mineral carbonation process resulting from, e.g., transportation, grinding and processing of by-products. Furthermore, the potential use of the formed products should also be taken into consideration.

There are abundant calcium and magnesium rich minerals available in the earth's crust. Although MgO and CaO are the most abundant alkali and alkaline metal oxides, they cannot be found as binary oxides in nature. Usually, they exist as hydroxides or silicate minerals. In a mineral carbonation process, these can also be used as feedstock to form carbonates that are chemically stable in a geological timeframe. Silicate minerals usually are richer in alkaline metal content such as magnesium, sodium, and calcium. Common silicate minerals suitable for carbonation are forsterite (Mg2SiO4), antigorite (Mg3Si2O5(OH)4) and wollastonite (CaSiO3) and their overall reaction conversions are given in Equations (3) to (5).

$$\rm Mg\_2SiO\_4(s) + 2CO\_2(g) + H\_2O(l) = 2MgCO\_3(s) + H\_4SiO\_4(aq) + 89kJ/mol\tag{3}$$

$$\mathrm{Mg\_3Si\_2O\_5(OH)\_4(s) + 3CO\_2(g) + 2H\_2O(l) = 3MgCO\_3(s) + 2H\_4SiO\_4(aq) + 64kJ/mol} \tag{4}$$

$$\text{CaSiO}\_3(\text{s}) + \text{CO}\_2(\text{g}) + 2\text{H}\_2\text{O}(\text{l}) = \text{CaCO}\_3(\text{s}) + \text{H}\_4\text{SiO}\_4(\text{aq}) + \text{90k}\text{l/mol} \tag{5}$$

The aforementioned mineral carbonation using slags as reactant also have the chance to be profitable if they are built closely to the steel production site in order to reduce production costs [9]. Furthermore, heat integration with the steel production plant can reduce overall energy costs significantly [9]. Carbonation of different types of slags are widely studied, one study by Georgakopoulos [4] suggests that blended hydraulic cement BHC slag has the highest conversion rate which is 68.3%.

Red gypsum usually exists in the form of calcium sulfate dihydrate (CaSO4·2H2O). Typically, red gypsum has a purity of 95% and has large resources in Malaysia [10]. Due to the large calcium content its carbonation is also a promising CCS option, i.e., one ton of red gypsum can stably bind 0.26 tons of gaseous CO2. A big advantage of red gypsum as carbonation resource is no mining cost is required since it exists in the form of fine powder which favors the carbonation reaction.

**Figure 1.** Reaction path of direct forsterite carbonation in aqueous solution.

Generally, the reaction path for the indirect carbonation of forsterite in an aqueous solution can be described by Equations (6) to (12) which is also illustrated in Figure 1 [11,12]. For simplicity, olivine consists only of Mg2SiO4, namely forsterite. First, gaseous carbon dioxide dissolves in the aqueous solution at a certain mass transfer rate as in Equation (6). Simultaneously, forsterite is dissolved in the aqueous solution (Equation (10)). In the aqueous solution, all species are assumed to be at equilibrium: Aqueous CO2 dissociates into bicarbonate, which further dissociates into carbonate (Equations (7) and (8)). Self-ionization of water is given by Equation (9). Aqueous silicic acid then precipitates as amorphous silica, which is a by-product, and lastly magnesium ions and carbonate form magnesite (Equations (11) and (12)).

$$\text{CO}\_2(\text{g}) \stackrel{r\text{CO}\_2}{\rightarrow} \text{CO}\_2(\text{aq})\tag{6}$$

$$\mathrm{HCO\_2(aq) + H\_2O(l) \stackrel{K\_{\odot}}{\leftrightarrow} HCO\_3^-(aq) + H^+(aq)}\tag{7}$$

$$\mathrm{HCO}\_{3}^{-}(\mathrm{aq}) \stackrel{K\_{\ominus}}{\leftrightarrow} \mathrm{CO}\_{3}^{2-}(\mathrm{aq}) + \mathrm{H}^{+}(\mathrm{aq})\tag{8}$$

$$\mathrm{H\_2O(l)} \overset{K\_{\mathbb{W}}}{\leftrightarrow} \mathrm{OH^-(aq)} + \mathrm{H^+(aq)}\tag{9}$$

$$\mathrm{Mg\_2SiO\_4(s)} + 4\mathrm{H^+(aq)} \stackrel{r\_{\mathrm{Mg\_2SiO\_4}}}{\rightarrow} 2\mathrm{Mg^{2+}(aq)} + \mathrm{H\_4SiO\_4(aq)}\tag{10}$$

$$\text{H}\_{\text{4}}\text{SiO}\_{4}(\text{aq}) \stackrel{r\_{\text{SiO}\_{2}}}{\rightarrow} \text{SiO}\_{2}(\text{s}) + 2\text{H}\_{2}\text{O}(\text{l})\tag{11}$$

$$\text{Mg}^{2+}\text{(aq)} + \text{CO}\_3^{2-}\text{(aq)} \stackrel{r\_{\text{MgCO}\_3}}{\rightarrow} \text{MgCO}\_3\text{(s)}\tag{12}$$

The particular process is characterized by several equilibrium and non-equilibrium reactions. The determination of process parameters such as temperature, pressure and pH for maximum overall conversion rates is elementary. Direct CO2 sequestration at high pressure with olivine as a feedstock

has already been performed in numerous studies at different temperatures and pressures with or without the use of additives such as carboxylic acid, and sodium hydroxide [11,12]. It is reported that optimal reaction conditions are in the temperature range of 150–185 ◦C and in the pressure range of 135–150 bar [10]. Additives are reported to have a positive influence on carbonation rate. Optimal addition of additives are reported by Bearat et al. [13] in studies about the mechanism that limits aqueous olivine carbonation reactivity under the optimum sequestration reaction conditions observed as follows: 1 M NaCl + 0.64 M NaHCO3, at 185 ◦C and P (CO2) about 135 bar. A reaction limiting silica-rich passivating layer forms on the feedstocks grains, slowing down carbonate formation and raising process costs. Eikeland [14] reported that NaCl does not have significant influence on carbonation conversion. The presented results show a conversion rate of more than 90% using a NaHCO3 concentration of 0.5 M, without adding of NaCl. Ideally, the solid phases exist as pure phases without growing together. In reality, different observations are made on the behavior of solid phases. Daval et al. [15] reported about high influence of amorphous silica layer formation on the dissolution rate of olivine at 90 ◦C and elevated pressure of carbon dioxide. This passivating layer may either built up from non-stoichiometric dissolution, precipitation of amorphous silica on forsterite particles or a combination of both. In contrast to that, Oelkers et al. [16] and Hänchen [17] observed stoichiometric dissolution and no build-up of a passivating layer except during start-up of experiments. Additionally, magnesite may precipitate on undissolved forsterite particles leading to a surface area reduction and therefore a reduction on forsterite dissolution rate, which was reported by Turri et al. [18]. Besides this undesired intermixing of solids, they observed pure particles of magnesite to be predominant in the smallest particle class, amorphous silica particles to be mainly present in the intermediate particle class and unreacted olivine particles to be predominant in the largest particle class. This knowledge may be of value for subsequent separation of products.

CO2 sequestration with olivine as a feedstock was performed in a rocking batch autoclave at 175 ◦C and 100 bars in an aqueous solution and a CO2-rich gas phase from 0.5 to 12 h. Turri showed maintainable recovery of separate fractions of silica, carbonates and unreacted olivine. Characterization of the recovered solids revealed that carbonates predominate in particle size range below 40 μm. The larger, residue fraction of final product after carbonation consisted mainly of unreacted olivine, while silica is more present in the form of very fine particles. An addition of sodium hydrogen carbonate at 0.64 M, oxalic acid at 0.5 M and ascorbic acid at 0.01 M was successfully applied in order to obtain maximal carbonation. The positive influence of the above-mentioned additives on the carbonation efficiency was reported by Olajire [19]. They studied the technology of CO2 sequestration by mineral carbonation with current status and future prospects, but the positive influence of additives was not explained in detail.

Formation of submicron magnesite during reaction of natural forsterite in H2O-saturated supercritical CO2 was studied between 35 and 80 ◦C and at pressure of 90 bars [20]. The magnesite particles formed under below-mentioned conditions exhibited an extremely uniform submicron grain-size and nearly identical rhombohedral morphologies at all temperatures. Then an evidence for carbonate surface complexation during forsterite carbonation in wet supercritical carbon dioxide was also considered. The effect of Fe on the measured rates of olivine carbonation and its role in the formation of Si-rich surface layers, which can significantly inhibit olivine dissolution and limit the extent of the carbonation reaction was considered by Saldi et al. [21]. A series of batch and flow-through reactor experiments was conducted in pure water at 90 and 150 ◦C and under a CO2 partial pressure of 100 and 200 bar, using both a natural sample of Fe-bearing olivine and a synthetic sample of pure forsterite. Experimental results show that Fe plays an ambivalent role in the carbonation.

The preparation of a magnesium hydroxy carbonate from magnesium hydroxide and carbon dioxide includes the formation of a magnesium hydroxide slurry and sparging CO2 gas through it. Various experimental conditions are evaluated in order to obtain the conditions that result in the formation of the magnesium hydroxy carbonate [22].

Our paper deals with the formation of magnesium carbonate using an Italian olivine (35.57 wt% MgO) and a synthetic reference material, mainly consisting of magnesia (97.56 wt% MgO) with special attention on the influence of the additives and different solid/liquid ratio. After carbonation, the settled solid fraction contained mainly carbonation products, which was studied by structure and composition analysis (X-Ray Diffraction XRD) and reactivity (Thermogravimetric analysis TGA and Differential Scanning Calorimetry DSC). The water solution was analyzed by Inductively Coupled Plasma Optical Emission Spectroscopy ICP-OES in order to determine the concentration of nickel, iron, magnesium and cobalt.

#### **2. Experimental Section**

The samples used were Italian olivine (Figure 2a) and a high-grade synthetic dead burned magnesia (Figure 2b). From its chemical composition apart from the chromite present at approx. 0.45 wt% and other inert minerals at trace levels, the olivine was considered as a mixed Mg-Ni and Fe silicate.

The olivine has been delivered with a particle size of below 200 μm, the used magnesia has a grain size between 10 and 30 mm. Particle Size Distribution PSD Analysis was performed used Mastersizer Hydro 2000G (Malvern PANalytical GmbH, Kassel, Germany)

The carbonation tests were planned for the three different particle size fractions <20 μm, 20–63 μm and 100–200 μm in order to evaluate the optimal process parameters and the use of additives.

**Figure 2.** (**a**) Photos of Italian Olivine; (**b**) Reference material after grinding.

The olivine with d90 = 100 μm (90% below 100 microns) as it is presented in the sieve analysis in Figure 3 has been sieved wet to produce the three grain size fractions. The magnesia sample has been crushed and milled in a lab-scale jaw crusher and has also been sieved wet into the required grain size fractions.

**Figure 3.** Sieve Analysis (Particle Size Distribution PSD) of olivine <200 μm.

The chemical composition of olivine and magnesia was analyzed by X-ray fluorescence XRF using Device PW2404 (Malvern Panalytical B.V., Eindhoven, Netherlands), such is presented in Table 1.


**Table 1.** Chemical composition of the investigated olivine and magnesia in wt%.

The planned experiments are shown in Table 2.

**Table 2.** Experimental plan (T = 175 ◦C, pCO2: 117 bar, 300 rpm, 4 h).


Carbonation tests have been carried out in the 250 mL autoclave from Parr Instrument Company (Moline, IL, USA), USA as shown in Figure 4 at 175 ◦C and 117 bars with pure grade CO2. An amount ranging from 10 to 30 g olivine has been added to 150 mL solution in different experiments. After reaction, the liquid had very low contents of metal cations, therefore characterization of the reaction products was restricted to the solid phase by TGA/DSC using Instrument STA 449F3 with Proteus Software (NETZSCH, Selb, Germany) and XRD-Analysis using Bruker D8 Advance with LynxEye detector (Bruker AXS, Karlsruhe, Germany). X-ray powder diffraction patterns were collected on a Bruker-AXS D4 Endeavor diffractometer in Bragg–Brentano geometry, equipped with a copper tube and a primary nickel filter providing Cu Kα1,2 radiation (λ = 1.54187 Å).

**Figure 4.** Parr Autoclave with maximum pressure of 200 bar and maximum temperature of 250 ◦C.

#### **3. Results and Discussion**

The characterization of products was performed using TGA, DSC, XRD and ICP-OES analysis (SPECTRO ARCOS, SPECTRO Analytical Instruments GmbH, Kleve, Germany) in order to confirm the formation of MgCO3. Additionally, the influence of additives on carbonation was discussed.

#### *3.1. Product Characterization–XRD Analysis of Product after Carbonation*

To evaluate the overall capability of the carbonation process, an experiment was performed on a synthetic reference material (>97 wt% MgO) considering the present mineralogical phases detected via XRD before and after the carbonation (Figures 5 and 6).

**Figure 6.** XRD analysis of reference sample after carbonation process.

The results prove the formation of magnesite (MgCO3) out of periclase (MgO) in the absence of any additives. Both XRD patterns show the existence of a single phase, which underlines the capability of both the reference material and the carbonation process. In the next step of the present study, the carbonation process was applied to an Italian olivine sample as an exemplary natural raw material aiming at a comparable MgCO3 formation as observed utilizing a synthetic reference material. XRD analysis of the initial olivine sample confirms the presence of forsterite, enstatite, clinoenstatite,

lizardite, spinel and tremolite, based on hydroxide and silicate of magnesium, calcium and iron (Figure 7).

Unfortunately, the chosen parameters (investigated fraction of 100–200 μm, solid/liquid ratio of 0.066 at 175 ◦C and 117 bar) did not contribute to the formation of magnesite as analyzed by XRD. Applying the same experimental conditions on an olivine sample ground to a particle size below 20 μm did not yield any magnesite formation as well. However, followed by the addition of NaHCO3, H2C2O4 and C6H8O6 the formation of MgCO3 can be proved via XRD (Figure 8) when using a fraction size of 20–63 μm.

**Figure 8.** XRD analysis of olivine sample after carbonation.

#### *3.2. Analysis of Water Solution after Carbonation of an Olivine*

The ICP-OES analysis of water solution after carbonation, as shown in Table 3 confirms very small dissolution of valuable elements such as cobalt. The most dominant species are magnesium and silicon in the order of magnitude of mg/L what corresponds to few percent of leaching efficiency. The pH value confirmed that the solution is a neutral medium.


**Table 3.** Chemical analysis of Si, Mg, Fe, Ni, Cr, and Al in solution (mg/L) after carbonation.

#### *3.3. Carbonation Extent*

Thermogravimetric analysis measurements were performed in order to establish the carbonation effect. The calculated carbonation was about 45% in the presence of additives, as shown in Figure 9.

**Figure 9.** TGA/DSC Analysis in Exp. 3 in the presence of additives.

The total analysis of thermal decomposition of formed (Mg,Fe)CO3 is shown in Table 4.


**Table 4.** Thermal Decomposition of samples before carbonation and after carbonation.

As shown in Table 4, the total weight loss for an initial sample of the Italian olivine in the interval between 25 and 1000 ◦C amounts 1.7%, what is the amount of bound water in the used sample. The total weight loss for the initial olivine material in the interval between 25 and 470 ◦C amounts 0.47% in comparison to 3.75% for this sample after carbonation. In difference for the sample without carbonation, in the temperature interval between 470 ◦C and 595 ◦C, the weight loss amounts 8.69%, with a thermal effect of 137 J/g, what confirms that this temperature range is most important for thermal decomposition of (Mg,Fe)CO3. Above 775 ◦C the change of weight is not significant, and thermal decomposition of sample is minimal, what means that the thermal decomposition is finished. In contrast to this experiment in the presence of additives in Exp. 3 with loss of weight of 15.2%, the overall weight losses of experiment 1 is very small (few percent) what confirms a low carbonation degree (as confirmed with XRD analysis in Figure 6). The weight loss for experiment 2 is about 3.5% (red line at Figure 10) which confirms very small carbonation rate in the absence of additives in comparison to the experiments with additives.

**Figure 10.** TGA and DSC Analysis in the experiments 1 and 2 (TGA and DSC for experiment 1- green color; TGA and DSC for experiment 2- red color).

This positive effect may be due to "reaction-driven cracking" in the presence of NaHCO3, formation of etch pits, and/or other processes that continually renew the reactive surface area of Mg2SiO4.

$$\text{NaHCO}\_3\text{ (aq)} \rightarrow \text{Na}^+ + \text{H}^+ + \text{CO}\_3 \text{2}^- \tag{13}$$

An addition of oxalic acid leads to formation of Mg-ions in solution, which react with carbonate ions forming magnesium carbonate.

$$2\text{ Mg}\_2\text{SiO}\_4 + 2\text{ H}\_2\text{C}\_2\text{O}\_4 \rightarrow 2\text{Mg}^{2+} + \text{C}\_2\text{O}\_4^{2-} + \text{H}\_4\text{SiO}\_4\tag{14}$$

$$\text{Mg}^{2+} + \text{CO}\_3^{2-} \rightarrow \text{MgCO}\_3 \text{ (s)}\tag{15}$$

The two analyses of thermal decomposition of products after carbonation of magnesia (97 wt%) confirmed total decomposition of the produced MgCO3.

**Figure 11.** TGA and DSC Analysis of formed product after carbonation of synthetic magnesia (20 ◦C/min, nitrogen, Exp. 4).

As shown at Figure 11, the weight loss of 52.34 % (TGA-red line) and 52.22 (TGA-blue line) with maximal DSC effect (1281 J/g; blue area) at Tmax = 699.8 ◦C confirm the formation of magnesium carbonate, what was compared with theoretical value, according to the Equation (16):

$$\text{MgCO}\_3(\text{s}) \xrightarrow{T} \text{MgO} + \text{CO}\_2(\text{aq}) \tag{16}$$

where: M (Mg) = 24.30 g/mol; M (MgCO3) = 84.30 g/mol, M (MgO)= 40.30 g/mol, M (CO2) = 44.0 g/mol.

Using a ratio between molar mass of magnesium carbonate and magnesium oxide, theoretical calculated loss of carbon dioxide amounts 52.55%, what is in good accordance with an experimental determined weight loss. Finally, it confirms the completed carbonation of synthetic magnesia under the chosen parameters and formation of magnesium carbonate, which particle size distribution was shown at Figure 12.

**Figure 12.** Particle size distribution of MgCO3 after carbonation of synthetic magnesia.

The measured values of produced magnesium carbonate d50 and d84 amount 15.066 and 37.066 μm, respectively.

#### **4. Conclusion**

Synthesis of magnesium carbonate was studied via carbonation of olivine using different size fractions (under 20 μm, between 20 and 63 μm, and between 100 and 200 μm) with different solid/liquid ratios of 1:15 and 1:5; at 175 ◦C and pressure of CO2 (117 bar) in an autoclave in the presence and in the absence of additives. The characterization of products using XRD, TGA, PSD and DSC analysis has confirmed the formation of MgCO3. In contrast to carbonation of olivine in the absence of additives the formation of magnesium carbonate is possible at high pressure and temperature with olivine (35.57% MgO) from Italy in the presence of sodium hydrogen carbonate, oxalic acid, and ascorbic acid at 175 ◦C, 117 bar in 4 h (Exp. 3). The maximum carbonation (more than 95%) was obtained at the same conditions for synthetic magnesia (97.56 wt% MgO) in the absence of additives. In order to validate the first results in 0.25 L autoclave, new scale up experiments will be performed in 1.0 and 10.0 L autoclaves. Especially, the influence of rotating speed, pH-values and different initial secondary materials such as slag and red mud shall be analyzed in our future work. The analysis of the obtained solution after carbonation revealed very small content of cobalt and chromium, but it will be also considered in our future work in the presence of pH buffering agents in order to increase an extraction efficiency. Especially, a life-cycle-assessment of the carbonation process in the presence of additives will be performed in our future work.

**Author Contributions:** S.S. conceptualized and managed the research, and co-wrote the paper. D.K. performed the preparation of the olivine materials (grinding, sieving) and co-wrote the paper. H.W. co- wrote the paper. S.N. has performed experiments in an autoclave. P.K. has performed of XRD and TGA-analysis of initial sample and obtained products. G.M. helped in discussion of TGA and XRD-analysis. C.D. analyzed the data and co-wrote the paper. S.E. supervised the XRF- and XRD-analyses and co-wrote the paper with R.T., P.K. conceptualized the research and provided industrial advice. D.R. managed the laboratory facilities for TGA and DSC analysis in Heidelberg Cement Technology Center. B.F. supervised the personnel, provided funding and co-wrote the paper.

**Acknowledgments:** We would like to thank the BMBF (Federal Ministry of Education and Research) in Berlin for the financial support for the CO2MIN Project (No. 033RCO14B) in period from 01.06.2017 to 31.05.2020. For a continuous support and cooperation we would like to thank Andreas Bremen, AVT and Hesam Ostovari, LTT, RWTH Aachen University.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Preparation of Spherical Mo5Si3 Powder by Inductively Coupled Thermal Plasma Treatment**

#### **Jang-Won Kang 1,2, Jong Min Park 1, Byung Hak Choe 2, Seong Lee 3, Jung Hyo Park 3, Ki Beom Park 1, Hyo Kyu Kim 1, Tae-Wook Na 1, Bosung Seo <sup>1</sup> and Hyung-Ki Park 1,\***


Received: 28 June 2018; Accepted: 1 August 2018; Published: 3 August 2018

**Abstract:** A method was developed to fabricate spherical Mo5Si3 powder by milling and spheroidizing using inductively coupled thermal plasma. A Mo5Si3 alloy ingot was fabricated by vacuum arc melting, after which it was easily pulverized into powder by milling due to its brittle nature. The milled powders had an irregular shape, but after being spheroidized by the thermal plasma treatment, they had a spherical shape. Sphericity was increased with increasing plasma power. After plasma treatment, the percentage of the Mo3Si phase had increased due to Si evaporation. The possibility of Si evaporation was thermodynamically analyzed based on the vapor pressure of Mo and Si in the Mo5Si3 liquid mixture. By this process, spherical Mo silicide powders with high purity could be fabricated successfully.

**Keywords:** Mo silicide; Mo5Si3; spheroidizing; powder; inductively coupled thermal plasma

#### **1. Introduction**

Refractory metal-based silicide alloys, which are also referred to as refractory metal in situ composites, currently receive a lot of attention as structural materials for ultrahigh temperature applications [1–3]. Among these materials, Mo silicide-based alloy [4,5] and Nb silicide-based alloy [6,7] have been intensively studied due to their excellent strength, creep resistance and oxidation resistance at ultrahigh temperature. Furthermore, Mo and Nb have a relative low density compared to other refractory metals, such as Ta and W [8].

Mo silicide-based alloys are composed of α-Mo and Mo silicide. Mo silicides are formed of three main phases: Mo5Si3, Mo3Si, and MoSi2 [9]. Of these, Mo5Si3 has the highest melting temperature of 2180 ◦C [10]. Therefore, many studies have evaluated the high-temperature creep and oxidation resistance of Mo5Si3.

Unfortunately, Mo silicide-based alloys have low fracture toughness at ambient temperature [11], along with low machinability due to the low thermal conductivity and brittle nature of Mo silicides [12,13]. Therefore, it is difficult to fabricate components of Mo silicide-based alloys by conventional casting and machining methods. In addition, since in the Mo–Si binary system the Mo3Si phase is in between the Mo solid solution and the Mo5Si3 phases, Mo silicide-based alloys composed of Mo and Mo5Si3 cannot be fabricated by casting. However, based on the powder metallurgy process, Mo-based silicide alloys, where the microstructures consist of Mo5Si3–Mo3Si and Mo–Mo5Si3–Mo3Si, could be fabricated.

Hence, powder metallurgy processes are an attractive way to fabricate components of Mo silicide-based alloys. Previous studies have attempted to fabricate Mo silicide-based alloys [14,15] and Mo silicide powders [16] by mechanical alloying. However, powders fabricated by mechanical alloying suffer from low productivity and oxygen contamination, as well as an irregular morphology. With regard to sintering, spherical powders are much more favorable than those with irregular shapes, as they offer higher packing density and fluidity [17].

It is therefore necessary to develop a method to fabricate high-purity spherical Mo silicide powders. To our knowledge, there is no previous work on the preparation of Mo silicide powders or pre-alloyed Mo silicide-based alloy powders by inductively coupled thermal plasma processing. Thus, in this study, we fabricated Mo5Si3 powders by pulverizing a Mo5Si3 ingot, utilizing its brittle nature. To improve the sphericity of the powders, they were spheroidized by an inductively coupled thermal plasma treatment. The effect of the plasma power on the morphology and phase balance of the powders was examined, and the evaporation behavior of Si during plasma treatment was analyzed thermodynamically.

#### **2. Experimental Procedures**

Mo5Si3 ingots with a chemical composition of 85.06Mo–14.94Si in wt% (62.5Mo–37.5Si in atom%) were fabricated by vacuum arc melting. For vacuum arc melting, the chamber was evacuated to a high vacuum (10−<sup>5</sup> torr) by oil diffusion pump and then high-purity argon gas was injected into the chamber until the pressure reached 400 torr. The ingot, which was 150 mm long × 75 mm wide × 10 mm high, was cast in a quadrangle-shaped cold copper crucible and its weight was 1100 g. To homogenize the composition, the ingot was remelted five times.

To analyze the chemical composition and oxygen concentration, the center of the ingot was cut to a cylinder 3 mm in diameter and 5 mm in height, and measurements were carried out five times in each sample. The chemical composition of the ingot and powders, as analyzed by inductively coupled plasma mass spectrometry (ICP-MS) (iCAP Q, Thermo Fisher Scientific, Waltham, MA, USA), is given in Table 1. To measure the concentration of Mo and Si in Mo silicide, ICP-MS was carried out following the procedure in [18].


**Table 1.** Mo and Si concentrations (in wt%) of the ingot and powders after spheroidizing at plasma powers 3–7 kW. The values given in parenthesis refer to standard deviation.

The oxygen concentration, as analyzed by an inert gas fusion infrared absorption method (LECO, 736 series), is given in Table 2.

**Table 2.** Oxygen concentrations (in wt%) of the ingot, powders after milling, and powders after spheroidizing at a plasma power of 6 kW. The values given in parenthesis refer to standard deviation.


For pulverization, two Mo5Si3 ingots with a weight of 2200 g were first crushed using a jaw crusher into particles with a size of less than 3 mm. Then, the particles were ball-milled in a stainless-steel container, using tungsten carbide balls with a diameter of 5 mm as milling media. The ball-to-powder ratio was 5:1 by weight and the steel container was purged with high-purity argon to prevent oxygen contamination during ball milling. The milling was performed for 5 h at a rotational speed of 200 rpm. After milling, the powders were sieved to the range of 38–75 μm, because the powders larger than 75 μm were not perfectly spheroidized by the thermal plasma system used in this study.

The sieved powders were spheroidized using an inductively coupled thermal plasma treatment (RFP-10, PLASNIX, Incheon, Korea). To investigate the spheroidizing behavior with respect to plasma power, plasma powers of 3–7 kW were examined. The other parameters were fixed as follows. The plasma oscillation frequency was 13.56 MHz, and the chamber pressure was 80 KPa. The powder carrier and center gas were high-purity argon, with flow rates of 5 and 2 slm (standard liter per minute), respectively. Since the maximum temperature in the chamber almost reaches 10,000 K during plasma processing, the sheath gas was injected along the chamber wall for cooling. The sheath gas was a mixture of argon and helium (argon/helium = 4:1) with a flow rate of 75 slm. To feed the powders into the plasma chamber, a vibrating feeder was used with a feeding rate of 300 g/h.

The morphology of the powders after sieving and spheroidizing was observed by field emission-scanning electron microscopy (FE-SEM, FEI, QUANTA FEG 250, Thermo Fisher Scientific, Waltham, MA, USA) along with energy-dispersive spectrometry (EDS, Octane Elite EDS, EDAX, Mahwah, NJ, USA). To investigate the phase balance, the powders were analyzed using X-ray diffractometry (XRD, Empyrean, PANalytical, Almelo, The Netherlands) with Cu Kα radiation, in the 2θ range of 25–75◦. For the XRD analysis, the samples were selected at random. The powder size distribution was investigated using a powder size analyzer (Mastersizer 3000, Malvern Panalytical, Malvern, UK). To understand the Si evaporation behavior during plasma treatment, thermodynamic parameters such as the standard Gibbs free energy change and activity coefficient were calculated by Thermo-Calc using the SSOL database.

#### **3. Results and Discussion**

Figure 1 is the microstructure of the cast Mo5Si3 ingot observed by back-scattered electron imaging. The ingot was composed of two phases. By EDS analyses, the Si concentrations of the dark area (Point 1) and bright area (Point 2) are 14.28 and 9.15 wt%, respectively, which correspond with the Si concentrations of Mo5Si3 and Mo3Si (14.94 and 8.89 wt%). This result indicates that the dark and bright areas were the Mo5Si3 and Mo3Si phases, respectively.

**Figure 1.** *Cont.*

**Figure 1.** (**a**) The microstructure of the cast Mo5Si3 ingot observed by back-scattered electron imaging, and (**b**) EDS area and point analyses result. Points 1 and 2 were determined as Mo5Si3 and Mo3Si, respectively.

Figure 2 shows the morphology of the powders before and after spheroidizing by inductively coupled thermal plasma treatment. As shown in Figure 2a, the powders after milling and sieving had an irregular shape. The powder size distribution of the powder after milling was examined by a powder analyzer; the *d*10, *d*50, and *d*<sup>90</sup> values were 47.9, 75.4, and 117.0 μm, respectively. Mo silicides, including Mo5Si3, have a brittle nature; grain boundary cracking occurs easily at room temperature [12]. Chu et al. [19] researched this by fabricating single crystal Mo5Si3 by the Czochralski method and evaluating the mechanical properties with respect to crystal orientation. The room temperature fracture toughness, which ranged from 2 to 2.5 MPa√m, was not severely affected by the crystal orientation. Therefore, the Mo5Si3 ingot was easily pulverized by jaw crushing and ball milling.

**Figure 2.** Morphology of the powders (**a**) after ball-milling and sieving, and (**b**–**f**) after spheroidizing by inductively coupled thermal plasma, with plasma powers of (**b**) 3 kW, (**c**) 4 kW, (**d**) 5 kW, (**e**) 6 kW, and (**f**) 7 kW.

After spheroidizing, the shape of the powders had changed from irregular to spherical. The sphericity of the powders was higher with increased plasma power. The powder size tended

to become smaller with increasing plasma power, and the *d*10, *d*50, and *d*<sup>90</sup> values of the powder spheroidized at 7 kW were 44.8, 62.2, and 86.3 μm, respectively.

Previously, the temperature profile during inductively coupled thermal plasma treatment was simulated by COMSOL Multiphysics. The simulation indicated that the maximum temperature of the chamber was almost 10,000 K during plasma processing [20,21] and the powder reached maximum temperature within 10 ms after injecting in the plasma chamber [22]. Therefore, when powders with an irregular shape are injected into the plasma chamber, they are fully melted and spheroidized to reduce the surface area.

To analyze the phases present, the powders were examined by XRD. Figure 3a shows the XRD patterns for 2θ = 25–75◦ of the powders before and after spheroidizing. The powders were composed of two phases, Mo5Si3 and Mo3Si. To investigate the percentage of Mo5Si3 and Mo3Si in the powders, the XRD patterns were analyzed by the Rietveld method, and the result is shown in Figure 3b. The percentages of Mo5Si3 and Mo3Si in the powder before spheroidizing were 92.5% and 7.5%, respectively. As the plasma power increased, the percentage of Mo5Si3 was gradually decreased in the spheroidized powders, reaching a minimum of 63.1% with the power of 7 kW.

**Figure 3.** (**a**) XRD patterns of the powders before and after spheroidizing by inductively coupled thermal plasma treatment. (**b**) The percentage of each phase in the powders was analyzed by the Rietveld method based on the XRD data in Figure 3a.

By the chemical stoichiometry, the concentration of Si is greater in Mo5Si3 than in Mo3Si (14.94 wt% vs. 8.89 wt%, respectively). The fact that the amount of the Si-rich phase reduces with increased plasma power suggests that Si evaporation may occur during plasma treatment. Therefore, the Si concentration of the powders was analyzed to examine this possibility.

Table 1 shows the Si concentration of the ingot and the powders after spheroidizing with different plasma powers. The weight concentration of Si in the Mo5Si3 having perfect stoichiometry is 14.94 wt%. The Si concentration of the ingot was 14.83 wt%, which is almost same as that of the Mo5Si3 phase. As the plasma power increased, the Si concentration gradually decreased, reaching a minimum of 11.15 wt% with the power of 7 kW. During the plasma treatment, the evaporation rate of Si would much higher than that of Mo; therefore, the Si concentration was decreased after spheroidizing.

The evaporation behavior of the element is determined by the vapor pressure [23,24]. An element with a higher vapor pressure will have a higher evaporation rate. The vapor pressure of a pure element *i* (*p<sup>o</sup> <sup>i</sup>* ) is calculated by the following Equation [25]: *<sup>p</sup><sup>o</sup> <sup>i</sup>* <sup>=</sup> *exp* −Δ*G<sup>o</sup> i RT* , where *R* is the gas constant, *T* is the temperature, and Δ*G<sup>o</sup> <sup>i</sup>* is the standard Gibbs free energy change of element *i* during the phase transformation from liquid to gas. Figure 4a shows the vapor pressure of pure Mo and Si with respect to the temperature. Both vapor pressures increase with increasing temperature. Furthermore, *p<sup>o</sup> Si* is higher than *p<sup>o</sup> Mo* over the whole temperature range, which means that the evaporation rate of Si is higher than that of Mo.

**Figure 4.** (**a**) Vapor pressure of pure Mo and Si with respect to temperature. (**b**) Vapor pressure of Mo and Si in molten Mo5Si3.

However, *p<sup>o</sup> <sup>i</sup>* is the vapor pressure for the pure element *i*. To more accurately analyze the evaporation behavior, the vapor pressures of Mo and Si in molten Mo5Si3 should be considered. The vapor pressure of element *i* in a liquid mixture (*pi*) is calculated by the following equation: *pi* <sup>=</sup> *exp* −Δ*G<sup>o</sup> i RT* ·*γi*·*Xi*, where *γ<sup>i</sup>* is the activity coefficient of element *i* in a liquid mixture and *Xi* is the molar fraction of element *i* [25]. To calculate the vapor pressure of Mo and Si in molten Mo5Si3 with respect to the temperature, activity coefficients of Mo and Si should be known. Therefore, the activity coefficients at 2500, 3000, 3500, 4000, 4500, 5000, and 5500 K of Mo and Si in the Mo–Si binary system were calculated by Thermo-Calc using the SSOL database. Then, the activity coefficients of Mo and Si in the composition of Mo5Si3 were extracted, and they were used to determine the *pMo* and *pSi*.

Figure 4b shows the vapor pressures of Mo and Si in molten Mo5Si3 with respect to temperature. Since the evaporation rate in a liquid mixture is affected by the activity coefficient, the vapor pressures of Mo and Si in molten Mo5Si3 were different to that for pure Mo and Si. However, the vapor pressure of Si was still higher than that of Mo, confirming that the amount of Si evaporation was greater than that of Mo during plasma treatment. Therefore, the decreasing amount of the Si-rich phase (Mo5Si3) with increasing plasma power (Table 1) was caused by Si evaporation during spheroidizing. Even though Si was evaporated during spheroidizing, any condensation of Si was not observed. Si nanoparticles could be nucleated during plasma treatment; however, they would be filtered by the cyclone system due to their small size and low weight.

To determine the extent of oxygen contamination during the fabrication of the spherical Mo silicide powder, the oxygen concentrations of the ingot, the powders after milling, and the powders after spheroidizing at 6 kW were analyzed. The oxygen concentrations of the three samples are shown in Table 2. The oxygen concentration of the ingot was 0.003%, which increased to 0.171% after milling. This could be caused by the increase in specific surface area by pulverizing, as well as the contamination during milling, as argon purging may not perfectly eliminate oxygen from the milling container. After spheroidizing at 6 kW, the oxygen concentration was decreased to 0.016%. This would be due to the reducing environment during the plasma treatment, due to the extremely high temperature and low oxygen partial pressure in the chamber [21]. Thus, while the oxygen concentration was higher in the spheroidized powder than in the ingot, it was still considerably low.

To examine the internal microstructure of the spherical powder, the powder spheroidized at 6 kW was observed by FE-SEM. Figure 5a shows the cross-sectional microstructure of the powder observed by back-scattered electron imaging. As expected from the XRD results (Figure 3), the powders were composed of two phases. To identify the phases, EDS area and point analyses were performed, and the result is shown in Figure 5b. The Si concentrations of the dark area (Point 1) and bright area (Point 2) are 14.57 and 9.10 wt%, respectively. Therefore, the dark and bright areas can be defined as Mo5Si3 and Mo3Si, respectively. The melting temperatures of Mo5Si3 and Mo3Si are 2453 and 2298 K, respectively [26]. As the powders cooled during plasma treatment, the Mo5Si3 phase solidified first, after which Mo3Si solidified. Therefore, the microstructure shown in Figure 5 would be formed by the first nucleation of Mo5Si3.

**Figure 5.** (**a**) Cross-sectional microstructure of the powder observed by back-scattered electron imaging, and (**b**) EDS area and point analyses result. Points 1 and 2 were determined as Mo5Si3 and Mo3Si, respectively.

#### **4. Conclusions**

The Mo5Si3 ingot was prepared by vacuum arc melting, after which it was pulverized to a powder by jaw crushing and ball milling. As the milled powders had an irregular shape, they were spheroidized by thermal plasma treatment. As the plasma power increased, the sphericity of the powders increased. They were perfectly spheroidized when the plasma power was higher than 6 kW. After plasma treatment, the ratio of Mo5Si3 to Mo3Si had decreased due to Si evaporation. Based on the thermodynamic analysis, Si has a higher vapor pressure than Mo in the Mo5Si3 liquid mixture. By this process, spherical Mo silicide powders with a low oxygen concentration of 0.016% could be fabricated successfully.

**Author Contributions:** Conceptualization, J.-W.K., J.M.P., B.-H.C. and S.L.; Data curation, K.B.P.; Formal analysis, J.M.P., J.H.P. and B.S.; Investigation, T.-W.N.; Methodology, H.K.K.; Writing original draft, H.-K. P.

**Acknowledgments:** This work was supported by a research fund of Korea Institute of Industrial Technology (KITECH EO-18-0012) and Agency for Defense Development (Project No. 811555-912515201).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*
