Synthesis of FeSi-Al 2 O 3 Composites by Autowave Combustion with Metallothermic Reduction

: Fabrication of FeSi-Al 2 O 3 composites with a molar ratio of FeSi/Al 2 O 3 ranging from 1.2 to 4.5 was conducted by the self-propagating high-temperature synthesis (SHS) method. The synthesis reaction involved metallothermic reduction of Fe 2 O 3 and SiO 2 by Al and the chemical interaction of Fe and Si. Two combustion systems were examined: one contained thermite reagents of 0.6Fe 2 O 3 + 0.6SiO 2 + 2Al, and the other had Fe 2 O 3 + 2Al to mix with different amounts of Fe and Si powders. A thermodynamic analysis indicated that metallothermic reduction of oxide precursors was sufficiently exothermic to sustain the combustion reaction in a self-propagating mode. The SHS reaction carrying out co-reduction of Fe 2 O 3 and SiO 2 was less exothermic, and was applied to synthesize products with FeSi/Al 2 O 3 = 1.2–2.5, while the reaction reducing only Fe 2 O 3 was more energetic and was adopted for the composites with FeSi/Al 2 O 3 = 2.5–4.5. Moreover, the former had a larger activation energy, i.e., E a = 215.3 kJ/mol, than the latter, i.e., E a = 180.4 kJ/mol. For both reaction systems, the combustion wave velocity and temperature decreased with increasing FeSi content. Formation of FeSi-Al 2 O 3 in situ composites with different amounts of FeSi was achieved. Additionally, a trivial amount of aluminum silicate was detected in the products of high FeSi contents due to dissolution of Si into Al 2 O 3 during the SHS process.


Introduction
Transition metal silicides have been the focus of many investigations due to their attractive high-temperature properties and a wide range of potential applications. Within the Fe-Si system, iron silicide compounds include Fe3Si, Fe2Si, Fe5Si3, FeSi, -FeSi2, and -FeSi2 [1]. According to the phase diagram, Fe3Si, FeSi, and -FeSi2 are stable at room temperature, while Fe2Si, Fe5Si3, and -FeSi2 are metastable. Depending on their crystal structures, these iron silicides exhibit magnetic, semiconducting, insulating, and metallic behavior [2][3][4]. For example, Fe3Si has a relatively high value of saturation magnetization and is a promising candidate as a ferromagnetic electrode [2]. FeSi has a high melting point, i.e., 1410 °C, good high-temperature structural stability, and excellent resistance to chemical corrosion, especially in harsh industrial environments [5]. Also, FeSi is a transition-metal Kondo insulator and acts as the host structure for the ferromagnetic semiconductor FexCo1−xSi [6]. Si-rich FeSi2 exists in two allotropic forms: orthorhombic semiconducting -FeSi2 at low temperatures and tetragonal metallic -FeSi2 at high temperatures. -FeSi2 is a potential optoelectronic and thermoelectric material, owing to its direct band gap of 0.87 eV, with a wide range of absorption spectrum and relatively large Seebeck coefficient [7][8][9].
Among various methods for fabricating transition metal silicides, combustion synthesis in the mode of self-propagating high-temperature synthesis (SHS) has the ad-vantages of energy efficiency, short processing time, low cost, and high purity of the product [10][11][12]. The SHS scheme has been extensively applied to synthesize many silicide compounds of the Ti-Si, Nb-Si, Zr-Si, Ta-Si, and Mo-Si systems from elemental powder compacts of their corresponding stoichoimetries [13][14][15][16][17]. However, weak combustion exothermicity restrains the preparation of iron silicides from the direct SHS reaction between Fe and Si powders. Despite the fact that the reactions are exothermic, Fe + Si  FeSi has an enthalpy of formation of Hf = −79.4 kJ/mol and an adiabatic temperature of Tad = 1650 K and Fe + 2Si  FeSi2 has Hf = −81.2 kJ/mol and Tad = 1310 K [18]. It has been empirically concluded that the autowave synthesis reaction will not be achievable unless the adiabatic combustion temperature is higher than 1800 K [19]. As a consequence, several activation approaches have been employed. By using high-energy ball milling to mechanically activate the self-sustaining combustion of Fe + 2Si powder mixtures, Gras et al. [20,21] obtained FeSi as the main product for the operation conditions of short duration milling (1-4 h) and moderate shock power (2.2 W), while -FeSi2 was formed in the postcombustion zone for those of long duration milling (6 h) or high shock power (4.8 W). The role of 20 wt% KNO3 as an additive to chemically activate the reactivity of Fe + 2Si mixtures was examined [22]. It was found that decomposition of KNO3 generated enough thermal energy to assist the combustion reaction in a stable and self-sustaining mode, and that the resulting product consisted of FeSi and -FeSi2 [22]. Moreover, the FeSi-Al2O3 composite was synthesized by the mechanical alloying of SiO2, Al, and Fe powders, and complete reduction of SiO2 by Al was reached after 45 h of milling [23].
As the first attempt, this study aims to fabricate FeSi-Al2O3 composites by the SHS process involving metallothermic reduction of metal oxides. By taking advantage of the highly-exothermic reaction of Fe2O3 with Al, this study employed two thermite reagents, Al-Fe2O3 and Al-Fe2O3-SiO2, blended with Fe and Si powders to serve as the reactant mixtures which were formulated to have a broad range of the starting composition. The effects of thermite mixtures and sample stoichiometries were investigated on the reaction exothermicity, combustion wave velocity and temperature, and phase compositions of the final products. In addition, the activation energy of the synthesis reaction was experimentally deduced from combustion wave kinetics.
where stoichiometric coefficients x and y are related to the amounts of Fe and Si powders in the reactant mixture, and also signify the molar ratio of FeSi/Al2O3 formed in the final product. The metallothermic reaction of Fe2O3 + 2Al  2Fe + Al2O3 is extremely exothermic with Tad = 3622 K, while exothermicity of aluminothermic reduction of SiO2 (Tad = 1760 K) is much weaker [24]. This suggests that Reaction (1) is less energetic than Reaction (2), since co-reduction of Fe2O3 and SiO2 occurs in Reaction (1). Moreover, because FeSi has low formation exothermicity (Tad = 1650 K), the increase of Fe and Si additions imposes a dilution effect on combustion. Accordingly, this study performed Reaction (1) with x varying from 1.2 to 2.5, within which stable and self-sustaining combustion was reached. It is apparent that the minimum value of x for Reaction (1) is 1.2, where no elemental Fe is added. When Reaction (1) was conducted with x > 2.5, it was found that, probably owing to inadequate reaction exothermicity, combustion was barely self-propagating and the phase conversion was incomplete.
On account of the reaction exothermicity, Reaction (2) is more favorable for the formation of higher contents of FeSi in the product. Stable and self-sustaining combustion for Reaction (2) occurs in the range of 2.5 ≤ y ≤ 4.5. For Reaction (2) with y = 2.0 (i.e., no addition of elemental Fe), violent combustion caused eruption of the reactants and massive melting of the sample, bringing about difficulty in recovering the end product. However, Reaction (2) with x > 4.5 appeared to be deficient in thermal energy to complete the phase conversion.
Calculation of Tad was performed for Reactions (1) and (2) under different x and y, based on the following energy balance equation [25,26] with thermochemical data taken from [18,27].
where Hr is the reaction enthalpy at 298 K, nj is the stoichiometric constant, Cp and L are the heat capacity and latent heat, and Pj refers to the product. The SHS experiment was conducted in a combustion chamber equipped with quartz viewing windows and filled with high-purity (99.99%) argon at 0.2 MPa. Reactant powders were well mixed in a ball mill, and then the mixture was uniaxially compressed in a stainless-steel mold at a pressure of 60-70 MPa to form cylindrical test specimens with 7 mm in diameter, 12 mm in height, and a relative density of 55%. The propagation velocity of combustion wave (Vf) was determined from the time series of recorded images. From the time derivative of combustion front trajectories, it was found the flame-front propagation velocity was slightly higher in the early stage right after the ignition, and then the velocity decelerated, finally reaching a nearly constant value. The relatively high propagation velocity in the beginning was attributed to the thermal energy supplied by the igniter. The constant velocity in the later stage represents the self-sustaining propagation of flame front and is reported in this study.
The thermocouple used in this study is R-type thermocouple (Omega Engineering Inc., Norwalk, CT, USA) with an alloy combination of Pt/Pt-13%Rh. The wire diameter is 62.5 m, bead size 125 m, and wire length 40 mm. The thermocouple bead was firmly attached on the sample surface at a position about 7 mm below the ignition plane. At this location, self-sustaining combustion was well developed, so that the measurement of combustion front temperature was justified. Details of the experimental setup were reported elsewhere [28]. Phase constituents of the final products were analyzed by X-ray diffraction (XRD) patterns from an X-ray diffractometer (Bruker D2 Phaser, Billerica, MA, USA) with CuK radiation.
The temperature dependence of combustion wave velocity offers a relationship to determine the activation energy (Ea) of the solid-state combustion reaction. The combustion wave velocity derived from an energy equation with a heat source can be expressed as [29,30] where  is the thermal conductivity,  the density, R the gas constant, Q the heat of reaction, Tc the combustion front temperature, and ko the Arrhenius rate constant. Thus, Ea can be obtained from the slope of a best-fitted linear line correlating ln(Vf/Tc) 2 with 1/Tc [30].

Thermodynamic Analysis
The calculated Hr and Tad of Reactions (1) and (2) under different stoichiometric coefficients are presented in Figure 1a,b, respectively. With the increase of x from 1.2 to 2.5 for Reaction (1), Figure 1a shows an increase in Hr from −667.4 to −760.2 kJ, but a decrease in Tad from 3355 to 2847 K. The increase of total reaction enthalpy is justified, because both the aluminothermic reaction and formation of FeSi release heat. However, due to weak exothermicity of the FeSi formation, the value of Tad decreases with increasing content of FeSi. For Reaction (2) with y increasing from 2.5 to 4.5, Figure 1b shows a rise in Hr from −975.8 to −1118.6 kJ and a fall in Tad from 3418 to 2872 K. Similar trends are observed for both reaction systems and Reaction (2) is more energetic than Reaction (1). At the same coefficients of x and y equal to 2.5, it can be seen that Tad of Reaction (2) is higher than that of Reaction (1). Thermodynamic analysis confirms that both reaction systems are thermally satisfactory under the selected conditions. Because the calculated adiabatic temperatures are higher than the criterion proposed by Merzhanov [19], self-sustaining combustion can occur.
Besides thermodynamic considerations, the synthesis of FeSi-Al2O3 composite via autowave combustion with metallothermic reduction has to overcome the kinetic limitation of the reaction. Kinetic restraints are caused by incomplete reactivity owing to the presence of diffusion barriers. It is believed that the reduction of Fe2O3 by Al to produce Fe and Al2O3 acts as the initiation step, followed by aluminothermic reduction of SiO2 in the case of Reaction (1). Subsequently, FeSi is produced through the interaction between Fe and Si.  Figure 2a,b illustrate two typical combustion sequences observed in this study and are respectively recorded from Reaction (1) of x = 1.5 and Reaction (2) of y = 4.5. Firstly of all, Figure 2 confirms the establishment of self-propagating combustion for Reactions (1) and (2). It is evident that for both SHS processes, a well-defined combustion front forms upon ignition and propagates along the sample compact in a self-sustaining manner. Secondly, Figure 2a,b signify different combustion behavior. As shown in Figure 2a, the flame propagation was fast and it took about 3.33 s for the combustion wave to traverse the sample. Moreover, combustion was in companion with obvious melting of the sample, suggestive of strong exothermicity of this test condition. On the other hand, Figure  2b exhibits a relatively slow combustion wave with a long spreading time of about 6 s and that the melting of the sample is considerably alleviated. Namely, a comparison between Figure 2a,b indicated that the increase of Fe and Si additions changed the combustion behavior from a rapid/molten mode to a moderate style and lowered the flame-front speed, due to a decrease in combustion exothermicity.   (2) with their corresponding stoichiometric coefficients, x and y. A significant decline in the combustion wave speed from about 3.9 to 1.1 mm/s was observed for Reaction (1) with increasing x from 1.2 to 2.5. The decrease of combustion front velocity was ascribed to the dilution effect of Fe and Si additions, because the formation of FeSi was less energetic than aluminothermic reduction of oxide reagents. Similarly, the combustion velocity of Reaction (2) decreased from 4.5 mm/s at y = 2.5 to 1.5 mm/s at y = 4.5. The combustion propagation rate is mostly governed by the layer-by-layer heat transfer from the reaction zone to unburned region. Namely, the reaction zone temperature has a great influence on combustion wave velocity.

Self-Sustaining Combustion Wave Kinetics
Typical combustion temperature profiles measured from the samples of Reactions (1) and (2) are depicted in Figure 4. The profile features a sharp rise to the highest point, followed almost immediately by a considerable decline owing to heat losses to the sur-roundings. This implies a thin reaction zone. The sudden increase signifies the rapid arrival of the combustion wave. The peak value represents the combustion front temperature (Tc). A decrease in Tc with stoichiometric coefficient was detected for both reaction systems. Figure 4a indicates that Tc of Reaction (1) decreases from 1710 °C at x = 1.2 to 1413 °C at x = 2.5, confirming the dilution effect on combustion by increasing Fe and Si additions. Likewise, as shown in Figure 4b, the increase of y from 2.5 to 4.5 lowers the Tc of Reaction (2) from 1826 to 1506 °C. Due to the fact that the thermite mixture of Fe2O3 + 2Al is more exothermic than that of 0.6Fe2O3 + 0.6SiO2 + 2Al, the value of Tc of Reaction (2) is higher than that of Reaction (1) under the same stoichiometric coefficient of 2.5. Moreover, the dependence of combustion front temperature on sample stoichiometry is in a manner consistent with that of combustion wave velocity.
The accuracy of fine-wire thermocouple measurement is essentially affected by the conduction cooling and radiation loss. According to Bradley and Matthews [31], the conduction loss from the thermocouple depends on the length of wire between the junction and the support. The dimension of the thermocouple used in this study explains why the measurement was practically unaffected by the conduction cooling [31]. The radiation loss from the thermocouple bead is another major source of error in the thermocouple measurement. An estimate of the radiation correction was performed, assuming that a steady state existed between the convective heat transfer to, and radiation loss from, the thermocouple bead [32,33]. Radiation corrections for the combustion front temperatures measured in this study are in the range from 28 to 36 °C, which is more significant in the regions of higher temperatures. After the radiation correction, it is believed that the accuracy of the combustion temperature measurement is within 10 °C. When compared with the calculated Tad, however, the measured Tc was much lower. This suggests that the burning samples experienced substantial heat losses to the surrounding gas by conduction and convection and to the chamber wall by radiation.   According to combustion wave kinetics [29,30], as mentioned above, the activation energy of solid-state combustion can be determined from the correlation between ln(Vf/Tc) 2 and 1/Tc. Figure 5 plots two sets of experimental data with their respective best-fitted linear lines. For Reaction (1), five data points from left to right were according calculated from Vf and Tc of x = 1.2, 1.5, 1.75, 2.0, and 2.5. Five data points of Reaction (2) from left to right are orderly associated with y = 2.5, 3.0, 3.5, 4.0, and 4.5. From the slopes of two straight lines, the values of Ea = 215.3 and 180.4 kJ/mol were deduced for Reactions (1) and (2), respectively. A smaller Ea for Reaction (2) means a lower kinetic barrier for the synthesis reaction to occur. Because the aluminothermic reduction is considered as the initiation step of the SHS process, co-reduction of Fe2O3 and SiO2 by Al is probably responsible for the large Ea obtained for Reaction (1). When compared with Ea = 145 kJ/mol reported for the Al-Fe2O3 thermite reaction [34], Reactions (1) and (2) of this study have higher kinetic barriers to conquer. This could be most likely caused by the fact that in addition to aluminothermic reduction, Reactions (1) and (2) involve chemical interactions between Fe and Si to form FeSi. These interactions could be complex and occur in both sequential and parallel manners during the SHS process.   Figure 6 might reflect the presence of a tiny amount of impurities. The impurity phase possibly existing in the final products is considered to be aluminum silicate (or called mullite). Mullite is a stable solid solution compound in the Al2O3-SiO2 system. The stoichiometry of mullite refers to Al4+2mSi2−2mO10−m with m varying between about 0.2 and 0.9 and the composition ranges from relatively silica-rich 3Al2O32SiO2 (3:2 mullite) to alumina-rich 2Al2O3SiO2 (2:1 mullite) [35]. Formation of aluminum silicate was attributable to dissolution of a small amount of Si into Al2O3 during the SHS process. To be shown below in Figure 7, the signature peaks of aluminum silicate are detectable, because the amount of Si is increased in the sample.

Phase Constituents of Synthesized Products
The XRD spectra of the final products synthesized from Reaction (2) with y = 2.5, 3.5, and 4.0 are shown in Figure 7a-c. Both FeSi and Al2O3 were identified and the peak intensity of FeSi relative to Al2O3 was augmented in the case of y = 4.0 when compared to those of y = 2.5 and 3.5. Besides, as indicated in Figure 7, aluminum silicate was detected in the final products. As can be seen in Figure 7, the quantity of aluminum silicate slightly increased with increasing FeSi content formed in the product, because the sample contained more Si.

Conclusions
The SHS process involving metallothermic reduction of Fe2O3 and SiO2 by Al and the elemental reaction of Fe with Si was conducted to fabricate FeSi-Al2O3 composites with a broad composition range. Thermite reagents made up of 0.6Fe2O3 + 0.6SiO2 + 2Al and Fe2O3 + 2Al were incorporated into the Fe-Si reaction system. Aluminothermic reduction of metal oxides generated Al2O3 and released a large amount of the reaction enthalpy which was adequate to sustain the combustion reaction in a self-propagating manner. The SHS reaction including the mixture of 0.6Fe2O3 + 0.6SiO2 + 2Al was less ex-othermic than that containing the reagent of Fe2O3 + 2Al. Moreover, the increase of Fe and Si additions for the production of a higher content of FeSi reduced the overall reaction exothermicity. Consequently, the former was adopted to produce the composites with a molar ratio of FeSi/Al2O3 from 1.2 to 2.5, within which a decrease in both the combustion wave velocity from 3.9 to 1.1 mm/s and reaction front temperature from 1710 to 1413 °C was observed. The latter was applied to fabricate the composites with FeSi/Al2O3 from 2.5 to 4.5; within this range, the increase of FeSi/Al2O3 lowered the reaction temperature from 1826 to 1506 °C and decelerated the combustion wave from 4.5 to 1.5 mm/s. Correlation of combustion wave kinetics indicated that the SHS reaction based on 0.6Fe2O3 + 0.6SiO2 + 2Al had a larger Ea of 215.3 kJ/mol in comparison to Ea = 180.4 kJ/mol for the reaction constructed out of Fe2O3 + 2Al. The difference in Ea could be ascribed to co-reduction of Fe2O3 and SiO2 by Al, in contrast to the sole reduction of Fe2O3. The formation of FeSi and Al2O3 in the final products was identified by XRD analysis. The increase of the FeSi content with increasing Fe and Si additions was confirmed. Phase conversion from the reactants to products was completely achieved, except for a trivial amount of aluminum silicate present in the end products with a high FeSi content.