**Green Process for Industrial Waste Transformation into Super-Oxidizing Materials Named Alkali Metal Ferrates (VI)**

#### **Ndue Kanari 1,\*, Etleva Ostrosi 2, Cécile Diliberto 3, Inna Filippova 1, Seit Shallari 4, Eric Allain 1, Frederic Diot 1, Fabrice Patisson <sup>5</sup> and Jacques Yvon <sup>1</sup>**


Received: 29 May 2019; Accepted: 17 June 2019; Published: 19 June 2019

**Abstract:** The investigation presented here features the design of a cleaner and greener chemical process for the conversion of industrial wastes into super-oxidizing materials. The waste of interest is the iron sulfate heptahydrate (FeSO4·7H2O) mainly generated through the sulfate route used for titanium dioxide industrial production. The products of this transformation process are alkali ferrates (A2FeO4, A = Na, K) containing iron in its hexavalent state and considered as powerful oxidants characterized by properties useful for cleaning waters, wastewaters, and industrial effluents. The proposed process includes two steps: (i) The first step consisting of the pre-mixing of two solids (AOH with FeSO4·xH2O) in a rotary reactor allowing the coating of iron sulfate in the alkali hydroxides through solid–solid reactions; and (ii) the second step involves the synthesis of alkali ferrates in a fluidized bed by oxidation of the single solid obtained in the first step in diluted chlorine. The chemical synthesis of alkali ferrates can be carried out within a timeframe of a few minutes. The usage of a fluidized bed enhanced the energy and mass transfer allowing a quasi-complete control of the ferrate synthesis process. The alkali ferrate synthesis process described here possesses many characteristics aligned with the principles of the "green chemistry".

**Keywords:** industrial waste; alkali ferrates; super-oxidizing materials; fluidized bed; green process

#### **1. Introduction**

Iron compounds are abundant in most nonferrous metal deposits where these metals often represent a small fraction. Although part of the ferrous components are separated during mining and primary mineral processing, a large part of iron goes through the extractive processes of these metals. Further, the iron is sometimes present in the natural mineral bodies (e.g., FeTiO3, CuFeS2, (Ni,Fe)9S8) of target metals that can be separated during the primary metal extraction. Therefore, considerable amounts of iron bearing co-products and wastes are inevitably generated from both hydro- and pyro-metallurgical operations on such raw materials.

One typical industrial example of generation of iron-containing waste is the extraction of titanium oxide (TiO2) from its bearing materials such as ilmenite, rutile, anatase, and slags. As described previously [1], "chloride" and "sulfate" routes are processes currently used for TiO2 production. The sulfate process, using ilmenite (FeTiO3) as a raw material, can lead to the generation of amounts of iron sulfate (FeSO4·7H2O—melanterite) as high as 6 tons of FeSO4·7H2O per ton of produced TiO2 [2], which is a real environmental drawback for the TiO2 industry. By 1993, all countries producing TiO2 through the sulfate process had to comply with the European directive [3] in order to avoid dumping industrial waste (such as melanterite) into the seawater. In the perspective of a circular economy and sustainable development, the ideal scenario should be to consider these wastes (e.g., iron sulfate) as an input for new chemicals and material synthesis, which are intrinsically non-hazardous, thus excluding the notion of undesirable by-product. Few recent investigations [4–7] address the usage of melanterite in various application and its treatment, such as for synthesizing slow-release fertilizers [5], cation-substituted LiFePO4 [6], and for Fe3O4 production through reductive decomposition using pyrite [7].

As a continuation of our previous research investigations [2,8–10], the present work aims at transforming industrial iron sulfate into alkali metal ferrates (A2FeO4, A = Na, K) considered as useful materials in different fields. One may note that the denomination "ferrate" is generally attributed to compounds containing iron at an oxidation state higher than Fe(III) and the ferrates(VI) seem to be the best known and most studied [2]. In an aqueous solution, the ferrate ion (FeO4 <sup>2</sup>−) is reduced, generating both Fe(OH)3 and nascent oxygen (Equation (1)). As summarized early [2], ferrates are used for water treatment due to their powerful oxidizing capacity (E(FeO4 <sup>2</sup>−/Fe3<sup>+</sup> = 2.2 V)) (oxidation of organic and mineral materials, bactericide agent) and because of the flocculating property of the evolved Fe(OH)3. The ferrates can replace chlorine in the pre-oxidation stage of water and can partially be used as a substitute of the iron and aluminum salts used as coagulating and flocculating agents. Furthermore, the decomposition of alkali ferrates generates basic medium favorable for the precipitation of heavy metals. These properties existing together in one single compound make ferrates an ongoing material particularly interesting for water treatment and effluent cleaning. Moreover, the oxidation products (iron oxy/hydroxides) (Equation (1)) are inoffensive to the environment.

$$2\text{FeO}\_4^{2-} + 5\text{H}\_2\text{O} \rightarrow 2\text{Fe(OH)}\_3 + 4\text{OH}^- + 3\text{O}^\bullet \tag{1}$$

The research works performed by Fremy [11–13] are frequently cited in the literature as the first ones to scientifically reveal the existence of iron in a hexavalent state and to effectively achieve ferrates synthesis. Since then, the preparation methods, developed mostly on a laboratory scale, have made little progress, and can be classified into three groups:


Synthesis of alkali ferrates (AF) by gas–solid reactions performed in a rotary reactor using chlorine as an oxidant showed that synthesis was achieved without external heat supply [2,8–10]. However, it was observed that the synthesis reactions were highly exothermic leading to a temperature rise in the reaction zone above 150 ◦C, when only 10 g of solids were used for the potassium and/or sodium ferrate synthesis. Experimental results indicated that the heat generation provoked the sample agglomeration and dramatically decreased the Fe(VI) efficiency of the synthesis process. The temperature increase phenomena became a real "obstacle" when higher amounts (100 g) of solids were used for the AF synthesis, although the reactor was cooled. Further, the kinetic of Na-ferrate synthesis is expected to be low compared with that of potassium ferrate [2]. In other words, it was concluded that the synthesis of alkali ferrates in a rotary reactor would not be considered as an appropriate route for an eventual AF large-scale production.

In this context, the goal of this research work is threefold: (i) Use the industrial iron sulfate as raw material for the AF synthesis; (ii) develop an appropriate process for AF preparation at room temperature; (iii) optimize the process by data analysis of various parameters affecting the efficiency of Fe(II,III) conversion into Fe(VI). The AF synthesis (mostly sodium ferrate preparation) is achieved in fluidized bed (FB) which can be considered as an appropriate reactor for handling processes that require high energy and mass transfers. It should be noted that the alkali metal ferrates (VI) manufacturing process, as developed in this work, is unique in its field.

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

Industrial ferrous sulfate (mainly monohydrate) is used for the synthesis of alkali ferrates. An examination by scanning electron microscope (Hitachi Ltd., Tokyo, Japan) coupled with energy dispersive spectrometry (Kevex Corp., Foster, CA, USA) (SEM-EDS), X-ray diffraction (XRD), chemical analyses, and Mössbauer spectroscopy (MS) suggested that the sample is free of heavy metals and that the quasi-totality of iron is in a bivalent state as FeSO4·H2O. Another sample of ferrous sulfate heptahydrate is also used after dehydration in an oven at about 150 ◦C leading to the formation of FeSO4·H2O and FeSO4·OH as the main iron-bearing phases. Commercial sodium hydroxide is used as pellets of 2 mm and pearls of about 1 mm. The oxidizing agent (chlorine) and diluting gas (nitrogen) were of high purity, whilst the air was supplied by a compressor.

The flowchart summarizing the features of the experimental protocol established for the synthesis of alkali ferrates is schematized in Figure 1. Accordingly, the synthesis process consisted of two main steps. The first step consists of a premixing of mostly FeSO4·H2O and/or FeSO4·OH with NaOH resulting in a single solid. This step is realized by using the experimental setup represented in Figure 2. The parameters studied for the solid premixing step are related to physical characteristics of iron sulfate and NaOH feed, their amounts, as well as the premixing time.

**Figure 1.** Schematic representation of the experimental procedure applied for the ferrates(VI) synthesis through a two-step process.

**Figure 2.** Apparatus assembly with a baffled reactor used for mixing iron sulfate with sodium hydroxide.

The second step of the synthesis process (Figure 1) consists of the reaction of the granulated and single solid, prepared in step 1, with chlorine diluted in air and/or N2 leading to the sodium ferrate preparation. A fluidized bed reactor equipped with a temperature-regulated water system is used for the experimental tests of alkali ferrate synthesis. A second fluidized bed is added in series for recycling the unreacted oxidant (chlorine) during AF synthesis taking place in the first fluidized bed. Several experimental parameters related to the ferrate synthesis step (particle size of solid, temperature, chlorine content, reaction time) were investigated. Other details of the experimental procedure will be introduced when describing the experimental results for both steps.

Solid synthesis products are subjected to visible microscopy, SEM-EDS, and XRD to examine the structure and the composition of the solid reaction products. The procedure of these characterization techniques is given in a recently published material [24].

Mössbauer spectroscopy was used to seek information about the oxidation state of iron as well as to evaluate the Fe(VI) content in the synthesis product. Details about this analysis method were given earlier by Jeannot et al. [25]. However, this method of examination is time consuming, i.e., the acquisition of a Mössbauer spectrum may take up to 24 h. In this context, a chemical analysis method is performed to quickly determine the Fe(VI) synthesis efficiency of the ferrate synthesis trials. This method is based on the chemical reaction of Fe(VI) with an excess of ferrous sulfate solution; then, the excess of Fe(II) is titrated with potassium bichromate.

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

#### *3.1. Concept of the Vibrating Fluidized Bed*

The fluidized bed technique is very attractive for different processes related to gas–solid reactions and for its easy extrapolation. However, a fluidized bed is not suitable for solids with important differences in particles sizes (or density), as it is the case of NaOH and iron sulfate. As a reminder, the mean particle sizes of NaOH, found in market, were about 1, 2, and 5 mm and that of FeSO4·H2O is less than 100 μm. A conventional fluidization of NaOH and FeSO4·H2O (for a given fluidization velocity) allows a heterogeneous distribution of the solids in the reactor. It was suggested that the vibrating fluidized bed would be a solution for the homogeneous fluidization of NaOH and iron sulfate. The distribution of particles in the bed through the respective conditions of the fluidization alone, vibration alone, and fluidization coupled with vibration are schematized in Figure 3. The choice of the optimal fluidization velocity of the reactive gases for the iron sulfate creates the distribution situation described in Figure 3a. Thus, iron sulfate is fluidized when the NaOH pearls creates a packed

bed at the bottom of the reactor. If the bed is only vibrated (Figure 3b), the pearls of NaOH move towards the top, while the iron sulfate accumulates at the bottom of the reactor. Meanwhile, the use of blow and vibration leads to an almost homogenous 'fluidization' of both solids (FeSO4 and NaOH) as described in Figure 3c. However, it seems that the vibrating fluidized bed is difficult to control on an industrial scale production, especially for delicate processes such as ferrate synthesis.

**Figure 3.** Schematic particles distribution of solids in a vibrated fluidized bed: (**a**) optimal gas fluidization velocity for iron sulfate; (**b**) only vibration of fluidized bed; (**c**) simultaneous blowing and vibration of fluidized bed.

#### *3.2. Idea of Premixing of Solids Prior to Fluidization*

The overall reaction of the sodium ferrate synthesis can be described by Equation (2), although the exact formula of this compound seems to be still undefined. It was already confirmed experimentally that the synthesis of Na-ferrate via Equation (2) is exothermic. The possible "two by two" reactions of the three substances (FeSO4·H2O, Cl2, and NaOH) could be represented by Equations (3)–(5). Iron sulfate monohydrate does not react with chlorine at room temperature. Recent experience in the field of gaseous chlorine reactivity with iron compounds showed that the oxidation of Fe(II) of wüstite (FeO) into Fe(III) takes place at temperatures higher than 200 ◦C [26]. Chlorine in the presence of oxygen can oxidize Cr(III) into Cr(VI), generating chromium oxychloride [27–29].

As could be expected, the reaction of chlorine with NaOH produces NaCl as a final reaction product involving heat with ΔH = −128 kJ/mol NaOH [30]. The most interesting reaction to be considered is that of iron sulfate with sodium hydroxide (Equation (5)) resulting in the formation of Na2SO4. The sodium sulfate is also an unavoidable product of the Na-ferrate synthesis (Equation (2)). Consequently, it was suggested to react iron sulfate with NaOH, allowing them to form a single solid (mixture of Fe(OH)2 and Na2SO4), which would be suitable for the subsequent fluidization in the FB.

The following paragraphs will describe the experimental results of the solid premixing concept.

$$\text{FeSO}\_4\cdot\text{H}\_2\text{O} + 2\text{Cl}\_2 + 8\text{NaOH} \rightarrow \text{Na}\_2\text{FeO}\_4 + \text{Na}\_2\text{SO}\_4 + 4\text{NaCl} + 5\text{H}\_2\text{O} \tag{2}$$

FeSO4·H2O + Cl2 → No evident reaction at room temperature (3)

$$\text{2NaOH} + \text{Cl}\_2 \rightarrow \text{2NaCl} + \text{H}\_2\text{O} + 0.5\text{O}\_2\tag{4}$$

$$\text{FeSO}\_4\cdot\text{H}\_2\text{O} + 2\text{NaOH} \rightarrow \text{Fe(OH)}\_2 + \text{Na}\_2\text{SO}\_4 + \text{H}\_2\text{O} \tag{5}$$

#### *3.3. Premixing of NaOH Pellets (2 mm) with Iron Sulfate*

Sodium hydroxide conditioned as 2-mm pellets was used for the experimental tests. Iron sulfate monohydrate was chosen as iron salt for the same tests. The reaction of these solids was carried out in a rotary reactor of 2.5 L without presence of any gas and using an apparatus of premixing step as illustrated in Figure 2. About 170 g of solids with a molar ratio Na/Fe close to 8 (to satisfy Equation (2)) were loaded in the reactor and rotated at a speed of 20 rpm. The premixing time was fixed at 30 min. The evolution of temperature during the reaction was recorded and the data are displayed as plots of temperature versus time in Figure 4.

**Figure 4.** Evolution of the temperature as a function of time for the reaction of FeSO4·H2O during their mixing in the rotary reactor (four trials run in same conditions).

As shown by this figure, the reaction is exothermic leading to a temperature of about 60 ◦C in the reaction zone. The NaOH pellets become gray-black, but they more or less keep their initial shapes. Images of initial substances NaOH pellets (visible microscopy (VM)) and FeSO4·H2O (SEM) are shown in Figure 5a,b, respectively. Figure 5a clearly shows that NaOH consists of spherical particles of a diameter lower than or equal to 2 mm. As a contrast, FeSO4·H2O is composed of grains of different shapes with an equivalent diameter lower than 30 μm (Figure 5b).

**Figure 5.** View of sodium hydroxide and ferrous sulfate samples: (**a**) Visible microscopy (VM) images of NaOH pellets; (**b**) SEM images of FeSO4·H2O.

Some relevant information about the reaction of NaOH with FeSO4·H2O was revealed by SEM-EDS investigation. A representative NaOH grain after reaction with iron sulfate (image SEM) and its elemental analyses (EDS) are grouped in Figure 6.

**Figure 6.** Scanning electron microscope coupled with energy dispersive spectrometry (SEM-EDS) examination results of a NaOH pellet reacted with FeSO4·H2O: (1) core of the NaOH pellet; (2) outer part of the NaOH pellet.

The examination of these results suggests the following deductions:


XRD analysis showed the presence of NaOH and NaOH·H2O as predominant crystallized phases in the mixture. The iron-bearing phase was not revealed by XRD. However, the chemical analysis of the obtained mixture indicated that this mixture contained about 93 g Fe/kg and that 54% of iron was in a three-valence state. The presence of Fe(III) was also confirmed by Mössbauer spectroscopy measurements.

These pre-mixing materials were subjected to the Na-ferrate synthesis in a fluidized bed using diluted chlorine as an oxidation reagent.

#### *3.4. Preliminary Fluidization Tests of the Premixing Materials in the Fluidized Bed*

As it is well known, the fluidization of a solid by gas depends on the physical characteristics of the solid such as the density, particle size, shape, and those of the gas (gas viscosity and density).

The Reynolds (*ReMF*) and Archimedes (*Ar*) numbers as well as the minimum velocity of fluidization (*UMF*) are calculated by using the relationships available in the literature. The correlation of Wen and Yu [31] is used in this work, which is considered as the most known relationship related to a narrow particle size distribution. Equations (6)–(8) describe the formulas for calculating *ReMF*, *Ar,* and *UMF*.

Reynolds number at minimum fluidization (*ReMF*):

$$\mathrm{Re}\_{MF} = \left[ 33.7^2 + 0.0408 \frac{d\_p^3 \rho\_\mathcal{S} (\rho\_\mathcal{s} - \rho\_\mathcal{S}) \mathcal{g}}{\mu\_\mathcal{g}^2} \right]^{1/2} - 33.7 \tag{6}$$

where <sup>μ</sup>*<sup>g</sup>* is gas viscosity (Pa·s(kg·s−1·m<sup>−</sup>1)); <sup>ρ</sup>*<sup>g</sup>* is gas density (kg·m<sup>−</sup>3); <sup>ρ</sup>*<sup>s</sup>* is particle density (kg·m<sup>−</sup>3); *dp* is particle size (m); and *<sup>g</sup>* is gravitational constant (m·s<sup>−</sup>2).

Archimedes number (*Ar*):

$$Ar = \frac{d\_p^3 \rho\_\mathcal{S} (\rho\_\mathcal{s} - \rho\_\mathcal{S}) \mathcal{g}}{\mu\_\mathcal{g}^2}. \tag{7}$$

Minimum velocity of fluidization (*UMF*):

$$\mathcal{U}L\_{\rm MF} = \mathcal{Re}\_{\rm MF} \frac{\mu\_{\mathcal{S}}}{d\_{\mathcal{P}} \rho\_{\mathcal{S}}} \tag{8}$$

The numeric substitution of the (μ*g*, ρ*g*, ρ*s*, *dp*, *g*) values showed that the mean minimum velocity of the fluidization of the premixing materials is about 1 m·s<sup>−</sup>1. However, these values are approximate ones. For example, the solid density is considered to be 2130 kg·m−<sup>3</sup> (density of NaOH), but it will be less if we consider that the NaOH particles become porous during the reaction with iron sulfate (see photo of Figure 6).

Based on these calculations, an FB of an internal diameter of about 1.75 cm was designed and constructed for the synthesis of Na-ferrate by using the NaOH conditioned as 2 mm pellets.

For the Na-ferrate synthesis, about 10 g of the premixed solids were loaded in the FB, creating a column of 9 to 10 cm. A total gas flow rate of 1100 L/h was necessary to ensure the fluidization of solids. The chlorine content of the used air + Cl2 and N2 + Cl2 gas mixtures was kept at 5.5%, whilst the synthesis time was fixed at 15 min. The temperature of the thermostated water varied from 25 to 55 ◦C.

Visual observations indicated a good fluidization of the solids in FB without dust formation. This reinforces the idea that iron sulfate was well cemented during premixing (NaOH + FeSO4·H2O) in the rotary reactor. The solids after treatment in FB were examined by VM. The images of VM are compared with those of NaOH and premixing solids as shown in Figure 7. The results confirmed that the solids before reaction with chlorine (Figure 7b) and after (Figure 7c) had similar obvious shapes. The purple color of the solid surfaces suggested the formation of Na-ferrate. Furthermore, the dissolution in water of the solids coming from FB gave evidence of iron presence in its hexavalent state. The experimental conditions, as well as the chemical analysis of products issued from the synthesis process, are summarized in Table 1.

The synthesis products contained about 10 to 21 g/kg of iron in a hexavalent state. The best results were obtained when nitrogen was used as diluting gas. This is probably due to the presence of moisture in the air, which can decompose the synthesized ferrate. A higher Fe(VI) yield was achieved when the water bath was regulated at 55 ◦C. These preliminary tests showed the possibility of the Na-ferrate synthesis in a fluidized bed using the mixture (NaOH + FeSO4·H2O) as raw materials. However, as the NaOH-pellets of 2 mm are no longer found in the market, it is suggested to test the synthesis of Na-ferrate using NaOH-pearls of about 1 mm (1000 μm).

(**c**)

**Figure 7.** Visible microscopy images of NaOH-pellets at different steps of the process (the small graduation of the paper in the background is 1 mm large). (**a**) Initial state of NaOH; (**b**) NaOH + FeSO4· H2O (premixing); (**c**) NaOH + FeSO4·H2O + Cl2 (fluidized bed).


**Table 1.** Experimental conditions and Fe(VI) efficiency of synthesis using NaOH pellets of 2 mm.

#### *3.5. Premixing of NaOH Pearls (1000* μ*m) with Iron Sulfate*

The tests carried out with pearls of sodium hydroxide and monohydrated ferrous sulfate in a reactor of 2.5 L (see Figure 2) showed that these reagents did not react like the NaOH pellets of 2 mm. Therefore, it was suggested to add FeSO4·7H2O in order to initiate the reaction. One may note that the particle size of FeSO4·7H2O is about 0.7 mm.

Figure 8 is a typical example of measured temperatures versus time in case of using ferrous sulfate with 3.5 and 4 mol of water (mixture of FeSO4·7H2O and FeSO4·H2O).

**Figure 8.** Evolution of the temperature as a function of time during the mixing of NaOH with iron sulfate (≈0.7 mm) containing 3.5 and 4 mol of water.

The molar ratio Na/Fe was fixed at 8 (to satisfy Equation (2) for the subsequent ferrate synthesis). This figure indicates that it was possible to mix about 325 and 465 g by using a hydration degree of 4 and 3.5 mol for the iron sulfate, respectively. The maximum temperature in the reactor did not exceed 80 ◦C. However, a close examination of the obtained mixture revealed that iron sulfate heptahydrate had not fully reacted with NaOH. This situation is presented in Figure 9 (visible microscope). Figure 9a showed the initial state of NaOH pearls, while Figure 9b represents the VM view of the mixture. As shown by Figure 9b, a large amount of FeSO4·7H2O is oxidized, agglomerated, and was difficult to separate by sieving.

*Iron sulfate oxidized and difficult to be separated*

**Figure 9.** Visible microscopy images of sodium hydroxide (pearls-1 mm): (**a**) Initial state; (**b**) after reaction with FeSO4·4H2O (≈0.7 mm).

To overcome the phenomenon observed above, it was planned to use iron sulfate heptahydrate in powder form. The use of 0.12 mol FeSO4·7H2O in powder form, 0.88 mol FeSO4·H2O, and 8 mol NaOH provided the best results for obtaining the mixture (NaOH + FeSO4·7H2O + FeSO4·H2O). The results of several solid premixing tests are plotted in Figure 10, as the evolution of reactor temperature against reaction time in the case where about of 500 g of solids are used. As it could be expected, the reaction was exothermic, and the maximum temperature oscillated between 80 and 100 ◦C. The output mixture was sieved at +850 μm and was used for the synthesis of Na-ferrate in a fluidized bed.

**Figure 10.** Evolution of the temperature as a function of time during mixing of NaOH with iron sulfate (powder) containing about 1.74 mol of water (four trials run in same conditions).

Figure 11a,b compares the VM images of the initial state of NaOH pearls and that of the obtained mixture according to trials mentioned above. The obtained mixture had a morphology similar to the initial NaOH particles.

**Figure 11.** VM views of sodium hydroxide at: (**a**) Initial state, ≤1 mm; (**b**) After reacting with powder iron sulfate.

#### *3.6. Synthesis of Na-Ferrate in Various Fluidized Beds*

A fluidized bed with a cross section of about 7.07 cm2 (Ø = 3 cm) was used for the synthesis of Na-ferrate based on the NaOH-pearls mixed previously with iron sulfate. It was interesting to follow the evolution of the temperature inside the fluidized bed as the reaction progressed; this was performed by placing a thermocouple in the fluidized bed and recording the temperature.

About 10 g of prepared mixture were loaded in the FB. Tests were performed at regulated water temperatures varying from 20 ◦C to 65 ◦C. The total flow rate of (air + Cl2) and (N2 + Cl2) gaseous mixture was 1800 L/h corresponding to the operational fluidization velocity for the particles in the FB of Ø 3 cm. The data obtained for a water temperature set at 35 ◦C are plotted in Figure 12a,b when air + Cl2 and N2 + Cl2 are used, respectively. The chlorine contents of both gaseous mixtures were fixed at 0.6% and 2.2% Cl2.

**Figure 12.** Plots of temperature evolution versus time during the synthesis of Na-ferrate in FB (Ø 3 cm) at 35 ◦C for: (**a**) air + Cl2; (**b**) N2 + Cl2.

These figures clearly show that the temperature inside the FB increased sharply during the first minute of treatment. Use of 2.2% Cl2 led to a higher maximum temperature level in both cases. The maximum temperatures observed were slightly lower when nitrogen was used instead of air. These observations are also valid over a water temperature range between 20 ◦C and 65 ◦C. The obtained products from these tests were examined by dissolving them in water. It was observed that most of the synthesized product did not show evidence of ferrate presence in the product. Data analyses suggested that the partial pressure of chlorine in the system was too low, making the oxidation of iron at high valence impossible. In order to have the flexibility of supplying high chlorine partial pressure, it was suggested to use a fluidized bed with a smaller diameter.

Tests in the FB with Ø 2 cm were carried out in a similar procedure as described for the previous case scenario. The total gas flow rate was 800 L/h corresponding to the operational fluidization velocity as in the case of FB with Ø 3 cm. The regulation temperature of water was varied from 18 ◦C to 55 ◦C, while the chlorine content was fixed at 2.5% and 5.0%. The evolution of the temperature as a function of the reaction time had a shape similar to that observed with FB Ø 3 cm.

The visual tests in water for the obtained product indicated that all the synthesis products at 2.5% chlorine were characterized by the absence of Na-ferrate. As a contrast, the synthesis products at 5.0% chlorine and the water regulated temperature higher than 18 ◦C showed some evidences of the presence of Na-ferrate. These results confirmed that the Na-ferrate synthesis is strongly affected by chlorine partial pressure.

Several synthesis products were subjected to a chemical analysis to determine the content in total iron and Fe(VI). A summary of the experimental conditions of the Na-ferrate synthesis in FB of Ø 2 cm and chemical analyses results is given in Table 2. These results indicated that the synthesis product (after 5 min of synthesis) contained between 72 and 74 g/kg of iron with about one half of iron as Fe(VI). Furthermore, the use of air instead of nitrogen decreased slightly the Fe(VI) efficiency.


**Table 2.** Experimental conditions and Fe(VI) efficiency of synthesis in FB of Ø = 2 cm.

Although the obtained results for the Na-ferrate synthesis efficiency were improved in the smallest fluidized bed (Ø = 2 cm), it was observed that the fluidization regime was better in the FB of Ø = 3 cm. For this reason, the fluidized bed of intermediate diameter (Ø = 2.5 cm) was also checked. The experimental procedures of the corresponding tests are similar to those developed for the FB of 3 and 2 cm.

Several experimental results are presented in Figure 13a–d plotted as the evolution of temperature inside the fluidized bed versus the reaction time in a half-logarithm scale. As in the previous case, the synthesis process seems to be completed in 5 min.

**Figure 13.** Plots of temperature evolution versus time during synthesis of Na-ferrate in FB (2.5 cm) using air + Cl2 for regulated water temperature at: (**a**) 25 ◦C; (**b**) 35 ◦C; (**c**) 45 ◦C; (**d**) 55 ◦C.

Table 3 groups the experimental conditions and chemical analysis for three chosen tests of the whole series of experiments performed with FB of Ø = 2.5 cm. The iron content of the synthesis process is about 75 g/kg, while the Fe(VI) yield varied between 43% and 54%. It seems that the Fe(VI) synthesis efficiency decreased when the regulated temperature exceeded 40 ◦C. The fluidization regime in FB of Ø = 2.5 cm is better (more homogenous) than in the case of Ø = 2 cm.


**Table 3.** Experimental conditions and Fe(VI) efficiency of synthesis in FB of Ø = 2.5 cm.

#### *3.7. Recycling of Non-Reacted Chlorine*

In optimum conditions of the fluidized bed tests, the content of chlorine in the gaseous mixtures (air + Cl2 and/or N2 + Cl2) was kept at 5%. As could be expected, only a certain fraction of chlorine is used for the ferrate synthesis. A part of unreacted chlorine was released through off-gases. Attempts were made to recycle this non-reacted chlorine. This was realized by connecting two fluidized beds in series. The off-gases of the first fluidized bed are used to supply the second FB.

About 10 g of the premixing material (single solid) were loaded in each FB. Gas mixture of N2 + Cl2 (5% Cl2), which was introduced in the first FB, also passed through the second FB. A good fluidization of the first FB was observed, while the fluidization of second FB was more difficult, and the material showed a tendency to agglomerate. This is due to the water released during the ferrate synthesis in the first FB and causing a material agglomeration in the second FB. Therefore, it was suggested that the non-reacted chlorine can be recycled after a drying process to eliminate any trace of humidity. This can be easily achieved on an industrial scale, guaranteeing that no gaseous effluent is generated during the ferrate synthesis through the proposed process.

#### *3.8. Environmental Considerations of the Process*

Although this research is devoted to the preparation of alkali metal ferrates(VI) at room temperature, only some selected literature reports [32–43] prove the attractive properties of Fe(VI) for different end-use applications. Adjectives such as "environmentally friendly oxidant", "green oxidant", "strong oxidant", "powerful oxidant", and "super-iron battery" are often reserved for this class of materials.

The synthesis of ferrates, as described by this process, should be considered as a "green chemistry" process. The following paragraphs list some supporting evidence of this by comparing the principles of green chemistry [44] with our suggested process.


#### **4. Conclusions**

In this research, the possibility of transforming an industrial waste into supper-oxidizing materials (alkali metal ferrates) containing iron in its hexavalent state is shown. The following conclusions may be drawn from this investigation:

The proposed process for the synthesis of alkali ferrates included two main steps: (i) Premixing of NaOH with iron sulfate (solid–solid reactions) leading to a single solid, and (ii) fluidization of the obtained mixture in diluted chlorine (gas–solid reactions).

The premixing of the solids was achieved in a rotary reactor and the overall reaction was exothermic. The number of solids, the NaOH particle size, the Na/Fe ratio, and the moisture of the input materials (FeSO4·H2O and/or FeSO4·7H2O) affected this step. Temperatures lower than 100 ◦C were required to obtain good results for the premixing.

The output materials of the premixing, a single solid, contains about 80 to 90 g of Fe/kg of iron depending on the experimental conditions. Both Fe(II) and Fe(III) are present in the obtained mixture.

Various fluidized beds (with different cross sections) were designed and tested for the fluidization of the already prepared mixture. The fluidization was realized by air + Cl2 and/or N2 + Cl2 assuring operational fluidization velocities.

The synthesis tests are carried out by varying parameters such as temperature of the thermostated water, partial pressure of chlorine, type of NaOH (2-mm pellets and 1-mm pearls), reaction time, etc.

The results suggest that fluidization is easy to achieve and that almost no dust is generated during the Na-ferrate synthesis in the fluidized bed, indicating that the iron sulfate is well embedded in the NaOH grains during the premixing step. The synthesis process is exothermic, and it is completed within a few minutes. Heat and water are rapidly evacuated from the reaction zone, leading to a dried Na-ferrate. This is a substantial advantage of performing the ferrate synthesis in a fluidized bed.

It seems that temperatures regulated close to 30 ◦C and temperatures in the fluidized bed lower than or equal to 70 ◦C (due to the exothermic reactions) provide the best results for the Na-ferrate synthesis. The Fe(VI) synthesis efficiency varied between 30% and 55% depending on other experimental parameters.

Ferrates obtained by this process could be used directly for different applications without any additional preliminary treatment (such as crushing), keeping in mind that the mean particle size of the ferrate produced by this invented process is close to 1 mm.

The proposed synthesis process meets numerous green chemistry and sustainable development principles. The synthesis product, alkali metal ferrates(VI), belongs to an advanced materials category with multipurpose functions for water and wastewater treatment as well as for cleaning various industrial effluents.

#### **5. Patents**

Kanari, N. Method of producing ferrates(VI). French patent, publication date: 14 March 2008, no. 2 905 609. Extension at international level: 13 March 2008, no. WO 2008/029046.

**Author Contributions:** Conceptualization, N.K.; formal analysis, C.D., I.F., S.S. and F.D.; investigation, N.K., E.O., C.D. and F.D.; visualization, E.O., I.F., S.S. and F.D.; resources, N.K., C.D. and J.Y.; writing—original draft, N.K., E.O., E.A. and F.P.; writing—review and editing, N.K., E.A., F.P. and J.Y.

**Funding:** A significant amount of this work is performed in the frame of contract no. G5RD-CT-2001-03011 (FP5-GROWTH) of the European Union. Further developments of the work have been supported by ANR program "Investissements d'avenir"—ANR-10-LABX-21-01/LABEX RESSOURCES21.

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

#### **References**


Available online: https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A31992L0112 (accessed on 18 May 2019).


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

## *Article* **Reactivity of Low-Grade Chromite Concentrates towards Chlorinating Atmospheres**

**Ndue Kanari 1,\*, Eric Allain 1, Lev Filippov 1, Seit Shallari 2, Frédéric Diot <sup>1</sup> and Fabrice Patisson <sup>3</sup>**


Received: 28 August 2020; Accepted: 7 October 2020; Published: 9 October 2020

**Abstract:** The most economically important iron-chromium bearing minerals is chromite. In natural deposits, iron(II) is frequently substituted by magnesium(II) while chromium(III) is replaced by aluminum(III) and/or iron(III) forming a complex chromium bearing material. The majority of mined chromite is intended for the production of ferrochrome which requires a chromite concentrate with high chromium-to-iron ratio. Found mostly in the spinel chromite structure, iron cannot be removed by physical mineral processing methods. In this frame, the present work deals with the reaction of chlorine and chlorine+oxygen with selected samples of chromite concentrates for assessing the reactivity of their components towards chlorinating atmosphere, allowing the preferential removal of iron, hence meeting the chromite metallurgical grade requirements. Isothermal thermogravimetric analysis was used as a reliable approach for the kinetic reactivity investigation. Results indicated a wide difference in the thermal behavior of chromite constituents in a chlorinating atmosphere when considering their respective values of apparent activation energy oscillating from about 60 to 300 kJ/mol as a function of the sample reacted fraction. During the chromite treatment by chlorine in presence of oxygen, chromium was recovered as liquid chromyl chloride by condensation of the reaction gas phase.

**Keywords:** chromite; chlorine; thermogravimetric analysis; isothermal treatment; apparent activation energy

#### **1. Introduction**

Chromium is part of the extended group of refractory metals offering beneficial properties for various end-uses and manufacturing applications. As displayed in Figure 1, part of these metals (Nb, W, V, Hf, Ta) belongs to the critical materials, according to the European Union criticality assessment [1]. Although chromium is in the cut-off level to be critical, it ranks third (after Mg and W) from the standpoint of economic importance.

Most refractory metals are found and extracted from their oxide bearing materials, likewise, the only economic source of chromium is chromite (FeCr2O4) ore. Nevertheless, in natural deposits, Mg(II) may substitute Fe(II), while Al(III) and Fe(III) often substitute Cr(III) resulting to a complex chromium bearing mineral with the fairly general formula (Fe,Mg)(Cr,Al,Fe)2O4, encompassing the main end-members such as FeO·Fe2O3 (magnetite), FeO·Cr2O3 (iron chromite), MgO·Cr2O3 (magnesiochromite), MgO·Al2O3 (true/regularly spinel). Being multiple and complete solid solution of the spinel group, the composition of chromite is no fixed, but varies largely and depends on the geographic and geochemical features of its deposits.

The major part of mined chromite goes to the ferrochrome (FeCr) manufacture [2–5] and in turn, the FeCr is intended to stainless steels and chromium bearing alloys production. It seems that chromium has no substitute for these industrial end-uses. According to available data [6], the average annual growth of the world stainless steels production is about 5.84% reaching 50.7 million metric tons in 2018. Selected reports [7–13] from numerous recent research works reported to the scientific journals *Materials* and *Metals* are focused on the chromium bearing steels and alloys showing the importance of these leading areas for chromium utilization driving the needs for the ferrochrome and chromite production. Chemical industry, foundry sands and refractory segments are other end-uses of chromite, but their weight relative to the total chromite demand is minor.

**Figure 1.** The 2017 criticality assessment of raw materials for the European Union.

It must be emphasized that the production of ferrochrome not only requires a chromite ore and/or concentrate with a high Cr2O3 content (46–48% Cr2O3), but also with a chromium-iron ratio above 2 (typically around 2.8). The physical processing of the mineral can be successful for the removal of the chromite ore gangue leading to a concentrate with a satisfactory Cr2O3 content meeting the metallurgical requirement. However, as the major part of iron is found in the lattice structure of chromite, only a chemical broken down of the chromite structure seems appropriate for the removal of iron from chromite. According to Nafziger [14], chemical techniques, such as hydrometallurgical methods, chlorination, roasting and leaching, as well as smelting, are required for increasing the chromium-to-iron ratio of lean chromite ores and concentrates. Recently, a carbo-thermic reduction followed by hydrochloric acid leaching was tried as an efficient method [15] for the extraction of iron from a poor chromium concentrate. Specific research works [16–21] regarding the use of various chlorination agents for chlorination of chromite ores and concentrate and their constituents were summarized earlier [2,3].

Our various research reports published previously [2,3,22–24] were focused on the carbochlorination, chlorination and oxychlorination of rich chromite concentrates having a chromium-to-iron ratio of around 3.2 suitable for ferrochrome manufacturing. However, no recent studies were disclosed in the literature regarding to the use of Cl2+O2 for processing lean chromite materials, i.e., having typically a chromium to iron ratio lesser than 2.8. In this regard, the present paper essentially describes the behavior of a poor chromite concentrate with a chromium-to-iron ratio near to 1.5, under Cl2 and Cl2+air atmospheres. The reactivity and behavior of chromite constituents were examined using thermogravimetric analysis (TGA) approach under isothermal conditions. For a better understanding of the processes, the experimental results are compared with those obtained for the treatment of a rich chromite concentrate (of metallurgical grade).

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

The first chromite concentrate sample (low grade) used for this investigation was provided by a European Union manufacturer. A second sample (high grade) provided by an Albanian chromite plant (Bulqiza, Albania) was also used, mostly for comparison purpose. The physico-chemical characterization of the chromite samples and of the reaction products was performed by diverse analytical methods such as chemical analysis (by inductively coupled plasma atomic emission spectroscopy "ICP-AES"), scanning electron microscopy-energy dispersive spectroscopy (SEM-EDS, HITACHI S-4800, Hitachi Ltd., Tokyo, Japan) and X-ray diffraction (XRD, Bruker D8 Advance device, Bruker, Karlsruhe, Germany). Their description was given in previous research works [2,25,26]; only the results will be reported hereafter.

Experimental tests of the reaction of chromite concentrates with chlorine were carried out in a vertical microbalance (model CAHN 1000, Cahn Co., Cerritos, CA, USA) operating at a sensitivity of 10 μg and designed to work under corrosive atmosphere. The equipment configuration with the accessory parts is shown schematically in Figure 2. Several milligrams (more often 40 to 50 mg) of sample were spread out in a silica crucible and the whole specimen was heated up to the desired temperature under nitrogen atmosphere. Subsequently, the nitrogen was replaced by Cl2 and/or Cl2+air (O2) and the evolution of the mass loss over time was recorded. The accessory parts of the setup illustrated in Figure 2 are the units for measuring and purifying the inlet gases as well as those for neutralizing the outlet gases.

**Figure 2.** Setup of the TG analysis experiment.

#### **3. Results**

#### *3.1. Physico-Chemical Characterization of Chromite Concentrate Samples*

The chemical composition in the five main constituents of the first chromite sample is given in Figure 3. As shown, this concentrate is characterized by a high iron content (26.9% wt expressed as FeO) and its chromium-to-iron ratio of 1.48 makes it unsuitable for the FeCr manufacturing. This sample is denoted as low-grade chromite concentrate (LGChC). The XRD patterns of the sample matched well with the (Fe,Mg)(Cr,Fe,Al)2O4 phase. Note that the simple constituents of chromite

(Fe3O4, FeCr2O4, MgCr2O4 and MgAl2O4) are isomorphs having analogous XRD profile making their individual identification difficult. Based on the chemical analysis and supposing a perfect stoichiometric composition without cation/anion deficiency and/or defect, the general formula of the chromite body of LGChC can be approximately represented as (Fe0.50,Mg0.50)(Cr1.20,Al0.60,Fe0.20)O4 and the chromite body can also be expressed as 15.1% Fe3O4, 37.6% FeCr2O4, 26.4% MgCr2O4 and 20.8% MgAl2O4 (Figure 4).

**Figure 3.** Chemical composition of the chromite concentrate samples.

**Figure 4.** Mineralogical composition of the chromite concentrate samples.

The second sample had a chromium-to-iron ratio close to 3.2 and is labelled as high-grade chromite concentrate (HGChC). Although a rich concentrate with 47.7% Cr2O3 (Figure 3), it contained high amount of gangue (close to 7% SiO2 belonging also to olivine and serpentine minerals). The chromite body was separated from the gangue by successive physical separations (using dense liquor) and was defined as (Fe0.30, Mg0.70)(Cr1.56,Al0.37,Fe0.07)O4 with the average composition of end-members 4.4% Fe3O4, 30.9% FeCr2O4, 51.0% MgCr2O4 and 13.7% MgAl2O4 (Figure 4). The mean particle size of both chromite samples used for this study is less than 100 μm. As the reactions of HGChC with Cl2+CO, Cl2 and Cl2+O2 gaseous mixtures were studied thoroughly in previous research work [2,3,22–24], its behaviour in the chlorinating atmosphere is used as a reference to explain the phenomena observed during the treatment of the low-grade chromite concentrate.

#### *3.2. Behavior of Chromite under Chlorine Atmosphere*

Envisaged reactions of the complex chromite constituents (Fe3O4, FeCr2O4, MgCr2O4 and MgAl2O4) and those of simple oxides (FeO, Fe2O3, Cr2O3, MgO and Al2O3) with chlorine may be represented by Equation (1) through (9). The value of standard free energy changes (ΔrG◦) at 900 ◦C is computed from HSC thermochemical database [27] and is indicated beside each reaction. According to these values, the reactions of chromite constituents with Cl2(g) (Equations (1)–(4)) proceed with ΔrG◦ > 0 indicating a nonspontaneous process in the forward direction; these reactions will absorb energy from its surroundings in order to take place. Among the reactions of simple metals oxides of the chromite with chlorine involving their respective chlorides (Equations (5) to (9)), only the reaction of FeO seems to be a spontaneous reaction from a thermodynamic point of view; Cr2O3 and Al2O3 are the most stable oxides in chlorine atmosphere.


An important point of the thermodynamic findings is that the thermodynamic reactivity of the complex constituents of the chromite in chlorine is decreasing according to the following sequence:

$$\rm Fe\_3O\_4 > FeCr\_2O\_4 > MgCr\_2O\_4 > MgAl\_2O\_4 \tag{10}$$

The evolution of the vapor pressure of chlorides of main chromite elements (Cr, Fe, Mg, Al, Si) is displayed in Figure 5 [28,29]. It indicates that in the case of the chlorination of chromite constituents, a selective separation of the obtained chlorides is feasible thanks to great differences in their vapor pressure in a selected temperature interval. A special case is the high volatility of chromyl chloride (CrO2Cl2), which will be discussed in Section 3.3.

**Figure 5.** Vapor pressure versus temperature for several chlorides likely to be produced during chromite chlorination.

Based on these thermodynamic predictions and on the work previously performed [2], experimental tests of LGChC were conducted with chlorine alone with a flow rate of 60 L/h. The recorded data are depicted in Figure 6 as percent mass loss (% ML) of the sample over the reaction time. The somewhat atypical curve shape is a first indication of the complexity of the reactions

of chromite with chlorine. As shown by Figure 6b, the first 50% of the sample was quickly chlorinated and volatilized, while the remaining sample appeared more refractory to chlorine. As an example, the isothermal data indicates that 32 min were sufficient to achieve 50% conversion at 950 ◦C while 75% conversion required 180 min at this temperature.

**Figure 6.** Mass change of the sample versus time for the treatment of LGChC in chlorine from 950 to 1040 ◦C: (**a**) General view of the obtained isotherms; (**b**) Zoom in on the graph up to 50% ML.

To get an insight about this change in the curve shape, it is helpful to compare several isothermal TGA plots for both samples, i.e., LGChC and HGChC, as illustrated in Figure 7. The chlorination of chromite concentrates at 800 ◦C (Figure 7a) tends to an asymptote of the %ML beyond 2 h of treatment corresponding to about 50% and 35% ML for LGChC and HGChC, respectively.

Gathering this data with the mineralogical composition of chromite concentrates (Figure 4) and the thermodynamic predictions allows us to hypothesize that only magnetite and iron chromite are chlorinated at 800 ◦C. It was evaluated that these compounds (Fe3O4+FeCr2O4) represent close to 52.7% and 35.3% of the LGChC and HGChC, respectively. With this evidence, one may also conclude that the TG analysis of chromite reactions with chlorine at low temperature can be an effective method for the fair determination of the amount of (Fe3O4+FeCr2O4) contained in the chromite ores and/or concentrates. Although less wide, the difference between the %ML obtained for the LGChC and HGChC is still evident at 950 ◦C (Figure 7b) and 1040 ◦C (Figure 7c).

The direct application of well-known kinetics models [30] for describing the reaction progress and rate expression faces certain difficulties related to successive reactions, altering of the physical and chemical reactivity of the remaining sample, inter-reaction between gaseous reaction products and the working sample, etc. Hence, it was suggested that the best way to evaluate the temperature impact on the chromite reactions with Cl2 was to calculate the reaction rate in increments of 5% mass losses in the interval ranging from 5.0% to 85.0% ML. This is performed for all isothermal data from 950 to 1040 ◦C. Shown in Figure 8 is the graphical representation of the processed data at 975 ◦C. Besides data linearization expressing the reaction rate, the correlation coefficient (R2) of data fitting, for each segment of 5% ML, is also shown.

**Figure 7.** Comparison of the thermal behavior of LGChC and HGChC in Cl2 atmosphere: (**a**) 800 ◦C; (**b**) 950 ◦C; (**c**) 1040 ◦C.

It was stated [2] that the reaction of chlorine with chromite concentrates generated metal chlorides having volatilization rate higher than their formation rate, indicating the %ML of the sample expresses directly the fraction of the sample (α) reacted. Therefore, the Arrhenius diagrams displaying the logarithm of the reaction rate plotted versus the inverse of the temperature for each 5% ML segment were drawn and values of the apparent activation energy (Ea) with standard error were computed.

Figure 9 gives Arrhenius' plots for the four chosen reacted fractions. Good fitting of the traced data is obtained with the value of Ea increasing from 58±5 kJ/mol at (0.10 ≤ α ≤ 0.15) to 285 ± 8 kJ/mol at (0.75 ≤ α ≤ 0.80) with a good confidence level. Furthermore, the reaction rate at 1000 ◦C for (0.75 ≤ α ≤ 0.80) is decreased by around 27 times with respect to the reaction rate for (0.10 ≤ α ≤ 0.15) at the same temperature, which seems unusual for the gas-solid reactions with gasification of reaction products and without formation of new solid products.

**Figure 8.** Plot of %ML versus time gathered with mean reaction rates calculated in increments of 5% ML during treatment of LGChC in chlorine at 975 ◦C: (**a**) 5–30% ML; (**b**) 30–55% ML; (**c**) 55–85% ML. The color markers are used to distinguish the segments of the %ML curves for which the linearization has been made for the calculation of the reaction rate.

**Figure 9.** Examples of the Arrhenius diagrams for the reaction of LGChC with Cl2: (**a**) 0.10 ≤ α ≤ 0.15 and 0.35 ≤ α ≤ 0.40; (**b**) 0.55 ≤ α ≤ 0.60 and 0.75 ≤ α ≤ 0.80.

An overall profile of the evolution of Ea as a function of the chromite conversion fraction is displayed in Figure 10. The beginning of the reaction proceeded with an Ea near to 78 kJ/mol and decreased to about 60 kJ/mol at 0.10 ≤ α ≤ 0.20. Such a trend suggests the formation of an intermediate species, unfortunately unknown, decreasing the potential barrier of the reaction. Based on the chemical and mineralogical composition of the LGChC the fraction converted at the beginning of reaction may be attributed to the reaction of Fe3O4 with Cl2 involving ferric chloride as final product as it is highly volatile at this temperature range (Figure 5). Thereafter, an increase of the Ea is observed reaching about 175 kJ/mol at (0.45 ≤ α ≤ 0.55). One may attribute globally this value of Ea to the reaction of iron chromite (FeCr2O4) with chlorine. Next, at α > 0.60, the apparent activation energy increased again reaching values as high as 304 kJ/mol for α ranging from 0.70 to 0.75. This conversion fraction corresponds most probably to the removal of MgCr2O4 from the chromite body. According to obtained apparent activation energy and reaction rate trends, the decreasing reaction rank of chromite constituents with chlorine follows the sequence:

$$\rm Fe\_3O\_4 > FeCr\_2O\_4 > MgCr\_2O\_4 > MgAl\_2O\_4 \tag{11}$$

**Figure 10.** Plot of the apparent activation energy (Ea) versus fraction reacted for the treatment of LGChC in chlorine atmosphere between 950–1040 ◦C.

In other words, a higher reaction rate was associated with a lower value of Ea and vice versa. It is hence concluded that the reactivity of chromite constituents towards chlorine is in good agreement with the apparent activation energy and the reaction rates are sufficiently different to achieve a selective elimination of one constituent without affecting the other constituents. Another point is also to be mentioned that this reaction sequence (Equation (11)) seems to match well with the sequence based on thermodynamic predictions and shown in (Equation (10)).

#### *3.3. Reactions of Chromite with Chlorine in Presence of Oxygen*

Having obtained information on the reaction of chromite with chlorine, it was useful to investigate the impact of oxygen on the reaction kinetic and involved products. The chemical reactions of the two main chromite constituents, FeCr2O4 and MgCr2O4, with chlorine +oxygen can be described by Equations (12) and (13), respectively. The values of ΔrG◦ are still positive, but they are much lower than those obtained for the chlorination with chlorine alone (Equations (2) and (3), respectively).


One reaction of particular interest in the system Cr-O-Cl is that of the chromium trioxide (Cr2O3) with chlorine in presence of oxygen with overall reaction described by Equation (14). As shown, the reaction consumes oxygen leading to the formation of CrO2Cl2 as final reaction product. The computed value of ΔrG◦ (900 ◦C) is 19.47 kJ/mol instead of 80.33 kJ/mol for the chemical reaction of Cr2O3 with chlorine solely generating CrCl3 at the same temperature (Equation (7)).

This thermodynamic assessment is completed with a kinetic study of the Cr2O3 and Cl2+O2 interaction using TG isothermal tests and varying the chlorine content from 0% to 100% Cl2. Three typical TGA curves at 50%, 80% and 100% Cl2 are plotted in Figure 11a as evolution of %ML versus reaction time. More than 160 min were required to reach 75% of the Cr2O3 sample reacted in Cl2 alone, while the reaction time is decreased to about 112 min when an equimolar (50% Cl2+50% O2) gas mixture was used and decreased again to about 85 min when the chlorine content in the Cl2+O2 was 80%. Data displayed in Figure 11b demonstrates that the initial reaction rate (0.05 ≤ α ≤ 0.40) of the Cr2O3 chlorination with Cl2+O2 had a maximum value at 80% Cl2 corresponding to Cl2 to O2 ratio equal to 4. Such a result matches well with the stoichiometric coefficients of the reaction described by Equation (14) resulting in chromyl chloride as final reaction product.

**Figure 11.** Treatment of Cr2O3 in Cl2+O2 at 800 ◦C: (**a**) Evolution of the sample mass loss versus time; (**b**) Dependency of the initial reaction rate on the chlorine content in the Cl2+O2 gas mixture.

According to this analysis, the LGChC is treated under a flowing gas (Cl2+air) with total flow rate of 61 L/h containing 28 L/h of chlorine and 33 L/h air (i.e., 26 L/h N2 and 7 L/h O2) corresponding to Cl2/O2 molar ratio equal to 4. Complementary data for choosing this gas mixture composition to enhance the chromium oxide reaction rate and to lower the reaction of iron oxides with the chlorinating gas mixtures are found in previously published research reports [31–34].

Shown in Figure 12 is the %ML versus time for the isothermal treatment of the LGChC under the above-mentioned atmosphere between 950 and 1040 ◦C for reaction time up to 200 min. As in the previous cases (Figures 5 and 6), there is an abrupt change on the curve shape after 50% ML, it is more obvious at low temperature and reflected on the reaction progress. For instance, the first 50% ML was reached with a reaction time of 36 min at 950 ◦C, while more than 410 min are needed to reach 75% ML of the sample.

**Figure 12.** Mass change of the sample versus time for the treatment of LGChC in Cl2+air.

For comparison, the acquired TG data for the reactions of LGChC and HGChC with chlorine in presence of oxygen are shown in Figure 13. The main observations are the high difference between %ML of HGChC and LGChC and the slope change of the %ML vs. time (ML rate), which is clearly more pronounced at 950 ◦C for both materials, although this evidence is also visible at 1025 ◦C. Combining the chemical and mineralogical composition of both concentrates with these isoconversion data lead to assign the kinetic changes to the substantial reaction of MgCr2O4 with chlorine for the conversion higher than 35% and 50 % for HGChC and LGChC, respectively. Both thermodynamic and kinetics reactivity may explain this particular behaviour of magnesiochromite in the Cl2+air gaseous mixture.

**Figure 13.** Comparison of the isothermal data for LGChC and HGChC treatment at 950 and 1025 ◦C under chlorine in presence of oxygen.

To have an idea about the evolution of the elemental and mineralogical composition of the treatment residue, the HGChC reacted at various α-values was examined by SEM-EDS (Figure 14). This method of analysis was preferred to XRD due to the presence of chromite crystalline isomorph phases.

**Figure 14.** SEM-EDS results of residue from the oxychlorination of HGChC at different α-values.

Distinct characteristic peaks of Cr, Mg, Al, Fe, Si and O are present in the SEM-EDS spectrum of the initial sample, in good agreement with the HGChC elemental composition afore-mentioned in Figure 3. The product corresponding to the fraction reacted α = 0.40 does not contain iron. As iron bearing compounds of the chromite body are FeO·Fe2O3 and FeO·Cr2O3, the SEM-EDS analysis of the fraction reacted at α-0.40 is an indirect confirmation for the removal of magnetite and iron chromite from the HGChC. Spectra at α = 0.60 and α = 0.80 with their decreasing chromium peak intensity reflect the evolution of the composition, down to a chromium-free residue at α = 0.88. Accordingly, magnesio-chromite (MgCr2O4) had reacted at 0.35 < α < 0.88, while the true spinel (α = 0.88) appeared more refractory to chlorine in presence of oxygen. Reasoning by analogy, the reaction pathways of the LGChC with chlorine in presence of oxygen will be similar, with the conversion fraction agreeing with its mineralogical composition (15.1% Fe3O4, 37.6% FeCr2O4, 26.4% MgCr2O4, 20.8% MgAl2O4) displayed in Figure 4. The reaction rates are higher due to the fast reaction of 52.7% (Fe3O4+FeCr2O4) with chlorine, hence exhibiting more reactive surface for the progress of the reaction.

The Arrhenius plot shown in Figure 15 for the reaction of LGChC with Cl2+air at various α-values illustrates large changes in the apparent activation energy, starting from about 130 kJ/mol at beginning of the reaction, followed by an average Ea of about 80–85 kJ/mol at 0.15 ≤ α ≤ 0.50 and by a final strong increase up to 300 kJ/mol.

**Figure 15.** Plot of the apparent activation energy (Ea) versus fraction reacted for the treatment of LGChC with Cl2+air at 950–1040 ◦C.

To help to interpret these peculiar changes, isothermal treatments under Cl2+O2 (Cl2/O2 = 4) of the main oxides (Fe2O3, Cr2O3 and MgO) of the chromite constituents (Fe3O4, FeCr2O4 and MgCr2O4) were performed up to 1025 ◦C. Note that ferrous oxide (FeO) is transformed into FeCl3 and Fe2O3 under a chlorinating atmosphere [32]. The experimental data showed that the reactivity of these oxides towards Cl2+O2 is widely different. As an example, 90% of the Fe2O3 sample was reacted for 10 min at 950 ◦C. This reaction time, for reaching the same reaction extent, was extended to 60 min and 270 min for Cr2O3 and MgO, respectively at 950 ◦C. As shown in Figure 16a, this trend of the TGA measurements is still evident during the treatment of these oxides at 1000 ◦C. Based on the isothermal data recorded, the Arrhenius plots for the reactions of the above-mentioned oxides with Cl2+O2 are reported in Figure 16b. The Ea values are about 148, 46 and 214 kJ/mol for the Fe2O3, Cr2O3 and MgO reactions, respectively. In addition, the decreasing reaction rate ranking of simple oxides with Cl2+O2 follows the sequence represented by Equation (15).

$$\text{Cr}\_2\text{O}\_3 > \text{Cr}\_2\text{O}\_3 > \text{MgO} \tag{15}$$

**Figure 16.** Treatment of Fe2O3, Cr2O3 and MgO in chlorine in presence of oxygen: (**a**) Evolution of the sample mass loss versus time at 1000 ◦C; (**b**) Arrhenius diagrams between 800 and 1025 ◦C.

Assuming that the chromite constituents (FeO·Fe2O3, FeO·Cr2O3, MgO·Cr2O3) are chlorinated only when both constituents (FeO and Fe2O3; FeO and Cr2O3; MgO and Cr2O3) are chlorinated and taking into account that the whole reaction rate of each chromite constituent is governed by the slowest reaction rate of its simple constituent, one may deduce that:


However, the energy of the chemical binding of the simple constituents in the chromite structure may affect the values of the inherent activation energy and the multistep reaction rates.

The comparison of the kinetic parameters for the reaction of chromite with chlorine and Cl2+air showed a difference in the apparent activation energy values (Figures 10 and 15) essentially for α between 0.20 and 0.55. Two factors may explain that. First, from a thermodynamic point of view, the reaction of Cr2O3 with Cl2 (Equation (7)) is less favourable than that with Cl2+O2 (Equation (14)); the Ea values accordingly appear higher for the FeCr2O4 chlorination with Cl2 alone. Second, the increasing and higher apparent Ea with Cl2 alone may reflect a MgCr2O4 reaction, characterized by a high Ea, starting before the FeCr2O4 reaction is finished.

In spite of these differences, the present study showed an atypical temperature impact on the chlorination of chromite due to the combination of different thermodynamic and kinetics aspects of chromite component reactions with chlorine in absence and/or presence of the oxygen.

As described in the previous sections, TGA tests were performed with small powder samples (40–50 mg) and high flowrates (e.g., 60 L/h Cl2) of reactive gases to attenuate the reaction starvation impact and to enhance mass and heat transfers. To be closer to the practical chlorination process, tests using tenth grams of chromite concentrates were also performed in a horizontal setup described earlier [2] under Cl2+air atmosphere from 700 to 1000 ◦C. The obtained data are depicted in Figure 17 as evolution of the chromium and iron content and chromium to iron ratio of the residues showing the function of the treatment temperature. These results show the preferential removal of iron from 700 ◦C; about 77% of iron and 18% of chromium were extracted during the treatment of HGChC at 900 ◦C for 2 h. The chromium and iron contents of the obtained residue at 900 ◦C were 35.6 and 3.2 wt%, respectively. As shown in Figure 17, the chromium to iron ratio increased rapidly from 3.9 at 700 ◦C to reach values as high as 11.1 at 900 ◦C.

**Figure 17.** Evolution of chromium and iron content and chromium to iron ratio versus temperature during treatment of HGChC in a Cl2+air gaseous mixture.

To gain understanding of the physical state of the chromium bearing phase synthetized during reaction of chromite with chlorine in presence of oxygen, a two-step cooling of the outlet gases was performed, first at room temperature and second at much lower temperature (−35◦ C).

The analyses of the solid condensate obtained at room temperature by SEM-EDS technique indicated the absence of chromium in the solid phase. However, a red liquid was isolated and collected (Figure 18) in a glassware vessel emerged in a refrigerated alcohol bath. This liquid produced reddish brown fumes in air and seems to correspond to chromyl chloride (CrO2Cl2) characteristics containing chromium at hexavalent state and it is characterized by a high vapor pressure at room temperature (Figure 5). The CrO2Cl2 synthesis was also reported in other research works [35–40] during thermal treatment of various chromium bearing materials.

**Figure 18.** Visual image of the CrO2Cl2(l) generated during the thermal treatment of chromite by Cl2+air and collected in liquid state at –35 ◦C.

By analogy, the ability of chlorine to oxidize iron (II,III) into iron(VI) in high alkali medium by gas-solid and solid-solid reactions was also demonstrated in recent investigations [41–46].

This research work gave several insights for the evolution of the (Fe,Cr,Mg,Al)-O-Cl system, in the case of chromite, at different temperatures. Nonetheless, more in-depth and detailed studies are needed to complete the current knowledge in such a complex system.

#### **4. Conclusions**

Thermogravimetric analysis technique provides valuable information to fairly evaluate the constituent composition of complex materials such as chromite and to analyze its reactions with Cl2 and Cl2+O2 gaseous mixtures.

The reactions of (Fe,Mg)(Cr,Al,Fe)2O4 constituents, for both chromite concentrates (low grade chromite concentrate-LGChC and high grade chromite concentrate-HGChC) with chlorine in isothermal conditions proceeded by gradual scheme starting by the reaction of iron oxides (Fe3O4) followed by interaction of iron chromite (FeCr2O4). Magnesio-chromite (MgCr2O4) appeared stable in Cl2 and Cl2+O2 atmosphere at temperatures equal to or lower than 800 ◦C.

The overall reaction of LGChC with chlorine is affected differentially by temperature at 950–1040 ◦C, resulting in an apparent activation energy strongly dependent on the degree of conversion, e.g., increasing sharply from about 60 kJ/mol to 300 kJ/mol for fractions reacted of 0.15 and 0.75, respectively. Having a low reactivity, the MgCr2O4 compound required high temperature for the reaction to occur. Similar trends were observed for the reaction of chromite with chlorine in presence of oxygen although the values of the apparent activation energy are somewhat different.

Thermodynamic analysis of the envisaged reactions of the chromite constituents with Cl2 and Cl2+O2 gave complementary elements for further clarifying this particular behavior of chromite in the chlorinating atmosphere.

The kinetics results of the simple oxides (Fe2O3, Cr2O3, MgO) reactions with Cl2+O2 in the interested temperature range was another insightful building block for better understanding the multistep process of chromite processing under chlorine in presence of oxygen.

Low temperatures and short times for the interaction chromite-chlorine favor the preferential removal of iron from the low-grade concentrate, giving a chromite with a chromium-to-iron ratio satisfactory for the ferrochrome production. The presence of oxygen in the system favors the synthesis of pure chromyl chloride.

**Author Contributions:** Conceptualization, N.K., E.A. and L.F.; Formal analysis, E.A., S.S. and F.D.; Investigation, N.K., E.A. and L.F.; Visualization, E.A., S.S. and F.D.; Resources, N.K., L.F. and F.P.; Writing—original draft, N.K., E.A., S.S., F.D. and F.P.; Writing—review and editing, N.K., E.A., L.F. and F.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** Some data used in this paper were part of the PhD Thesis of Ndue Kanari. Another part of this development work has been supported by the French National Research Agency through the national program "Investissements d'avenir" with the reference ANR-10-LABX-21-01 / LABEX RESSOURCES21.

**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* **High-Temperature Mechanical Behaviors of SiO2-Based Ceramic Core for Directional Solidification of Turbine Blades**

#### **Jiangwei Zhong and Qingyan Xu \***

Key Laboratory for Advanced Materials Processing Technology (MOE), School of Materials Science and Engineering, Tsinghua University, Beijing 100084, China; zhongjw16@mails.tsinghua.edu.cn **\*** Correspondence: scjxqy@tsinghua.edu.cn

Received: 13 August 2020; Accepted: 10 October 2020; Published: 14 October 2020

**Abstract:** The high-temperature mechanical behaviors of SiO2-based ceramic cores for the directional solidification of turbine hollow blades were investigated. Isothermal uniaxial compression tests of ceramic core samples were conducted on a Gleeble-1500D mechanical simulator with an innovative auxiliary thermal system. The stress–strain results and macro- and micro- structures of SiO2-based ceramic cores were investigated experimentally. The microstructures were characterized by the scanning electron microscope (SEM). Based on the experimental data, a nonlinear constitutive model for high temperature compressive damage was established. The statistical results of Weibull moduli show that the stability of hot deformation increases with the increase of temperature. The fracture type of the SiO2-based core samples is brittle fracture, but when the temperature exceeds 1400 ◦C, the mechanical behavior exhibits thermo-viscoelastic and viscoplastic property. Under high-temperature (>1400 ◦C) and stress conditions, the strength of the ceramic core is weakened owing to the viscous slip of SiO2, which is initially melted at the temperature of 1400 ◦C. The comparison results between the predictions of nonlinear model and experimental values indicate that the model is applicable.

**Keywords:** superalloy; ceramic core; high temperature; mechanical behavior; auxiliary thermal system

#### **1. Introduction**

In response to the increasing worldwide need for reliable, low-cost, and environmentally compatible generation of energy, the new generation of H-class gas turbines (GT) is developed [1,2]. Ni-based single-crystal (SX) superalloy turbine blades, which are the key hot-end assemblies of the gas turbine engines, can be produced by using the directional solidification (DS) technology [3]. The complex inner cavity formed by the ceramic core provides the possibility for the development of the hollow blade cooling technology. Nowadays, due to the complex thermal stress–strain interactions during DS, the size of blades appears imprecise, and the ceramic core even appears cracked. As a result, the performance of the SiO2-based ceramic core directly affects the dimensional accuracy of the SX hollow turbine blade. Therefore, the high-temperature mechanical properties of the SiO2-based ceramic core are crucial for the preparation of SX hollow turbine blade.

There are some studies focused on the high-temperature mechanical behaviors of SiO2-based ceramic cores [4]. Xu et al. [5] investigated the flexural strength of silica-based ceramic cores doped with different silica nanopowders at 1540 ◦C. The results showed that large quantities of cristobalite were crystallized at 1540 ◦C, which might provide enhanced mechanical property in the casting. Niu et al. [6] found that the ceramic cores with 3 wt% mullite fibers showed excellent properties, such as flexural strength being 22.3 MPa at 1550 ◦C, owing to fiber reinforcing. At the same time, there are two opposite conclusions about the high-temperature properties of SiO2-based ceramic core. For instance, Kazemi et al. [7] found that the increase of zircon content could result in the decrease of cristobalite

formed in situ, owing to cristobalite crystallized on the surface of fused silica particles during heat treatment. This result is contrary to Wang and Hon's report [8], but is in good agreement with the result of Wilson et al. [9]. On the other hand, there are many examples in the literature exploring the static and dynamic hot deformation behaviors of other types of ceramic materials in-depth, such as ZrB2 [10], SiC [11], Al2O3 [12], and ceramic composite material [13]. The methods presented in these articles can be applied to the study of SiO2-based ceramic core.

During the directional solidification process, the ceramic cores will be subjected to mechanical loading at high temperature for a long time. In order to prevent the core fracture, it is very necessary to investigate the high temperature behavior of ceramic core. In this study, an auxiliary thermal system is employed to carry out the thermal compression tests of ceramic core. Constitutive modeling and various characterization methods are used to understand the high-temperature mechanical property of SiO2-based ceramic core.

#### **2. Experimental Procedure**

#### *2.1. Experimental Methods and Design*

The characteristics of fused silica and zircon used as raw materials are illustrated in Table 1. According to the formulation used in actual production, the composition of the samples was 60 wt% fused silica and 40 wt% zircon. Porous silica-based ceramic cores were prepared by using ceramic injection molding. After a series of procedures, such as mixing, ball milling, adding adhesive, and drying, the green bodies were obtained. The sintered samples were subsequently subjected to heat treatment at 1500 ◦C for 30 min and then were removed from the furnace at 1000 ◦C, to the atmosphere at 25 ◦C, for simulating the realistic rapid cooling process during the directional solidification from the heating zone into the cooling zone.

**Table 1.** Characteristics of the used fused silica and zircon as raw materials.


Thermal process of ceramic shell/core during directional solidification is shown in Figure 1a. Since the tendency of core breakage is mainly concentrated at temperatures above 800 ◦C, the hot compression temperatures of ceramic cores were set at 700 ◦C (ST700), 1100 ◦C (ST1100), and 1400 ◦C (ST1400). The sample ST25 was tested at 25 ◦C. The average size of the ceramic cores is 14.77 mm in diameter and 15.25 mm in length (ϕ14.77 mm <sup>×</sup> 15.25 mm). The stain rate of 0.001 s−<sup>1</sup> was chosen. After taking into consideration the inhomogeneity of the ceramic core, we tested three parallel samples for each deformation temperature. The whole process can be represented by Figure 1 and Table 2.

#### *2.2. High-Temperature Experimental System for Mechanical Behaviors*

The Gleeble system has been widely employed in the research of material constitutive model [14,15]. It mainly includes three parts: heating system, mechanical system, and computer control system. The heating system forms a current loop with a loaded metal sample (as a resistor) to heat the metal sample. The heating rate and heating temperature are varied by controlling the current in the sample. Therefore, the Gleeble system is generally unable to measure the high-temperature mechanical properties of non-conductive materials, such as ceramic materials. In order to realize the high-temperature mechanical measurement of non-conductive materials on the Gleeble simulator, we designed and developed an auxiliary thermal device that could expand the measurement material range of the Gleeble simulator [16]. The schematic diagram of the auxiliary thermal system of the Gleeble testing is shown in Figure 2. The measured temperature can reach 1600 ◦C. The temperature control accuracy is ±4 ◦C.

**Figure 1.** (**a**) Thermal process of directional solidification; and (**b**) the schematic diagram illustrating the compression process of SiO2-based ceramic cores.

**Table 2.** Heat treatments of ceramic cores to simulate the directional solidification process and test conditions.


**Figure 2.** Schematic diagram of the auxiliary thermal system (-1 ceramic sample, -2 compression bars, -3 silicon carbide screw tube, -4 corundum tube, -5 insulation fiber box, -6 S-type thermocouple for temperature control, -7 temperature S-type thermocouple for temperature calibration, and -8 temperature control cabinet).

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

#### *3.1. High-Temperature Mechanical Properties*

Figure 3 shows the stress–strain curves variation ranges of the samples ST25, ST700, ST1100, and ST1400, from which can be found ST25, ST700, and ST1100 are all brittle fractures, while ST1400 shows thermo-viscoelastic and viscoplastic property. The average compressive strengths of ST25, ST700, and ST1100 are 51.83, 55.82, and 80.46 MPa, respectively, as shown in Figure 4. It is worth noting that the compressive strength of ST1100 is higher than that of ST25 and ST700. This is mainly due to the conversion of α-cristobalite to β-cristobalite. The densities of α-cristobalite and β-cristobalite

are about 2.32 and 2.22 g/cm3, respectively, and the expansion of the β-cristobalite volume makes the sample more compact [17]. More α-cristobalite is transformed into β-cristobalite at 1100 ◦C, and the strength of β-cristobalite is stronger than that of α-cristobalite [18]. At the same time, Zener pinning would be more pronounced at higher temperatures [9].

**Figure 3.** Stress–strain experimental data zone of isothermal uniaxial compression tests: (**a**) ST25, (**b**) ST700, (**c**) ST1100, and (**d**) ST1400.

**Figure 4.** Compressive strength and elastic modulus of samples (ST25, ST700, ST1100, and ST1400).

In the elastic stage, the stress–strain curve of sample ST25 has a good linear-elastic regime. With the increase of temperature, the stress–strain curves of ST700 and ST1100 have a large amplitude and demonstrate a certain degree of dispersion. The elastic moduli of ST25, ST700, ST1100, and ST1400 are 2726, 2259, 2316, and 1442 MPa, respectively. The elastic modulus of the SiO2-based ceramic core shows a decreasing trend when the temperature is increased. There is a small change in the range of 25~1100 ◦C, while the elastic modulus decreases rapidly at the range of 1100~1400 ◦C. The stress–strain curve of ST1400 at the viscoplastic stage is narrow, indicating that the high-temperature experimental result achieves high repeatability and reproducibility. However, the overall high-temperature mechanical property of the sample ST1400 decreases significantly.

On the other hand, it can be found from Figure 4 that the dispersion properties of compressive strengths vary greatly with the increase of temperature. The samples ST25 and ST700 demonstrate large dispersion, while samples ST1100 and ST1400, especially sample ST1400, show less dispersion. To quantify this result, a universal empirical model called Weibull approach is introduced. The three-parameter Weibull distribution function can be simplified to a two-parameter Weibull distribution function without affecting its accuracy [10,19]:

$$P(\mathbf{x}) = 1 - \exp\left[-\left(\sigma/\sigma\_0\right)^m\right] \tag{1}$$

where *P*(*x*) is the failure probability, σ<sup>0</sup> is shape factor, and *m* is Weibull modulus. Generally speaking, the larger *m* indicates the material is more uniform and less dispersion.

From Figure 5, it can be found the Weibull moduli of ST25, ST700, ST1100, and ST1400 are 12.24, 10.87, 18.65, and 38.39, respectively. Obviously, ST1400 demonstrates the largest modulus, indicating that the hot deformation of sample at 1400 ◦C tends to exhibit certain stable and repeatable characteristic. At the same time, as the moduli of ST700 and ST25 are relatively small, it can be concluded that the deformation at 700 and 25 ◦C will be quite unstable. In fact, these results furtherly confirm the difference of stress–strain curves in Figure 3 from the angle of data.

**Figure 5.** Weibull distribution of high-temperature compressive strengths for SiO2 ceramic cores: (**a**) ST25, (**b**) ST700, (**c**) ST1100, and (**d**) ST1400.

#### *3.2. Microstructures Evolution*

As is well-known, the mechanical properties are always associated with microstructures evolution. Figure 6 exhibits the macrostructural investigations of the samples (ST25, ST700, ST1100, and ST1400) fracture surfaces. The difference in the fracture patterns of ST25, ST700, ST1100, and ST1400 can be clearly distinguished (Figure 6a–d). ST25, ST700, and ST1100 are crushed brittle fractures. As the temperature increases, the average size of residual pieces increases. When the strain of ST1400 reaches 0.04, the sample can maintain a substantially complete shape, and only minor fragments occur on the cylindrical surface. At the same time, a sliding plane of approximately 45◦ with the bottom surface of the cylindrical sample can be clearly seen. It is this kind of viscous slip that causes the stress plateau of ST1400 after the strain is greater than 0.02 at 1400 ◦C. The main reason of this sticky slip can be explained in the microstructure morphology.

**Figure 6.** Macrostructural investigation of the sample fracture: (**a**) ST25, (**b**) ST700, (**c**) ST1100, and (**d**) ST1400.

Figure 7 shows the XRD patterns of various samples after hot compression at different temperatures. It can be found that all samples are composed of zircon and α-cristobalite. With the increase of deformation temperature, the peak intensities of two phases have a little change. In order to clarify the phase distribution in the microstructure, EDS point elemental analysis was employed and the results are shown in Figure 8. As shown in Figure 8, the point 1 with gray-black color is SiO2 and the point 2 with bright white color is ZrSiO4 in the image of backscattered electron (BSE).

**Figure 7.** XRD patterns of ceramic samples: (**a**) ST25, (**b**) ST700, (**c**) ST1100, and (**d**) ST1400.

**Figure 8.** EDS analysis of ST1100: (**a**) the image of BSE, (**b**) EDS result of point 1, and (**c**) EDS result of point 2.

Figure 9a,c,e,g shows the images of secondary electrons (SE), and Figure 9b,d,f,h shows the images of the BSE. In the BSE images, the major component of the white-gray phase is ZrSiO4, and the major component of the black-gray phase is SiO2. From the micro-topography of ST25, ST700, and ST1100, it can be seen that the SiO2 particles mainly undergo cleavage fracture, the ZrSiO4 particles mainly undergo dimple fracture. Most penetrated cracks are distributed in larger SiO2 particles. However, there are almost no cleavage fractures in SiO2 particles of the ST1400, and there are little dimple-like ZrSiO4 sections. The surfaces of the small SiO2 particles have a smooth curved shape. The occurrence of material fracture usually has great uncertainly, which is also the reason for the divergence of the stress–strain curves. As mentioned before, the Weibull modulus of ST1400 is much larger than that of ST25, ST700, and ST1100, meaning that the deformation of ST1400 is more stable. Therefore, the microstructure observation results of samples above are relatively consistent with the stress–strain curves.

In the cross-section of ST25, ST700, and ST1100, the surfaces of the large SiO2 particles are clean, and almost no adhesion of fine SiO2 particles is observed. However, in addition to penetrating cracks on the surface of the large SiO2 particles in ST1400, a large number of smooth and fine SiO2 particles are attached to the surface (Figure 9h). It is generally believed that the melting temperature of β-cristobalite is 1720 ± 10 ◦C [17,20]. However, some studies have shown that, when the temperature reaches 1400~1450 ◦C, it will slowly melt on the surface of the SiO2, and the presence of other elements or impurities may reduce the β-cristobalite transformation temperature [21,22]. Therefore, in the high-temperature environment of 1400 ◦C, the main reason for the viscous slip of the SiO2-based ceramic core samples is that the surfaces of the fine SiO2 particles are initially melted, which plays a role in lubrication between large particles. The SiO2, which is initially melted at the temperature of 1400 ◦C, adheres to the surface of the large SiO2 particle. When the temperature drops further to the room temperature, it combines with the large particles, to form a unitary body.

#### *3.3. Nonlinear Constitutive Models for High-Temperature Compressive Damage*

It can be seen from Figures 4 and 5 that the compressive strength and modulus are basically negatively correlated with temperature. The macro-effect of temperature on the properties of ceramic core includes two aspects. On the one hand, the intermolecular forces decrease with increasing temperature. On the other hand, the change of structure caused by the variation of temperature will greatly affect the properties of the material, such as thermal mismatch. Therefore, the thermal damage *D(T)* is employed to describe the temperature effect on the property [23]:

$$D(T) = 1 - E\_T / E\_0 \tag{2}$$

where *E*<sup>0</sup> is elastic modulus at room temperature, and *ET* is the elastic modulus at *T*. The elastic modulus denoted by thermal damage is expressed as follows:

$$E\_T = [1 - D(T)]E\_0 \tag{3}$$

According the analysis of the experimental results at different temperatures, the thermal damage value, *D(T)*, at each temperature point is calculated, as shown in Figure 10. Through data fitting, the expression of thermal damage with temperature variation can be written as follows:

$$D(T) = -0.0328 + 0.00125T - 2.136 \times 10^{-6}T^2 + 1.068 \times 10^{-9}T^3 \tag{4}$$

**Figure 9.** SEM images of the sample fracture surface (red arrows = dimple fracture; yellow arrows = cleavage fracture; blue arrows = high temperature viscous slip; white-gray = ZrSiO4, and black-gray = SiO2). SE: (**a**) ST25, (**c**) ST700, (**e**) ST1100, and (**g**) ST1400. BSE: (**b**) ST25, (**d**) ST700, (**f**) ST1100, and (**h**) ST1400.

**Figure 10.** Thermal damage values of ceramic cores under different temperatures.

According to the author's previous research [24], the continuous damage constitutive model based on Weibull distribution method, at room temperature, is summarized as follows:

$$
\sigma\_1 = E \varepsilon\_1 \exp\left[-\left(\frac{\varepsilon\_1}{\varepsilon\_0}\right)^m\right].\tag{5}
$$

where ε<sup>1</sup> is the axial strain, and ε<sup>0</sup> is a constant. In order to obtain the nonlinear constitutive model for high-temperature compressive damage, the *E* in Equation (5) can be substituted by *ET*, and the formula can be rewritten as follows:

$$
\sigma\_1 = [1 - D(T)] E\_0 \varepsilon\_1 \exp\left[-\left(\frac{\varepsilon\_1}{\varepsilon\_0}\right)^m\right] \tag{6}
$$

The experiment results of typical compression stress–strain of SiO2-based ceramic core and the simulation results based on thermo-visco damage model are presented in Figure 11.

**Figure 11.** The comparison between the experimental results and nonlinear constitutive model prediction: (**a**) ST700, (**b**) ST1100, and (**c**) ST1400.

From Figure 11, it can be found that the nonlinear constitutive model has a good generalization property. In other words, this model could reflect the uniaxial compression behaviors of ceramic cores deformed at various temperatures.

#### **4. Conclusions**


and viscoplastic properties, which mainly can be ascribed to the initial surface melting of SiO2 fine particles.

(3) Nonlinear constitutive model for high-temperature compressive damage is established to predict the hot deformation of ceramic core. The comparison results between the nonlinear model predictions and experimental values indicate that the model is applicable.

**Author Contributions:** Conceptualization, J.Z.; data curation, J.Z.; formal analysis, J.Z.; funding acquisition, Q.X.; validation, J.Z.; writing—original draft, J.Z.; writing—review and editing, Q.X. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Science and Technology Major Project (2017-VI-0003-0073 and 2017-VII-0008-0101) and the Project to Strengthen Industrial Development at the Grass-Roots Level (TC160A310/18).

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

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 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/).

#### MDPI

St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Materials* Editorial Office E-mail: materials@mdpi.com www.mdpi.com/journal/materials

Academic Open Access Publishing

www.mdpi.com ISBN 978-3-0365-8030-2