**Ultrasound-Assisted Leaching Process—Part 1: Process Design and E**ffi**ciencies**

#### **Ferdinand Kießling 1, Srecko Stopic 2,\*, Sebahattin Gürmen <sup>3</sup> and Bernd Friedrich <sup>2</sup>**


Received: 30 April 2020; Accepted: 27 May 2020; Published: 1 June 2020

**Abstract:** The treatment of industrial polycrystalline diamond (PCD) blanks in aqua regia at atmospheric pressure between 333 K and 353 K was performed via the ultrasound-assisted leaching process to investigate whether the influence of ultrasound is beneficial. Cobalt content in the solution and in the blanks was monitored as well as the effects of leaching temperature, solid-to-liquid ratio, and PCD blank size. The use of intermittent and permanent ultrasound helped reduce the leaching time and thus energy consumption by up to 50%. In all trials with ultrasound, higher temperature only has a slight effect. Solid-to-liquid ratio does not have a positive or negative impact. A new process design was tested using an innovative experimental setup for ultrasound-assisted leaching aiming at maximum cobalt and diamond recovery from PCD and final reuse of fine PCD for cutting and polishing other hard materials in different important industrial applications.

**Keywords:** polycrystalline diamond; leaching; cobalt; ultrasound

#### **1. Introduction**

Cobalt is a ferromagnetic transition metal which is located between iron and nickel in the periodic table of elements and mostly available in lateritic ores [1–3]. Cobalt is used as a solvent catalyst in the production of polycrystalline diamond (PCD) that would otherwise take even more pressure and a higher temperature to achieve. [4]. The wide use of cobalt relative to other metals is associated with the high solubility of carbon in its melt during thermobaric treatment [5]. Unfortunately, at temperatures above 800 ◦C, which are developed during operation of a polycrystalline diamond tool, the cobalt promotes the formation of microcracks and results in significant heat-resistance reduction and subsequent reduction in the abrasion resistance of the polycrystalline diamond [6]. Thus, the reduction of the cobalt content in the sintered polycrystalline diamond (PCD) greatly improves the performance of a superhard composite. Finally, the removal of cobalt from the used PCD is the most important request in the industry of wire-drawing tools made from diamond materials and tungsten carbide.

Therefore, recycling is a chosen strategy for cobalt recovery in contrast to traditional primary metallurgy [7]. Diamond, to this date, is the hardest material that is put to use on commercial and industrial scales. It appears as the working edges of cutting tools or the grains in the hardest abrasives, among other uses. Due to advancements in the field of high strength steels and super alloys, the demand for hard cutting and forming tools will be increased in the future [8]. Industrially used diamonds can be found in the form of a naturally grown and mined crystals or as a man-made products with mono- or

poly-crystalline microstructures. Die blanks are usually made at high temperatures and high pressure processes which require cobalt (Co) as the solvent catalyst [9]. Cobalt is incorporated in the final product resulting in a multiphase compound (PCD). This multiphase characteristic renders the PCD vulnerable to thermal stress since cobalt and diamond have different thermal expansion coefficients. The only way to make these PCDs stable for industrial applications at high temperatures, such as hot forming of metals, is to remove the inclusions from the cavities in the framework of diamond grains. Because of its strong oxidizing properties, aqua regia, or more precisely the forming nitrosyl chloride (NOCl), was used for the leaching of cobalt. This research also acknowledges the importance of using cobalt responsibly, possibly in a closed-loop recycling, not only because of price and toxicity but rather socioeconomic problems and difficulties linked to the conflict mineral columbite–tantalite, often referred to as 'coltan' [10,11].

Since the middle of the twentieth century, there have also been investigations into whether ultrasound can increase the chemical turnover in leaching processes. Many researchers have found that the reaction rate as well as the overall leaching efficiency can be increased significantly [12–14]. This positive effect is attributed to cavitation and specifically the Kelvin impulse [15–17]. Ultrasound is capable of creating such dynamic pressure changes in liquids that they will pass the phase boundary to their gaseous phase. When located near a solid surface, these cavities collapse and create an energetic jet pointed at the surface, because there is less instreaming liquid from the direction of the surface and overall impulse has to be conserved. These microjets have the ability to pierce through diffusion layers, effectively constantly renewing them. The positive influence of ultrasound was confirmed using ultrasound for the synthesis of nanosized particles by ultrasonic spray pyrolysis (USP). Due to its easy feasibility, flexibility and cost-efficiency, the USP method is an important alternative to the chemical vapor deposition (CVD) and other synthesis methods. Cobalt nanoparticles were successfully prepared from cobalt nitrate solution formed after an acidic treatment of the cemented tungsten carbide using the ultrasonic spray pyrolysis method [18,19]. An increase of ultrasound from 0.8 to 2.5 MHz decreases an aerosol diameter of cobalt nitrate to 2.2 μm that leads to the formation of submicron cobalt particles after drying and precipitation above 500 ◦C in a furnace using a hydrogen reduction atmosphere.

Shortening the process time for the cobalt removal from PCD in the presence of ultrasound is the main motivation for this study. In solid–liquid reactions, there is always an obstacle in the form of a diffusion layer where the mass transfer is inhibited because adsorption reaction and desorption take additional time and energy; even more so if a phase transition is involved. What's more, the component's concentrations can be quite different from the bulk solution, hindering further reaction. One way to mitigate this effect is to mechanically agitate the solution and decrease the thickness of this layer. For this reason, a stirrer was used to create sufficient turbulence in the reaction vessel.

An increase in temperature has an effect on most chemical reactions as it measures a substance's inner energy that is potentially available for reactions, changes the substance's activity or simply aids the mass transfer processes. In the leaching process, it has been found that increased temperature can accelerate the reaction speed [20–22]. The dissolution process with an acid needs a high activation energy to begin this solid–liquid reaction. Since this study is carried out at ambient pressure with the leaching solution described, care has to be taken because increasing the temperature also increases vapor pressures of the liquids involved [23].

Finally, the main aim of this work is optimizing leaching of cobalt from polycrystalline diamond blanks with grain size 5 μm in the presence of ultrasound. The influence of temperature and ultrasound on the leaching of cobalt will be studied in one ultrasound leaching assisted process. This study will take a look at the effects of ultrasound on the leaching efficiency but also at the penetration depth into the PCD, proposing a new experimental setup. Being capable of making any statements about the metallic constituents of PCD without breaking the blanks is an advantage and the reason why this method has been implemented. A challenge of this work is to maximize efficiency of cobalt leaching and diamond recovery in a shorter time than traditional hydrometallurgical methods.

*Metals* **2020**, *10*, 731

#### **2. Experimental**

#### *2.1. Material*

As shown in Figure 1, this study is centered on the polycrystalline diamond (PCD) blanks made by Redies GmbH & Co. KG, Aachen, Germany, a manufacturer of wire drawing dies.

**Figure 1.** Scanning electron microscopy (SEM) image of raw diamond powder 5 μm class before the high temperature and high pressure (HTHP) process.

The SEM (Scanning electron microscopy) and EDS (Energive Dispersive Spectroscopy) analysis of the PCD surface after being polished is shown in Figure 2 and Table 1.

**Figure 2.** SEM image of ground and polished PCD surface, 5 μm class (dark grey areas are the bridged diamond grains with cavities where also traces of cobalt show up in lighter shades).

In contrast to the cobalt content of 0.1 wt.% in nickel lateritic ore, its average content in PCD is about 5–20 wt.%, which makes it a very promising material for the recycling of cobalt. The maximal value of cobalt in analyzed sample amounts is 1.67 wt.%, as shown in Table 1.


**Table 1.** Energy dispersive X-ray spectroscopy image of ground and polished PCD surface, also 5 μm class.

Different types of samples were used in our work as shown in Table 2. The columns "volume", "surface area", and "surface-to-volume ratio" in Table 2 do not contain measured but calculated values. The values for volume should be seen as the apparent outer volume of a porous body, not solid volume. The same applies to surface area. The measured values in the columns "diameter" and 'height' were obtained by taking a sample of thirty PCDs of the same type and averaging the values. The value for weight was obtained by weighing 100 PCDs and dividing the measurement by 100. This approach was chosen since the blanks did not have critical deviations dimension-wise. Assuming a homogeneous density, this average is sufficiently accurate. The same applies to weighing the batches as a whole after treatment. Figure 3 shows the deviations in the weights of individual PCDs. All the weight measurements were taken into a stock plot to depict actual variations and uncertainties. The deviations of surface-to-volume ratio were obtained by relating the largest surface to the smallest volume and vice versa.

**Table 2.** Dimensions of PCD samples with grain size of 5 μm.


#### *2.2. Procedure*

The experiments were carried out in two glass reactors simultaneously set up in a fume cabinet, using argon gas (Linde Gas AG, Höllriegelskreuth, Germany) a flow meter, type Rota Yokogawa (Yokogawa Deutschland GmbH, Ratingen, Germany), and a bottle with sodium hydroxide (Merck KGaA, Darmstadt, Germany), as shown at Figure 4. The reactor vessels were three-necked round bottom flasks with a capacity of 500 mL, the necks with standard ground joints 29/32 served as couplings for a stirrer seal, two gas hose couplers, and as access points for sampling, respectively, as shown at Figure 5. For sampling, the gas inlet coupler had to be removed temporarily. The stirrer unit consisted of a motor unit with a drill chuck, type "IKA Eurostar digital" (IKA®-Werke GmbH & Co.KG, Staufen, Germany) IKA®-Werke GmbH & Co. KGIKA®-Werke GmbH & Co. KGIKA and a Polytetrafluorethylen (PTFE) coated impeller including a PTFE stirrer seal, as shown at Figure 4. The standard ground joints on the sides were sealed with a high-viscosity, silicone-based lubricant.

**Figure 3.** Deviations in weight of individual D14 and D18 PCD blanks.

**Figure 4.** The description and picture of the leaching reactor.

**Figure 5.** Innovative experimental set-up for leaching of PCD samples where: a—argon, b—valve; c—flow meter; d—mixer, e—reactor, f—bottle with dissolved sodium hydroxide, e—exhaust system.

The depth of immersion was chosen so that the surface level of the stirred liquid was as high as the water level in the heated ultrasound bath. The means of hindering evaporation can also be seen in this image. First, styrofoam beads were added to minimize the surface area available for evaporation, and secondly an acrylic lid made of two parts and an improvised cable tie hinge was used. For additional safety, each connection was clamped. For the purpose of temperature value measurement a "Testo 720" digital thermometer (testo SE & Co. KGaA, Lenzkirch, Germany) with "PT100" thermocouple (Temperatur Messelemente TMH, Hettstedt GmbH, Maintal, Germany) was used.

Below the aforementioned setup ultrasonic baths, "Bandelin Sonorex RK 52H" (BANDELIN electronic GmbH &Co. KG, Berlin, Germany) types were placed on lab jacks so they could be lowered for sampling and batch changes. This arrangement made disassembly easier and did not require readjusting the upper structure with every batch change. These ultrasound baths have a nominal frequency of 35 kHz and put out 60 W effectively, while output peaks can occur up to 240 W. The output level was fixed and ultrasound irradiation was altered by using it intermittently. In this case, they were filled with tap water to maximum capacity. The water volume was about 1.1 L due to the volume displaced by the reaction vessels.

As shown in Table 3, the parameters for the leaching experiments were proposed using our previous hydrometallurgical experience and previously performed experiments, reported in the literature [20].


**Table 3.** Parameters for the leaching of cobalt from polycrystalline diamond blanks.

The abbreviations in the first column are read as successive week number and R1 and R2 representing reactors 1 and 2, respectively. These codes also served as stems for sample identification. The column header S/L is short for solid-to-liquid ratio in units of grams per liter. Constant parameters were stirring speed and batch time. The duration of each batch was planned to be between 90 and 100 h. Sample names D06, D14, D15 and D18 are abbreviated product names of Mant® MSD-06-005, MSD-14-005, MSD-15-005, and MSD-18-005, all self-supported PCD blanks with diamond grain sizes of 5 μm.

#### 2.2.1. Preparation of Samples

The PCD blanks were weighed and measured with the teslameter (Projekt Elektronik GmbH, Berlin, Germany), beforehand to obtain the important '100%' reference value for evaluation. The aqua regia was mixed from three parts fuming hydrochloric acid 37% Emsure ACS/ISO quality (Merck KGaA, Darmstadt, Germany) and one part nitric acid 65% ISO analysis quality PanReac ApplicChem (Chicago, IL, USA). 240 mL of hydrochloric acid and 80 mL nitric acid were prepared in covered beakers for each reactor. Meanwhile, the ultrasound baths were filled with tap water and their heaters were set to 333 K and 353 K, respectively, to ensure that the bath temperature was nominal from the beginning of the experiment. Gas tightness of the apparatus was checked daily.

#### 2.2.2. Conduct of Experiments

In weeks 1 through 3, the ultrasound baths were only used as heated water baths. In weeks 4–6 the ultrasound was intended to be switched on for eight hours per day. In addition to refilling water, the time of ultrasound irradiation had to be noted. In the remaining three weeks the ultrasound was switched on permanently. Concerning the forced gas flow, a current of around 0.25 L/min was sufficient to ensure the flow in one direction only. Argon was chosen over nitrogen or pressured air because oxygen and nitrogen might have skewed the equilibria with NO2, NxOx or formed combustible mixtures with chlorine gas or hydrogen gas.

#### 2.2.3. Sampling

Throughout the experiments, only cobalt content in the solution and changes in magnetic properties of the PCD were sampled each day. First, a few milliliters of the solution were pumped from each reactor using a plastic syringe and PTFE tube and transferred into a small beaker. From there, 1 mL of solution was taken with a pipette and added to a 50 mL round glass flask which was then filled with deionized water up to the 50 mL mark, resulting in a 1 in 50 dilution. This sample solution was transferred again into a 50 mL sample vial, and analyzed by the chemistry department at the IME, RWTH Aachen University (Aachen, Germany) by inductively coupled plasma–optical emission spectrometry (ICP-OES) (SPECTRO ARCOS, SPECTRO Analytical Instruments GmbH, Kleve, Germany). The solid sample was analyzed by X-ray fluorescence (Axios FAST, Malvern Panalytical GmbH, Germany).

The first indication that CoCl2 was formed could be seen when taking samples from the solution. Especially towards Thursday and Friday of any experimental week, the solution taken from the reactor had a dark greenish teal color that changed to pink after a few seconds in the beaker, indicating the typical drying salt color change from the dihydrate (CoCl2·2H2O) to the hexahydrate (CoCl2·6H2O) form of CoCl2 when cooling below approximately 308 K. Liquid samples were analyzed with the method of inductively coupled plasma optical emission spectrometry (ICP-OES).

#### 2.2.4. Weighing of Sample

For the purpose of determining the mass balances, PCDs were weighed before and after each batch on the same scale in the laboratory, type LA620P (Sartorius AG, Göttingen, Germany). Before the experiments, the PCDs were weighed as they were delivered. After an experiment they were wet with aqua regia, so they had to be rinsed with distilled water at least twice. In between stages, the PCDs were left in fresh distilled water for about 15 min. After the last rinse they were shaken with a little ethanol to assist in the drying process. Then, after two days at 353 K in the laboratory dryer. They were weighed while still warm. This was to ensure that no humidity would skew the results of weighing. According to Redies GmbH & Co. KG (Aachen, Germany), Aachen, a loss of around 20 weight percent due to extraction of cobalt from PCD is to be expected.

#### 2.2.5. Observation of Changes in Magnetic Properties of PCD

Changes in magnetic properties were observed using a teslameter, type FM 205 (Projekt Elektronik GmbH, Berlin, Germany), with a reference neodymium magnet. The apparatus for measuring of magnetic properties is shown in Figure 6.

**Figure 6.** Measuring magnetic properties of a PCD. a: Magnet support, b: ∅ 5 mm by 8 mm Nd alloy magnet, c: PCD blank, d: probe, e: handheld teslameter.

It consists of a teslametric probe that is kept at a fixed distance from the reference magnet, such as Nd alloy. The probe will display a value at any time representing the current magnetic situation. To diminish disrupting effects from the surroundings, the probe was offset by the nearby reference magnet. The actual measurement was always a sum of the magnetic surroundings, reference magnet and the subject in between. With only air and plastic between magnet and probe, the reading on the display was 583. A measurement was taken before every run with readings varying above 600. The value displayed for air as a subject is taken as 0% as a reference value. The value for the unleached PCD marks the 100% value for each batch and reactor, respectively. This method allows for a normalization of values and a plot of inferred Co content in the PCD relative to its initial content. In the course of the experiment, this value dropped towards the value of normal air, enabling an estimate of the progress of relative cobalt content in the PCD.

Since all measurements were always compared against the offset value and normalized to the pre-experiment value being 100%, units cancel out. It has to be noted that this offset value was registered before each individual PCD measurement because it changed between 583 and 584—perhaps due to other magnetic influences, or the magnet may have been placed in a way that resulted in a measurement on the threshold between 583 and 584. To ensure consistency, the magnet was not moved until after the experiments. An important caveat is the fact that the relative Co content is inferred, not measured. The change in the magnetic field of the probe consists of more than just the effect due to the presence of cobalt. In this case, analyses have shown that there are oxygen and iron impurities present in the raw PCD. Fe(II), Fe(III), Co(II), and Co(III) oxides have magnetic properties that naturally differ from pure Co. The overall effect of magnetic metals is measured and the Co content is concluded from these values.

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

#### *3.1. Mass Balance*

The measured concentration of cobalt in the samples was plotted against the time when the samples were taken to visualize the accumulation in solution, as shown in Figure 7 (from 1 and 9 week).

**Figure 7.** Cobalt concentration versus time in weeks 1 and 9.

These plots were chosen as the extreme cases, with the other results very similar and within their range. Without ultrasound and at the low end of solid-to-liquid ratios, the concentration of cobalt developed very similarly regardless of temperature, as shown at Figure 8. However, towards the end of week nine, at a high solid-to-liquid ratio and with full ultrasound, a difference emerged where the Co content in reactor 2 seemed to go into saturation. It appeared from week 4 onwards, so it may be linked to the higher temperature and use of ultrasound.

**Figure 8.** Concentration of cobalt in solution in time between 333 K and 353 K with full-time ultrasound.

When comparing the two reactors independently at Figure 8, this discrepancy becomes more obvious. With ultrasound used full-time, the solution at 353 K accumulated several percent less cobalt in total than the cooler reactor. However, the initial increase happened faster. To see the effect of ultrasound itself, the concentration curves from low and high solid-to-liquid ratios were plotted for both reactors, as shown at Figure 9. Interestingly, full-time ultrasound did not seem to achieve the highest cobalt yields. At high solid-to-liquid ratios, it did not even seem to do any better than the experiments without ultrasound.

**Figure 9.** Influence of ultrasound on concentration of cobalt in the solution over time.

As stated above, the PCDs were weighed to gain insight into whether the chosen parameters, especially ultrasound, aided the leaching process. These measurements were used to compare the results of all runs, shown individually for D14 and D18 blanks in Figure 10. At first glance, the most obvious fact is that the larger PCDs were not leached to completion within the 90 to 100 h timeframe. The maximum values for lost PCD weight were 21.04% and 13.86% for D14 and D18, respectively. For D14, the most influential parameter appears to be ultrasound. There is a strong influence of bath temperature without it. However, already intermittent ultrasound is enough to drive the leaching efficiency towards the expected value. In the runs with ultrasound, higher temperature only has a slight effect. Solid-to-liquid ratio does not have a positive or negative impact. For D18, the influence of ultrasound is measurable, but is not as strong as it is for the smaller D14. The difference between intermittent and permanent ultrasound irradiation seems negligible, but higher solid-to-liquid ratios have an adverse effect in the runs with intermittent ultrasound. With permanent ultrasound, this inhibition is apparently gone. Higher bath temperature, on the other hand, influenced leaching in a positive way. The secondary axis "expected leaching efficiency" has to be viewed with caution. The 100% mark refers to the expected 20% weight percent of leachable substance in the PCD. In fact, this value has only been orally confirmed by the manufacturer and the only analysis is the surface SEM and EDS, as shown at Figures 1 and 2. The leaching efficiency was calculated using Equation (1). The obtained results with D14 show that this value is reasonably accurate.

$$\text{Leaching efficiency} \, [\, \% \,] = (\Delta \text{m}\_{\text{PCD}} \text{ or } \text{c(Co)} \text{s} \, \text{Cl}) \, (0.2 \, (\text{m}\_{\text{O}} \text{C} \text{D}) \, \text{s} \tag{1}$$

where ΔmPCD is the lost PCD weight after the experiment, c(Co)SOL is the cobalt content in solution after the experiment, 0.2 is the given factor of initial cobalt content in PCD and m0,PCD is the initial PCD weight.

**Figure 10.** Lost weight for different PCD samples after leaching with different parameters described in the plot. 35 kHz ultrasound varied from "none", "ultrasound on" for a third of the run time, to "full-time ultrasound", as shown from left to right.

Weight data was used to calculate the average PCD weight lost per day, as shown at Figures 10 and 11. The difference between initial weights and final dry weights was divided by the actual batch time in hours. This was done to renormalize the values to a certain time interval because the batches had different run times. Though differences in the diagrams above may only be small, the diagrams below are truly adequate for a comparison. The negative effect of the solid-to-liquid ratio on leaching efficiency of cobalt from D18 blanks with intermittent ultrasound became clearer, as shown at Figure 11. A less obvious difference is the fact that the highest columns in the diagrams below now truly are the runs with the most extracted weight percentage per time.

**Figure 11.** Lost weight for different PCD samples averaged per day. 35 kHz ultrasound varied from "none", "ultrasound on" for a third of the run time, to "full-time ultrasound", as shown from left to right.

In weeks 5, 8, and 9 the volume of remaining liquid was measured. The weight of dissolved cobalt was calculated by combining these volumes with the Co (II) concentrations from chemical analysis. A comparison between the weight of dissolved Co (II) versus lost PCD weight during leaching is presented in Figure 12. If pure metallic cobalt was used and remained in its metallic state during the

making of the PCD, there should be no difference between these values. All the material that is leached from the PCD should be cobalt and end up in the solution as dissolved ions.

**Figure 12.** Weight of dissolved Co (II) versus lost PCD weight.

As can be clearly seen, there is a significant difference, ranging between +3% and +10%, among the compared values. On account of the EDS data (Table 1), this difference probably stems from cobalt oxides and iron impurities rather than fluctuations or measuring errors and uncertainties. Finally, the obtained solution using an aqua regia (Figure 13—left) from polycrystalline drawing die blanks via the ultrasound-assisted leaching process is shown in Figure 13—right. Cobalt powder was obtained from this solution using the precipitation method.

**Figure 13.** (**Left**): Aqua regia approximately 9 min after mixing and stirring. (**Right**): View of the reactor during the experiment, containing the PCD and cobalt bearing solution. In this case, it was running at 333 K, with full-time ultrasound and 45 g/L solid-to-liquid ratio.

#### *3.2. Results Regarding Process Optimization*

The results from this study suggest that the leaching of D14 does not require a 353 K bath temperature but can be done at 333 K. Ultrasound can accelerate the leaching process to the extent that the PCD can reach a desaturated state with less than 10% of metallic inclusions remaining after three to four days if they are leached at low solid-to-liquid ratios close to 15 g/L. The depth of ultrasound penetration into PCD, meaning the depth at which the disruptive effects of ultrasound are no longer able to outrun diffusion, was determined to be in the region of 1.4–1.7mm.

If the emphasis were put on just shortening the leaching time, one way could be to leach at 353 K, replacing the solution after three days to reset the concentration gradient and refresh the active compounds in solution. As there is more research needed regarding the kinetics and mechanism of the dissolution process, the same applies to process safety and possible replacement of aqua regia with other less harmful leaching agents. The potentiometric aspects of leaching of cobalt from PCD also deserve considerable attention in order to ensure controlled potential cobalt leaching as a selective way for total cobalt removal in a short time. The kinetics and mechanisms of the studied ultrasound-assisted leaching process from polycrystalline diamond blanks will be reported in Part 2 [24] of this research in detail.

#### *3.3. Possible Recycling Routes for Cosolution in Order to Produce Cobalt Powder and Its Compounds*

After the experiments, a highly acidic aqua regia solution laden with divalent cobalt remains. This cobalt content is very valuable and should not be discarded. One simple (but hardly elegant) way to reuse the cobalt chloride would be the complete evaporation of liquids using the remaining cobalt chloride as drying salt. Instead, there are ways to selectively extract Co from the solution with DEHPA2, for example, and then precipitating or electrolytic winning of the metal powder. As stated earlier, there also is the possibility to make cobalt nanopowder using the ultrasonic spray pyrolysis method and the chemical reduction method in the aqueous solution, which would be a very versatile substance to be used in battery technology as well as catalyst applications. The production of cobalt hydroxide shall be reached using sodium hydroxide as precipitation agent. A goal-oriented refining process such as solvent extraction as a traditional hydrometallurgical method could be imagined depending on the desired metal powder and its compounds.

#### **4. Conclusions**

This study was designed for the recovery of pure demetallized PCD and cobalt from raw PCD. In nine experimental runs with a 5 day duration, cobalt containing PCD was leached in aqua regia at atmospheric pressure between 333 K and 353 K. Using two reactors in parallel, temperature, ultrasound irradiation time, solid-to-liquid ratio, and PCD size were varied to find out which parameters are beneficial and could possibly accelerate this process. PCD weights and cobalt content in solution were also monitored. It was found that aqua regia accumulated more dissolved cobalt at 333 K than at 353 K probably due to volatile reagents being less available over time. The ultrasound treatment increases the leaching efficiency. With added ultrasound (even at just a third of total run time) and at a low S/L ratios close to 15 g/L, the leaching time for D14 to reach the 90% leached mark was reduced to three days, which is a significant shortening of leaching time. PCD type D18 with a thickness of 3.5 mm was not leached to completion within five days. The leaching temperature had more impact on the results than ultrasound. These findings were reinforced by the mass balance in which a small discrepancy was found. The PCD lost a fraction of weight that could not be explained by the weight of dissolved cobalt. From EDS data and the nature of PCD, this fraction probably consisted of oxygen from oxides in the PCD or single diamond grains that were broken off by the impact of ultrasound. Advances in synthesis of metallic powders using the ultrasound-assisted leaching process from polycrystalline diamond blanks can be used for the cemented tungsten carbide in order to estimate a scale-up of this process in future.

#### **Recovery of Diamond and Cobalt Powders from Polycrystalline Drawing Die Scraps via Ultrasound-Assisted Leaching Process—Part 2: Kinetics and Mechanisms**

The kinetic models were used for the study of cobalt dissolution from polycrystalline diamond blanks via a measurement of declining ferromagnetic properties over time. For a better understanding of this leaching process, thermochemical aspects were included in this work. The lowest free Gibbs energy corresponds to a low solid/liquid ratio and fully used ultrasound in the process. A transition from a reaction-controlled to a diffusion-controlled shrinking core model was found for PCD with a thickness larger than 2.8–3.4 mm. Intermittent ultrasound doubles the reaction rate constant and fully using of ultrasound causes a further increase with a factor of 1.5. The obtained activation energy between 333 K and 353 K is 20 kJ/mol, and small for all diamond blanks with a diameter size of 5 μm, which corresponds to the diffusion-controlled process.

**Author Contributions:** F.K. and S.S. conceptualized and managed the research. S.S. cowrote the paper. S.G. contributed the SEM and EDS analysis of the PCD surface. B.F. supervised personnel, coordinated resources, and co-wrote the paper. F.K. performed the experiments and wrote the paper. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Projektträger Jülich (PtJ), Grant Number 005-1902-0147.

**Acknowledgments:** We would like to thank Redies Deutschland GmbH & Co. KG (Aachen, Germany) for providing PCD samples as well as additional equipment.

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

#### **References**


© 2020 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* **Advances in Thermochemical Synthesis and Characterization of the Prepared Copper**/**Alumina Nanocomposites**

#### **Marija Kora´c 1, Željko Kamberovi´c 1, Zoran Andi´ ¯ c <sup>2</sup> and Sre´cko Stopi´c 3,\***


Received: 30 April 2020; Accepted: 26 May 2020; Published: 28 May 2020

**Abstract:** This paper presents thermochemical synthesis of copper/alumina nanocomposites in a Cu-Al2O3 system with 1–2.5 wt.% of alumina and their characterization, which included: transmission electron microscopy: focused ion beam (FIB), analytical electron microscopy (AEM) and high resolution transmission electron microscopy (HRTEM). Thermodynamic analysis was used to study the formation mechanism of desirable products during drying, thermal decomposition and reduction processes. Upon synthesis of powders, samples were cold pressed (2 GPa) in tools dimension 8 × 32 × 2 mm and sintered at temperatures within the range 800–1000 ◦C for 15 to 120 min in a hydrogen atmosphere. Results of characterization showed that dispersion-strengthened compacts could be produced by sintering of thermo-chemically prepared Cu-Al2O3 powders with properties suitable for material application, such as a contact material exhibiting high strength and high electrical conductivity at the same time. Additional research was carried out in order to analyze the application of the obtained nanocomposite powders for the synthesis of copper/alumina nanocomposites by a new method, which is a combination of a thermochemical procedure and mechanical alloying. The measured values of an electric conductivity and hardness were compared with ones in literature, confirming an advantage of the proposed combined strategy.

**Keywords:** synthesis; oxide; nanocomposites; characterization; copper; alumina; thermochemistry

#### **1. Introduction**

Research of metal matrix composite (MMC) materials has considerably intensified since its first mention in the 1950s [1]. Various combinations of base (Al, Cu, Ni, Mg, Ti, Fe, Co, etc.) and reinforcing material (e.g., oxides, borides, carbides, fibers, tubes) have been studied [2]. Through selection of appropriate combinations and ratios of materials, a wide spectrum of properties can be achieved, followed by extensive industrial applications.

Copper is considered the most significant base material for industrial applications, due to its good electrical and heat conductivity. Disadvantages of copper are its mechanical properties, such as high ductility, low wear resistance and thermodynamic instability at elevated temperatures. One of the possibilities for overcoming poor mechanical properties is reinforcement by dispersion strengthening; i.e., the introduction of fine ceramic particles. By the dispersion strengthening of copper, significant increases in mechanical properties can be achieved, with low adverse impacts on its electrical and heat conductivity. The main requirements for dispersed particles are higher thermodynamic stability at elevated temperatures; higher hardness, strength and wear resistance; and low solubility in base metal. Appropriate size and even distribution of dispersed particles also contribute to a positive effect via dispersion strengthening [3]. Finer particles with homogenous distribution and low volume fraction of dispersed particles in the total volume of the base metal will act as obstacles to dislocation motion, even at elevated temperatures without significant effects on conductivities, both thermal and electrical [4–7].

One of the most widely used oxides is alumina, which fulfills all the requirements for the dispersed particles and has low cost at the same time [8]. Additionally, alumina can increase the temperature of recrystallization of the copper matrix and demonstrates excellent strength at elevated temperature by pinning grain and sub-grain boundaries of the matrix. Finally, alumina particles add to strengthening by blocking the movement of dislocations [7,9,10]. The usual amount of alumina used for dispersion strengthening is 0.5–5.0 wt.% [11], but significant results regarding particle size can be achieved even with higher amounts, such as 50 wt.% of Al2O3 [12].

There are numerous routes for the synthesis of metal–matrix composites, but nowadays two main routes are the mechanical alloying and thermochemical route. Mechanical alloying is extensively used method for synthesizing of nanocrystalline materials by severe plastic deformation on the powder using high-energy ball milling technique [13–16]. This technique commonly employed for the prevention of formation of clusters and agglomerates enables the production of uniformly dispersed fine particles in a metal–matrix. On the other hand, using the thermochemical method [17] where input materials are in a liquid state enables production of finer particles and much more homogeneous structure of the final powder, which further contributes to the increase in the mechanical properties for the final product through various strengthening mechanisms.

Authors have also developed a new synthesis route based on the combination of routes mentioned above [18]. This route may be regarded as a new strategy for materials in the Cu-Al2O3 system, even though some phases of this process have been previously investigated by the authors [19–21]. Additionally, previous attempts have been made by authors for application of similar process in the system Cu-Ag-Al2O3, where a three-component system was produced by mechanically alloying the thermo-chemically-synthesized Cu-Al2O3 and Cu-Ag powder [22].

The main aim of this work was to investigate a thermochemical synthesis of metallic particles and nanocomposite with a microstructure and strengthening mechanism of copper with finely dispersed alumina particles. A novelty of this synthesis is a decreased reduction temperature for chemical reaction of the powder in a hydrogen atmosphere at 350 ◦C, which is an advantage in contrast to 820 ◦C for 1 h to produce the final Cu–Al2O3 nanocomposite powder, as described by Seyedraoufi et al. [23]. It can be very important point for decreasing production costs. A thermodynamic analysis of the reduction, spray drying and synthesis reactions was performed in order to predict a chemical behavior of the compounds. Amirjan et al. [24] have used artificial neural networks to predict Cu-Al2O3 properties. In order to prepare copper based composites, copper powder with four different amounts of Al2O3 reinforcement (1, 1.5, 2, 2.5 wt%) were mechanically alloyed, and the consolidated compacts of prepared powders were sintered in five different temperatures of 725–925 ◦C at seven several sintering times of 15–180 min. Guevara et al [25] have studied the synthesis of copper-alumina composites by mechanical milling via an analysis of materials and manufacturing processes. Ha et al. [26] studied the fabrication of Al2O3 dispersion strengthened copper alloy by spray in-situ synthesis casting process above 1250 ◦C as a new method. Mohammadi, E. et al. [27] used a combustion method for the synthesis of Cu-Al2O3, which take place in a short time at temperatures higher than 1000 ◦C. Generally, our synthesis method offers a cost-friendly process for the synthesis of Cu-Al2O3 in comparison to other processes [27].

These powders could be used for production of sintered materials with properties suitable for material applications, such as contact material exhibiting high strength and high electrical conductivity at the same time.

Some comparative results for different synthesis methods are presented, indicating that by mechanical alloying of atomized copper powders with produced composites, followed by thermo-mechanical treatment, sintered materials with improved properties could be produced.

#### **2. Experimental**

Water soluble copper and aluminum nitrates, Cu(NO3)2·3H2O and Al(NO3)3·9H2O, were used to synthesize a two-component nanocomposite Cu-Al2O3 powder by the thermochemical procedure.

The synthesis was carried out through four stages, as presented in Figure 1.

**Figure 1.** Flowsheet of the synthesis of Cu-Al2O3 nanocomposite powder by the thermochemical procedure [28].

Process temperatures are derived from thermodynamic consideration of the process and the following six chemical reactions:

Spray drying:

$$\text{Cu(NO}\_3\text{)}\_2\text{-}6\text{H}\_2\text{O}=\text{Cu(NO}\_3\text{)}\_2 + 6\text{H}\_2\text{O} \tag{1}$$

$$\text{Al(NO}\_3\text{)}\_3 \cdot 6\text{H}\_2\text{O} = \text{Al(NO}\_3\text{)}\_3 + 6\text{H}\_2\text{O} \tag{2}$$

Heat treatment:

$$\text{Cu(NO}\_3\text{)}\_2 = \text{CuO} + \text{N}\_2\text{O}\_5 \tag{3}$$

$$2\text{Al(NO}\_3\text{)}\_3 = \text{Al}\_2\text{O}\_3 + 3\text{ N}\_2\text{O}\_5 \tag{4}$$

Reduction:

$$\text{CuO} + \text{H}\_2 = \text{Cu} + \text{H}\_2\text{O} \tag{5}$$

$$\text{Al}\_2\text{O}\_3 + 3\text{H}\_2 = 2\text{Al} + 3\text{H}\_2\text{O} \tag{6}$$

Using HSC Chemistry® software package 6.12 (Outotec, Espoo, Finland), chemical and thermodynamic parameters of the processes for synthesis Cu-Al2O3 composites were analyzed. As shown at Figure 2, the calculated values of Gibbs energy of reactions versus temperature (up 1000 ◦C) for reactions (1)–(6) have positive and negative values; negative values confirmed the possibility for the beginning of these chemical reactions at the studied temperature. Because of the high positive values of Gibbs energy (more than 800 kJ/mol), reduction of aluminum oxide with hydrogen (as shown with Equation (6)) is not possible between 25 and 1000 ◦C.

**Figure 2.** Gibbs energy of reactions versus temperature.

The first stage is the preparation of 50 wt.% aqueous solutions of Cu(NO3)2·3H2O and Al(NO3)3·9H2O (the quantities of salt were set so that the requested composition of a Cu-Al2O3 nanocomposite system with 5 wt.% of alumina could be produced). The second phase is spray drying of nitrate solution using Mini Spray Dryer B-290 Advance (BÜCHI Labortechnik GmbH, Essen, Germany) for producing the precursor powder, with inlet/outlet temperature 190/143 ◦C and a solution flow rate of 10% pump power. The third stage is oxidative calcination of the precursor powder in an air atmosphere at 900 ◦C for 1 h to form copper oxide and the phase transformation of Al2O3 up to the thermodynamically stable α-Al2O3 phase. Final fourth stage was the reduction of thermally treated powders in hydrogen atmosphere flow rate 20 L/h at 350 ◦C for one hour, where copper oxide was transformed into elementary copper, while Al2O3 remained unchanged.

All temperatures were below the temperatures the melting temperature of Cu (1085 ◦C) [29] and Al2O3 (2072 ◦C) [30].

In previous work of authors procedures [22,29] and process parameters [21,31] are fully described for the synthesis of two-component nanostructured composite materials.

The obtained powders were cold-pressed (force 500 kN, calculated pressure 2 GPa) in tools with dimensions of 8 × 32 × 2 mm and sintered at temperatures within 800–1000 ◦C for 15 to 120 min in a hydrogen atmosphere. Kinetics of the sintering process were determined and presented elsewhere [21].

Characterization of compacted powders after sintering at 900 ◦C for 2h included transmission electron microscopy: analytical electron microscopy (AEM), high resolution transmission electron microscopy (HRTEM) and focused ion beam (FIB) at e-beam 5.00 kV. The powder to be tested is suspended in a liquid (water, ethanol or butanol) with the aid of an ultrasonic device. Depending on particle size, requirements and type of examination, the powder is first ground. By means of a pipette a drop of suspension is taken up and placed on a carbon carrier net. The liquid is then allowed to evaporate (dry) under a lamp, resulting in a C-carrier net with the powder on top. After this powder preparation, our sample was studied by TEM Analysis. HRTEM analysis was performed using Philips CM200/FEG (FEI Company, Hillsboro, OR, USA).

Mechanical properties of sintered samples were also investigated and are presented in authors' previous research [28].

Ames Portable Hardness Tester was employed for hardness measurements using a 1/16" ball with an applied load of 60 kg. For electrical conductivity measurement, SIGMATEST 2.069 (FOERSTER, Pittsburgh, PA, USA) operating at 120 kHz and with an 8 mm electrode diameter was used.

Values of hardness and electrical conductivity represent the mean values of at least six measurements conducted on the same composite.

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

In previous work of the authors [22], the results of determination of fluidness, pouring density and specific area of the obtained nanostructured composites with different amounts of Al2O3 dispersed in the copper matrix showed that all the investigated powders are not fluid and that mean values of pouring density and specific area are the same for different contents of Al2O3 up to the 5% investigated.

Additionally, in some previous studies of the authors [21,22,32] the results of differential thermal and thermogravimetric analysis (DTA-TGA) and scanning electron microscopy can be found, which show the flow of phase transformations during the process of oxidation, and the morphologies of the obtained powders.

Only peaks corresponding to the nitrates of copper and aluminum were identified in the structure during XRD examination of the precursor powder produced by spray drying an aqueous solution of copper and aluminum nitrates, which is in accordance with the experiment set-up [20–22]. X-ray diffraction analysis after annealing of dried powder exhibited peaks corresponding to CuO and Al2O3, and one unidentified peak. According to Lee [33] this peak corresponds to a third phase, Cu*x*Al*y*O*z*, which appears in the structure due to the eutectic reaction of (Cu + Cu2O) with Al2O3.

The produced powders were analyzed by AEM with corresponding EDX, as shown in the previous works of authors [34]. Based on AEM analysis, particles 20–50 nm in size, are clearly noticeable, as is the presence of agglomerates >100 nm. Particles are irregularly shaped; there are nodular individual particles with rough surface morphology. EDS analysis of marked spot show that the identified peaks correspond to Cu, Al and O. The intensities of peaks correspond to demanded compositions of the examined systems; therefore, the peak corresponding to copper is considerably higher than the peaks corresponding to aluminum and oxygen.

In order to identify the third phase, the authors performed additional research through the synthesis of Cu-50 wt.% by a thermochemical procedure. X-ray diffraction analysis of the obtained sample, presented in [18], shows the presence of copper peaks and CuAl2O4 compounds, which may represent a metastable phase that developed in the microstructure during the process of powder synthesis, thermal treatment and reduction on the surface of the contact between Cu and Al2O3 and is a seed for the development of the third phase during the sintering process.

FIB analysis of the sintered Cu-Al2O3 system based on the powders obtained by the thermochemical procedure, as shown in Figure 3, is characteristic for the final stage of sintering. FIB analysis did not indicate even at considerably higher magnifications, the existence of a phase rich with alumina. The bright fields are identified, i.e., a phase rich with copper, and gray fields, which can lead to a possible existence of the third Cu*x*Al*y*O*z* phase identified by X-ray diffraction analysis [18]. The formation of this phase is thermodynamically possible on Cu-Al contact surfaces. During eutectic joining of copper and Al2O3, the eutecticum formed by heating up to the eutectic temperature expands and reacts with Al2O3 creating Cu*x*Al*y*O*z*, which is compatible with both phases on the inter-surface. According to [35,36], the process of formation of the third phase is developed through the following reactions: 2CuO + H2 → Cu2O + H2O, Cu2O + Al2O3 → 2CuAlO2 and/or CuO + Al2O3 → CuAl2O4. CuAlO2 is stable in air with the temperature range from 800 ◦C to 1000 ◦C, while CuAl2O4 is transformed into CuAlO2 at the temperature of approximately 1000 ◦C. However, the presence of Cu*x*Al*y*O*<sup>z</sup>* phase demands a detailed characterization by using high resolution apparatus. Additionally, from micrographs of the examined samples, homogenous distribution of the present phase is clearly noticeable and the size of microstructural constituents in the range of 50–250 nm (Figure 3).

**Figure 3.** FIB image of compacted Cu-5wt.% Al2O3 composite sinter.

Additionally, the FIB analysis confirmed the analysis of structural stabilization of the system based on the values of the specific electric resistance of sintered samples (<sup>ρ</sup> = 0.061 <sup>×</sup> 10−<sup>6</sup> <sup>Ω</sup>·m). Results show also that the hardness of sintered samples (HRB 10/40 (average = 124.7)) was very high for the achieved density of the sample yet lower than expected. The results of examining density, relative change of volume and the electrical and mechanical properties of sintered systems based on nanocomposite Cu-Al2O3 powders synthesized by the thermochemical process have been presented in previous papers by the authors [21,32].

Typical microstructure of Cu-Al2O3 5 wt.% is presented in Figure 4. In BF (bright field)–DF (dark field) pair, it can be seen that a copper crystal exhibits annealing twins. Twins are slightly curved, a typical feature of deformation twinning, but in the presented case, it could be a consequence of a high temperature sintering stage.

**Figure 4.** TEM analysis of sample after sintering: BF (bright field) and DF (dark field) images showing nano-twinning on Cu crystal (**a**,**b**), homogenous distribution of Al2O3 particles (**c**)

In Figure 4a,b, a typical TEM pair bright field (BF)-centered dark field (CDF) of nanocomposite Cu-5wt.% Al2O3 sintered system is shown, where the well-developed crystals of copper are exposed to twinning, despite their small size. In addition, detailed analysis indicates the detection of fine Al2O3 individual particles or aggregates. Conditions for twinning are accomplished when a great number of obstacles, such as homogeneously distributed Al2O3 particles, are created in the crystal which hamper dislocation mobility, dislocation plaits or already present twins. Since dislocations are piled up at the obstacles, in such local regions internal tension is increased, which, along with external tension, provokes creation of twins. Decreasing of dislocation mobility represents a condition for creating twin embryos; therefore, in Figure 4a,b, the clearly noticeable presence of twins indicates a decreased mobility of dislocations, i.e., stabilization of dislocation substructure, which is an elementary precondition for improving mechanical properties; i.e., reinforcing of metal materials.

Fine dark spots noticeable in the BF image (Figure 4c) present Al2O3 particles, size range 5–20 nm, dispersed in the copper matrix. Additionally, Figure 4c shows a homogenous distribution of Al2O3 particles, which is one of the requirements for dispersion of strengthened copper composites, to retain electrical conductivity of the base metal.

In a second set of TEM BF-DF pair images (Figure 5a,b) dispersion of alumina particles is also visible. Furthermore, TEM results in Figure 5a show the presence of dislocation density (the upper-right region of the grain) in a copper matrix surrounding the alumina particles, additionally increasing the strength of the material. Additionally, in Figure 5b, Moire fringes could be observed. According to [37] inside a single copper crystal, the clusters of Al2O3 particles could considerably alter the surrounding lattice structure, enough to prompt formation of Moire fringes.

**Figure 5.** TEM analysis of a single copper grain containing a fine dispersion of alumina particles, dislocations and Moire fringes.

Figure 6a,b shows selected area diffraction patterns (SADPs), where both single spots and a Debye–Scherrer ring pattern can be observed. Single spots in SADP correspond to crystalline copper along the [111] axis, while the Debye–Scherrer rings in Figure 6b correspond to alumina Al2O3 particles.

In Figure 6a, besides the spots corresponding to copper, additional diffraction spots are visible, indicating presence of a solid solution. During sintering stage, formation of a third phase is possible, during eutectic reaction under suitable thermodynamic conditions at the Cu-Al2O3 interphase containing all three elements in a very narrow region. Existence of the third phase in the structure remains to be proven by further indexing and calculations. Composition of this phase could be, according to the literature, CuAlO2 or CuAl2O4 [11,38] due to presence of an O-rich interface with larger adhesive energy [39]. Presence of this phase additionally reinforces the copper matrix by blocking the grain and sub-grain boundaries.

**Figure 6.** Selected area diffraction pattern (SADP) taken inside the grain in Figure 4: (**a**) single diffraction spots for copper along the [111] axis, (**b**) Debye–Scherrer rings for the pure Al2O3 particles and (**c**) inset showing the simulated Debye–Scherrer rings for γ-Al2O3.

Ratios of measured D-values from the ring diffraction pattern are in good agreement with the calculated ratios of the corresponding g-vectors for the γ-alumina (Figure 6c).

There is a subtle difference between the crystal structure of γ-alumina and μ-alumina, yet the performed characterization via TEM did not provided a definite proof of that. It is more likely that the structure is that of γ-alumina as the original structure is boehmite and its transformation sequence does not include μ-alumina in accordance with [40].

Successful application of synthesized of the nanocomposite Cu-Al2O3 powders obtained by the thermochemical procedure in mechanical alloying of atomized copper powders is in detail presented in [18]. Because of high strength and electrical properties, this material can be used as electrode material for lead wires, relay blades, different contact materials and various switches, and especially for electrode materials for spot welding due to high conductivity of copper and high hardness and excellent thermal stability of aluminum.

The obtained nanocomposite powders, with structure basically preserved with the final product, provided a significant reinforcement effect in the produced sintered system. This is a consequence of homogenous distribution of the elements in the structure, accomplished during synthesis of powder and presence of the third phase which causes stabilization of dislocation substructure, accomplishing a relevant reinforcing effect and achievement of a good combination of mechanical–electric properties of the sintered systems. Comparative analyses of mechanical properties of produced composites were derived from the previous work of the authors [18–22,28,34] regarding thermo-chemical synthesis, mechanical alloying and the new synthesis method. Sintering of copper/alumina nanocomposite

powders was performed at 875 ◦C in 60 min in laboratory electro resistant furnace in a hydrogen atmosphere in order to avoid oxidation of samples. From the presented results in Table 1, it could be concluded that use of obtained powders for mechanical alloying followed by plastic deformation have the same level of hardness with a much lower amount of Al2O3, which has a direct consequence through higher values of electrical conductivity. The maximal values of electrical conductivity and hardness were obtained for the sample based on 1.0 wt.% percent of Al2O3 in structure. Regarding to the same chemical composition of copper/alumina nanocomposite in comparison to Amirjan [24], the values of electrical conductivity are higher in all cases, what confirms an advantage for our studied combined strategy.


**Table 1.** Comparison of the electrical conductivity and hardness values (HRF) after sintering at 875 ◦C and 60 min in our work with values of Amirjan [24].

\* IACS (International Annealed Copper Standard).

After annealing at 800 ◦C, hardness and electrical conductivity amounted to 58 HRF and 61.78% IACS, respectively.

The proposed strengthening mechanism is presented in Figure 7. The mechanism combines the strengthening in thermo-chemically synthesized composites and strengthening during mechanical alloying.

**Figure 7.** Microstructure transformation induced by following technological steps: mechanical alloying, heat treatment, plastic deformation and sintering.

Throughout thermo-chemical synthesis, copper base strengthening is achieved by dispersion of fine particles of Al2O3, and strengthening by grain boundaries, as presented in this paper. As reported by Amirjan [24] with respect to strengthening mechanism of Orawan, with increasing reinforcement amount, the distances between particles in the microstructure will decrease. Therefore, the dislocations can encompass the particles easily and lead to lower values of hardness. We assume the grain size of the composite matrixes microstructure increases with increasing sintering time. According to Hall–Petch effect, larger grain size in microstructure leads to a decrease in hardness values.

During mechanical alloying of atomized copper particles, copper-alumina composites are built into its surface. Along with the process of sintering occurs the formation of the compact structure and formation of the third phase on the grain boundary, which causes strengthening on the grain boundaries. Due to the plastic deformation, the deformation strengthening occurs, and after heat treatment, the strengthening by annealing occurs. The annealing treatment increases the system's strength by reducing dislocation emission sources and improves material ductility through strengthening grain boundaries' resistance to intergranular cracks.

#### **4. Conclusions**

Characterization of produced nanostructured composites in system Cu-Al2O3 showed the possibility of their synthesis via a thermochemical route. The mechanism of formation of copper-alumina nanocomposite was studied using thermodynamic analysis of drying, thermal decomposition, reduction step and homogenization. The reduction of thermally treated powders was performed in hydrogen with flow rate 20 L/h at 350 ◦C for one hour, where copper oxide was transformed into elementary copper, while Al2O3 remained unchanged.

By AEM analysis it was confirmed that homogenous distribution of Al2O3 particles was achieved by the thermochemical route followed by cold pressing and sintering, a necessary requirement for retaining electrical conductivity of the base metal.

Increasing of the strength of the material was achieved by presence of dislocation density in a copper matrix surrounding the alumina particles, confirmed by TEM analysis.

Additionally, the selected area diffraction pattern showed the possible presence of a third phase formed during the sintering stage at interphase containing all three elements in a very narrow region, which additionally reinforces the copper matrix by blocking the grain and sub-grain boundaries. The existence of the third phase in the structure remains to be proven by further indexing and calculations. The proposed strengthening mechanism combines the strengthening in thermo-chemically synthesized composites and strengthening during mechanical alloying. The maximal values of electrical conductivity and hardness were obtained for the sample based on 1.0 wt.% percent of Al2O3 in structure. Regarding the same chemical composition of copper/alumina nanocomposite in comparison to literature values by Amirjan [24], the values of electrical conductivity and hardness are higher in all cases, which confirms an advantage for our studied combined strategy.

The future study can be focused on the kinetics of the thermochemical synthesis of the studied nanocomposites. The economic size (cost effect) of this method shall be calculated via partial operations. This is a practical way for manufacturing these composites for powder metallurgy based on different applications.

**Author Contributions:** M.K. conceptualized and managed the research, and co-wrote the paper together with the other co-authors. Ž.K. ensured financial support, supervised M.K. and co-wrote the paper. Z.A. participated in analysis and discussion of the obtained results and co-wrote this paper. S.S. helped in discussion of the results and co-wrote this paper. All authors have read and agreed to the published version of the manuscript.

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

**Acknowledgments:** Ministry of Education, Science and Technological development of Republic of Serbia through project TR34033 financially supported the research presented within this paper.

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

#### **References**


© 2020 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* **Reaction Mechanism and Process Control of Hydrogen Reduction of Ammonium Perrhenate**

**Junjie Tang 1,2, Yuan Sun 2,\*, Chunwei Zhang 2,3, Long Wang 2, Yizhou Zhou 2,\*, Dawei Fang <sup>3</sup> and Yan Liu <sup>4</sup>**


Received: 17 April 2020; Accepted: 13 May 2020; Published: 15 May 2020

**Abstract:** The preparation of rhenium powder by a hydrogen reduction of ammonium perrhenate is the only industrial production method. However, due to the uneven particle size distribution and large particle size of rhenium powder, it is difficult to prepare high-density rhenium ingot. Moreover, the existing process requires a secondary high-temperature reduction and the deoxidization process is complex and requires a high-temperature resistance of the equipment. Attempting to tackle the difficulties, this paper described a novel process to improve the particle size distribution uniformity and reduce the particle size of rhenium powder, aiming to produce a high-density rhenium ingot, and ammonium perrhenate is completely reduced by hydrogen at a low temperature. When the particle size of the rhenium powder was 19.74 μm, the density of the pressed rhenium ingot was 20.106 g/cm3, which was close to the theoretical density of rhenium. In addition, the hydrogen reduction mechanism of ammonium perrhenate was investigated in this paper. The results showed that the disproportionation of ReO3 decreased the rate of the reduction reaction, and the XRD and XPS patterns showed that the increase in the reduction temperature was conducive to increasing the reduction reaction rate and reducing the influence of disproportionation on the reduction process. At the same reduction temperature, reducing the particle sizes of ammonium perrhenate was conducive to increasing the hydrogen reduction rate and reducing the influence of the disproportionation.

**Keywords:** ammonium perrhenate; rhenium; disproportionation reaction; hydrogen reduction

#### **1. Introduction**

Rhenium as an important rare metal is widely used in metallurgy and the aerospace industry [1]. The plasma method, electrolysis method, vapor deposition method, and powder metallurgy are the main processes for the preparation of rhenium [2,3]. Jurewicz et al. [4] prepared a high-purity nanometer rhenium powder by the plasma method. Leonhardt et al. [5] used plasma spray spheroidization to control the microstructure of rhenium and obtained spherical rhenium powders. Schrebler et al. [6] also prepared spherical rhenium powder from a mixture of rhenic acid and sodium sulfate by electrolysis. Liu et al. [7] prepared small particles of superfine spherical rhenium powder by vapor deposition. The rhenium powders prepared by the above processes have a uniform particle size distribution and large specific surface area, and the pure rhenium materials prepared from these rhenium powders have high densities. However, due to the complexity of the above processes and high equipment

requirements, none of the above processes have been industrialized. The hydrogen reduction of ammonium perrhenate is a commonly used process to prepare rhenium powder in industry, which has the characteristics of a simple process flow, low production cost, and low equipment requirements [8,9]. The preparation process flow chart of rhenium ingot is shown in Figure 1. However, the rhenium powders that are produced by this preparation technology have an uneven particle size distribution, small specific surface area, and the rhenium ingots produced have poor compactness. Moreover, due to the low efficiency of the mass and heat transfer in the traditional process, the reaction with hydrogen is not complete and the second reduction step at a high temperature is required [10–12].

**Figure 1.** Industrial preparation of rhenium ingots.

In order to optimize the process of preparing rhenium by a high-temperature reduction, we can look into work dealing with the preparation of other metal powders by a high-temperature reduction. In recent years, there have been many studies on the preparation of metal powders by a high-temperature hydrogen reduction [13–21]. For instance, Wang et al. [22] proposed a novel route to synthesize Mo powders via a carbothermic prereduction of molybdenum oxide followed by a reduction by hydrogen; they removed oxygen from the samples by a secondary reduction. Kanga et al. [23] prepared nanosized W and W-Ni powders by applying ball milling and a hydrogen reduction of oxide powders. Gua et al. [24] prepared Mo nanopowders through a hydrogen reduction of a combustion-synthesized foam-like MoO2 precursor. All of the above studies are based on a reducing substance pretreatment, which provides a certain reference experience for this study. However, the hydrogen reduction reaction of ammonium perrhenate is a complicated process. This process not only involves reduction, but also a disproportionation reaction. Colton [25] pointed out that the disproportionation of ReO3 occurred above 300 ◦C to produce Re2O7 and ReO2. This disproportionation reaction is the main reason why ammonium perrhenate cannot be completely reduced to rhenium at a low temperature. There are few reports on the preparation of a high-quality Re powder by a hydrogen reduction at present. Bai et al. [26] reduced volatile rhenium oxide to prepare Re powder. However, due to the high equipment requirements of this process, it cannot be used for industrial production.

#### *Chemical Reaction Considerations*

Ammonium perrhenate decomposes into oxides with different valence states when reduced at a high temperature, and the main oxides are Re2O7, ReO3, and ReO2 [27]. The total equation for the reduction of ammonium perrhenate in hydrogen is represented by Equation (1). Ammonium perrhenate is decomposed by heat to Re2O7, which is gradually recombined with hydrogen and finally reduced to Re, as shown in Equations (2)–(4). ReO3 is very reactive; ReO3 is disproportionated at a high temperature to produce ReO2 and Re2O7, as shown in Equation (5).

$$2\text{NH}\_4\text{ReO}\_4\text{ (s)} + 7\text{H}\_2\text{ (g)} = 2\text{Re (s)} + 8\text{H}\_2\text{O (g)} + 2\text{NH}\_3\text{ (g)}\tag{1}$$

$$\text{Re}\_2\text{O}\_7\text{ (s)} + \text{H}\_2\text{ (g)} = 2\text{Re}\text{O}\_3\text{ (s)} + \text{H}\_2\text{O}\text{ (g)}\tag{2}$$

$$\text{ReO}\_3\text{ (s)} + \text{H}\_2\text{ (g)} = \text{ReO}\_2\text{ (s)} + \text{H}\_2\text{O (g)}\tag{3}$$

$$\text{ReO}\_2\text{ (s)} + 2\text{H}\_2\text{ (g)} = \text{Re (s)} + 2\text{H}\_2\text{O (g)}\tag{4}$$

$$\text{CaReO}\_3\text{ (s)} = \text{Re}\_2\text{O}\_7\text{ (s)} + \text{ReO}\_2\text{ (s)}\tag{5}$$

In the present work, we determined the reduction mechanism of ammonium perrhenate through a differential thermal analysis, and innovatively proposed that the disproportionation reaction in the reduction process was the main reason affecting the complete reduction of ammonium perrhenate. This work also determined an innovative process for reducing the particle size and reduction temperature of rhenium powder, aiming to produce a high-density rhenium ingot, and ammonium perrhenate is completely reduced by hydrogen at a low temperature. This is the technological innovation point of this paper. Moreover, the optimization and innovation of this process is based on the already industrialized hydrogen reduction process to produce rhenium ingot, which makes it easy to realize as an industrialized production process.

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

#### *2.1. Instrument*

Instrument: The following instruments were used herein: RE-2000A rotary evaporator, Qiqiang instrument manufacturing co. LTD, Shanghai, China; SSX-550 scanning electron microscope, Shimadzu, Osaka, Japan; PW3040/60 X-ray diffractometer (XRD), Panalytical Company, Almelo, Netherlands; Escalab250 250 X-ray photoelectron spectroscopy (XPS), Hewlett-Packard Company, Palo alto, CA, USA; VEP223 high-temperature vacuum sintering furnace, Beizhen Vacuum Technology Co. Ltd., Shenyang, China; HYL-1076 laser particle size analyzer, Haoyu Technology Co. Ltd., Dandong, China; the self-developed recrystallization reactor, Institute of metals, Chinese academy of sciences, Shenyang, China; STA PT1600 Differential Thermal Analysis (DTA), Linseis Co. Ltd., Selb, Germany.

#### *2.2. Materials*

Materials: NH4ReO4 (99.99%, Re ≥ 69.4%) from Halin Chemical Co. LTD, Weifang, China.

#### *2.3. Analytical Methods*

XRD detection: The light tube type was a Cu target, ceramic X light tube. λ = 0.15406 nm, scan range was 10–90 degrees, the scanning speed was 2 degrees/min.

DTA detection: The temperature ranged from 300 to 700 ◦C, and the gas atmosphere was nitrogen.

Particle size distribution detection: the test medium was ethanol, the optical model was Mie, and the distribution type was volume distribution.

The parameters of SEM: The electron acceleration voltage was 20.0 KV, the working distance was 21.8 mm, and the magnification was 15,000 times.

XPS analysis parameters: The fitting software was XPSPEAK (XPSPEAK.41, Hewlett-Packard Company, Palo alto, CA, USA). The read base pressure was 2.4 <sup>×</sup> <sup>10</sup>−<sup>8</sup> Pa, utilizing monochromatic Al Kα radiation operating at 1486.6 eV. At the pass energy of 50 eV, with a 0.1 eV step, the high-resolution scans were performed. At the pass energy of 100 eV and a step size of 1 eV, the survey spectra were acquired. The reproducible C (1 s) binding energy of all samples was 284.6 eV and the charge neutralization was achieved using low-energy electrons and argon ions. The spin–orbit splitting was 2.4 eV, and the spin orbital split intensity ratio of 4f7/<sup>2</sup> and 4f5/<sup>2</sup> was 4:3.

#### *2.4. Experimental Procedure*

As illustrated by Figure 2 was the experimental flow chart of a hydrogen reduction of ammonium perrhenate in this study. Ammonium perrhenate (99.99%, Re ≥ 69.4%) was prepared with deionized water into a saturated solution at room temperature, the room temperature was about 25 ◦C. Then the saturated solution of ammonium perrhenate at room temperature was placed in an RE-2000A rotary evaporator, and part of the water was evaporated to form a hot saturated solution at 120 ◦C [28]. The thermally saturated solution of ammonium perrhenate was introduced into the recrystallization condensation reactor; the stirring speed and cooling temperature were adjusted and recrystallized at 5 ◦C for 3 h. Finally, the cooled solid–liquid mixture was filtered and dried to obtain ammonium perrhenate crystals. The SEM diagrams of the recrystallized ammonium perrhenate are shown in Figure 3 [29], and the median diameters (D50) and specific surface area are shown in Table 1. The recrystallization ammonium perrhenate particles (60 g) were reduced with the different temperatures (300–900 ◦C) in the high-temperature vacuum sintering furnace; the hydrogen flow rate was 500 mL/min, and the heating rate was 10 ◦C/min. After 3 h of reduction, the reduction product of ammonium perrhenate was obtained.

**Figure 2.** The experimental working procedure diagram.

**Figure 3.** SEM images of recrystallized ammonium perrhenate particles at different agitation intensities and ammonium perrhenate raw material particles (**a**) ammonium perrhenate raw material particles, (**b**) 100 rpm stirring strength recrystallized particles, and (**c**) 200 rpm stirring strength recrystallized particles) [29].

**Table 1.** Specific surfaces and median diameters (D50s) of the NH4ReO4 particles.


Preparation of rhenium ingots: 20 g rhenium powder was put into the powder press mold (the height was 32.00 mm and the inner diameter of the mould was 16.60 mm) to press. The pressure of the powder press was 74,000 N, the sintering furnace temperature was 2300 ◦C, and the sintering time was 3 h. The theoretical density of rhenium is 21.04 g/cm3.

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

#### *3.1. Reduction Mechanism Analysis*

In order to research the reaction mechanism of ammonium perrhenate in the reduction process, the decomposition products of ammonium perrhenate (Re, ReO2, and ReO3) were analyzed by a differential thermal analysis (DTA), and the differential thermal analysis curve is shown in Figure 4. It can be seen that an obvious endothermic peak appeared at 350 to 400 ◦C. Re is stable at a high temperature, and the decomposition temperature of ReO2 is 700 ◦C. Therefore, the generation of this endothermic peak can only be due to the disproportionation of ReO3. The disproportionation reaction products of ReO3 are ReO2 and Re2O7, and the reaction equation is shown in Equation (5). In the hydrogen reduction process, the ammonium perrhenate is firstly decomposed into Re2O7, the Re2O7 reacts with hydrogen to form ReO3, and ReO3 reacts with hydrogen to form ReO2 until they are reduced to Re. In the process of a hydrogen reduction of ammonium perrhenate, if the disproportionation reaction and reduction reaction exist simultaneously, the disproportionation reaction will be the main reason affecting the complete reduction of ammonium perrhenate at a low temperature.

**Figure 4.** The differential thermal analysis curve of Re, ReO2, and ReO3.

In order to research the effect of disproportionation on the hydrogen reduction of ammonium perrhenate, the hydrogen reduction experiments of ammonium perrhenate were carried out at the same reduction time and at different reduction temperatures. The reducing substance was recrystallized ammonium perrhenate at a 200 rpm stirring strength (D50 was 71.17 μm, specific surface was 26.93 m2/kg). The XRD patterns of the reduction products of ammonium perrhenate at different temperatures are shown in Figure 5. The characteristic peaks of the reduction products were complex at lower temperatures (300–600 ◦C), and the diffraction peaks of the reduction products indicated Re and ReO2, and ReO3. The characteristic peaks of Re did not change obviously from 400 to 600 ◦C. However, in the range of 300 to 400 ◦C, the characteristic peaks of ReO3 were enhanced. In the range of 400 to 600 ◦C, the characteristic peaks of ReO2 were enhanced, while that of ReO3 were weakened. When the temperature reached 700 ◦C, the characteristic peaks of the reduction products were Re, and other crystal peaks were not observed. This result suggested that the contents of ReO3 in the reduction products increased in the range of 300 to 400 ◦C. However, within the temperature range of 400 to

600 ◦C, the content of Re in the reduction products did not increase significantly, while the content of ReO2 increased significantly.

**Figure 5.** XRD patterns of the reduction products of ammonium perrhenate (D50 was 71.17 μm, specific surface was 26.93 m2/kg) at different temperatures. Figure (**a**) was the XRD diffraction pattern of the reduced product at 300 to 400 ◦C; Figure (**b**) was the XRD diffraction pattern of the reduced product at 500 to 700 ◦C.

In order to further clarify the influences of reduction temperatures on an ammonium perrhenate hydrogen reduction, X-ray photoelectron spectroscopy was used for the rhenium atomic quantitative analysis in different valence states. According to the references [30–32], a rhenium atom has a split energy level (f), where the spin–orbit splitting was 2.4 eV, and the spin orbital split intensity ratio of Re 4f7/<sup>2</sup> and Re 4f5/<sup>2</sup> was 4:3. The background was a mixed Shirley background (Shirley + straight line) [33], and the slope of the line was eight. The peak positions and fit paramters for all samples are given in Table 2. The X-ray photoelectron spectroscopy of the reduced products at the different reduction temperatures are shown in Figure A1.


**Table 2.** The peak positions and fit paramters for Re 4f7/<sup>2</sup> and Re 4f5/2.

The X-ray photoelectron spectroscopy of the reduced products at different reduction temperatures are shown in Figure A1. According to the peak areas of the different valence states of rhenium, the content of each valence state was calculated, as shown in Table 2. The percent of rhenium atom in each valence state was calculated, as shown in Table 3.

**Table 3.** The percent of rhenium atom in each valence state of the reduction products of ammonium perrhenate (D50 was 71.17 μm, specific surface was 26.93 m2/kg) at different temperatures.


According to the results in Table 3, the percents of rhenium atom in each valence state of the reduced products at different temperatures were plotted, as shown in Figure 6. Combined with Table 2 and Figure 6, it can be concluded that the percent of Re6<sup>+</sup> in the reduction products increased in the range of 300 to 400 ◦C, and within the temperature range of 400 to 600 ◦C, the percent of Re in the reduction products did not increase significantly, while the content of Re4<sup>+</sup> increased significantly. These results showed that the ReO3 content in the reduction products increased in the temperature range of 300–400 ◦C and decreased with the increase in the reduction temperature; the Re content decreased in the temperature range of 300–400 ◦C and increased with the increase in the reduction temperature. In order to explain this rule, the following conclusions were obtained by combining the XRD and DTA detection results: the disproportionation reaction of ReO3 occurred between 350 and 400 ◦C, ReO3 decomposed into ReO2 and Re2O7, and Re2O7 was converted to ReO3 by a hydrogen reduction. Therefore, in the temperature range of 350 to 400 ◦C, the effect of disproportionation hindered the normal reduction process, and this is why ammonium perrhenate cannot be reduced to more rhenium at 350 to 400 ◦C in the same reduction time. This phenomenon shows that the disproportionation of ReO3 hindered the normal reduction reaction in a certain temperature range, but with the increase in the temperature, the reduction reaction rate gradually increased, and the content of rhenium in the reduction products also increased. At the reduction temperature of 300 ◦C, the disproportionation of ReO3 did not occur, and only the reduction reaction was carried out. Therefore, the content of Re in the reduction products was higher. As discussed above, in the process of a hydrogen reduction of ammonium perrhenate, the disproportionation of ReO3 decreased the rate of the reduction reaction, and the increase in the temperature can increase the reduction reaction rate and reduce the effect of disproportionation on the reduction process. Increasing the reduction time and lowering the reduction temperature below the temperature at which the disproportionation of ReO3 occurs can also increase the content of Re in the reduction products.

**Figure 6.** Oxidation state atomic content changes of ammonium perrhenate (D50 was 71.17 μm, specific surface was 26.93 m2/kg) hydrogen reduction reaction at different temperatures.

#### *3.2. Influence of Particle Size on Reduction*

In the solid-state reaction system, the reaction rate is related not only to the temperature but also to the diffusion rate of the reactant [34]. The particle size of the reactant decreases, which is conducive to increasing the diffusion rate [35]. In order to research the effects of particle sizes of ammonium perrhenate on the reduction effect, the recrystallized ammonium perrhenate (D50 81.05 μm) was used for the reduction experiments under the same operating conditions. The XRD patterns of the reduction products of ammonium perrhenate at different temperatures are shown in Figure 7. The characteristic peaks of the reduction products were complex at lower temperatures (300–700 ◦C), and the diffraction peaks of the reduction products indicated Re and ReO2, and ReO3. The characteristic peaks of Re did not change obviously from 500 to 800 ◦C. However, in the range of 300 to 400 ◦C, the characteristic peaks of ReO3 were enhanced. In the range of 500 to 800 ◦C, the characteristic peaks of ReO2 were enhanced, while that of ReO3 were weakened. When the temperature reached 900 ◦C, the characteristic peaks of the reduction products were Re and ReO2, and other crystal peaks were not observed. This result suggested that the contents of ReO3 in the reduction products increased in the range of 300 to 400 ◦C. However, within the temperature range of 500 to 800 ◦C, the content of Re in the reduction products did not increase significantly, while the content of ReO2 increased significantly.

**Figure 7.** XRD patterns of the reduction products of ammonium perrhenate (D50 was 81.05 μm, specific surface was 21.72 m2/kg) at different temperatures. Figure (**a**) was the XRD diffraction pattern of the reduced product at 300 to 400 ◦C; Figure (**b**) was the XRD diffraction pattern of the reduced product at 500 to 900 ◦C.

In order to further clarify the influence of particle sizes on an ammonium perrhenate hydrogen reduction, X-ray photoelectron spectroscopy was used to quantitatively analyze this feature. The X-ray photoelectron spectroscopy of the reduced products at different reduction temperatures are shown in Figure A2. The percent of rhenium atom in each valence state was calculated, as shown in Table 4. According to the results in Table 4, the percents of rhenium atom in each valence state of the reduced products at different temperatures were plotted, as shown in Figure 8. Combined with Table 4 and Figure 8, it can be concluded that the content of Re in the reduction products increased with the decrease of ReO3 between 350 and 800 ◦C. At the reduction temperature of 300 ◦C, the disproportionation of rhenium trioxide did not occur and the content of rhenium was higher. The composition law of the reduction products was the same as that of previous reduction experiments, which proved the accuracy of the reduction mechanism analysis. When the reduction temperature reached 800 ◦C, rhenium trioxide still existd. It can be concluded that the disproportionation of ReO3 existed in the range of 350 to 800 ◦C. Compared with the previous reduction experiment, the temperature range at which the disproportionation reaction occurs had increased. This was because the particle size of ammonium perrhenate increased, the solid reaction diffusion rate decreased, and the contact between the reactant and hydrogen was not ideal. Therefore, under the same operating conditions, reducing the particle size of ammonium perrhenate can increase the hydrogen reduction diffusion rate and reduce the influence of the disproportionation on the reduction process.


**Figure 8.** Oxidation state atomic content changes of ammonium perrhenate (D50 was 81.05 μm, specific surface was 21.72 m2/kg) hydrogen reduction reaction at different temperatures.

#### *3.3. Analysis of the Products of the Rhenium Ingots*

The Re powders were prepared by unrecrystallized ammonium perrhenate and recrystallized ammonium perrhenate with a different D50, and the rhenium ingots were prepared from these rhenium powders. The relevant physical parameters of the products are shown in Table 5. The particle size distribution of the rhenium powder is shown in Figure 9. It can be seen from Table 5 and Figure 9 that the rhenium powder prepared by ammonium perrhenate with a small particle size is also of relatively small particle size; the particle size distribution uniformity of rhenium powder prepared by recrystallized ammonium perrhenate was improved obviously, and the rhenium ingot prepared has a higher density. The theoretical density of rhenium is 21.04 g/cm3, and the density of the rhenium ingot prepared by D50 71.17 μm ammonium perrhenate reached 20.106 g/cm3. The rhenium ingot prepared under this condition was close to the theoretical density. The SEM images of the surface defects of the rhenium ingots are shown in Figure 10. It can be seen that the rhenium ingot prepared by the small particle size rhenium powder not only has a high density but also has a small surface hole defect. Therefore, reducing the particle size of the rhenium powder is a key factor in preparing a high-density rhenium ingot.

**Table 5.** The rhenium materials prepared from ammonium perrhenate of different particle sizes.


51


**Figure 9.** The particle size distribution of rhenium powder prepared by ammonium perrhenate with different particle sizes ((**a**) prepared by 81.05 μm D50 ammonium perrhenate, (**b**) prepared by 71.17 μm D50 ammonium perrhenate, (**c**) prepared by 123.90 μm D50 ammonium perrhenate).

**Figure 10.** SEM images of the surface defects of the rhenium ingots ((**a**) prepared by 52.15 μm D50 Re powders, (**b**) prepared by 34.04 μm D50 Re powders, (**c**) prepared by 19.74 μm D50 Re powders).

#### *3.4. Proposed Flow Sheet*

Based on the experimental results, the flow sheet for the processing of rhenium ingots by ammonium perrhenate was tentatively suggested, which is shown in Figure 11. The ammonium perrhenate particles were refined by homogeneous recrystallization, and the D50 of the ammonium perrhenate particles was refined from 123.90 to 71.17 μm. The recrystallized ammonium perrhenate was completely reduced by hydrogen at 700 ◦C for 3 h, and the D50 of 19.74 μm rhenium powder was obtained. The density of the rhenium ingots pressed by these rhenium powders was 20.106 g/cm3. The theoretical density of rhenium is 21.04 g/cm3, and a rhenium ingot that reaches the theoretical density of more than 90% is a high-quality product. In this study, the rhenium ingot density was 95.56% of the theoretical density, which reached the high-quality product standard. The optimized production process of rhenium ingot not only realized a low-temperature reduction, but also increased the density of the rhenium ingot, which can provide a theoretical basis and practical experience for industrial production.

**Figure 11.** Proposed flow sheet for the preparation of high-density rhenium ingot by homogeneous recrystallization of ammonium perrhenate by a hydrogen reduction.

#### **4. Conclusions**

In this study, the influences of disproportionation on the hydrogen reduction of ammonium perrhenate were investigated and the following conclusions were drawn:

(1) In the process of a hydrogen reduction of ammonium perrhenate, the disproportionation of ReO3 decreased the rate of the reduction reaction, and the increase in the reduction temperature was conducive to increasing the reduction reaction rate and reducing the influence of disproportionation on the reduction process.

(2) At the same reduction temperature, reducing the particle sizes of ammonium perrhenate was conducive to increasing the hydrogen reduction rate and reducing the influence of the disproportionation on the reduction process.

(3) The rhenium ingot prepared by the small particle size rhenium powder not only has a high density but also has a small surface hole defect. Therefore, reducing the particle size of the rhenium powder is a key factor in preparing high-density rhenium ingot.

(4) It is feasible to increase the density of rhenium ingot by reducing the particle size of the rhenium powder. The particle size of ammonium perrhenate was reduced to a rhenium powder with a D50 of 19.74 μm and a specific surface area of 163.70 m2/kg, which was pressed into a rhenium ingot with a density of 20.106 g/cm3, close to the theoretical density of rhenium.

**Author Contributions:** J.T.: Writing—original draft; Data curation; Formal analysis; Investigation; Funding acquisition. Y.S.: Data curation; Formal analysis; Funding acquisition; Resources. C.Z.: Data curation; Software. L.W.: Resources; Writing—review and editing. Y.Z.: Investigation; Supervision; Funding acquisition; Resources. D.F.: Writing—review and editing; Resources. Y.L.: Writing—review and editing; Funding acquisition. All authors have read and agreed to the published version of the manuscript.

**Funding:** Liaoning provincial department of science and technology doctoral research initiation fund, grant number 2019-BS-130; Open project fund of key laboratory of ecological metallurgy of polymetallic symbiosis in ministry of education, Northeastern University, grant number NEMM2019003.

**Acknowledgments:** School of Biomedical & Chemical Engineering, Liaoning Institute of Science and Technology and the analysis and test center of institute of metals, Chinese academy of sciences undertook the sample test for this study. The authors are grateful for these supports.

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

#### **Appendix A**

**Figure A1.** X-ray photoelectron spectroscopy of the reduction products of ammonium perrhenate (D50 was 71.17 μm, specific surface was 26.93 m2/kg) at different temperatures.

**Figure A2.** X-ray photoelectron spectroscopy of the reduction products of ammonium perrhenate (D50 was 81.05 μm, specific surface was 21.72 m2/kg) at different temperatures.

#### **References**


© 2020 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* **Electrochemical Deposition of Al-Ti Alloys from Equimolar AlCl3** + **NaCl Containing Electrochemically Dissolved Titanium**

**Vesna S. Cvetkovi´c 1,\*, Nataša M. Vuki´cevi´c 1, Ksenija Mili´cevi´c-Neumann 2, Sre´cko Stopi´c 2, Bernd Friedrich <sup>2</sup> and Jovan N. Jovi´cevi´c <sup>1</sup>**


Received: 14 November 2019; Accepted: 31 December 2019; Published: 4 January 2020

**Abstract:** Al-Ti alloys were electrodeposited from equimolar chloroaluminate molten salts containing up to 0.1 M of titanium ions, which were added to the electrolyte by potentiostatic dissolution of metallic Ti. Titanium dissolution and titanium and aluminium deposition were investigated by linear sweep voltammetry and chronoamperometry at 200 and 300 ◦C. Working electrodes used were titanium and glassy carbon. The voltammograms on Ti obtained in the electrolyte without added Ti ions indicated titanium deposition and dissolution proceeding in three reversible steps: Ti4<sup>+</sup> - Ti3<sup>+</sup>, Ti3<sup>+</sup> - Ti2<sup>+</sup> and Ti2<sup>+</sup> - Ti. The voltammograms recorded with glassy carbon in the electrolyte containing added titanium ions did not always clearly register all of the three processes. However, peak currents, which were characteristics of Al, Ti and Al-Ti alloy deposition and dissolution, were evident in voltammograms on both working electrodes used. A constant potential electrodeposition regime was used to obtain deposits on the glassy carbon working electrode. The obtained deposits were characterized by SEM, energy-dispersive spectrometry and XRD. In the deposits on the glassy carbon electrode, the analysis identified an Al and AlTi3 alloy formed at 200 ◦C and an Al2Ti and Al3Ti alloy obtained at 300 ◦C.

**Keywords:** Al-Ti alloy; electrochemical co-deposition; chloroaluminate melt; XRD

#### **1. Introduction**

Intermetallic materials based on a combination of aluminium and titanium, which possess high specific strength and low weight and required stiffness and excellent oxidation resistance at elevated temperatures (particularly over 600 ◦C), are of increasing importance as new structural materials in aerospace industry and medicine [1–3]. Due to its ability to increase the temperature of titanium allotropic transformation, aluminum is the main alloying element for titanium. The density of aluminum is less than the density of titanium, so the addition of aluminum increases the specific strength of Ti alloys. High functional properties make the Ti-Al system the foundation of many titanium alloys. The presence of thermodynamically stable intermetallic phases in titanium-aluminum composite materials allows for and significantly enhances physical and mechanical characteristics of these systems [4].

Over the last thirty years, various processing methods have been studied to fabricate these intermetallic materials [3]. The most prominent methods for fabricating Al-Ti alloys are rapid solidification [2], sputter deposition [5], ball milling [6], mechanical alloying [7], spark plasma sintering [8], reaction sintering of elemental powders, etc. [3].

In general, Al-Ti alloys could be electrodeposited from electrolytes containing Ti(II) species [2,3,9–11]. Electrodeposition synthesis of aluminium-titanium-based materials is a tempting process that shows the potential to replace processing methods identified earlier [2,7]. The fundamental aspects of chemistry and electrochemistry of titanium ions in molten salt electrolytes have been investigated, but data on the electrochemical behaviour of titanium ions in molten chloride/fluoride salt electrolytes are scarce and contradictory. The main barrier for successful development of an electrochemical route for aluminium-titanium alloy production is associated with the existence of different oxidation states of dissolved titanium species, namely Ti(II), Ti(III) and Ti(IV) [9,12–15].

Electrochemical deposition of aluminium-titanium intermetallics has been investigated from either an Lewis acidic chloroaluminate molten salts electrolyte made of 2:1 AlCl3-NaCl or AlCl3-1-ethyl-3-methylimidazolium chloride ionic liquid (IL) [1–3,7,10]. In Lewis acidic 2:1 AlCl3-NaCl electrolyte systems, authors particularly studied the influence of Ti2<sup>+</sup> concentration on the alloy composition and found that, with low Ti2<sup>+</sup> concentrations, alloy composition depended on current density. For example, an Al3Ti alloy containing 25% atomic fraction of titanium was deposited only at low current densities [1]. With an increase in current density, the titanium content in the alloys decreased [2]. The concentration limit of titanium in the alloy composition was proposed to be due to a mechanism, by which an Al-Ti alloy forms through the reductive decomposition of a divalent species—Ti(AlCl4)2. The electrochemical reduction of Ti2<sup>+</sup> ions to metallic Ti was not observed at potentials more positive than that required for aluminium deposition, but an Al3Ti alloy was deposited onto a copper working substrate under specific deposition conditions [2].

In comparison to other molten salt electrolytes systems, the electrochemical behaviour of titanium ions in chloride melts is different because of the stability of various oxidation states of titanium ions, which is caused by the influence of electrolyte composition [12–15] and temperature [14,15].

However, to our knowledge, there is no information published that addresses electrodeposition of Al-Ti alloys from an equimolar chloroaluminate AlCl3-NaCl molten electrolyte on glassy carbon (GC) at temperatures below 300 ◦C. Equimolar AlCl3-NaCl electrolytes have been characterised in the following ways: (a) lower vapour pressure above an equimolar melt at the same temperature applied than on an acidic AlCl3-NaCl electrolyte [16,17]; (b) the aluminium deposition potential from AlCl− <sup>4</sup> ions in an equimolar and acidic AlCl3-NaCl electrolyte was more negative than the aluminium deposition potential from Al2Cl<sup>−</sup> <sup>7</sup> ions in an acidic melt, which provided a larger potential distance to the titanium deposition potential [18,19]; (c) the deposition current density of aluminium was greater for the reaction: AlCl− <sup>4</sup> + 3e<sup>−</sup> → Al + 4Cl<sup>−</sup> , than for the reaction: 4Al2Cl<sup>−</sup> <sup>7</sup> + 3e<sup>−</sup> → Al + 7AlCl<sup>−</sup> 4 for the same value of an overpotential (exceeding −60 to −80 mV) recorded with the same AlCl3 concentration in the melt [1,3,18,19].

The aim of the present paper is to study titanium and aluminium co-deposition from an equimolar chloroaluminate molten salt containing Ti ions introduced by electrochemically dissolved Ti metal. This novel electrodeposition route consisting of anodic dissolution of Ti and co-deposition of Ti and Al may be a useful route for Al-Ti alloy production.

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

An equimolar mixture of AlCl3 and NaCl served as a base electrolyte [20], and preparation of the electrolyte was identical to those described in previous articles [21,22].

Electrochemical measurements and electrodeposition processes were carried out at 200 and 300 ◦C in a three-electrode electrochemical cell. In the cell used for titanium ion introduction into the equimolar AlCl3-NaCl molten salt, the working electrode (WE, an anode) was a titanium plate (Ti 99.99% Alfa Aesar, Haverhill, MA, USA), the counter electrode was titanium and the reference electrode was an aluminium rod with a diameter of 3 mm (Al 99.999% Haverhill, MA, USA). In the cells used for aluminium and titanium deposition and co-deposition, the cathode was a glassy carbon (GC, Alfa Aesar, Haverhill, MA, USA) cylinder, a titanium plate was used as the counter electrode and the reference electrode was an aluminium rod with a diameter of 3 mm.

All the reported potentials of WEs in this work were measured relative to the equilibrium potential of the aluminium reference electrode in the melt used under given conditions.

Before the experiment, the GC WE was polished with 0.05 μm alumina powder (Merck & Co., Kenilworth, NJ, USA) and cleaned several times by sonication in ethanol and Milli-Q water with each duration of 3 min.

The aluminium electrodes were etched in solutions made of 50 vol% HF, 15 vol% H2O, 25 vol% ammonia solution and 5 vol% H2O2. Thereafter, the electrodes were rinsed with deionised water and absolute ethyl alcohol and dried before use.

The Ti WEs were etched in a mixture of HF + H2O + H2O2 (volume ratio: 1:20:1), rinsed with distilled water and dried before use.

Argon flow was maintained in the cell, and the electrolyte was not stirred during experiments. After the electrodes were introduced into the electrolyte, the system was left for 5 to 10 min to achieve thermal equilibrium.

The study started with recording potentiodynamic polarization curves for the Ti WE in an equimolar AlCl3-NaCl melt without previously added titanium at 200 or 300 ◦C. The potential (measured relative to the aluminium reference electrode) was scanned from a starting value, *EI* = 0.0 V, to a final value, *EF* <sup>=</sup> 1.200 V, with a scan rate of 1 mV·s<sup>−</sup>1.

In linear sweep voltammetry (LSV) experiments when Ti was used as a WE, and an electrolyte without previously added titanium, the potential was changed with a scan rate of 20 mV s−<sup>1</sup> from a potential slightly more negative than the open-circuit potential (OCP) of Ti to a different cathodic end potential (EC), then back to anodic potential (EA) and finally to the starting potential. In order to examine the anodic part of the voltammograms, LSV experiments were performed, starting from the open-circuit potential to the final anodic end potential (EA), and back to slightly more negative potential than the OCP.

Ti ions were introduced into the electrolyte by electrochemical dissolution of titanium metal with constant potentials: at 200 ◦C, the potential was maintained at 0.500 V; and at 300 ◦C, the potential was maintained at 0.450 V.

The voltammograms obtained on the GC WE in the equimolar chloroaluminate molten salts with Ti ions present started from a potential *EI*, usually 0.050 V more negative than the GC OCP (measured against the aluminium reference electrode), changed to a cathodic potential limit, *EF*, and back to *EI* with various sweep rates.

Controlled electrodeposition onto the GC electrode in the electrolyte with previously added titanium ions was initiated 5 min. after insertion of the WE into the melt in order to allow for thermal equilibrium. Titanium and aluminium were electrodeposited at a constant overpotential at two different temperatures (200 and 300 ◦C). After potentiostatic deposition, the WE was taken out of the cell, washed thoroughly with absolute ethanol (Zorka-Pharma, Šabac, Serbia) in order to remove any melt residue and dried in a desiccator furnished with silica gel. The morphology and the composition of the samples deposited were explored by a scanning electron microscope (VEGA 3 model; TESCAN, Brno, the Czech Republic), equipped with an energy-dispersive spectrometer (Oxford INCA 3.2, Oxford Instruments, High Wycombe, UK) and an optical microscope (VH-Z100R model; Keyence, Osaka, Osaka Prefecture, Japan).

The deposit collected from the GC WE obtained at 200 ◦C was analyzed by XRD on a Philips PW1050 powder diffractometer at room temperature with Ni-filtered Cu Kα radiation (λ = 1.54178 Å) and a scintillation detector within a 2θ range of 20–85◦ in steps of 0.05◦ and a scanning time of 5 s per step, and the deposit was obtained at 300 ◦C by SmartLab® X-ray diffractometer (Rigaku Co., Tokyo, Japan) using Cu Kα radiation (λ = 1.542 Å). The patterns were collected within a 2θ range of 10–90◦ at a scan rate of 0.5◦/min with a divergent slit of 0.5 mm, operated at 40 kV and 30 mA. The phases

formed during the deposition were identified by a comparison of the recorded diffraction peaks with the references from the Joint Committee on Powder Diffraction Standards (JCPDS) database.

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

#### *3.1. Dissolution of Titanium*

The composition of a solvent-fused salt has a dramatic influence on electrodeposition process of titanium [14,15]. The published works on melts used for titanium deposition (inorganic and organic melts and ILs) emphasize problems encountered with the control of electrolytes made by titanium salts dissolution. These involve titanium salt sublimation at elevated temperatures, titanium ions unwanted disproportionation and titanium oxide deposition onto electrodes including passivation. To avoid most of the mentioned problems, it was decided to introduce titanium into an equimolar AlCl3-NaCl melt by electrochemically controlled dissolution. Figure 1 exhibits voltammograms obtained with a titanium electrode in an equimolar AlCl3-NaCl melt at 200 ◦C.

**Figure 1.** Voltammograms with a Ti working electrode in an equimolar AlCl3-NaCl melt (*T* = 200 ◦C, *v* = 20 mV·s<sup>−</sup>1): (**a**) potential changes during the first cycle from a starting point (*EI* = 0.250 V) to an anodic potential limit (*EA* = 1.100 V) and then back to 0.250 V and potential changes during the second and third cycles starting from *EI* = 0.250 V to different cathodic potential limits EC, then to EA and finally to a value *EF* = 0.250 V; (**b**) potential changes during cycles from a starting point (*EI* = 1.000 V) to different cathodic potential limits and back to a final value (*EF* = 1.000 V).

It was found that in the presence of aluminium ions in the equimolar chloroaluminate molten salt electrolyte, the electrochemical reduction of titanium ions to metallic titanium was complicated by the formation of intermediate oxidation states of Ti4<sup>+</sup>, Ti3<sup>+</sup> and Ti2<sup>+</sup> [1,3]. These were recorded as cathodic peaks IC (Ti<sup>4</sup>+/Ti3<sup>+</sup>), IIC (Ti<sup>3</sup>+/Ti2<sup>+</sup>) and IIIC (Ti<sup>2</sup>+/Ti) and their respective anodic counterparts IA, IIA and IIIA, shown in Figure 1. These observations were similar to the results reported on Pt and Ti electrodes from an AlCl3 + N-(n-butyl)pyridinium chloride (mole ratio: 2:1) melt at 25 ◦C [23]. The potentials related to these processes can be read from voltammograms. Their values greatly depend on temperature, the composition of electrolytes and concentrations (amounts) of dissolved titanium ions in the melt [1,3,10,13–15,24]. In the equimolar AlCl3-NaCl melt used, the average recorded values of the reversible potential for the pairs Ti<sup>4</sup>+/Ti3<sup>+</sup>, Ti3+/Ti2<sup>+</sup> and Ti<sup>2</sup>+/Ti were approximately 0.410, 0.190 and 0.149 V. However, they can be identified also from the potentiodynamic polarization curve of titanium recorded in the used melt at 200 ◦C (Figure 2). It is apparent that the potentials designated as *ETi*2+/*Ti* ≈ 0.200 V, *ETi*3+/*Ti*2<sup>+</sup> ≈ 0.240 V and *ETi*4+/*Ti*3<sup>+</sup> ≈ 0.370 V are in reasonably good agreement with the reversible potentials determined by the peak pairs IC/IA, IIC/IIA and IIIC/IIIA from the voltammograms in Figure 1.

**Figure 2.** Potentiodynamic polarization curve of the Ti working electrode in the equimolar AlCl3-NaCl melt at *T* = 200 ◦C, *EI* = 0.0 V and *EF* = 1.200 V.

Two important features of titanium in the equimolar AlCl3-NaCl melt at 200 ◦C should be mentioned:

(1) The changes of the peak shape that the reaction (IC/IA) exhibits when recorded with different sweep rates are presented in Figure 3. It was proposed [12–15] that, in all alkali chloride melts, this pair reflects redox reaction Ti<sup>4</sup>+/Ti3+. Using the analysis of the relationship between the peak maximum current densities (for both cathodic and anodic currents shown in Figure 3) and the square root of a scan rate used, it was found that the relationship is linear, which confirmed that the process is a simple diffusion-controlled reversible process [25]. The positions of the other cathodic and anodic peak currents on the voltammograms (namely Ti<sup>3</sup>+/Ti2<sup>+</sup> and Ti<sup>2</sup>+/Ti) show the reversibility of the process as well. However, they were not defined well enough in order to conduct the same analysis.

(2) The reversible potential of Ti2<sup>+</sup> - Ti in the equimolar AlCl3 -NaCl melt was recorded at ≈ 0.200 V, which is a potential positive to that required for aluminium deposition. This finding was similar to the findings observed in different electrolytes [1,3].

**Figure 3.** (**a**) Voltammgrams of the Ti working electrode in the equimolar AlCl3-NaCl melt at T = 200 ◦C with different sweep rates; (**b**) plots of anodic and cathodic peak current densities vs. square root of scan rate calculated from (**a**).

In most molten chloride/fluoride electrolytes, there are equilibria between metallic titanium and Ti2+, Ti3<sup>+</sup> and Ti4<sup>+</sup> ions [12–15]. According to some studies in chloride electrolyte systems, metallic titanium is usually in equilibrium with two different titanium species Ti2<sup>+</sup> and Ti3<sup>+</sup> [9,14]. The presence of different oxidation states of titanium ions in molten chloride electrolytes and the tendency for reoxidation or disproportionation reactions mostly cause poor current efficiency and deposited product quality [9,15]. Furthermore, the melt temperature has a significant influence on equilibrium and electrodeposition processes of titanium in aforementioned electrolytes [14,15].

Taking into account the voltammograms obtained in the system used (Figure 1), the electrodissolution of titanium was done potentiostatically at 0.500 V and 200 ◦C and at 0.450 V and 300 ◦C (Figure 4). The chosen anodic potentials were sufficient enough compared to the reversible potential of the Ti3+/Ti4<sup>+</sup> redox couple to sustain titanium dissolution at a current density of around 1 mA·cm–2. The alternate rise and drop of the dissolution current recorded in Figure 4 is due to processes of dissolution-precipitation reactions on the electrode surface at a working potential applied. Similar effects at potentials, which were anodic but close to the Ti2+/Ti3<sup>+</sup> reversible potential, were addressed in the literature [1,3] and were attributed to the precipitation of a Ti3<sup>+</sup> product, while the breakdown of a passive film was positioned at potentials, where Ti4<sup>+</sup> species generation started, i.e., at potentials close to the Ti3+/Ti4<sup>+</sup> reversible potential.

**Figure 4.** Anodic dissolution of the Ti working electrode at 0.500 V for 3.6 h in the equimolar AlCl3–NaCl melt at *T* = 200 ◦C.

Thus, a melt that was equimolar in AlCl3 and NaCl and contained ≈ 0.1 M titanium was prepared to be used in experiments involving titanium and aluminium electrodeposition on GC. The Ti molarity was calculated from the Ti anode mass lost during controlled potentiostatic dissolution (Faraday's law applied to Ti <sup>→</sup> Ti2<sup>+</sup>) and from the slopes in Figure 3, following the procedure proposed in a similar system using Randles-Sevcik equation [3,26]:

$$\mathbf{i}\_{\mathbb{P}} = 0.4463 \frac{\mathbf{F}^3}{\mathbf{R} \mathbf{T}} \Big| \stackrel{\scriptstyle \mathbb{P}}{\text{n}^\natural} \, \text{AD}\_0^{\frac{1}{2}} \, \text{C}\_0^\* \, \upsilon^{\frac{1}{2}} \, \tag{1}$$

where *ip* is the peak current in amperes; *<sup>v</sup>* is the sweep rate in V·s<sup>−</sup>1; *<sup>C</sup>*<sup>∗</sup> <sup>0</sup> is the concentration in mol·cm<sup>−</sup>3, *D*<sup>0</sup> is the diffusion coefficient in cm2·s−1; *A* is the area of an electrode in cm2, *n* is the number of electrons, *F* is the Faraday's constant; *R* is the gas constant; *T* is the temperature. Both methods showed that the concentration of titanium in the melt used was around 0.1 M. A titanium anode was used to replace (by its dissolution) Ti ions reduced to titanium metal from the electrolyte. Thus, the Ti ions concentration was kept close to a wanted value during experiments.

#### *3.2. Deposition of Titanium and Aluminium onto GC*

The LSV results of the GC WE in the equimolar AlCl3-NaCl melt used with anodically dissolved titanium and recorded at 200 and 300 ◦C are presented in Figures 5 and 6. The obtained voltammograms were very similar to those obtained on W, Cu, mild steel and Pt in chloride and fluoride/chloride inorganic melts and ILs with different titanium concentrations and working temperatures [1,3,10,15,23].

**Figure 5.** Voltammograms of the glassy carbon (GC) working electrode in the equimolar AlCl3-NaCl melt containing anodically dissolved Ti, obtained at different cathodic potential limits and at *T* = 200 ◦C with different sweep rates: (**a**) <sup>ν</sup> <sup>=</sup> 5 mV·s<sup>−</sup>1; (**b**) <sup>ν</sup> <sup>=</sup> 20 mV·s<sup>−</sup>1; (**c**) <sup>ν</sup> <sup>=</sup> 5 mV·s<sup>−</sup>1.

**Figure 6.** Voltammograms obtained on the GC working electrode in the equimolar AlCl3-NaCl melt containing anodically dissolved Ti at *T* = 300 ◦C on different conditions: (**a**) potential change: *EI* = 1.000 V to *EF* = 0.0 V with different sweep rates; (**b**) at different cathodic potential limits at a constant sweep rate <sup>ν</sup> of 50 mV·s<sup>−</sup>1.

The voltammograms in Figures 5 and 6 do not exhibit well-defined cathodic and anodic sides of the voltammograms, as was the case with the titanium WE in the same electrolyte with a much lower titanium concentration (Figure 1). At a lower temperature (Figure 5), the cathodic side of the voltammogram was better defined than its anodic counterpart and tentatively suggests three current increases reflecting all three steps of Ti4<sup>+</sup> ions being reduced to Ti metal at potentials more positive than the aluminium reversible potential. It appears that the peak potentials, although not always easily identified, approached the values recorded for the same reactions on the Ti WE (see Figure 1). At a higher temperature, the peak current density structures were recognizable for both the cathodic and anodic sides of the voltammograms. However, the anodic side of the voltammograms was better defined, showing all three expected oxidation peaks after the applied cathodic potential limit was made more negative than −0.020 V.

When the cathodic potential limit was pushed further to an aluminium overpotential region, pronounced cathodic currents were recorded in the electrolyte with 0.1 M of titanium ions, independent of the temperature applied. The anodic response to the entrance into the aluminium overpotential region showed peaks, suggesting dissolution of Ti and Al and most probably dissolution of an Al-Ti alloy. The charge under the curve of the corresponding anodic peaks, however, did not always equal those under the curves of the cathodic counterparts.

The data obtained by LSV were used to define the potentials needed for the electrodeposition of an Al-Ti alloy, which was the primary goal of this work. The chronoamperometic response in the form of *i* = f(*t*) to an overpotential of −0.085 V applied to the GC WE for two hours at 200 ◦C in the equimolar AlCl3-NaCl melt containing 0.1 M of titanium ions is presented in Figure 7a. When the falling part of the transient in Figure 7a was transformed into the form of *i* = f(*t* <sup>−</sup>1/2), a linear relationship became obvious [25], shown in Figure 7b, suggesting that after approximately 400 s, the Al and Ti deposition was proceeded under diffusion control. With all other conditions being the same, the chronopotentiograms recorded at 300 ◦C on the GC electrode were very similar in shape, with deposition current densities being about two times greater than the same potential applied at 200 ◦C.

**Figure 7.** Potentiostatic deposition on the GC working electrode from the equimolar AlCl3-NaCl melt containing anodically dissolved Ti at −0.085 V and 200 ◦C for two hours: (**a**) current–time transient of the deposition; (**b**) current as a function of *t* <sup>−</sup>1/<sup>2</sup> for the falling part of the transient in (a); (**c**,**d**) SEM photographs of the deposit obtained with energy-dispersive spectroscopy (EDS) results embedded.

For both temperatures applied, thick but nonuniform deposits were obtained (Figure 7c,d and Figure 8). At a higher magnification, grains of different sizes similar to those obtained from AlCl3-BMIC ILs on mild steel published recently [27] can be observed. The energy-dispersive spectroscopy (EDS) analysis made from a larger portion of the same deposits (approximately 400 μm2) reported 36.7 wt. % of Al and 20.4 wt. % of Ti (Figure 7c). In Figure 7d, a result of EDS analysis for one of the larger grains in the deposit is presented numerically (in the inserted circle), and it suggests presence of 50.2 wt. % of Al and 33.2 wt. % of Ti. The deposits obtained at 300 ◦C and −0.020 V after two hours showed a larger average grain size than the deposit obtained after two hours at −0.085 V and 200 ◦C (Figure 8).

**Figure 8.** Optical microscopy image of the deposit obtained on the GC electrode from the equimolar AlCl3-NaCl melt containing anodically dissolved Ti at −0.020 V and 300 ◦C for two hours.

Data acquired from the XRD analysis of the deposits are shown in Figure 9.

**Figure 9.** XRD patterns of the deposits obtained potentiostatically on the GC electrode on different conditions: (**a**) at −0.085 V and 200 ◦C for two hours; (**b**) at −0.020 V and 300 ◦C for two hours.

The analysis for the deposit obtained at 200 ◦C (Figure 9a) exhibited diffraction peaks at 2θ = 35.9◦ with reflection (200), 38.95◦ with reflection (002), 41.036◦ with reflection (201) and 71.97◦ with reflection (203), which are characteristics of a hexagonal AlTi3 alloy (JCPDS No. 03-065-7534). Several stronger diffraction peaks at 2θ = 38.47◦, 44.73◦, 65.13◦, 78.22◦ and 82.43◦ with the respective reflections (111), (200), (220), (311) and (222) should be attributed to face-centered cubic Al (JCPDS No. 00-004-0787). The spectrum indicates no evidence of additional peaks, implying that the deposit produced by the electrochemical deposition method in this study was relatively pure.

According to the data from XRD analysis of the deposit obtained on GC at a higher temperature (300 ◦C), dominating alloy appeared to be Al2Ti (Figure 9b). The spectrum gives diffraction peaks at 2θ = 22.68◦, 29.37◦, 38.99◦, 45.66◦, 66.55◦ and 79.94◦, with the respective reflections (101), (008), (116), (200), (220) and (316), which are characteristics of body-centered tetragonal Al2Ti (JCPDS No. 00-052-0861). Peaks in the spectrum at 2θ = 39.15◦, 54.32◦ and 84.15◦ with the respective reflections (112), (211) and (224) can be attributed to body-centered tetragonal Al3Ti (JCPDS No.03-065-2667). The spectrum also indicates characteristics of a hexagonal AlTi3 alloy, of which the peaks are at 2θ = 35.76◦, 40.84◦ and 62.69◦ with the respective reflections (200), (201) and (103) (JCPDS No. 03-0-052-0859). The peak at 2θ = 35.15◦ can be associated with the trace of Al2O3 (JCPDS No: 00-046-1212), implying that it was not possible to handle a sample without exposing it to the atmosphere.

Each of the peaks attributed to a certain alloy were chosen from a group of five or seven highest peak intensities as defined by the JCPDS database for the mentioned alloy. However, due to the fact that the Al-Ti binary phase diagram is not fully understood [4,28,29] and the fact that we only presented initial experimental results, there is a space for improvement in attribution of the peak positions in XRD analysis of the obtained electrodeposits in the future.

The AlTi3 alloy obtained on the GC WE from the equimolar AlCl3-NaCl melt with titanium ions added by electrodissolution of metal Ti is a finding different from a TiAl3 alloy predominantly produced electrochemically on Cu and mild steel from electrolytes reported in the literature [1,2,7,10,11]. The AlTi3 alloy is hexagonal in structure, and in the Al-Ti phase diagram, it appears in the composition region of 13–25 wt. % Al (i.e., 75–87 wt. % Ti) at temperatures below 1200 ◦C [30]. The AlTi3 intermetallic compound is largely accepted as having a variable composition, with a wide homogeneity domain and as an intermetallic compound formed by order-disorder transformation (αTi) ↔ (AlTi3). This aluminide is emerging as a revolutionary material for high-temperature applications and aeronautical industry [31, 32].

Al3Ti is an intermediate phase of a tetragonal structure and appears in the binary phase diagram in a composition region, where the Al mass is between 75% and 100% [28]. The Al3Ti alloy has a great potential application in aerospace and automobile as a high-temperature structural material, but it has poorest ductility among three typical Al-Ti alloys (AlTi3, AlTi and Al3Ti), which limits its engineering applications [33].

The Al2Ti compound is considered stable up to 1216 ◦C, existing with a very narrow Al range between 60 and 67 at % [29]. Al2Ti is one of the four intermetallic phases in the Ti-Al binary system that are stable below 1150 ◦C [34]. Al2Ti is a very promising material for elevated-temperature applications.

However, although the Al-Ti binary phase diagram has been intensively studied, it still cannot be considered fully reliable [4,28,29]. It seems that there are 12 intermetallic compounds recognised [35]. According to the Ti-Al phase diagram, there can be up to seven stable intermetallic phases. The most stable intermetallic phases that increase the physico-mechanical properties of titanium aluminide are γ-TiAl, α2-Ti3Al and γ-TiAl + α2-Ti3Al. Lately, it was pointed out that AlTi3 and AlTi intermetallic compounds are largely accepted as having variable compositions, with a wide homogeneity domain, while Al2Ti and Al5Ti2 are accepted as having constant compositions [29]. According to the same authors, Al3Ti is treated as an intermetallic compound with a variable composition or with a constant composition [29]. If thermal procedures are used, all of the abovementioned intermetallics can be produced at temperatures above 1110 ◦C.

When electrochemical co-deposition is used as a procedure for binary Al-Ti alloy generation, such high temperatures are not required. The mechanism of alloy formation in this case includes nanoscale relationships between adatoms of co-depositing metals on the substrate, such as interdiffusion in a solid state [18,36]. Interdiffusion phenomena were investigated in the Ti-Al system (a Ti region from 25 to 100 at %), but only at elevated temperatures (between 516 and 1200 ◦C). In a temperature range between 516 and 642 ◦C, metallic Ti, as well as Ti-Al alloys, was coated with a solid layer of metallic Al, and interdiffusion was studied. A well-adhered layer of TiAl3 was formed, while no other intermetallic compounds were observed and no solid solution of Al in Ti was recorded [37]. In another study, interdiffusion phenomena were investigated in a Ti-Al system but at higher temperatures (between 768 and 972 ◦C) [38]. In the Al-rich part of the diagram, the Ti2Al5 phase was identified. It was found that, at temperatures between 768 and 865 ◦C, Ti was the more mobile element in the Ti3Al phase whereas in the Al-richer compounds Al was the more mobile element at temperatures between 784 and 972 ◦C. The results of the studies in our work indicated appreciable interdiffusion between the co-deposited Al and Ti even at 200 and 300 ◦C, which led to the formation of Al3Ti, Al2Ti and AlTi3.

#### **4. Conclusions**

The electrodeposition of Al-Ti alloys from an equimolar AlCl3-NaCl melt on a GC electrode was successfully performed. It was shown that there is a novel way to obtain Al-Ti alloys, such as AlTi3, Al2Ti and Al3Ti, in a very controlled manner under favorable and technologically suitable conditions.

The voltammograms generated from a system of a Ti WE in the equimolar AlCl3-NaCl melt at 200 and 300 ◦C without introducing Ti ions indicated titanium deposition and dissolution proceeding in three reversible steps: Ti4<sup>+</sup> - Ti3<sup>+</sup>, Ti3<sup>+</sup> - Ti2<sup>+</sup> and Ti2<sup>+</sup> - Ti, occurring at potentials more positive than the reversible potential of Al. The reversible potential of titanium in the equimolar AlCl3-NaCl melt was identified as ≈ 0.200 V, and the starting deposition potential of titanium onto Ti was ≈ 0.020 V at 200 and 300 ◦C.

The titanium deposition starting potential on the GC electrode in the electrolyte made of an equimolar AlCl3-NaCl melt containing ≈ 0.1 M of titanium ions appeared to be between 0.050 and 0.0 V for both temperatures applied (200 and 300 ◦C). However, we did not succeed in depositing pure titanium without aluminium, because their deposition potentials were very close.

The XRD analysis of the deposits revealed that AlTi3, Al2Ti and Al3Ti alloys were generated on the GC electrode, with AlTi3 dominating at a lower temperature and Al2Ti dominating at a higher temperature.

The results obtained in this work suggest new possibilities of aluminium-titanium alloys formation (including AlTi3) using low temperatures via a better controlled process.

**Author Contributions:** V.S.C. designed and managed the research and participated in the manuscript preparation; N.M.V. performed most of the experiments and participated in the manuscript preparation; K.M.-N., S.S. and B.F. helped with the corrections of the manuscript; J.N.J. supervised the experiments and the manuscript writing. S.S. and B.F. from RWTH Aachen University provided funding for publication. All authors discussed the results and commented on the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** Part of the research was supported by the funds of the bilateral research project (ID: 451-03-01971/2018-09/4) supported by the Ministry of Education, Science and Technological Development of the Republic of Serbia and German Academic Exchange Service (DAAD).

**Acknowledgments:** Vesna S. Cvetkovi´c and Nataša M. Vuki´cevi´c acknowledge the financial support for the investigation received from the Ministry of Education, Science and Technological Development of the Republic of Serbia.

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

#### **References**


© 2020 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*

### **Application of the Flotation Tailings as an Alternative Material for an Acid Mine Drainage Remediation: A Case Study of the Extremely Acidic Lake Robule (Serbia)**

**Nela Petronijevi´c 1,\*, Srđan Stankovi´c 1, Dragana Radovanovi´c 2, Miroslav Soki´c 1, Branislav Markovi´c 1, Sre´cko R. Stopi´c <sup>3</sup> and Željko Kamberovi´c <sup>4</sup>**


Received: 25 November 2019; Accepted: 18 December 2019; Published: 20 December 2019

**Abstract:** Flotation tailings rich in carbonate minerals from the tailings deposit of the copper mine Majdanpek (Serbia) were applied for neutralization of the water taken from the extremely acidic Lake Robule (Bor, Serbia). Tests conducted in Erlenmeyer flasks showed that after neutralization of the lake water to pH 7, over 99% of aluminum (Al), iron (Fe), and copper (Cu) precipitated, as well as 92% of Zn and 98% of Pb. In order to remove residual Mn and Ag, the water was further treated with NaOH. After treatment with NaOH, all concentrations of the metals in the lake water samples were below discharge limits for municipal wastewater according to the national legislation of the Republic of Serbia. The results of this work suggest that mining waste could be used for active neutralization of the acid mine drainage. The use of the mining waste instead of lime could reduce the costs of the active treatment of the acid mine drainage.

**Keywords:** acid mine drainage (AMD); flotation tailings; AMD neutralization; metals' precipitation; polluted site remediation; synergy of processes

#### **1. Introduction**

Environmental pollution by acid mine drainage (AMD) is a widespread problem in mining impacted areas [1,2]. One such area is located near the town of Bor, in the eastern part of Serbia. Mining activities in the region of Bor began in 1903. Approximately 7 <sup>×</sup> 108 tons of mining waste (overburden and flotation tailings) have been deposited in the close proximity of the town [3]. An extremely acidic water body named Lake Robule was formed at the foot of the overburden deposit, just a few kilometers from the center of the town. The length of the lake is approximately 400 m, and the width at its widest part is approximately 130 m. The water is characterized by a low pH (approximately 2–2.5) and very high concentrations of ferric iron, causing the deep red color of the lake. High concentrations of Cu, Zn, Al, Mn, and other metals were also detected in the lake water [4–6]. Approximately 10 L s−<sup>1</sup> of acidic water flows from the lake to the Bor River through the drainage pipe [7]. One of the attempts to find an environmentally and economically feasible method to treat this accumulated AMD was conducted by Pavlovi´c et al. [8]. They used a laboratory scale cascade line system with three reactors

for selective precipitation of metals such as iron, copper, nickel, and arsenic from the synthetic solution that resembled acidic effluent from the open pit mine from the Bor area. The neutralizing agent was 1 M NaOH. Recently, Masuda et al. [9] applied hydrated lime (Ca(OH)2) for neutralization of acidity and selective precipitation of metals from the water collected from Lake Robule using semi-industrial scale equipment for active AMD treatment with two chemical reactors. Stopic et al. [10] used red mud from Greece and Germany firstly for neutralization of AMD (pH value of 2.3) from South Africa and for precipitation of copper. Mwewa et al. [11] performed precipitation of poly-alumino-ferric sulfate coagulant for wastewater treatment using AMD solution from South Africa.

There are active and passive methods in the treatment of AMD [12,13]. The conventional method for active neutralization of the acid mine drainage is the application of Ca(OH)2. This process is fast and efficient; hydrated lime increases the pH of the solution, resulting in precipitation of the dissolved metals in the form of metal hydroxides [14]. The main disadvantage of this technology is generation of the voluminous sludge. The price of the lime and need for the dehydration, solidification, and stabilization of the sludge after treatment increase the capital and operational costs of the AMD neutralization process [14–16]. More recently, passive methods have been attracting more attention due to reduced energy and maintenance costs [13]. With a view toward sustainable development, many studies have focused on finding alternative materials for AMD neutralization. Kaur et al. [17] investigated the application of the waste material from the alumina refining industry (Bayer liquor and precipitates formed by sea water neutralization of the Bayer liquor) to treat AMD from mine pit water. Kefeni et al. [18] reviewed technologies for AMD treatment, including alternative approaches for AMD neutralization by waste materials, such as the application of coal combustion by-products (fly ash, bottom ash, flue gas desulphurization materials), recycled concrete aggregates, and cryptocrystalline magnesite tailings from the gold extraction industry. Moodley et al. [19] also reviewed alternative methods for AMD remediation with a focus on industrial by-products, such as by-products from the paper mill and steel mill industries, meat industry, tire manufacturing, and phosphate waste rock.

The aim of the research is to investigate the ability of flotation tailings, a significant voluminous waste of the same industry that causes the formation of AMD in the Bor area, to neutralize and purify water collected from the acidic Lake Robule. Flotation tailings are a waste material generated during the production of mineral concentrate by froth flotation. Approximately 99 wt. % of the processed ore becomes flotation tailings [20]. Flotation tailings are consisted mostly of gangue minerals, but a considerable amount of carbonate minerals (calcite and dolomite) identified in particular flotation tailings samples collected from the dump of Copper Mine Majdanpek indicates a possible acid neutralization potential of the material. The motivation of the research is not only to investigate the ability of flotation tailings to neutralize and purify water collected from the acidic Lake Robule, but also to remove iron in order to prepare a solution for recovery of valuable elements such as copper, gold, aluminum, and silver in our future work. In order to predict the behavior of metals and especially their precipitation, the geochemical software will be tested and discussed with experimental results.

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

#### *2.1. Sampling*

Copper Mine Majdanpek is a part of the mining and smelting industry in Bor, Serbia. The mining operations started in 1963; and since then, 378 million tons of flotation tailings have been deposited in a large dump. Flotation tailings samples were dug out from cubes of a 0.04 m<sup>3</sup> volume (0.2 m <sup>×</sup> 0.2 m × 1 m) at 25 equidistant spatial locations (the distance between locations was 50 m), forming a square shaped network, and put in the separate bags. In the laboratory, the composite sample was prepared by the quartering method.

A total of 10 water samples was collected in March 2018 from the pipe that drains water from Lake Robule. The main physicochemical parameters (pH, Eh, temperature) were measured on the spot (Hanna Instruments HI98196, Hanna Instruments Deutschland GmbH, Vöhringen, Germany). Containers (1 L, high density polyethylene) were rinsed with concentrated HNO3 (Tehnohemija, Belgrade, Serbia) and deionized water before collecting samples. The filled containers were immediately transported to the laboratory the same day. Containers were stored at 4 ◦C in the refrigerator. A schematic geographic map of Serbia with the location of Copper Mine Majdanpek and the extremely acidic Lake Robule is given in Figure 1.

**Figure 1.** Schematic geographic map of Serbia with the location of Copper Mine Majdanpek and the extremely acidic Lake Robule.

#### *2.2. Characterization of the Flotation Tailings*

The chemical characterization of the flotation tailings was performed by dissolving a sample of the material in aqua regia and measuring the concentration of the selected elements by the Atomic Absorption Spectroscopy (AAS) method using Perkin Elmer Aanalyst 300 (PerkinElmer, Inc, Norwalk, CT, USA).

The concentration of the selected chemical elements in the water samples was measured by the Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES) method (Spectro Genesis, Spectro Analytical Instruments, Kleve, Germany).

The sulfate concentration in the lake water sample was determined gravimetrically with BaCl2. Carbonates and hydrogen carbonates were determined by titration with 0.1 M HCl using phenol phthalein and methyl orange (Tehnohemija, Belgrade, Serbia) as indicators.

Mineralogical characterization of the tailings was performed by the microscopy and XRD (X-Ray Diffraction, Zeiss-Jena, Oberkochen, Germany) methods. The polarizing microscope Carl-Zeiss, Model "JENAPOL-U" equipped with 10×, 20×, 50×, and 100× (oil immersion) objectives and a system for photomicrography ("Axiocam105 color" camera and "Carl Zeiss Axio Vision SE64 Rel. 4.9.1." software package with Multiphase module), was used for microscope investigations in reflected light. The XRD patterns were obtained using a Philips PW-1710 automated diffractometer with a Cu tube operated at 40 kV and 30 mA. The instrument was equipped with a diffracted beam curved graphite monochromator and an Xe filled proportional counter. The diffraction data were collected in the 2θ Bragg angle range of 4–65◦, counting for 1 s. Semi-quantitative analysis of the data obtained by XRD was performed by "Powder Cell" computer software [21], (Federal Institute for Materials Research and Testing, Berlin, Germany).

#### *2.3. Acid Neutralization Capacity*

Acid Neutralization Capacity (ANC) represents the ability of a material to neutralize acid and remain stable under the external effluence of the environment [22]. In order to investigate and compare the ANC of flotation tailings before and after the treatment with the lake water, both samples of the tailings (fresh and treated) were subjected to the ANC test described by Stegemann and Cote [23]. The test procedure consisted of a series of leaching tests with increasing concentrations of nitric acid (HNO3) and measuring the pH values of the leachates. Samples and solutions with different nitric acid

contents, in the range from 0 (distilled water) to 2 H<sup>+</sup> eq/kg, with a solid to liquid ratio of 10, were rotated in polyethylene bottles (*V* = 50 mL) for 48 h prior to centrifugation and measurement of pH. Titration curves were obtained by plotting measured pH values versus acid content.

#### *2.4. Determination of the Optimal Quantity of Flotation Tailings Required for Neutralization of the Lake Water Samples and Metal Precipitation*

The neutralization test was performed with increasing quantities of the flotation tailings in order to identify the optimal solid to liquid ratio for neutralization of the lake water. The tests were conducted in eight 100 mL Erlenmeyer flasks containing 50 mL of water with increasing pulp density: 1, 3, 5, 10, 15, 20, 25, 30, and 40%. The Erlenmeyer flasks were put in an incubated orbital shaker (Heidolph Unimax 1010, Heidolph, Shwabach, Germany) at a temperature of 25 ◦C and a shaking speed of 250 rpm. After two hours of shaking, the samples were put to rest for seven days at room temperature.

After determination of the optimal quantity of the flotation tailings for the neutralization of the lake water, the precipitation of metal cations as a function of time and pH was determined. Nine 100 mL Erlenmeyer flasks were filled with 50 mL of the lake water, and 7.5 g of the flotation tailings were added to each flask. Flasks were placed in an orbital shaker under the same conditions as in the previous test. After agitation was stopped, the solution was left to stand for 30 min in order to let the particles settle out. Samples were collected after 5, 10, 15, and 30 min and 2, 4, 24, 72, and 168 h. The collected samples were filtered by using filter paper and analyzed by ICP-OES.

#### *2.5. Treatment of the Lake Water Samples with Hydrated Lime and NaOH*

After treatment of lake water samples with flotation tailings, the water was further treated with hydrated lime Ca(OH)2 in order to increase its pH to 10, as is shown in Figure 2. A flask with 50 mL of the lake water previously treated with flotation tailings was put on a magnetic stirrer; the pH of the water was measured during the experiment by a laboratory pH meter (Hach Sension + MM340). Hydrated lime was carefully added until the pH of the solution reached a value of 10.

In order to compare the results obtained by lake water neutralization with flotation tailings, the lake water was also neutralized with NaOH. Two Erlenmeyer flasks with 50 mL of the lake water were put on a magnetic stirrer and treated with NaOH and flotation tailings until the pH value of the water reached 7.

#### *2.6. Simulation of Metal Precipitation Using PHREEQC Software*

PHREEQC is a computer program designed to perform a variety of aqueous geochemical calculations [24], and in numerous research works [12,13,25,26], it was used to explain and support the experimental results addressed to the behavior of contaminants during a specific treatment. Here, the PHREEQC software (USGS, Reston, VA, USA) was applied for the modelling of the aqueous speciation of the water from Lake Robule and the saturation index calculations of the solid phases formed during the treatment of lake water by flotation tailings based on the pH and concentrations of elements in the solutions. An individual simulation was made for each defined time during the experiment according to the measured values of pH and the concentration of metals as input values for the calculations. Water volume, solid to liquid ratio, and temperature, as constant parameters during the experiment,

were also input values for the calculation. Saturation Indices (SI) are defined as the logarithm of the ratio of the Ion Activity Product (IAP) to the solid solubility product (*Ks*), Equation (1).

$$\text{SI} = \log(\text{IAP}/\mathbb{K}\_s) \tag{1}$$

A positive SI indicates precipitation of the mineral, while a negative value of SI indicates mineral dissolution. Values of SI ≥ 0 were chosen to specify the controlling mineral for the constituent element's precipitation.

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

#### *3.1. Physical and Chemical Properties of the Water Collected from Lake Robule*

The physical and chemical properties of the water collected from Lake Robule are presented in Table 1. The acidic nature of the Lake Robule water was confirmed by a pH value of 2.47. The most dominant ions in the water samples were sulfate ions (7.5 g/L), Al2<sup>+</sup> (1017.62 mg/L), and Fe ions (287 mg/L), almost completely in the form of Fe3<sup>+</sup> (286.9 mg/L).


**Table 1.** Physico-chemical properties of the water collected from Lake Robule.

#### *3.2. Characterization of the Flotation Tailings' Sample from Copper Mine Majdanpek*

The chemical composition of the flotation tailings' composite sample collected from Copper Mine Majdanpek is presented in Table 2, while the results of the X-ray diffraction analysis are presented on Figure 3.

The most abundant minerals in the analyzed flotation tailings sample identified by X-ray diffraction analysis were quartz, pyrite, and carbonates (calcite and dolomite), followed by feldspar, clay minerals, mica, and illite, which were less abundant. The results of the semi-quantitative analysis of the data obtained by XRD were: quartz ≈ 50–55%, total carbonates ≈ 20–25%, kaolinite ≈ 5–10%, pyrite ≈ 5%, and illite ≈ 5%. High pyrite content supported the results of chemical analysis and also indicated the high potential of exposed tailings to generate AMD. The content of carbonate minerals (20–25%) indicated the acid neutralization capacity of the material. The concentrations of Mn (0.138%), Cu (0.072%), Zn (0.086%), and Pb (0.0079%) indicated the possibility of leaching of these metals from tailings in contact with acidic water from lake Robule.

**Table 2.** Chemical composition of the flotation tailings from Copper Mine Majdanpek.


**Figure 3.** Diffractogram of the flotation tailings sample from Copper Mine Majdanpek.

#### *3.3. Results of the Acid Neutralization Capacity Test*

Remediation of the acid mine drainage is based on neutralization of the solution's acidity, which results in the precipitation of the metal cations, mostly in the form of the insoluble hydroxides and carbonates [8,14,27–30]. The acid neutralization capacity of mining waste, such as flotation tailings, depends on the relative reactivity of the contained minerals to interact with H<sup>+</sup> ions and undergo the neutralization process. According to Sverdrup [31], minerals can be divided into several groups based on their relative reactivity in the acid neutralization process: dissolving (calcite, dolomite, magnesite), fast weathering (anorthite, nepheline, forsterite, olivine, garnet, jadeite, leucite, spodumene, diopside, wollastonite), intermediate weathering (sorosilicates, pyroxenes, amphiboles, phyllosilicates), slow weathering (plagioclase, kaolinite), very slow weathering (K-feldspar), and inert (quartz). The most relevant minerals as potential neutralization agents contained in flotation tailings are carbonates, hydroxides and silicates [32,33]. The results of the ANC test of fresh and treated flotation tailings are shown in Figure 4. Common to both materials is a wide plateau, from 0.2 to 1.4 H<sup>+</sup> eq/kg, at pH values between 5 and 6 associated with the dissolution of carbonate minerals [34]. Carbonates, primarily calcite, are the main minerals involved in the acid neutralization reaction described by Equation (2) at pH > 6.3 and Equation (3) at pH < 6.3 [33].

$$\text{CaCO}\_3 + \text{H}^+ \Leftrightarrow \text{Ca}^{2+} + \text{HCO}\_3^- \tag{2}$$

$$\text{CaCO}\_3 + 2\text{H}^+ \Leftrightarrow \text{Ca}^{2+} + \text{H}\_2\text{CO}\_3 \tag{3}$$

An interesting observation is that the treatment did not significantly reduce the buffering capacity of the material and that the amount of acid that the carbonates could neutralize was in the same range for both samples. Carbonate consumption for acid neutralization was reflected in the initial pH values after leaching with distilled water (0 H<sup>+</sup> eq/kg): 8.6 for fresh tailings and 7.5 for treated. Products of the treatment and their effect on the ANC were reflected in the position of the last plateau on the titration curves at a higher H<sup>+</sup> concentration for treated tailings (pH <sup>≈</sup> 4.5) than for fresh tailings (pH ≈ 3.0). This was associated with the formation of solid hydroxide minerals of metal cations with a valence of 3+ (Fe3<sup>+</sup> and Al3<sup>+</sup>) during the treatment. These metal hydroxides hydrolyzed with high ionic potential, thus representing important buffers controlling the pH at ≈ 4.3 (Al(OH)3) and ≈ 3.5 (Fe(OH)3) [22,33,34]. It can be concluded from the results of the ANC test that flotation tailings from Copper Mine Majdanpek had significant acid neutralization capacity and that these flotation tailings could be used multiple times for AMD neutralization. Furthermore, an important finding was that the buffering capacity of the tailings actually increased after treatment, due to the hydrolysis of the precipitated metal hydroxides.

**Figure 4.** Results of the Acid Neutralization Capacity (ANC) test. Untreated: flotation tailings before neutralization of the lake water. Treated: flotation tailings after neutralization of the lake water.

#### *3.4. Treatment of Water from Lake Robule by Using Flotation Tailings*

3.4.1. Determination of the Optimal Quantity of Flotation Tailings Required for Neutralization of the Lake Water Samples

The next step was to find the optimal solid to liquid ratio (S:L) for neutralization of the water from Lake Robule. The results of the neutralization test with gradually increased amounts of flotation tailings are presented in Table 3.


**Table 3.** Determination of the optimal quantity of flotation tailings required for neutralization of the lake water samples to pH 7.

The results of the experiment presented in Table 3 showed that following the increase in pulp density, the pH value dropped faster, but after seven days of settling, the difference in the final pH values between flasks with pulp densities of 15%, 20%, 25%, 30%, and 40% was not significant. Based on these results, a pulp density of 15% was identified as optimal for further experiments.

#### 3.4.2. Experimental Results of the Treatment

Changes in the concentrations of the selected metals in solution during the treatment are presented in Table 4. Beside concentrations, the removal percentage of each element was calculated according to Equation (4).

$$\text{Removal} = \frac{\text{final concentration}}{\text{starting concentration}} \times 100\% \tag{4}$$


**Table 4.** Concentration of elements in solution during treatment time.

The + sign in the removal row means that the concentration of the certain metal increased after the neutralization experiment.

Experimental results showed that the concentration of all elements, except Mn and Cd, decreased during the treatment with the increasing value of pH. After 168 h of treatment, when the pH of the solution reached 7.01, the concentrations of all elements, except Mn, were below the discharge limits for municipal wastewaters prescribed by the national legislation of the Republic of Serbia [35], with over 99% of Fe, Al, and Cu precipitation, 98% removal of Pb, and 92% of Zn. Unlike all other metals, the concentrations of Mn and Cd in the solution increased during the treatment, indicating that these metals might be leached from the tailings.

#### *3.5. Results of the PHREEQC Software Simulation of the Water Treatment*

The results of the treatment simulation obtained by the PHREEQC program are given in Table 5. The results were based on the measured concentrations of the elements and pH values of the obtained solutions over time, as well as on the defined water volume (50 mL), solid to liquid ratio (15%), and temperature (18 ◦C) as constant values during the experiment. The results are presented in the form of calculated Saturation Indices (SI) of the corresponding mineral phases. The saturation indices of certain mineral phases, noted in the Table 5, increased and became positive at specific pH values. Such SI denote minerals whose constituent elements achieved thermodynamic equilibrium concentration in the solution during the neutralization reaction (SI = 0), after which they began to precipitate (SI > 0). The results of PHREEQC modeling partially complemented the experimental results and gave an additional explanation of the mechanism of metals' removal during the treatment.


**Table 5.** Results of the PHREEQC software simulation.


**Table 5.** *Cont*.

#### *3.6. Mechanism of Acid Neutralization and Precipitation of Metals during the Treatment*

Figures 5–7 present the changes in the concentrations of the metal cations as a function of the pH. After two hours of mixing, the pH of the solution increased to 4.78, and after seven days of settling, the pH of the solution further increased to 7. In samples collected from Lake Robule, over 99% of the total iron was in the ferric state. Deeper anoxic sections of the lake, which have not been disturbed nor sampled, could have iron in the ferrous (Fe2<sup>+</sup>) state. During the first 5 min of the treatment, the pH value of the solution increased due to the carbonate dissolution and acid neutralization process and reached a value of 4.3, when the solution was supersaturated with Fe3<sup>+</sup> ions that began to precipitate. The removal of iron was relatively rapid with a sharp decline in iron concentration in the pH range between 2 and 4. Approximately 99.8% of iron from the water samples was removed. Iron in the ferric state should readily precipitate as oxyhydroxide compounds (FeO(OH)), as shown in Equation (5), at pH values greater than 3.5 [17].

$$\text{Fe}^{3+}\text{(aq)} + 2\text{OH}^-\text{(aq)} \leftrightarrow \text{FeO(OH)}\_{\text{(s)}} + \text{H}^+\text{(aq)}\tag{5}$$

Equation (5) includes the dynamic information about the formation of goethite, but the analysis of the software included the equilibrium information, where the maximal possibility for precipitation was in the case of the hematite. Generally, as shown in Table 5, an increase of time and pH-values increased the possibility of the precipitation of iron hydroxide, goethite, and hematite with different values of calculated Saturation Indices (SI). In the case of the other elements such as zinc, cadmium, lead, and manganese, these SI-values were negative with small possibilities for precipitation.

**Figure 5.** Changes in the concentrations of Al and Fe as a function of the pH.

**Figure 6.** Changes in the concentrations of Mn, Zn, and Cu as a function of pH.

**Figure 7.** Changes in the concentrations of Cd, Ni, and Pb as a function of pH.

Park et al. [36] concluded that in synthetic multimetal solutions, ferric iron precipitated as an amorphous ferric oxyhydroxide and then was transformed by a slow dehydration reaction and internal atomic rearrangement to hematite. These two transformation processes compete with each other, and one of them is dominant depending on the reaction conditions [37].

$$\text{Fe}^{2+} \quad \xrightarrow{\quad} \xrightarrow{\quad} \quad \text{Fe}^{3+} \quad \quad \xrightarrow{\quad} \xrightarrow{\quad} \quad \text{a-t-FeO(OH)} \quad \xrightarrow{\quad} \xrightarrow{\quad} \quad \text{a-t-Fe}\_2\text{O}\_3$$

Oxidation Precipitation Dehydration After 30 min of the process, the pH value of the solution was 4.7, at which the Al3<sup>+</sup> ions began to precipitate in the form of gibbsite (Al(OH)3). This was consistent with the fact that in the presence of carbonate ions, the trivalent metals began to precipitate in the form of their hydroxides, Fe at pH <sup>≥</sup> 3.5 and Al at pH <sup>≥</sup> 4.5 [10]. Park et al. [36] also found that Al3<sup>+</sup> from complex multimetal AMD solutions precipitated as amorphous AlOHSO4. In the experiment presented in this paper, aluminum precipitated in the pH range from 3.5 to 5 (Figure 5).

The experimental results showed a significant decrease in Zn and Cu concentrations in the lake water samples during the treatment, although simulation results complemented only the precipitation of Zn in the form of smithsonite (ZnCO3) and hydroxide (Zn(OH)2) at pH values above 7, which was not achieved during the treatment. The majority of copper cations precipitated in the pH range between 5 and 7, and over 99% of copper was removed. Software simulation did not predict the precipitation of copper. Precipitation of copper from complex solutions depends on the interactions with other ions. Park et al. [36] investigated the precipitation of metals from quaternary mixtures Fe/Al/Cu/Zn and Fe/Al/Cu/Ni and discovered that metals precipitated in the order Fe-Al-Cu-Zn or Ni. In these solutions, copper probably forms brochantite in the pH range from 5.0 to 6.6 and transforms to tenorite (CuO) above a pH of 6.6 [36,38]. Therefore, in this experiment, the forms of Cu precipitates were expected to be brochantite and tenorite. Theoretically, brochantite can be transformed to tenorite as follows [39]:

$$\rm{^4Cu^{2+} + 6OH^- + SO4^{2-} \leftrightarrow Cu\_4(SO\_4)(OH)\_6} \tag{6}$$

$$2\text{ Cu}\_4\text{(SO}\_4)\text{(OH)}\_6 + 2\text{OH}^- \leftrightarrow 4\text{CuO} + \text{SO4}^{2-} + 2\text{H}\_2\text{O} \tag{7}$$

$$2\text{ Cu}\_4\text{(SO}\_4\text{)}\_\text{aq} + 2\text{ OH}^- \leftrightarrow \text{CuO} + \text{SO4}^{2-} + \text{H}\_2\text{O} \tag{8}$$

During the neutralization experiment, the concentration of zinc increased between pH 3.29 and 6.6 and then dropped to a minimal value of 1.4 mg L<sup>−</sup>1, leading to the removal of 92% of Zn. The increase in zinc concentration was probably the consequence of the leaching of this metal from flotation tailings. Park et al. [36] concluded that in solutions that mimic AMD, zinc probably precipitates as hydrozincite Zn5(CO3)2(OH)6. The removal of Zn and Cu from multimetal solutions can be also attributed to the effect of co-precipitation or adsorption onto amorphous Fe and Al hydroxides [40]. At pH values between 6 and 7, the PHREEQC simulation predicted a precipitation of Pb in the form of carbonate mineral cerussite (PbCO3) since Pb(OH)2 precipitates at higher pH values between 9 and 10; as a result, 98% of lead precipitated at a pH value of approximately 4.8.

According to the software simulation, manganese should precipitate as MnCO3, but no precipitation of manganese occurred during neutralization with flotation tailings. Instead, the concentration of manganese actually increased during the experiment probably due to leaching of Mn from the flotation tailings. The possible explanation is that the formation of MnCO3 was not thermodynamically favorable under the experimental conditions and that also manganese hydroxide (Mn(OH)2) precipitated at a pH above 10.

The concentration of cadmium also increased after neutralization because flotation tailings probably released some cadmium into the solution. The software simulation did not predict the precipitation of cadmium, which is an accordance with the experimental results. Fifty percent of nickel precipitated during the neutralization test. Park et al. [36] found that in the quaternary solution, Ni completely precipitated at pH 8.4. Precipitation of nickel started at pH 7, and the pH required for complete precipitation of Ni was not reached in this experiment. Nickel probably did not precipitate as Ni(OH)2 because nickel hydroxide precipitates at approximately pH 10, but more likely as NiCO3.

#### *3.7. Comparison of the Lake Water Neutralization with Flotation Tailings and NaOH*

In order to determine if other mechanisms of metal removal were included, such as adsorption of metal cations on the surface of the minerals that constituted flotation tailings, neutralization of the lake water to pH 7 was performed by application of flotation tailings and NaOH. The most significant differences were observed for the concentrations of Mn and Ag (Table 6). The concentrations of these metal cations were higher in water treated with flotation tailings than in water treated with NaOH. The most probable explanation is that these metals were leached from flotation tailings during neutralization. Changes in metal concentrations during neutralization with flotation tailings were the result of the increased pH of the solution; adsorption on the surface of the minerals had no significant effect on the metal removal efficiency.


**Table 6.** Comparison of the metals' precipitation efficiency after treatment with Flotation Tailings (FT) and NaOH.

#### *3.8. Post-Treatment of the Lake Water Samples with Hydrated Lime*

In order to purify water from residual Mn and Ag, the sample of water that was treated with flotation tailings was further treated with hydrated lime, and the pH of the water was increased to 10. The concentrations of manganese and silver were reduced to 0.062 mg L−<sup>1</sup> and 0.013 mg L−<sup>1</sup> (Table 7). After this treatment, the concentrations of all metals in the water were below the discharge limits for municipal wastewaters according to national legislation of the Republic of Serbia [35].

**Table 7.** Results of the treatment of the lake water sample with Ca(OH)2 after neutralization with flotation tailings. Results were compared with national discharge limits for municipal wastewater of the Republic of Serbia [35].


/: not defined.

#### *3.9. Characterization of the Solid Residue after the Treatment of Water from Lake Robule with Flotation Tailings*

Results of the XRD analysis of the flotation tailings residue after the neutralization experiment with 15% pulp density are presented in Figure 8. The content of carbonate minerals (calcite and dolomite) only slightly decreased after water treatment. The results of the semi-quantitative analysis of the XRD data were: quartz 55–60%, total carbonates 15–20%, mica/illite ≈ 10%, pyrite ≈ 5%, and kaolinite ≈ 5%. These results confirmed that the same sample of the flotation tailings could be used multiple times for neutralization of the water from Lake Robule.

**Figure 8.** Diffractogram of the flotation tailings residue after neutralization of the lake water sample.

#### **4. Conclusions**


**Author Contributions:** Conceptualization, investigation, and experimental design, N.P. and S.S.; methodology and editing, B.M.; formal analysis and editing, D.R.; writing, review and editing, S.R.S.; supervision, conceptualization, and editing, M.S. and Z.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was realized in the frame of the projects TR 34023 and TR 34033, supported by the Ministry of Education, Science and Technological Development of the Republic of Serbia.

**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/).
