Clean magnesium production using concentrated solar heat in a high-temperature cavity-type thermochemical reactor

Journal of Cleaner Production 232 (2019) 784e795

Contents lists available at ScienceDirect

Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro

Clean magnesium production using concentrated solar heat in a high-temperature cavity-type thermochemical reactor
phane Abanades a, * Srirat Chuayboon a, b, Ste a b

Processes, Materials and Solar Energy Laboratory, PROMES-CNRS, 7 Rue du Four Solaire, 66120, Font-Romeu, France Department of Mechanical Engineering, King Mongkut's Institute of Technology Ladkrabang, Prince of Chumphon Campus, Chumphon, 86160, Thailand

a r t i c l e i n f o

a b s t r a c t

Article history: Received 12 March 2019 Received in revised form 6 May 2019 Accepted 30 May 2019 Available online 1 June 2019

The synthesis of magnesium from the corresponding oxide via a solar carbo-thermal and methanothermal reduction process using high-temperature concentrated solar heat was investigated. The reduction of magnesium oxide (MgO) was experimentally demonstrated in a directly-irradiated prototype solar reactor at reduced pressure and temperature up to ~1650
C. The solar reactor was successfully operated with a variety of reducing agents (carbon and CH4) in batch and continuous modes under atmospheric and low pressure conditions (0.1e0.9 bar), thus representing the first process demonstration of MgO carbothermal reduction with continuous reactant injection in vacuum condition. A parametric study regarding operating pressure, carbon feedstock type, and C/MgO molar ratio was conducted to emphasize their effect on products yield (Mg and CO) and solar reactor performance. MgO conversion, reduction rate, and CO yield increased with decreasing pressure, in agreement with thermodynamic analysis. Utilizing activated charcoal as reducing agent showed the highest MgO conversion and CO yield. High MgO conversion over 99% was demonstrated with maximal CO yield up to 24.59 mmol/gMgO, closely approaching theoretical maximum value (24.81 mmol/gMgO). Employing methane as a reducing agent was also shown to be an alternative option to produce Mg, although methane cracking occurred simultaneously at the elevated reaction temperature. Mg recovery in the outlet products was identified as one of the most critical process challenges because of the pyrophoric property of the produced nanopowder and its strong oxidation reactivity with air. © 2019 Elsevier Ltd. All rights reserved.

Keywords: Concentrated sunlight Carbothermal reduction Vacuum Magnesium Syngas production Solar reactor

1. Introduction Solar energy is the largest unlimited-sustainable energy source among all the carbon-neutral energies. After optical concentration, it can be utilized as an external source of process heat to drive high
, temperature thermochemical reactions (Abanades and Andre 2018; Kodama, 2003; Koepf et al., 2017; Liu et al., 2018). Among the relevant thermochemical reactions, the ones based on metal oxide redox pair systems for the production of synthesis gas (syngas) and metals are of major interest. In general, syngas can be further used as the chemical building block for a wide variety of synthetic hydrocarbon fuels such as methanol and liquid fuels via FischereTropsch process (Klerk, 2013). However, the thermal reduction of metal oxides in two-step redox cycles usually requires high operating temperatures (Abanades and Flamant, 2006),

* Corresponding author. E-mail address: [email protected] (S. Abanades). https://doi.org/10.1016/j.jclepro.2019.05.371 0959-6526/© 2019 Elsevier Ltd. All rights reserved.

leading to significant heat losses and reactor materials issues. Consequently, the utilization of carbonaceous materials such as biomass (Chuayboon et al., 2018a), solid carbon (Koepf et al., 2015; Steinfeld and Fletcher, 1991; Villafan-Vidales et al., 2016), and methane (CH4) (Chuayboon et al., 2019; Nair and Abanades, 2016) as reducing agents in a redox cycle can significantly lower the reduction temperature, thereby alleviating the current process limitations. Various metal oxide redox cycles have been explored and can be mainly classified into two groups: “non-volatile” and “volatile” oxide cycles (Agrafiotis et al., 2015). On the one hand, the nonvolatile oxides such as ferrite (Kodama et al., 2002) and ceria (Fosheim et al., 2019) remain in the condensed state during the entire process, bypassing dramatic structural and phase changes at the expense of low fuel productivity or deactivation issues (sintering). On the other hand, the volatile oxide cycles such as ZnO/Zn ^que and Abanades, 2015), and (Murray et al., 1995), SnO2/SnO (Leve MgO/Mg (G
alvez et al., 2008) exhibit higher fuel production extents as a result of stoichiometric reactions; however at the expense of a

S. Chuayboon, S. Abanades / Journal of Cleaner Production 232 (2019) 784e795

recombination issue with oxygen (O2) during thermal reduction. The use of solid or gaseous reductants eliminates the O2 production (CO or CO/2H2 are instead produced when using solid carbon or CH4 in the reduction step, respectively). Solar-produced Mg is an attractive metal with potential application in H2/CO production (obtained from the reaction of Mg with H2O/CO2 (G
alvez et al., 2008)), power generation in magnesiumbased combustion engines and incendiary weapons, and automobile manufacturing such as magnesium-based alloys (Alam et al., 2011; Mordike and Ebert, 2001). It is highly flammable, especially in the forms of powder or thin strips. The Mg is conventionally produced by a Pidgeon, Magnetherm, or electrolytic method. The Pidgeon and Magnetherm methods require the reduction of calcined dolomite ore with ferrosilicon at high temperatures (1700
C), while the electrolysis requires the reduction of molten magnesium chloride in the temperature range 680e720
C (AbuHamed et al., 2007). They proceed with additional materials (silicothermic process) and require very high energy consumption from either electricity or fossil fuel to drive the reactions. For these reasons, the solar carbo-thermal or methano-thermal reduction of MgO offers a more attractive approach regarding both the utilization of solar energy to drive the chemical reactions, CO2 emission mitigation and the possible operation as cycle towards the coproduction of both Mg and syngas. The carbothermal and methano-thermal reduction reactions of MgO are written as: MgO þ C / Mg þ CO

(1)

MgO þ CH4 / Mg þ CO þ 2H2

(2)

A possible side reaction may occur during MgO reduction with either carbon or methane: C þ CO2 4 2CO

(3)

Previous studies mainly focused on the kinetics of the MgO carbothermal reduction with the utilization of various solid carbonaceous materials such as graphite (Chubukov et al., 2016; Hong et al., 2003; Rongti et al., 2003) and charcoal (Chubukov et al., 2016) via thermogravimetry analysis (TGA). A two-step thermochemical process based on MgO/Mg with charcoal and petcoke for syngas production was thermodynamically and experimentally
lvez et al., 2008). examined via TGA under atmospheric pressure (Ga The reactions kinetics were also studied and compared (the Ea values of MgO reduction with charcoal, 468.4 kJ/mol and petcoke, 419.1 kJ/mol were higher than those reported in previous studies (Hong et al., 2003)). Despite employing carbonaceous materials for decreasing the reduction temperature, the carbothermal reduction of MgO still requires high temperatures above 1600
C at atmospheric pressure for complete conversion (Hong et al., 2003; Rongti et al., 2003). Two main approaches can be considered to avoid gaseous products recombination and allow metal product recovery: rapid quenching of the vapor or dissolving the magnesium directly in a suitable metal solvent before reversion can occur (Brooks et al., 2006). Alternatively, decreasing CO partial pressure also improves Mg yield (Hischier et al., 2016). Vacuum operation has been studied theoretically and experimentally with the MgO carbothermal reduction to lower the temperature (Chubukov et al., 2017, 2016; ^que and Abanades, 2015; Tian et al., 2015, Halmann et al., 2011; Leve 2014; Xie et al., 2016; Xiong et al., 2018; Yang et al., 2014). This concept offers possibilities not only to alleviate heat losses and reactor materials issues but also to increase MgO reduction rate and conversion, at the expense of pumping energy requirement. Chubukov et al. (2016) studied pressure dependent kinetics of MgO

785

carbothermal reduction with carbon black in the temperature range 1350e1650
C and pressure range 0.1e100 kPa in a graphite furnace. They found that the rate of MgO carbothermal reduction increased inversely with pressure, and a transition between conversions of 20e35% occurred after the reaction rate was maximal at 10 kPa. Recently, the same group (Chubukov et al., 2017) studied the kinetic enhancement of MgO carbothermal reduction with petroleum coke by the means of catalysis, milling, and vacuum operation in the same furnace. Most previous reported works were focused on the kinetic rate and pressure-dependence of MgO carbothermal reduction with different carbon types and carried out with an electrical furnace in a batch experiment. Only one research has been reported on the combination of simulated solar energy with electricity to drive the reduction of MgO near 1277
C via an electrolytic process (Sheline et al., 2013), and no prior researches have been performed on real solar process demonstration with continuous reactant injection in a solar reactor under vacuum operation. Using solar energy for process heat in place of fossil fuels is a means to reduce the dependence on conventional energy resources and to avoid emissions of CO2 and other pollutants, while upgrading the calorific value of the feedstock for the production of high-value chemical fuels (both Mg and CO) via the storage of intermittent solar energy. The new solar process is thus a sustainable alternative option outperforming the conventional processes currently used for Mg production. In this study, a thermodynamic and experimental investigation of MgO carbothermal/methanothermal reduction was performed. The main objective of this study was to demonstrate the feasibility of the Mg production process with high MgO conversion, by focusing on the design, operation, parametric study and performance assessment of a new prototype solar reactor. The Mg production was carried out in a solar vacuum reactor driven by real concentrated sunlight and operated in either batch or continuous modes. The results provide insights into the influence of pressure, carbonaceous feedstocks (different solid carbon types or CH4), and C/MgO molar ratio on CO production rate, products yield (CO and Mg), and reactor performance, and the feasibility of MgO carbothermal reduction with continuous reactant injection under vacuum condition is demonstrated. 2. Experimental set-up and methods 2.1. Materials MgO (particle size: 1e2 mm, 99.8% purity) and Mg (particle size: 1e2 mm, 99.8% purity, used for a calibration method) powders were obtained from Alfa Aesar. Concerning solid carbon sources (physical properties given in Table S1 in Supporting Information (SI)), synthetic graphite powder (99.9% purity) and activated charcoal (AC, 99.9% purity) were purchased from Sigma Aldrich, while carbon black (CB) was supplied by Asahi Carbon (Japan). Solid carbon and magnesium oxide powders were mechanically mixed with C:MgO molar ratios of 1.5 and 2 for MgO carbothermal reduction experiments. 2.2. Reactor prototype A new 1.5 kWth prototype solar vacuum reactor was designed and constructed, based on the concept of directly-irradiated cavitytype solar reactor (Fig. 1). This reactor can be operated in either batch or continuous modes under vacuum pressure conditions at high reduction temperatures up to ~1650
C. The reactor is composed of an inner cylindrical cavity receiver made of alumina surrounded by a layer of porous ceramic insulation (Fig. S1), thereby enabling rapid solar heating to the desired temperature.

786

S. Chuayboon, S. Abanades / Journal of Cleaner Production 232 (2019) 784e795

Fig. 1. Schematic diagram of the 1.5 kWth directly irradiated prototype solar reactor and external components.

The bottom of the cavity is sealed with a circle alumina plate pierced at its center so that a small alumina tube can be inserted for the injection of nitrogen (N2) carrier gas and methane (CH4) to the cavity receiver. N2 protective gas is also directly injected into the window area and then subsequently enters downwardly the cavity through the aperture before exiting with the product gases through the outlet port, thereby protecting the transparent window from products deposition. A packed bed of inert alumina particles (2 mm diameter) is placed at the bottom of the cavity receiver above a layer of alumina wool to support the reacting powder at the cavity center. The top of the cavity receiver is closed by an alumina cap with a 17 mm-diameter aperture where concentrated solar radiation enters, and a protective graphite plate (2 mm-thick) with a 15 mm-diameter aperture is then placed on top of the alumina cap to protect it (Fig. S1). A hemispherical transparent glass window is lastly attached to the front flange edge of the reactor shell to operate under controlled atmosphere. During reactor operation, the temperature within the cavity receiver was measured with a type-B thermocouple (T1), and the uppermost sample surface temperature was measured with a solarblind optical pyrometer (operating at 4.8e5.2 mm in a H2O absorption band). The cavity pressure (P) was measured by a pressure transducer while N2 and CH4 flow-rates (purity of 99.999%) were regulated by electronic Mass Flow Controllers (MFC, Brooks Instruments model SLA5850S, range 0e5 Nl/min ±0.2% of full scale). In case of continuous operation (Fig. S2), the reactor was equipped with an automatic particle delivery system consisting of a hopper and a screw feeder driven by an electrical motor (Chuayboon et al., 2018b). All the product gases including N2 carrier gas exit the reactor through a single outlet port (20 mm indiameter alumina tube) at the upper cylindrical sidewall of the cavity and subsequently flow into a ceramic filter where the main

solid products (Mg) are deposited. Indeed, the Mg formed at the reaction temperature is gaseous (Mg boiling point: 1091
C), it is thus entrained by the carrier gas out of the reactor and it subsequently condenses as small particles upon gas cooling. Mg particles are thus collected in the outlet deposits on the inner tube walls (zone A) and in the filter (zone B). 2.3. Procedure The solar vacuum reactor is positioned at the focus of a vertical axis parabolic dish solar concentrator with a solar concentration ratio up to 10,551 suns (2 m diameter, 0.85 m focal distance, peak flux density of ~10.5 MW/m2 for a DNI of 1 kW/m2) at CNRSPROMES, Odeillo (France). The incident concentrated solar power is controlled manually by opening a shutter placed above the suntracking heliostat that vertically reflects incident solar radiation towards the facing down concentrator. Prior to the experimental tests, homogeneous reacting samples are prepared by mechanical mixing between MgO and solid carbon (activated charcoal (AC), carbon black (CB), and graphite) at C/MgO molar ratios of 1.5 and 2. They are subsequently placed directly in the cavity receiver (batch mode) or in the hopper (continuous mode). After complete reactant sample loading, the solar reactor is initially flushed with both N2 flow and vacuum pumping to purge residual air from the system (O2 below 10 ppm). Subsequently, the reactor is gradually heated by highly concentrated sunlight (Fig. 2). During solar heating, the cavity pressure is maintained at ~0.87 bar (Patm ¼ ~0.85 bar at site elevation 1500 m above sea level), and the N2 carrier gas (0.2 Nl/ min) and N2 protective gas (1 or 2 Nl/min) are provided to the reactor cavity and window area, respectively. In case of continuous operation, another N2 carrier gas (0.2 Nl/min) is provided to prevent backflow through the screw feeder.

S. Chuayboon, S. Abanades / Journal of Cleaner Production 232 (2019) 784e795

787

Fig. 2. On-sun experimental testing of MgO carbothermal reduction under vacuum operation (a) during solar run and (b) after experiment: (1) transparent window; (2) reactor shell; (3) type-B thermocouple; (4) ceramic filter; (5) reactor frame to adjust the reactor position with respect to the focal point; (6) filter outlet; (7) connector; (8) protective graphite plate; (9) reactor aperture; (10) alumina tube plug (batch test) or particle feeder position (continuous test); (11) water cooling; (12) cavity pressure tube.

After reaching the temperature of ~900
C, the reactor is evacuated with a rotary vane vacuum pump (Alcatel) to the targeted pressure (0.11e0.16 bar), and meanwhile, a secondary membrane pump connected to the outlet filter is turned on to suck a bypass stream of product gases to an on-line syngas analyzer (GEIT 3100, uncertainty <±0.1% of full scale) for continuous gas analysis. The reaction starts in a temperature range 1000e1200
C (depending on total pressure), reflected by an increase of the CO concentration measured by the gas analyzer. During MgO reduction reaction, solar heat supply rate was kept constant (full shutter opening) and DNI remained stable throughout the tests. Product gases exit the reactor along with Mg vapor and subsequently flow into the ceramic filter. All the measured data are continuously recorded by an automated data acquisition system (BECKHOFF). After the reaction is complete (evidenced by CO production rate approaching zero), the shutter is closed, leading to temperature cool down. Finally, the solid products contained in the removable outlet components (outlet alumina tube and filter) are collected with a hands-in-bag atmospheric chamber once the reactor gets to room temperature for alleviating sample oxidation issues and safe handling of pyrophoric powders. The collected Mg particles were analyzed by calibrated X-ray diffraction (XRD) for phase identification (Philips PW, 1820 diffractometer) with the Cu Ka radiation (aCu ¼ 1.5418 Å, angular range ¼ 20e100
in 2-Theta, step size of 0.02
, recording time ¼ 2 s). Particle morphology analysis was carried out using a field emission scanning electron microscope (FESEM, Hitachi S4800) and Mg powders reactivity was assessed via thermogravimetric analysis (TGA, Setaram Setsys Evolution). 3. Results and discussion 3.1. Thermodynamic analysis The thermodynamic stability domain of chemical species for various MgO reduction reactions either with or without carbonaceous materials is represented by the Gibbs free energy change (DG
) in Fig. 3. According to the DG
variations, the thermodynamically favorable reactions are MgO þ C, MgO þ CH4, and C þ CO2 that proceed at above 1800, 1500, and 700
C, respectively. In contrast, the other reactions (in particular, MgO þ H2 and MgO þ CO) are not able to produce Mg below 2000
C, according to thermodynamics. Direct thermal reduction of MgO to Mg(g) and O2(g) would require temperature in excess of 3430
C, thus making the process not feasible in practice. Note that the formation of CO2

Fig. 3. Variations of DG
for various MgO reduction reactions as a function of temperature.

can only be obtained from Boudouard equilibrium (CO disproportionation is favored below 700
C, Eq. (3)), since the reaction of MgO with CO is not favorable. In addition, thermochemical equilibrium compositions were calculated for both MgO/C and MgO/CH4 systems (using HSC Chemistry). The calculated equilibrium compositions of MgO carbothermal reduction as a function of temperature and pressure are shown in Fig. 4. At 1 bar, the MgO þ C reaction (Fig. 4a) starts at 1200
C and reaches completion at 1600
C, thereby yielding an equimolar gas mixture of Mg(g) and CO(g). For a pressure decreased to 0.1 bar, the MgO þ C reaction starts at 1000
C, and reaches completion at 1400
C, confirming that reduced pressure can theoretically shift the chemical equilibrium to lower temperature. Fig. 4b further validates that lower temperatures require larger pressure reduction to reach Mg formation (below 0.4 bar at 1500
C vs. below 0.1 bar at 1400
C). When employing methane as reducing agent at 1 bar (Fig. 5a), CH4 starts thermally decomposing into both solid carbon and H2 at above 500
C. Subsequently, MgO reduction with produced solid carbon starts at above 1300
C and reaches completion at above 1600
C, thereby yielding a gas mixture of 1 mol Mg(g), 2 mol H2(g), and 1 mol CO(g). Similar to the MgO carbothermal reduction reaction, decreasing pressure theoretically lowers the temperature of the MgO methano-thermal reduction (Fig. 5b). For example, the

788

S. Chuayboon, S. Abanades / Journal of Cleaner Production 232 (2019) 784e795

Fig. 4. Thermodynamic equilibrium composition of MgO carbothermal reduction as a function of (a) temperature (at 0.1 and 1 bar) and (b) pressure (at 1400 and 1500
C).

MgO þ CH4 reaction requires pressure below 0.45 bar to produce Mg at 1500
C, versus below 0.17 bar at 1400
C. The outputs of this thermodynamic analysis were used to settle the favorable operating conditions for the experimental study.

3.2. Solar reactor performance assessment Table 1 summarizes both the operating conditions and experimental results for the experimental tests of the prototype vacuum solar reactor for producing Mg and syngas. For this study, experiments were carried out to evaluate the solar reactor performance under the following range of parameters: P ¼ 0.11e0.90 bar, m_N₂ ¼ 1.2e2.2 Nl/min, Q_ solar ¼ 1.25e1.47 kWth, reducing agents including solid carbon (AC, CB, and graphite) or gaseous CH4, and C/ MgO molar ratio ¼ 1.5e2. The ranges of maximum temperatures were 1437e1668
C for T1 and 1464e1622
C for TPyrometer, confirming homogeneous temperature inside the cavity receiver. As a result, the carbon conversion (XC) was found in the range 60.9e73.1% for C/MgO ¼ 1.5 and 47.2e55.5% for C/MgO ¼ 2, while the CH4 conversion (XCH₄) was 91.7% for MgO methano-thermal reduction. This clearly indicates that a lower C/MgO molar ratio resulted in an enhanced XC because a higher C/MgO excess ratio means a higher amount of unreacted carbon remaining inside the cavity (Run. No 5e8 in Table S4). However, an excess of carbon with respect to stoichiometry (C/MgO ¼ 1) is necessary to both favor CO yield and completely convert MgO to Mg. The MgO conversion

Fig. 5. Thermodynamic equilibrium composition of MgO methano-thermal reduction as a function of (a) temperature at 1 bar and (b) pressure at 1400 and 1500
C.

(XMgO) was calculated by an oxygen mass balance:

XMgO ¼

nCO þ 2nCO2 : nMgO

(4)

XMgO was found in the range 68.6e99.9% (28.9% for methanothermal reduction), and the CO selectivity was in the range 0.98e1.00 (the amount of CO2 stemming from CO disproportionation was negligible). The global material mass balance was also performed to quantify the amount of reactant actually converted into the products, which are syngas productions, Mg deposits and solid residues (details are given in SI). The energy upgrade factor (U) represents the ratio of chemical energy content in the products to the calorific value of the carbon feedstock. It represents the fraction of solar energy stored in the chemical products (CO and Mg), according to Eq. (5):

U¼

ðLHVCO ,m_ CO Þ þ m_ Mg ,DHMgþ0:5O2 /MgO LHVcarbon ,m_ carbon


(5)

where LHV represents the Lower Heating Value (J/kg), m_ the mass flow rate (kg/s), and DH the standard enthalpy change (J/kg). U values were in the range 1.1e1.9, demonstrating successful solar energy storage in the chemical products. The highest U value (1.9), which approaches the maximal theoretical U (2.25) calculated from Eq. (1), was found with the following conditions: P ¼ 0.11 bar (the lowest pressure) and C/MgO ¼ 1.5 (Run. No.1), demonstrating a

S. Chuayboon, S. Abanades / Journal of Cleaner Production 232 (2019) 784e795

789

Table 1 Operating conditions and solar reactor performance assessment. Run No.

C/ P C type MgO (bar)

_ N2 (Nl/ T1 m min) (
C)

Tpyrometer Q_ solar (kW) CO yield (mmol/ gMgO) (±2%) (
C)

CO2 yield (mmol/ gMgO) (±2%)

Global mass balance (%)

hsolar-to-fuel XC U XMgO (±2%) (±2%) (±2%) (±2%)

CO selectivity

1 2 3 4a 5a 6 7a 8 9 10b

1.5 1.5 1.5 1.5 2 2 2 2 e 1.5

1.2 2.2 2.2 1.2 1.2 1.2 2.2 2.2 1.2 1.4

1517 1464 1622 1576 1508 1528 1538 1545 1299 1551

0.23 0.22 0.12 0.04 0.02 0.10 0.17 0.25 0.06 0.39

84.2 78.3 67.5 81.6 82.7 85.6 52.5 64.5 72.7 84.8

1.9 1.7 1.6 1.5 1.1 1.4 1.3 1.4 0.6 1.4

0.99 0.99 0.99 1.00 1.00 1.00 0.99 0.99 0.99 0.98

a b c

0.11 0.16 0.90 0.11 0.11 0.11 0.16 0.16 0.09 0.11

AC AC AC CB CB AC AC Graphite CH4 AC

1551 1546 1668 1437 1534 1538 1568 1581 1314 1501

1.32 1.25 1.47 1.26 1.29 1.32 1.39 1.39 1.10 1.47

23.80 20.21 20.10 20.76 21.16 24.59 21.21 16.53 7.05 21.11

1.7 1.8 1.1 1.2 1.2 1.5 1.6 1.2 3.0 7.8

73.1 64.3 62.4 60.9 47.9 55.5 52.1 47.2 91.7c 59.5

97.8 83.2 82.0 84.0 85.5 99.9 86.8 68.6 28.9 88.2

Auto-combustion occurred. Continuous test. CH4 conversion.ð1  m_ unreacted CH4 =m_ CH4 Þ

significant beneficial influence of pressure and C/MgO molar ratio on U. The solar-to-fuel energy conversion efficiency (hsolar-to-fuel) indicates how well solar energy is utilized as energy source for process heat:

hsolartofuel ¼

ðLHVCO ,m_ CO Þ þ m_ Mg ,DHMgþ0:5O2 /MgO Q_ þ ðLHV ,m_ Þ solar

carbon


(6)

carbon

where Q_ solar is the total solar power input (W) during reaction (CO evolution period). The hsolar-to-fuel for batch tests was found in the range of 1.1e3.0%, implying high heat losses due to a long non-isothermal period. This issue can be tackled by operating the process in an isothermal continuous mode, which directly lowers the processing duration and heat losses, thereby improving hsolar-to-fuel (7.8% for a continuous test).

measured CO and CO2 flow rates (Figs. 6 and 7) over the entire reaction duration as a function of total pressure for different carbon types and C/MgO molar ratios (1.5 and 2). As expected, the CO yield at both AC/MgO molar ratios increased considerably with decreasing total pressure. For example, the CO yield increased from 21.21 mmol/gMgO at 0.16 bar to 24.59 mmol/gMgO at 0.11 bar (AC/ MgO molar ratio of 2), thus reaching 99.1% of the theoretical maximal CO yield (24.81 mmol/gMgO) (1 mol CO per mol MgO according to Eq. (1)). The enhanced CO yield means that the recombination of Mg and CO was lowered when decreasing the pressure (according to Le Ch^ atelier's principle). In addition, CO2 yields were found in a negligible amount (e.g. 0.10e0.17 mmol/gMgO for a AC/ MgO molar ratio of 2) and the selectivity to CO was thus above 99%. This low CO2 production may presumably occur from the Boudouard equilibrium (Eq. (3)), since the solidegas reaction between MgO and CO (MgO þ CO/Mg þ CO2) is not thermodynamically favorable (Fig. 3). Thus, the MgO reduction reaction chiefly occurs via the solidesolid reaction (Eq. (1)).

3.3. Parametric study of MgO carbothermal reduction 3.3.1. Influence of pressure The influence of total pressure on CO and CO2 production rates and yields for the MgO carbothermal reduction with solid carbon was experimentally investigated in a batch mode (the MgO/C mixture was thus preloaded in the reactor before heating). Activated charcoal (AC) was employed as reducing agent with fixed AC/ MgO molar ratios of 1.5 and 2. The total pressures varied from slightly over-atmospheric (0.9 bar) to vacuum pressures (0.16 and 0.11 bar), while the MgO reduction occurred during non-isothermal heating (heating rate controlled by adjusting the shutter opening). The transient CO and CO2 production rates for carbothermal reduction of MgO over AC are plotted in Fig. 6 (AC/MgO ¼ 1.5) and in Fig. 7 (AC/MgO ¼ 2). The flow rate of gas species (Fi ) was calculated from their measured mole fraction (yi) and the known inlet flow rate of N2, (FN₂):(Fi ¼ FN2 ,yi =yN2 ). The effect of pressure was obvious with respect to the peak rates of CO production and CO evolution durations. For example, the peak CO production rate was 20 mL min1 g1 MgO (50 min duration) at 0.9 bar compared to 26 mL min1 g1 MgO (40 min duration) at 0.11 bar. Therefore, the kinetic rate of MgO carbothermal reduction reaction was enhanced by decreasing the total pressure, in agreement with thermodynamic analysis. The peak rate of CO production was observed before reaching the maximal temperature (below 1600
C) at reduced pressure (0.16 and 0.11 bar), whereas the reaction required temperatures above 1600
C at atmospheric pressure to reach completion. Fig. 8 shows the CO and CO2 yields calculated by integrating the

3.3.2. Influence of solid carbon type The influence of solid carbon type (AC, CB, and graphite) was experimentally investigated at C/MgO molar ratios of 1.5 and 2, while total pressure was kept constant at 0.11 bar. The progress of CO and CO2 production rates for AC and CB are shown in Fig. S3 (C/ MgO ¼ 1.5) and Fig. S4 (C/MgO ¼ 2). The highest reaction rates were found for AC at both C/MgO molar ratios, as evidenced by the higher peak of CO production rate for AC compared to that of CB, while the reaction duration for AC seemed to be lower than for CB, demonstrating faster reduction rate. Fig. 8 also confirms that utilizing AC as reducing agent led to the highest CO and CO2 yields at both C/MgO molar ratios when compared to CB. This can be explained by the significantly higher available surface area accessible for particle contact between MgO and AC, favoring the solid-solid reaction and thereby leading to higher CO yield. To further confirm the effect of carbon type, the carbothermal reduction of MgO with graphite was also tested and compared with that of AC. As expected, remarkable higher reaction rates were obtained for AC (Fig. S5). The peak CO production rate was 1 25 mL min1 g1 g1 MgO for AC compared to 10 mL min MgO for graphite at 0.16 bar (C/MgO ¼ 2). Additionally, the higher reaction rates when using AC led to a substantially higher CO yield (21.21 mmol/gMgO for AC compared to 16.53 mmol/gMgO for graphite,
lvez et al., 2008). Fig. S6), in agreement with TGA experiments (Ga The use of AC as solid reducing agent is thus the most suitable option for solid-to-solid contact between MgO and AC, which promotes reaction rate and extent, thereby leading to higher CO

790

S. Chuayboon, S. Abanades / Journal of Cleaner Production 232 (2019) 784e795

Fig. 7. CO and CO2 production rates along with reactor temperatures for the carbothermal reduction of MgO over AC at different total pressures (AC/MgO ¼ 2).

and 0.16 bar). The transient CO and CO2 production rates are plotted for C/MgO molar ratios of 1.5 and 2 in Fig. S7 (AC, 0.11 bar), Fig. S8 (CB, 0.11 bar), and Fig. S9 (AC, 0.16 bar). The influence of C/MgO molar ratios was nearly unnoticeable as evidenced by both similar peak CO production rates that occurred at the same temperatures and similar durations for both AC and CB. As expected, similar CO and CO2 yields (Fig. 8) between the C/ MgO molar ratios of 1.5 and 2 were noticed for both solid carbon types, even though the CO yields at C/MgO ¼ 2 were slightly higher. Thus, increasing C/MgO molar ratio in excess from 1.5 to 2 did not significantly influence the CO and CO2 yields.

Fig. 6. CO and CO2 production rates along with reactor temperatures for the carbothermal reduction of MgO over AC at different total pressures (AC/MgO ¼ 1.5).

and Mg yields. 3.3.3. Influence of C/MgO molar ratio The influence of C/MgO molar ratio was investigated with different solid carbon types (AC and CB) and total pressures (0.11

3.3.4. MgO methano-thermal reduction In order to compare the use of either solid carbon or CH4 as reducing agent, MgO reduction with CH4 was carried out using the same vacuum solar reactor operated in a batch mode at 0.09 bar. Before heating, pure MgO (2.0064 g) was loaded in the solar cavity receiver directly exposed to the solar radiation. After heating the reactor at ~1300
C in N2, CH4 was injected at a constant flow-rate of 0.1 Nl/min along with a N2 flow-rate of 0.2 Nl/min. The H2 production rate increased considerably (up to 85 mL min1 g1 MgO) while the CO production rate also increased (up to 18 mL min1 g1 MgO) due to MgO reduction (Fig. 9), in agreement with thermodynamic

S. Chuayboon, S. Abanades / Journal of Cleaner Production 232 (2019) 784e795

791

Fig. 8. CO and CO2 yields for MgO carbothermal reduction with AC and CB as a function of total pressure at C/MgO molar ratios of 1.5 and 2.

analysis (Fig. 5). Subsequently, the CO production rate decreased continuously and eventually approached zero, while the H2 production rate decreased slightly and tended to be stable after 15 min. The obtained H2, CO, and CO2 yields were 57.33, 7.05, and 0.06 mmol/gMgO, respectively (theoretical CO yield of 24.81 mmol/gMgO from Eq. (2)). This indicates that the targeted reaction (Eq. (2)) occurred fairly (as evidenced by the moderate CO yield), while the side CH4 cracking reaction was prevailing (as evidenced by the large H2 yield), which can be ascertained by the stable pattern in the H2 profile after CO declined. Thus, most of injected methane was converted into H2 and solid carbon, and the solid carbon minimally reacted with MgO, leading to high H2 and low CO production rate. Thus, methane cracking reaction was much more favorable than MgO methano-thermal reduction. For this reason, the use of CH4 as reducing agent is less suitable than solid carbon for MgO reduction because of the favored CH4 cracking reaction.

Fig. 9. Syngas species production rates along with reactor temperatures for MgO reduction with CH4 (P ¼ 0.09 bar).

3.4. Characterization of Mg products The solid products were collected after each experiment in the removable outlet components of the solar reactor where solid products are condensed and deposited (detailed information are provided in SI in the section of solid products analysis, Figs. S10eS13 and Tables S4eS5). Mg was recovered in the outlet products as a pyrophoric fine powder having strong reactivity with air. All the samples were inevitably exposed to ambient air prior their analysis. Fig. 10 shows the XRD patterns of the collected solid products in zones A (outlet tube) and B (filter) after experiments at (a) AC/MgO ¼ 1.5 and (b) AC/MgO ¼ 2. The products were well crystallized with low contamination of carbon. Phases ascribed to Mg and MgO were identified in both zones, suggesting partial Mg recombination at the outlet (with CO) or oxidation after sample collection (with air). Indeed, the condensed Mg can be oxidized during its inevitable exposure to air when opening the reactor or transferring the powder for analysis, leading to a subsequent artificial increase of MgO content in the solid products after their collection from the reactor (auto-ignition observed at room temperature). Nevertheless, the high MgO conversion (99.9%, Table 1) was ascertained from the measured CO production, which confirmed that a high Mg yield was achieved in the reactor. The influence of pressure on Mg yield is noticeable at both AC/MgO molar ratios, especially in zone B (filter). The amount of Mg phase is higher when decreasing pressure, consistently with the higher CO yield in Fig. 8, thereby confirming that decreasing pressure favored CO and Mg production. Regarding the influence of solid carbon type, the XRD patterns of the solid products collected in each zone at both molar ratios were not significantly different between AC and CB (Fig. S14). However, higher Mg intensity was observed in zone B when compared with zone A. In contrast, very small Mg intensities of the XRD peaks were observed when employing graphite as reducing agent (Fig. S15), due to the weak reaction rate, in agreement with the low CO yield (Fig. S6). The XRD patterns between C/MgO ¼ 1.5 and C/MgO ¼ 2 for both solid carbon types were similar (Figs. S16 and S17), denoting that MgO and carbon were sufficiently in contact and confirming the low influence of the C/MgO ratio. In

792

S. Chuayboon, S. Abanades / Journal of Cleaner Production 232 (2019) 784e795

Fig. 10. XRD patterns of the collected solid products in zone A and zone B for the reduction of MgO with AC at (a) AC/MgO ¼ 1.5 and (b) AC/MgO ¼ 2.

addition, the XRD pattern of solid products when using CH4 (Fig. S18) shows that a small amount of Mg was produced, in agreement with low CO yield. The particle morphology was analyzed by FESEM (Fig. 11 at nanoscale and Fig. S19 at microscale) regarding the condensed solid products in the outlet (zone A) and filter (zone B) for two pressures (0.11 and 0.9 bar) and two solid carbon types (AC and CB). Overall, the condensed Mg was produced as nanoparticles (60e300 nm, Fig. 11), with morphology exhibiting spherical shapes arising from the droplet condensation during cooling (Mg melting point: 650
C). The size of condensed Mg from zone A appeared larger than zone B (e.g. Fig. 11e compared to Fig. 11f). The presence of condensed Mg in zone A was less obvious than in zone B (Fig. 11a) because of the oxidation with air issue. The presence of solid carbon and MgO product was also observed in the filter (Fig. 11d). Besides, a noticeable increase of the condensed Mg was observed when decreasing pressure to 0.11 bar (Fig. 11b compared to Fig. 11d). Employing CB as reducing agent resulted in smaller sizes of the produced Mg (Fig. 11f). Furthermore, the high reactivity of solar Mg powders with CO2 for additional CO production was evidenced by TGA, thus demonstrating their high oxidation capability for cycle closure (Figs. S20 and S21). The reaction was composed of a fast initial step (about 70% Mg conversion reached after the first 5 min period) followed by a slower reaction regime due to diffusion limitation in the packed powder layer, and complete Mg conversion was achieved at 380
C regardless of the starting Mg material, thus producing additional CO and solid MgO that can be recycled to the solar step.

3.5. Continuous process demonstration A proof-of-concept MgO carbothermal reduction experiment was performed to demonstrate the feasibility of solar reactor operation with continuous reactant injection under vacuum and isothermal conditions. A homogeneous mixture of MgO (10 g) and AC (4.5 g) resulting in a C/MgO molar ratio of 1.5 was prepared and then placed into the hopper equipped with a screw feeder (Fig. S2). The reacting powder was injected through the inlet path exiting into the reactor cavity with a constant feeding-rate of ~0.7 g/min (21 min injection duration) at a constant temperature of 1500
C, while the pressure was maintained at ~0.11 bar over the entire run. Overall, product gases were produced continuously until finishing injection under isothermal and vacuum conditions (Fig. 12). A narrow gap of the stable temperature between T1 and Tpyrometer throughout the test was observed, indicating homogenous reactor cavity temperature and isothermal condition. However, a fluctuating pattern in the CO production rate was observed. This can be attributed to slight particle feeding rate instabilities and to reactant accumulation during injection occurring when the reactant feeding rate was higher than the rate of MgO carbothermal reduction. Noticeably, the experimental duration (25 min) was slightly higher than the expected duration (21 min), thereby confirming the reactant accumulation issue. The CO and CO2 production rates decreased progressively until reaching zero at the end of reaction, in close agreement with batch tests. After experiment, the amount of unreacted carbon remaining inside the cavity receiver (~1.3 g) corresponded roughly to the excess of fed carbon, and the presence

S. Chuayboon, S. Abanades / Journal of Cleaner Production 232 (2019) 784e795

793

Fig. 11. SEM micrographs of solid products from the outlet tube (zone A) and filter (zone B) during (a,b) Run No. 3, (c,d) Run No. 1, and (e,f) Run No. 4.

Fig. 12. CO and CO2 production rates along with reactor temperatures for MgO reduction with AC during continuous reactant injection under vacuum pressure (P ¼ 0.11 bar).

of MgO particles was not observed in the cavity. The CO and CO2 yields (Run No.10) were compared with those obtained from a batch test (Run No.1) under the same pressure (0.11 bar) and C/MgO molar ratio (1.5), according to Table 1. The CO yield was somewhat lower in comparison with a batch test, presumably because of a weak particle entrainment by the gas flow before reacting inside the cavity, thus explaining the slightly lower global MgO conversion. The feasibility of continuous solar Mg production with high yields was evidenced. The energy content of the feedstock (MgO and C) was upgraded by the solar power input in the form of both CO and Mg and a solar-to-fuel energy conversion efficiency of ~8% was achieved with continuous solid reactants injection. The hsolarto-fuel increased during continuous isothermal operation when compared to batch tests, due to enhanced MgO reduction rate and more efficient utilization of solar energy input. This study is the first solar process demonstration with continuous reactant injection in a high-temperature solar reactor under vacuum operation.

794

S. Chuayboon, S. Abanades / Journal of Cleaner Production 232 (2019) 784e795

4. Conclusion Solar carbothermal and methano-thermal MgO reduction was investigated in a prototype vacuum reactor operated with concentrated sunlight up to 1650
C. Mg production was successfully achieved using different solid and gaseous reducing agents and operating parameters (pressure and C/MgO molar ratio) in continuous and batch modes under atmospheric or vacuum conditions (0.1e0.9 bar). The feasibility of vacuum MgO carbothermal reduction with continuous reactant injection was demonstrated for the first time. Decreasing pressure significantly enhanced the rate of MgO carbothermal reduction, MgO conversion, and CO yield, in agreement with thermodynamics. The highest CO yield (24.59 mmol/ gMgO) obtained for activated charcoal at the lowest pressure closely approached the theoretical yield (24.81 mmol/gMgO), demonstrating nearly complete MgO conversion. MgO conversion up to 99.9% was achieved (at P ¼ 0.11 bar). High energy upgrade factor (up to 1.9) revealed efficient storage of solar energy into chemical fuels. Utilizing activated charcoal as reducing agent showed the best MgO conversion due to effective solid-to-solid contact. Employing methane as reducing agent was less suitable than solid carbon, because of the adverse impact of methane cracking reaction. No significant impact of excess C/MgO molar ratios (between 1.5 and 2) on MgO conversion was noticed. Mg product was produced as condensed nanosized particles, which highly favors their reactivity and oxidation when exposed to air (auto-ignition observed at room temperature). The new prototype vacuum reactor is expected to be flexible in processing different carbonaceous feedstocks for MgO reduction in both batch and continuous modes under vacuum condition. Future work should be focused on the continuous reactant injection under vacuum pressure in a scaled-up solar reactor. Declarations of interest None. Acknowledgement The King Mongkut's Institute of Technology Ladkrabang, Thailand and the Franco-Thai scholarship program are gratefully acknowledged for fellowship granting. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.jclepro.2019.05.371. References
, L., 2018. Design and demonstration of a high temperature Abanades, S., Andre solar-heated rotary tube reactor for continuous particles calcination. Appl. Energy 212, 1310e1320. https://doi.org/10.1016/j.apenergy.2018.01.019. Abanades, S., Flamant, G., 2006. Thermochemical hydrogen production from a twostep solar-driven water-splitting cycle based on cerium oxides. Sol. Energy 80, 1611e1623. https://doi.org/10.1016/j.solener.2005.12.005. Abu-Hamed, T., Karni, J., Epstein, M., 2007. The use of boron for thermochemical storage and distribution of solar energy. Sol. Energy 81, 93e101. https://doi.org/ 10.1016/j.solener.2006.06.012. Agrafiotis, C., Roeb, M., Sattler, C., 2015. A Review on Solar Thermal Syngas Production via Redox Pair-Based Water/carbon Dioxide Splitting Thermochemical Cycles. Renew. Sustain. Energy Rev. 42, 254e285. https://doi.org/10.1016/j.rser. 2014.09.039. Alam, M.E., Han, S., Nguyen, Q.B., Salem Hamouda, A.M., Gupta, M., 2011. Development of new magnesium based alloys and their nanocomposites. J. Alloy. Comp. 509, 8522e8529. https://doi.org/10.1016/j.jallcom.2011.06.020. Brooks, G., Trang, S., Witt, P., Khan, M.N.H., Nagle, M., 2006. The carbothermic route to magnesium. J. Occup. Med. 58, 51e55. https://doi.org/10.1007/s11837-006-

0024-x. Chuayboon, S., Abanades, S., Rodat, S., 2019. Syngas production via solar-driven chemical looping methane reforming from redox cycling of ceria porous foam in a volumetric solar reactor. Chem. Eng. J. 356, 756e770. https://doi.org/10. 1016/j.cej.2018.09.072. Chuayboon, S., Abanades, S., Rodat, S., 2018a. Solar chemical looping gasification of biomass with the ZnO/Zn redox system for syngas and zinc production in a continuously-fed solar reactor. Fuel 215, 66e79. https://doi.org/10.1016/j.fuel. 2017.11.021. Chuayboon, S., Abanades, S., Rodat, S., 2018b. Comprehensive performance assessment of a continuous solar-driven biomass gasifier. Fuel Process. Technol. 182, 1e14. In: https://doi.org/10.1016/J.FUPROC.2018.10.016. Chubukov, B.A., Palumbo, A.W., Rowe, S.C., Hischier, I., Groehn, A.J., Weimer, A.W., 2016. Pressure dependent kinetics of magnesium oxide carbothermal reduction. Thermochim. Acta 636, 23e32. https://doi.org/10.1016/j.tca.2016.03.035. Chubukov, B.A., Palumbo, A.W., Rowe, S.C., Wallace, M.A., Weimer, A.W., 2017. Enhancing the rate of magnesium oxide carbothermal reduction by catalysis, milling, and vacuum operation. Ind. Eng. Chem. Res. 56, 13602e13609. https:// doi.org/10.1021/acs.iecr.7b03175. Fosheim, J.R., Hathaway, B.J., Davidson, J.H., 2019. High efficiency solar chemicallooping methane reforming with ceria in a fixed-bed reactor. Energy 169, 597e612. https://doi.org/10.1016/j.energy.2018.12.037.
lvez, M.E., Frei, A., Albisetti, G., Lunardi, G., Steinfeld, A., 2008. Solar hydrogen Ga production via a two-step thermochemical process based on MgO/Mg redox reactionsdthermodynamic and kinetic analyses. Int. J. Hydrogen Energy 33, 2880e2890. https://doi.org/10.1016/j.ijhydene.2008.04.007. Halmann, M., Frei, A., Steinfeld, A., 2011. Vacuum carbothermic reduction of Al2O3, BeO, MgO-CaO, TiO2, ZrO2, HfO2 þ ZrO2, SiO2, SiO2 þ Fe2O3, and GeO2 to the metals. A thermodynamic study. Miner. Process. Extr. Metall. Rev. 32, 247e266. Hischier, I., Chubukov, B.A., Wallace, M.A., Fisher, R.P., Palumbo, A.W., Rowe, S.C., Groehn, A.J., Weimer, A.W., 2016. A novel experimental method to study metal vapor condensation/oxidation: Mg in CO and CO2 at reduced pressures. Sol. Energy 139, 389e397. https://doi.org/10.1016/j.solener.2016.10.024. Hong, L., Sohn, H.Y., Sano, M., 2003. Kinetics of carbothermic reduction of magnesia and zinc oxide by thermogravimetric analysis technique. Scand. J. Metall. 32, 171e176. https://doi.org/10.1034/j.1600-0692.2003.00639.x. Klerk, A., 2013. FischereTropsch Process. Kirk-Othmer Encycl. Chem. Technol., Major Reference Works. https://doi.org/10.1002/0471238961.fiscdekl.a01. Kodama, T., 2003. High-temperature solar chemistry for converting solar heat to chemical fuels. Prog. Energy Combust. Sci. 29, 567e597. https://doi.org/10.1016/ S0360-1285(03)00059-5. Kodama, T., Shimizu, T., Satoh, T., Nakata, M., Shimizu, K.-I., 2002. Stepwise production of CO-rich syngas and hydrogen via solar methane reforming by using a Ni(II)eferrite redox system. Sol. Energy 73, 363e374. https://doi.org/10.1016/ S0038-092X(02)00112-3. Koepf, E., Alxneit, I., Wieckert, C., Meier, A., 2017. A review of high temperature solar driven reactor technology: 25years of experience in research and development at the Paul Scherrer Institute. Appl. Energy 188, 620e651. https://doi.org/10. 1016/j.apenergy.2016.11.088. Koepf, E.E., Advani, S.G., Prasad, A.K., Steinfeld, A., 2015. Eeperimental investigation of the carbothermal reduction of ZnO using a beam-down, gravity-fed solar reactor. Ind. Eng. Chem. Res. 54, 8319e8332. ^que, G., Abanades, S., 2015. Investigation of thermal and carbothermal reducLeve tion of volatile oxides (ZnO, SnO2, GeO2, and MgO) via solar-driven vacuum thermogravimetry for thermochemical production of solar fuels. Thermochim. Acta 605, 86e94. https://doi.org/10.1016/j.tca.2015.02.015. Liu, T., Liu, Q., Lei, J., Sui, J., Jin, H., 2018. Solar-clean fuel distributed energy system with solar thermochemistry and chemical recuperation. Appl. Energy 225, 380e391. https://doi.org/10.1016/j.apenergy.2018.04.133. Mordike, B.L., Ebert, T., 2001. Magnesium: properties d applications d potential. Mater. Sci. Eng. A 302, 37e45. https://doi.org/10.1016/S0921-5093(00)01351-4. Murray, J.P., Steinfeld, A., Fletcher, E.A., 1995. Metals, nitrides, and carbides via solar carbothermal reduction of metal oxides. Energy 20, 695e704. https://doi.org/ 10.1016/0360-5442(95)00032-C. Nair, M.M., Abanades, S., 2016. Tailoring hybrid nonstoichiometric ceria redox cycle for combined solar methane reforming and thermochemical conversion of H2O/CO2. Energy Fuels 30, 6050e6058. https://doi.org/10.1021/acs.energyfuels. 6b01063. Rongti, L., Wei, P., Sano, M., 2003. Kinetics and mechanism of carbothermic reduction of magnesia. Metall. Mater. Trans. B 34, 433e437. https://doi.org/10. 1007/s11663-003-0069-y. Sheline, W., Matthews, L., Lindeke, N., Duncan, S., Palumbo, R., 2013. An exploratory study of the solar thermal electrolytic production of Mg from MgO. Energy 51, 163e170. https://doi.org/10.1016/j.energy.2012.11.053. Steinfeld, A., Fletcher, E.A., 1991. Theoretical and experimental investigation of the carbothermic reduction of Fe2O3 using solar energy. Energy 16, 1011e1019. https://doi.org/10.1016/0360-5442(91)90061-P. Tian, Y., Xu, B., Yang, B., Yang, C., Qu, T., Liu, D., Dai, Y., 2015. Magnesium production by carbothermic reduction in vacuum. J. Magnes. Alloy. 3, 149e154. https://doi. org/10.1016/j.jma.2015.04.001. Tian, Y., Xu, B., Yang, C., Yang, B., Qu, T., Liu, H., Dai, Y., Liu, D., 2014. Analysis of magnesia carbothermic reduction process in vacuum. Metall. Mater. Trans. B 45, 1936e1941. https://doi.org/10.1007/s11663-014-0106-z. Villafan-Vidales, H.I., Abanades, S., Montiel-Gonzalez, M., Romero-Paredes, H., 2016. Carbo- and methanothermal reduction of tungsten trioxide into metallic

S. Chuayboon, S. Abanades / Journal of Cleaner Production 232 (2019) 784e795 tungsten for thermochemical production of solar fuels. Energy Technol. 5, 692e702. https://doi.org/10.1002/ente.201600455. Xie, W., Chen, J., Wang, H., Zhang, X., Peng, X., Yang, Y., 2016. Kinetics of magnesium preparation by vacuum-assisted carbothermic reduction method. Rare Met. 35, 192e197. https://doi.org/10.1007/s12598-014-0275-6. Xiong, N., Tian, Y., Yang, B., Xu, B.-Q., Liu, D.-C., Dai, Y.-N., 2018. Volatilization and

795

condensation behaviours of Mg under vacuum. Vacuum 156, 463e468. https:// doi.org/10.1016/j.vacuum.2018.08.014. Yang, C., Tian, Y., Qu, T., Yang, B., Xu, B., Dai, Y., 2014. Analysis of the behavior of magnesium and CO vapor in the carbothermic reduction of magnesia in a vacuum. J. Magnes. Alloy. 2, 50e58. https://doi.org/10.1016/j.jma.2014.02.003.