in an Isothermal Redox Cycle Based on Ceria - American Chemical

Article pubs.acs.org/EF

Efficient Splitting of CO2 in an Isothermal Redox Cycle Based on Ceria Luke J. Venstrom,†,§ Robert M. De Smith,† Yong Hao,‡,∥ Sossina M. Haile,‡ and Jane H. Davidson*,† †

Department of Mechanical Engineering, University of MinnesotaTwin Cities, 111 Church Street South East, Minneapolis, Minnesota 55455, United States ‡ Materials Science, California Institute of Technology, 1200 California Boulevard, Pasadena, California 91125, United States ABSTRACT: An isothermal thermochemical cycle to split CO2 based on nonstoichiometric reduction and oxidation of ceria is demonstrated. Carbon monoxide is produced via an oxygen partial pressure swing by alternating inert sweep gas and CO2 flows over the ceria. The rates of reduction and oxidation at 1500 °C in a porous ceria particle bed are measured for sweep gas and CO2 flow rates from 50 to 600 mL min−1 g−1 and analyzed to identify cycle operating conditions (gas flow rates and reduction and oxidation durations) that maximize process efficiency. For a solar reactor assumed to operate at 3000 suns concentration and with 90% of the sensible heat of the gases recovered, the optimal cycle uses 150 mL min−1 g−1 sweep gas and 50 mL min−1 g−1 CO2 at reduction and oxidation periods of 100 and 155 s, respectively. This cycle is demonstrated in an IR imaging furnace over 102 cycles, yielding a stable average rate of CO production of 0.079 μmol s−1 g−1 and a projected reactor efficiency of 4%. The optimal conditions apply at large scale if the flow rates are scaled in proportion to the ceria mass.

1. INTRODUCTION The cerium dioxide (ceria) thermochemical partial redox cycle has been proposed by a number of research groups for the efficient production of hydrogen and synthesis gas from water and carbon dioxide using solar thermal energy.1−15 In reduction, an endothermic reaction, a fraction of oxygen in the nonstoichiometric ceria is liberated using concentrated sunlight as the source of high-temperature process heat: 1 CeO2 − δox = CeO2 − δrd + ΔδO2 (1) 2 where Δδ = δrd − δox, δrd is the extent of nonstoichiometry at the initiation of the subsequent oxidation step, and δox is that at its completion. In oxidation, an exothermic reaction, H2 and/or CO are produced from H2O and/or CO2 as the reduced ceria is reoxidized to what we define as the final state. For oxidation by CO2: CeO2 − δrd + ΔδCO2 = CeO2 − δox + ΔδCO

Figure 1. State diagram for undoped ceria16 with isotherms of the equilibrium state for a range of oxygen partial pressures. The equilibrium states for oxidation with 20% and 100% CO2 (solid black lines) and 20% and 100% H2O (dashed black lines) at a total pressure of 1 atm as a function of temperature are also plotted. Where these lines cross the isotherms depicts the bound on the extent to which the reduced ceria can be reoxidized.

(2)

The net effect of reactions 1 and 2 is the splitting of CO2 into CO and O2. ΔδCO2 = ΔδCO +

Δδ O2 2

(3)

chemical thermodynamics predicts that a swing from reduction at 1500 °C and pO2 = 10−2 atm to oxidation at 800 °C yields Δδ = 0.015, irrespective of the oxidant (H2O or CO2) or, within reasonable bounds, its concentration. The temperature-swing cycle has been implemented in a prototype solar reactor in which an inert sweep gas was used to achieve low oxygen partial pressure during reduction.3,6,9 The average solar-to-chemical energy efficiency, defined as the energy of the produced fuel divided by the required solar input, was 1.7% for CO

The present work addresses splitting of CO2 but reactions analogous to 2 and 3 can be written for splitting H2O. Carbon monoxide is produced in proportion to the change in nonstoichiometry between reduction and oxidation, Δδ, which is a function of temperature and the oxygen partial pressure as shown by the thermodynamic data in Figure 1.16 These data point to the possibility of producing CO (and/or H2) via either a temperature-swing redox cycle1−13 or an oxygen pressure-swing cycle at a f ixed temperature.14,15,17 Prior work has focused on the temperature-swing approach because large changes in nonstoichiometry are favored when ceria is reduced at high temperature and low oxygen partial pressure and oxidized at lower temperature. For example, © 2014 American Chemical Society

Received: December 17, 2013 Revised: March 12, 2014 Published: March 14, 2014 2732

dx.doi.org/10.1021/ef402492e | Energy Fuels 2014, 28, 2732−2742

Energy & Fuels

Article

redox cycle. Data obtained for a wide range of flow rates of inert sweep gas and CO2 are interpreted to determine optimal flow rates and cycling periods and to provide a realistic assessment of the process efficiency. Prior theoretical estimates of efficiency assume optimistically low gas flow rates and thermodynamic fuel production.14 The present study uses a packed bed of porous particles to promote favorable rates of CO production, mass transfer, and heat transfer without imposing a significant pressure drop and hence parasitic flow work loss. To reduce the number of experiments required to identify the gas flow rates and cycling periods that lead to the most efficient production of CO, we develop a technique to analyze reaction rate data collected in cycles of extended duration to predict fuel production for shorter cycles. Our approach is also applicable to the production H2 or the simultaneous production of CO and H2. Similitude is applied to extrapolate the operating conditions identified in a benchtop reactor for efficient CO production to operation of prototype or industrial-scale reactors.

production in a batch process after the duration of reduction was selected for maximum efficiency.9 Thermodynamic analysis of the process shows that higher efficiencies are possible with recuperation of the sensible heat of the sweep and oxidizing gas streams as well as of the sensible heat of the solid ceria.7 Although approaches to achieve continuous solid phase heat recuperation have been proposed,13,18,19 they have either had limited success or they have not been implemented in a working reactor. Recovery of the sensible heat of the gases is also a challenge due to the material constraints for the heat exchanger at the reduction temperatures required for appreciable fuel production. Less gas phase heat recuperation is required if the amount of sweep and oxidizing gas can be reduced without also decreasing the change in nonstoichiometry achieved in the cycle. The isothermal pressure-swing cycle eliminates the need for solid phase heat recuperation14,15,17 at the penalty of increased need for gas phase heat recovery. This penalty arises because the oxidant must be heated to a higher temperature than in the case of the temperature-swing cycle and because a greater swing in gas composition is required to produce a given change in nonstoichiometry than in the case of the temperature-swing cyclethat is, a lower oxygen partial pressure is required in the reduction step and/or a higher concentration of the oxidant is required in the oxidation step (Figure 1). For example, to achieve the same change in nonstoichiometry in the isothermal cycle at 1500 °C when CO is produced from 100% CO2 as that of the temperature-swing cycle from 1500 °C and pO2 = 10−2 atm to 800 °C, the oxygen partial pressure during reduction must be pO2 = 1.4 × 10−4 atm. If the CO is produced from 20% CO2 instead of 100% CO2, a further reduction in the oxygen partial pressure during reduction to pO2 = 6.9 × 10−5 atm is required. If H2 is produced from 20% H2O, the oxygen partial pressure required must be even further reduced to pO2 = 3.7 × 10−5 atm because H2O has a lower oxygen activity than CO2 at temperatures above 810 °C (the temperature at which the changes in Gibbs free energy of the dissociation of CO2 and H2O are equal). To maintain low oxygen partial pressure during reduction, a sweep gas and/or operation below atmospheric pressure is required. Operation at less than atmospheric pressure has yet to be demonstrated in the literature. Without 100% gas phase sensible heat recovery (an idealization), the challenge for efficient fuel production is to reduce the amount of sweep gas and CO2 used in the cycle while maintaining a rapid rate of fuel production. Rate data are thus required over a range of gas flow rates and cycling periods (the durations of reduction and oxidation) to select the operating conditions that maximize efficiency. The viability of isothermal redox cycling of ceria for hydrogen production has been demonstrated in an initial, proof of concept experimental study.15 Though the range of conditions considered was insufficient for optimization of the gas flow rates and cycling periods, it was possible to infer from the data that an average rate of H2 production comparable to that of the temperature-swing cycle would be achievable when the temperature-swing cycle is subject to appropriate finite heating and cooling rates.15 In the present study, we demonstrate that isothermal redox cycling is also capable of producing CO at rates comparable to the temperature-swing cycle. The present study presents the first measurements of CO production rates from CO2 splitting in an isothermal ceria

2. APPROACH Reactive Substrate Configuration. Porous ceria particles, 3−5 mm on edge, were prepared by cutting a porous ceria monolith. The monolith was synthesized in a chemical solution approach beginning with cerium nitrate to obtain nanocrystalline ceria. The fine ceria was mixed with isopropanol to create a paste, which was then placed into a cylindrical alumina mold and heat-treated at 1500 °C (6 h hold) to yield a sintered structure. The microstructure of a typical particle cut from such monoliths, Figure 2, reveals an overall porosity of ∼65% and

Figure 2. Scanning electron microscopy image of the porous particle surface revealing its microstructure to be a randomly oriented network of pores with diameters on the order of micrometers.

pores with diameters on the order of a micrometer implying a volume-specific surface area of ∼106 m−1. Computed and/or estimated thermophysical properties of the packed bed of porous particles and of the parent monolith are listed in Table 1, including the permeability and Forchheimer coefficient, the stagnant thermal conductivity, the extinction coefficient, the specific surface area, and the porosity. References are noted in the table. The use of a packed bed of porous particles provides several benefits for thermochemical cycling. Most significant of these is a minimal pressure drop. Given the 3 orders of magnitude 2733

dx.doi.org/10.1021/ef402492e | Energy Fuels 2014, 28, 2732−2742

Energy & Fuels

Article

⎛ d 2 ⎞ p ⎟nCO MW = ⎜⎜ ⎟ ̇″ a D 36 eff c ⎠ ⎝

Table 1. Thermophysical Properties of the Packed Bed of Porous Particles and the Monolith Microstructurea property

packed bed of porous particles

monolith microstructure

ϕ, % a, m−1 K, m2 CF kc, W m−1 K−1 β, mm−1

80 3 × 106 9 × 10−9 0.9 0.15 0.18

65 4 × 106 3 × 10−12 0.6 0.20 11

(5)

Deff is the effective diffusion coefficient computed according to Deff =

ϕ DCO−CO2 τ̅

(6)

where ϕ is the particle porosity and τ ̅ is the tortuosity. The latter is assumed to be four in the absence of experimental data.24 The Wagner modulus is estimated conservatively using the maximum rate of CO production to be 0.01−0.05 for particle diameters of 3−5 mm. Being much less than 0.15, the Wagner modulus indicates that reactions proceed uninhibited by gas phase diffusion within the particles.22,23 Gas phase mass transport limitations on the scale of the packed bed (i.e., from particle-to-particle), however, are not addressed by the Wagner modulus and are observed in the present study. Heat transfer in the packed bed is another essential consideration and is estimated in the absence of fluid motion considering radiation and conduction. The radiative conductivity of the particle bed is computed using the Rosseland approximation:25

a

Estimates of the packed bed properties assumed spherical particles with a diameter of 5 mm and a packing fraction of 0.4 using the Ergun equation for the permeability and Forchheimer coefficient and geometric optics for the extinction coefficient. Data for the monolith microstructure are from ref 21.

difference in the permeability of the packed bed computed from the Ergun equation20 (9.2 × 10−9 m−1) and the permeability of the particle microstructure estimated from a tomography study of a similar structure21 (3 × 10−12 m−1), gas flows primarily in the void space around the particles rather than through their internal porosity. Pressure drop is thus estimated with extended Darcy’s law using the permeability and the Forchheimer coefficient of the packed bed: μ CF ΔP = f G″ + (G″)2 Δx ρf K ρf K (4)

kr =

16n2σT 3 3β

(7)

where the extinction coefficient, β, is calculated from the theory of geometric optics26 and the index of refraction, n, is taken from ref 27. The stagnant thermal conductivity is treated by applying the geometric mean model at the particle porosity and bed porosity scales

where the variables (and all subsequently introduced variables) are defined in the Nomenclature. As shown in Figure 3a, the pressure drop is less than 105 Pa m−1 for N2 sweep gas fluxes less than 1.1 kg m−2 s−1 at 1500 °C. The optimal sweep gas flux identified in the present study is 0.04 kg m−2 s−1. The configuration of the reactive substrate further ensures that the material surface is fully accessible to the process gas. That is, while gases flow around the particles, diffusion in their pores effectively bathes the internal surfaces with reactant (or, conversely, removes products). This situation can be concluded from the following. The rate of gas phase diffusion through the internal porosity and the rate of reaction are compared in the dimensionless Wagner modulus,22,23

ϕ

kc = [(k CeO2)1 − ϕ (k CO2)ϕ ]1 − ϕbed k CObed2

(8)

using the thermal conductivity of ceria reported in ref 28 and the conductivity of CO2.29 The overall conductivity (the sum of the conduction and radiation conductivities) is shown in Figure 3b as a function of temperature. At 1500 °C, the overall conductivity is 18.5 W m−1 K−1, 120 times that of thermal conduction (Table 1). Radiation dominates heat transfer at 1500 °C because the void space between particles reduces attenuation (the extinction coefficient of the packed bed is small relative to the monolith microstructure, Table 1). The enhancement in conductivity due to radiation in the packed bed mitigates particle-to-particle temperature variations, an important consideration for isothermal operation. Rapid volumetric thermal energy generation within an individual particle due to chemical reaction can potentially result in undesirable thermal gradients, even in the absence of particle-to-particle temperature variations. The relative rate of heat transfer and thermal energy generation is evaluated from the dimensionless heat transfer modulus Mht =

|Δr H |(n i̇ ″a)R p2 6kc + rTs

(9)

where kc+r is the overall conductivity of the particle microstructure. When Mht ≪ 1, heat transfer within the particle results in a uniform particle temperature. Using the most rapid reaction rate measured in the present study (2.25 μmol g−1 s−1), the heat transfer modulus is estimated from an energy balance on the particle to be Mht ≈ 10−5. Composition gradients within the solid phase of the particles are likewise estimated to be negligible during cycling. The

Figure 3. (a) Pressure drop as a function of the flux of N2 sweep gas and (b) overall conductivity (conduction plus radiation) of the packed bed of millimeter-scale porous particles. 2734

dx.doi.org/10.1021/ef402492e | Energy Fuels 2014, 28, 2732−2742

Energy & Fuels

Article

Figure 4. Sketch (not to scale) of the isothermal pressure-swing reactor featuring a packed bed of millimeter-scale porous particles.

Oxidation was initiated by switching to a flow of CO2 and continued until the ceria returned to its initial equilibrium state. Equilibration required less than 6 min. Optimization of Cycling. The reduction and oxidation rate data, collected over 20 and 6 min, respectively, were analyzed to predict fuel production as a function of both gas flow rate and cycle duration. This approach allows one to optimize the cycle for efficient fuel production without independently varying experimental reduction and oxidation times, a prohibitively time-consuming task due to the number of possibilities. The average molar rate of fuel production per unit mass of ceria for steady cycling is

oxygen chemical diffusivity through ceria at close to 1000 °C is ∼10−4 cm2 s−1,30 implying a characteristic diffusion time through 1 μm of just 0.4 ms. For a single set of cycling conditions, oxygen release and hydrogen production rates are enhanced during isothermal water splitting when Rh catalyst is applied to the surface of ceria,15 indicating a process limited by surface reaction rather than solid-state oxygen diffusion. Although the present study considers CO, rather than H2, production, the extremely high diffusivity makes it unlikely that solid-state oxygen diffusion will become the rate-limiting step. Compositional uniformity within the solid phase of the particles allows a straightforward analysis of the fuel productivity data and ready identification of optimal cycling conditions. Measurement of Reduction and Oxidation Rates. Experiments to explore the impact of inert sweep gas and CO2 flow rates on the rates of O2 and CO production were conducted at 1500 °C. Figure 4 is a sketch of the experimental apparatus. The porous particles (1.0245 g) were packed in a high-density, high-purity 13 mm o.d., 9.5 mm i.d. alumina tube over a length of 10 mm into a bed with a void fraction of ∼45%. The alumina tube containing the particles was heated in an IR imaging furnace. No reaction was observed between the alumina tube and the ceria. Two alumina-sheathed Pt/Pt−Rh thermocouple probes with exposed junctions were placed in contact with the upstream and downstream faces of the packed bed to monitor temperature. Sweep gas (10.6 ± 0.1 ppm O2 in N2) and CO2 (99.99% pure) were delivered using highaccuracy (±1% of reading) mass flow controllers. Rapid switching ( 100 when backmixing is avoided.23 2740

dx.doi.org/10.1021/ef402492e | Energy Fuels 2014, 28, 2732−2742

Energy & Fuels

Article

U = average gas velocity, m s−1 V̇ = volumetric flow rate, evaluated at 25 °C and 1 bar, mL min−1 x = position coordinate in the streamwise direction, m X = mole fraction of gas species β = extinction coefficient, m−1 ΔrH = enthalpy change of reaction, J mol−1 Δδ = change in ceria nonstoichiometry δ = ceria nonstoichiometry ϵ = heat recuperator effectiveness ηth = thermal efficiency μ = dynamic viscosity, kg m−1 s−1 ρ = density, kg m−3 σ = Stefan−Boltzmann constant, W m−2 K−4 τ = duration in time, s τ ̅ = tortuosity of particle porosity ϕ = porosity of ceria particle

recovery systems at the elevated temperature required for the ceria cycle is imperative.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Present Addresses

§ Department of Mechanical Engineering, Valparaiso University, 1900 Chapel Dr., Valparaiso, IN 46383, USA. ∥ Institute of Engineering Thermophysics, Chinese Academy of Sciences, 11 Beisihuanxi Rd., Beijing, 100190, People’s Republic of China.

Funding

The research was supported by the U.S. Department of Energy, through ARPAe Contract DE-AR0000182, and the University of Minnesota Initiative for Renewable Energy and the Environment.

Subscripts

Notes

bed = particle bed chem = dissociation of CO2 in the isothermal ceria cycle eq = equilibrium f = fuel (CO or H2) or fluid i = gas species loss = solar reactor convection losses ox = oxidation, oxidizer, or the sensible heating of the oxidizer flow p = particle R = solar reactor rd = reduction ref = reference state at the inlet of the packed bed reradiation, refers to reradiation from the solar reactor cavity s = surface sg = sweep gas or the sensible heating of the sweep gas flow solar = concentrated solar input to the cycle total = total molar flow of a gas mixture

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank Bernard Kim for preparing the porous ceria monoliths and Nicholas D. Petkovich for the scanning electron micrograph of the porous ceria microstructure.

■

NOMENCLATURE a = specific surface area, m2 m−3 C = concentration ratio c = concentration, mol m−3 CF = Forchheimer coefficient d = diameter, m Da = Damköhler number Deff = effective diffusion coefficient, m2 s−1 DCO−CO2 = binary diffusivity of a CO−CO2 mixture, m2 s−1 F = convective loss factor of reactor G″ = superficial mass flux, kg m−2 s−1 HHV = higher heating value, J mol−1 h̅ = molar enthalpy, J mol−1 I = direct normal solar irradiance, W m−2 0 = aspect ratio K = permeability, m2 kc+r = overall conductivity, W m−1 K−1 kr = effective radiative conductivity, W m−1 K−1 L = length of packed bed, m M = molecular weight, g mol−1 Mht = heat transfer modulus MW = Wagner−Weisz−Wheeler modulus m = mass, g n = index of refraction ṅ = molar production rate or molar flow rate, mol s−1 ṅ″ = surface reaction rate, mol s−1 m−2 P = pressure, Pa pO2 = oxygen partial pressure, atm PeL = Peclet number q = average rate of thermal energy transfer per unit mass of ceria, W g−1 R = radius, m T = temperature, K Tox,out = temperature of oxidizer leaving recuperator, K Tsg,out = temperature of sweep gas leaving recuperator, K t = time coordinate, s tf = time at which ceria reaches equilibrium, s

Superscripts

■

o = at the inlet of the packed bed ′ = per unit mass of ceria − = overbar indicates an average value over a cycle

REFERENCES

(1) Abanades, S.; Flamant, G. Solar Energy 2006, 80, 1611−1623. (2) Chueh, W. C.; Haile, S. M. Philos. Trans. R. Soc., A 2010, 368, 3269−3294. (3) Chueh, W. C.; Falter, C.; Abbott, M.; Scipio, D.; Furler, P.; Haile, S. M.; Steinfeld, A. Science 2010, 330, 1797−1801. (4) Diver, R. B.; Miller, J. E.; Siegel, N. P.; Moss, T. A. ASME Int. Conf. Energy Sustainability, Proc., 4th, 2010 ES 2010 2010, 97−104. (5) Petkovich, N. D.; Rudisill, S. G.; Venstrom, L. J.; Boman, D. B.; Davidson, J. H.; Stein, A. J. Phys. Chem. C 2011, 115, 21022−21033. (6) Furler, P.; Scheffe, J. R.; Steinfeld, A. Energy Environ. Sci. 2012, 5, 6098−6103. (7) Lapp, J.; Davidson, J. H.; Lipiński, W. Energy 2012, 37, 591−600. (8) Venstrom, L. J.; Petkovich, N.; Rudisill, S.; Stein, A.; Davidson, J. H. J. Sol. Energy Eng. 2012, 134, No. 011005. (9) Furler, P.; Scheffe, J.; Gorbar, M.; Moes, L.; Vogt, U.; Steinfeld, A. Energy Fuels 2012, 26, 7051−7059. (10) Meng, Q.; Lee, C.; Shigeta, S.; Kaneko, H.; Tamaura, Y. J. Solid State Chem. 2012, 194, 343−351. (11) Rudisill, S. G.; Venstrom, L. J.; Petkovich, N. D.; Quan, T.; Hein, N.; Boman, D. B.; Davidson, J. H.; Stein, A. J. Phys. Chem. C 2013, 117, 1692−1700. (12) Siegel, N. P.; Miller, J. E.; Ermanoski, I.; Diver, R. B.; Stechel, E. B. Ind. Eng. Chem. Res. 2013, 52, 3276−3286. 2741

dx.doi.org/10.1021/ef402492e | Energy Fuels 2014, 28, 2732−2742

Energy & Fuels

Article

(13) Ermanoski, I.; Siegel, N. P.; Stechel, E. B. J. Sol. Energy Eng. 2013, 135, No. 031002. (14) Bader, R.; Venstrom, L. J.; Davidson, J. H.; Lipiński, W. Energy Fuels 2013, 27, 5533−5544. (15) Hao, Y.; Yang, C.; Haile, S. M. Phys. Chem. Chem. Phys. 2013, 15, 17084−17092. (16) Panlener, R. J.; Blumenthal, R. N.; Garnier, J. E. J. Phys. Chem. Solids 1975, 3611, 1213−1222. (17) Muhich, C. L.; Evanko, B. W.; Weston, K. C.; Lichty, P.; Liang, X.; Martinek, J.; Musgrave, C. B.; Weimer, A. W. Science 2013, 341, 540−542. (18) Diver, R. B.; Miller, J. E.; Allendord, M. D. J. Sol. Energy Eng. 2008, 130, No. 041001. (19) Lapp, J.; Davidson, J. H.; Lipiński, W. J. Sol. Energy Eng. 2013, 135, No. 031004. (20) Ergun, S. Chem. Eng. Prog. 1952, 48, 89−94. (21) Haussener, S.; Steinfeld, A. Materials 2012, 5, 192−209. (22) Weisz, P. B.; Hicks, J. S. Chem. Eng. Sci. 1962, 17, 265−275. (23) Levenspiel, O. Chemical Reaction Engineering; Wiley: New York, 1999. (24) Davis, M. E.; Davis, R. J. Fundamentals of Chemical Reaction Engineering; McGraw-Hill: Boston, 2002. (25) Rosseland, S. Theoretical Astrophysics: Atomic Theory and the Analysis of Stellar Atmosphere and Envelopes; Clarendron Press: Oxford, U.K., 1936. (26) van de Hulst, H. C. Light Scattering by Small Particles; John Wiley & Sons: New York, 1957. (27) Oh, T.; Tokpanov, Y. S.; Hao, Y.; Jung, W.; Haile, S. M. J. Appl. Phys. 2012, 112, No. 103535. (28) Chekhovskoy, V. Y.; Stavrovsky, G. I. Conf. Therm. Conduct., 9th 1970, 295−298. (29) Yaws, C. Yaws’ transport properties of chemicals and hydrocarbons; Knovel: New York, 2010. (30) Millot, F.; De Mierry, P. J. Phys. Chem. Solids 1985, 46, 797− 801. (31) Scheffe, J. R.; Steinfeld, A. Energy Fuels 2012, 26, 1928−1936. (32) Lovegrove, K.; Luzzi, A. Solar Thermal Power Systems; In Encyclopedia of Physical Science & Technology, 3rd ed.; Meyers, R. A., Ed..; Academic Press: New York, 2003; pp 223−235. (33) Fletcher, E. A.; Moen, R. L. Science 1977, 197, 1050−1056. (34) Meier, A.; Bonaldi, E.; Cella, G. M.; Lipinski, W.; Wuillemin, D. Sol. Energy 2006, 80, 1355−1362. (35) Pal, U. B.; MacDonald, C. J.; Chiang, E.; Chernicoff, W. C.; Chou, K. C.; Van, D. A.; Molecke, M. A.; Melgaard, D. Metall. Mater. Trans. B 2001, 32, 1119−1128. (36) Hirai, S.; Sumita, E.; Shimakage, K.; Uemura, Y.; Nishimura, T.; Mitomo, M. J. Am. Ceram. Soc. 2004, 87, 23−28. (37) Mills, A. F. Mass transfer; Prentice Hall; Upper Saddle River, NJ, USA, 2001. (38) Bulfin, B.; Lowe, A. J.; Keogh, K. A.; Murphy, B. E.; Lübben, O.; Krasnikov, S. A.; Shvets, I. V. J. Phys. Chem. C 2013, 117, 24129− 24137.

2742

dx.doi.org/10.1021/ef402492e | Energy Fuels 2014, 28, 2732−2742