Experimental Demonstration of the Thermochemical Reduction of Ceria in a Solar Aerosol Reactor.

We report on the experimental demonstration of an aerosol solar reactor for the thermal reduction of ceria, as part of a thermochemical redox cycle for splitting H2O and CO2. The concept utilizes a cavity-receiver enclosing an array of alumina tubes, each containing a downward gravity-driven aerosol flow of ceria particles countercurrent to an inert sweep gas flow for intrinsic separation of reduced ceria and oxygen. A 2 kWth lab-scale prototype with a single tube was tested under radiative fluxes approaching 4000 suns, yielding reaction extents of up to 53% of the thermodynamic equilibrium at 1919 K within residence times below 1 s. Upon thermal redox cycling, fresh primary particles of 2.44 μm mean size initially formed large agglomerates of 1000 μm mean size, then sintered into stable particles of 150 μm mean size. The reaction extent was primarily limited by heat transfer for large particles/agglomerates (mean size > 200 μm) and by the gas phase advection of product O2 for smaller particles.

Introduction

Thermochemical cycles based on metal oxide redox reactions are capable of splitting H2O and/or CO2 into H2 and/or CO using concentrated solar energy.[1,2] Nonstoichiometric ceria is currently considered the benchmark among the nonvolatile redox materials because of its rapid kinetics and crystallographic stability.[3−6] The two-step cycle proceeds according to the following reactions:During the first solar-driven endothermic step (eq ), ceria is thermally reduced at temperatures generally above 1573 K using solar process heat. The equilibrium reduction extent, given by the final nonstoichiometry δf, is dependent on the reduction temperature Treduction and the oxygen partial pressure pO of the surrounding gas atmosphere.[7] Following reduction, the reduced ceria is reoxidized to its initial nonstoichiometry δi with H2O (eq ) and/or CO2 (eq ) to form H2 and/or CO at an oxidation temperature Toxidation generally lower than 1273 K. Several solar reactor concepts have been proposed for performing the redox cycle with nonvolatile metal oxides. These include a circulating fluidized bed for the reduction of NiFe2O4/m-ZrO2 particles,[8] which offers good heat transfer and kinetics due to small particles while the window imposes a limitation on the scale-up potential, a counter-rotating set of rings for reducing Fe3O4 or CeO2 that enables solid–solid heat recovery,[9,10] a similar rotary-type reactor for reducing CeO2 or Ni–Mn–ferrite (Ni0.5Mn0.5Fe2O4) that alternates between reduction and oxidation,[11] and a rotating cylinder made from porous CeO2 with an additional inert counter rotating cylinder on the inside for heat recovery.[12] All of these concepts benefit from the implemented solid–solid heat recovery and are able to produce a continuous fuel output, while facing the challenge of a window and moving reactor component at high operating temperatures. A cylindrical cavity containing an array of tubular reactive elements operated isothermally in batch or continuous mode inherently avoids the need for solid–solid heat recuperation but is limited to low fuel concentrations introducing additional downstream processes.[13] A moving packed bed of ceria particles incorporating solid–solid heat recovery and pressure separation of reduction and oxidation zones was proposed, which has the potential to decouple the reduction from the oxidation.[14] An additional concept is a cavity-receiver lined with porous ceria monoliths or felts.[15,16] The latter was also tested with a reticulated porous structure made of ceria for enhanced heat and mass transfer, reaching a solar-to-fuel energy conversion efficiency of 1.73% (without heat recovery).[17,18] Recently, we proposed a novel solar reactor concept based on a cavity-receiver enclosing an array of alumina tubes, each containing a downward aerosol flow of ceria particles counter to an inert sweep gas flow.[19] Owing to the countercurrent arrangement of the particle and gas flows, the local pO gradually decreases while the particle temperature increases, resulting in an intrinsic separation of the reduced metal oxide from the produced oxygen. This separation is critical to avoid recombination. Furthermore, this concept offers enhanced heat and mass transfer, good scalability because of the modular multiple-tube configuration, and can be operated isothermally and continuously, thus decoupling the two redox steps both in space and time for round-the-clock fuel production.[20,21] The disadvantages are associated with the complexity of handling large flows of particles across the hot reaction zone in short residence times, and the limitations imposed by the materials of construction of the solar tubular absorber such as the maximum operating temperature, thermal conductivity, resistance to thermal shocks, and inertness to oxidation by air.

In this work, we report on the experimental performance of a solar aerosol reactor prototype fabricated with a single tube and tested for the first time under realistic concentrated radiation conditions. Additionally, we study the effect of cycling on the particle morphology and the subsequently measured reduction extent, and elucidate the dominant controlling mechanisms via heat and mass transfer modeling.

Experimental Methods

The solar reactor is schematically depicted in Figure . It consists of a vertical Al2O3 tube (Haldenwanger, ALSINT 99.7, Dout = 50 mm, Din = 42 mm, L = 850 mm) which withstands the high operating temperatures and it is inert to air. The tube is positioned inside a cavity-receiver (200 × 86 × 86 mm3) lined with 40 mm-thick alumina/silica based insulation (65% Al2O3, 35% SiO2, Insulform 1600) and supported by a stainless steel shell. Concentrated solar radiation enters the windowless cavity through its 30 mm-diameter circular aperture. Because of multiple internal reflections and absorptions, the cavity approaches a blackbody absorber and efficiently captures the incoming concentrated solar radiation as high-temperature heat. A refrigerated polished-Al compound parabolic concentrator (CPC) of 45° half acceptance angle is incorporated at the aperture to boost the solar concentration ratio and mitigate reradiation losses.[22,23] Ceria particles (Dv50 = 70 μm, Chempur, 99.9% purity) were supplied to the top of the alumina tube by a particle feeder featuring an adjustable orifice operated by an electrical motor (Micro motors E192.24.25) and carried to the reaction zone by gravity. At the bottom of the alumina tube, the particles were collected on a balance (Mettler Toledo XS1003S) for online measurement of the ceria mass flow rate ṁCeO. Argon purge gas was delivered upward through the balance housing and downward through the feeder by electronic mass flow controllers (Bronkhorst, F-201CV and F-201C). Gases exited through a lateral outlet above the reaction zone shortly below the feeder and were analyzed online using a mass spectrometer (MS, Pfeiffer Vacuum, OmniStar GSD 320) and an electrochemical-based oxygen sensor (Setnag, trace oxygen analyzer JC24 V-R). The residual oxygen partial pressure pO within the system was measured prior to an experiment. The nominal cavity temperature Tcavity was measured by a type-B thermocouple located at the front of the alumina tube. Because of multiple internal absorptions, reflections, and re-emissions by the inner cavity walls, no large temperature gradients were expected. The particle temperature, which increased from ambient to its maximum value before exiting the hot reaction zone, was not measured. Thus, Tcavity set an upper limit for the actual particle-flow temperature. The reduction extent was calculated using the measured values of ṁCeO and pO. Prior to each experimental run, reduced particles from the previous experiment were exposed to air at 573 K for more than 8 h to ensure a fully oxidized state (δi = 0), as verified by thermogravimetric analysis. Experiments were performed at the High-Flux Solar Simulator (HFSS) of ETH Zurich: an array of seven Xe arcs, close-coupled to truncated ellipsoidal specular reflectors, provided an external source of intense thermal radiation, mostly in the visible and IR spectra, closely approximating the radiative heat transfer characteristics of highly concentrating solar systems such as solar towers and dishes.[24] The radiative power input Psolar was determined optically with a calibrated CCD camera and verified by calorimeter measurements.

Figure 1

(a) Schematic of the solar receiver indicating the main components and material flows. (b) Scheme of the reaction zone alongside qualitative profiles of the particle temperature Treduction and oxygen partial pressure pO.

During a typical solar experimental run, the cavity-receiver was heated by concentrated radiation delivered by the HFSS to the desired Tcavity, as high as 1919 K. The radiative power input was in the range Psolar = 2.28–2.8 kW, resulting in mean solar concentration ratios over the aperture in the range 3226–3954 suns (1 sun = 1 kW m–2). A constant argon flow was delivered at a rate V̇Ar in the range of 2–4 L min–1 (SLPM, gas flow rates calculated at 273.15 K and 101 325 Pa) through the tube and 0.1 L min–1 through the feeder. Residual oxygen concentrations were below 200 ppmv. Once approximate isothermal conditions were reached, a constant ṁCeO was supplied from the top. An electrically heated setup (Carbolite, STF 16/-/450) was also used to assess key operational parameters under more controlled conditions and a longer heated section of the tube. The electric setup was previously described, and its key features are summarized here.[19] Similar to the solar reactor, it also featured a single alumina tube, but of dimensions Dout = 25.4 mm, Din = 19.1 mm, L = 1300 mm. For both setups, ceria particles (Dv50 = 2.44 μm, Sigma-Aldrich, 99.9% purity) were fed at an ṁCeO of up to 0.7 g s–1 for at least 60 s. Particle morphology was characterized by scanning electron microscopy (SEM, Hitachi TM-1000), and particle size distribution was analyzed by laser scattering (Horiba LA-950) and image processing software (ImageJ v1.48). Particle sieves were used to control their size distribution.

Results and Discussion

Solar Experimentation

The steady-state ceria mass flow rate and the corresponding reduction extent are shown in Figure . Figure a as a function of time for three operating conditions (cases 1–3), while the corresponding average values are summarized in Table and shown in Figure b. Error bars indicate the 95% confidence interval. Operating conditions were as follows. Case 1 (baseline): ṁCeO = 0.15 g s–1, V̇Ar = 2 L min–1, Psolar = 2.28 kW, Tcavity = 1902 K. Case 2: ṁCeO = 0.17 g s–1, V̇Ar = 4 L min–1, Psolar = 2.28 kW, Tcavity = 1919 K. Case 3: ṁCeO = 1.5 g s–1, V̇Ar = 2 L min–1, Psolar = 2.80 kW, Tcavity = 1886 K. For all three cases, pO = 2 × 10–4 atm. For case 1 (solid lines), the average δ measured was 0.0162, which corresponded to a reaction extent of 20% of the thermodynamic equilibrium. Since the equilibrium was calculated based on Treduction = Tcavity and pO = 2 × 10–4 atm, the value of the reaction extent is a conservative estimate as Tcavity sets the upper particle temperature limit and pO sets the lower oxygen concentration limit. For case 2 (large dashes), in which the gas flow rate was doubled while the ceria mass flow rate was kept constant within the feeder accuracy, δ was almost tripled to 0.0462 with a reaction extent of 53%. For case 3 (small dashes), in which ṁCeO was increased by a factor of 10 while Psolar was increased from 2.28 to 2.8 kW to maintain Tcavity > 1873 K, δ decreased to 0.0042 and the reaction extent reached 6%. Typical residence time in the reaction zone for free falling particles is below 1 s.

Figure 2

(a) Temporal evolution of ceria mass flow rates and corresponding reduction extents for three solar experimental runs. (b) Average reduction extents as a function of average ceria mass flow rate. Error bars indicate the 95% confidence interval. Operating conditions: Case 1) (baseline) ṁCeO = 0.15 g s–1, V̇Ar = 2 L min–1, Psolar = 2.28 kW, Tcavity = 1902 K; Case 2) ṁCeO = 0.17 g s–1, V̇Ar = 4 L min–1, Psolar = 2.28 kW, Tcavity = 1919 K; Case 3) ṁCeO = 1.5 g s–1, V̇Ar = 2 L min–1, Psolar = 2.80 kW, Tcavity = 1886 K. For all three cases, pO = 2 × 10–4 atm.

Table 1

Operating Conditions and Main Results of the Three Solar Experimental Runs of Figure

case	ṁCeO2 [g s–1]	V̇Ar [L min–1]	Tcavity [K]	Psolar [kW]	δ [−]	ηsolar-to-fuel [%]
1	0.15 ± 0.0015b	2	1902 ± 8a	2.28	0.0162 ± 0.0004b	0.18
2	0.17 ± 0.0026	4	1919 ± 8	2.28	0.0463 ± 0.0012	0.56
3	1.50 ± 0.0167	2	1886 ± 8	2.80	0.0042 ± 0.0001	0.37

Given by manufacturer.

95% confidence interval.

The average solar-to-fuel energy conversion efficiency ηsolar-to-fuel is defined aswhere ṅCeO is the molar flow rate of ceria, ṅinert is the molar flow rate of the inert gas, HHVH is the higher heating value of hydrogen (assuming complete oxidation eq ) and Einert is the energy required to separate the inert gas (assumed 20 kJ mol–1[25]). Sensible heat recovery from the exiting hot particles and gases was not realized in this study. Particles were fed at room temperature instead of the more realistic oxidation temperature, thus resulting in a conservative estimate for the efficiencies. It is further implicitly assumed that complete reoxidation with stoichiometric amounts of H2O can be achieved, as corroborated experimentally.[26] Results are summarized in Table . Overall, ηsolar-to-fuel increased with ṁCeO rate and V̇Ar. Specifically, ηsolar-to-fuel increased from 0.18 to 0.37% when increasing ṁCeO from its baseline by a factor of 10, even though δ dropped by almost a factor of 4. The drop in δ is attributed to radiative heat transfer effects, as a higher ṁCeO results in a larger extinction coefficient, which in turn results in a lower particle temperature because of the optically thicker medium. An increase of V̇Ar from its baseline by a factor of 2 resulted in an increase of ηsolar-to-fuel to 0.56% because the local pO was lower, thus shifting the thermodynamic equilibrium to higher δ values at same temperatures. Recycling Ar at these gas flow rates requires its separation from O2, which introduces an energy penalty as indicated in eq . For our experimental conditions with V̇Ar in the range 2–4 L min–1, Einert∫ṅinert dt = 0.7–1.3% of Psolar (2.28 kW).

Particle Morphology

To study the effect of repeated cycling on the particle morphology, we subjected a 200 g batch of ceria powder to eight consecutive redox cycles using the electrically heated setup. Each cycle consisted of several individual experiments with varying feeding rates until the entire 200 g batch was exhausted, for a total of 15 experiments per batch carried out at Tt = 1873 K. Results are shown in Figure a, where δ is plotted as a function of ṁCeO for cycles 1, 3, 5, and 7. δ was generally shown to be inversely proportional to ṁCeO. For example, in cycle 7, δ was 0.039 at ṁCeO = 15 mg s–1 and 0.026 at ṁCeO = 130 mg s–1, reaching respectively 42% and 29% of the thermodynamic limit for Treduction = 1873 K and pO = 5 × 10–5 atm. Furthermore, δ increased substantially from cycle 1 to 3, but decreased slightly between cycle 5 and 7. This trend can be seen more clearly in Figure b, where average reduction extents δav are shown versus cycle number for a narrow range of ṁCeO between 50 and 120 mg s–1. The trend is in the direction of decreasing δ with subsequent redox cycles. In a previous study using an infrared furnace, ceria was shown to be stable for 500 redox cycles, after an initial stabilization period of about 100 redox cycles.[3]

Figure 3

(a) Experimentally measured reduction extents as a function of the ceria feeding rate using the electrically heated setup. Experimental conditions: V̇Ar = 0.5 L min–1, pO < 5 × 10–5 atm, Tt = 1873 K. (b) Evolution of the mean reduction extent, δav, over the first eight cycles. Error bars correspond to 95% confidence intervals. Experimental conditions: ṁCeO = 50 and 120 mg s–1, V̇Ar = 0.5 L min–1, pO < 5 × 10–5 atm, Tt = 1873 K.

Image-based particle size distributions measured in their oxidized state revealed the presence of large particle agglomerates. Results are indicated in Figure a for cycles 1, 3, 5, and 7. During the first three cycles, the agglomerate size decreased, followed by a small increase during the remaining cycles. This correlates well with the behavior of δav shown in Figure b where larger agglomerates correspond to lower δav. Laser scattering measurements shown in Figure b enable insight on the particle history. Fresh particles are shown to have the smallest size because primary particles were measured rather than agglomerates. Upon cycling, the primary particle size increased and a second peak could be detected at ∼160 μm. Even though the fraction of these larger particles gradually increased, no further increase in their size was detected. This growth was likely a result of particle agglomerates sintering, ultimately causing a transition from Geldart class C to class A particles.[27] In addition to the size variation between cycles, the agglomerates also exhibited a variation in their surface morphology, as revealed by SEMs in Figure for cycles 1, 3, 5, and 7. We selected agglomerates of similar size to emphasize their structural differences. During the first three cycles the agglomerates became denser, resulting in smoother surfaces. By cycle 5 these agglomerates were covered with smaller primary particles. This process continued during the remaining cycles, apparently resulting in a porous structure made up of small primary particles (Figure d).

Figure 4

Agglomerate sizes obtained from image processing (a) and corresponding primary particle size distributions obtained by laser scattering (b) for cycles 1, 3, 5, and 7.

Figure 5

SEM images of ceria particles as used during reduction cycles 1, 3, 5, and 7 are shown in panels a, b, c, and d, respectively.

Heat and Mass Transfer Analysis

A 1-D 2-phase coupled heat and mass transfer model is formulated for the reaction zone consisting of a cylindrical tube section (length Lreac = 0.35 m, diameter Dt = 0.191 m) containing a particle flow at a rate ṁCeO falling by gravity and countercurrent to an argon flow at a rate V̇Ar. The particles were assumed spherical with a surface area Ap, falling at terminal velocity based on Schiller–Nauman–Drag through a reaction zone.[28] The solid-phase energy balance for the particle stream considered the incoming radiative flux from the hot tube wall, reradiation, and convective heat transfer with the gas phase.[29,30] For simplicity, the reaction enthalpy was neglected as it typically corresponded to less than 11% of the total energy transferred. Because the optical thickness κ = 0.06 (≪ 1), the near-transparent media approximation was applied. The incident radiative flux was approximated based on the view factor Ft-p(z) to a single particle falling through the tube centerline multiplied by the total radiative flux emitted by the tube wall and the particle concentration N. Normal total emittance of ceria was taken from literature εCeO = 0.75.[31] The total energy incident on the particle becomes independent of the tube wall emittance as a result of the small particle to wall surface ratio. The gas-phase energy balance considered convective heat transfer between gas flow and tube wall and between gas flow and particle stream, with the corresponding heat transfer coefficients ht-g and hp-g calculated using Nu correlations.[29,32] Conduction by particle–particle contact was neglected.[33] Conduction through the particles was neglected as justified by the Biot number <0.07. Note that the gas and particle flows were countercurrent and entered the reaction zone from opposite ends at 300 K. The oxygen source resulted from the change in δ for a given ṁCeO, assuming thermodynamic equilibrium at pO and Ts.[26] The governing 1D steady-state energy conservation equations for the solid (particles) and gas phases arewhere the subscripts s and g correspond to the solid and gas phases, respectively. Additionally, subscripts t and p denote a single particle and the inner tube wall. The governing 1-D mass conservation equations for both gaseous species O2 and Ar arewhere c denotes concentration and v(z) denotes the gas velocity calculated based on the local gas temperature and the ideal gas law. Note that eqs and 5 are coupled to eqs and 7 through Ts(z) and v(z). The boundary conditions are Ts(0) = 300 K, Tg(Lreac) = 300 K, Tt = 1873 K, pO (Lreac) = 100 ppmv, while ṁCeO, V̇Ar, and dp are the operational parameters taken equal to the experimental conditions. Thermal properties (cp, k) of solid and gases were taken from the literature.[34,35]

The modeled particle and gas temperature profiles along the reaction zone are shown in Figure a for particle diameters dp = 100, 150, and 200 μm. The modeled O2 concentration and reaction extent profiles along the reaction zone are shown in Figure b for the corresponding particle diameters. Small particles of dp = 100 μm reached maximum temperatures of 1872 K after only 0.07 m, intermediate particles of dp = 150 μm needed 0.25 m, while larger particles of dp = 200 μm reached a maximum 1679 K before exiting the reaction zone. This temperature difference has a large effect on the final reduction extent due to the strong temperature dependence on the thermodynamic equilibrium as seen from Figure b.[7] Thus, for the investigated solar reactor geometry (Lreac = 0.35 m, Dt = 0.191 m), particle diameters smaller than 200 μm are required to avoid heat transfer limitations. Since the majority of the cycled particles were larger (c.f., Figures and 5), δ was primarily limited by heat transfer. Tg closely follows Ts for dp = 100 μm as a result of their high specific surface area. For a constant ṁCeO, this surface area decreases with increasing dp, resulting in a larger difference between Tg and Ts. The reduction extents shown in Figure b increase significantly after a particle temperature of 1200 K is reached. An additional increase is observed at the end of the reaction zone were the Ar flow enters. For both particle diameters of 150 and 100 μm, which reach Tt, the reduction extent is limited by the local pO resulting from the released O2 and the purge flow. For example, for a particle diameter of 150 μm the maximum particle temperature is reached after a distance of 0.3 m while the reduction continues until unfavorable oxygen partial pressures are reached at 0.34 m.

Figure 6

(a) Modeled particle temperature profiles (Ts, solid lines) along the reaction zone for various particle sizes alongside the corresponding gas phase temperature profiles (Tg, dashed lines). (b) Modeled reduction extent (δ, solid lines) and oxygen partial pressure profiles (pO dashed lines) along the reaction zone. Conditions: ṁCeO = 100 mg s–1, V̇Ar = 0.5 L min–1.

Experimental Validation

To further investigate the effect of different operating conditions (ṁCeO, V̇Ar, and Tt) on reduction extents, particles after cycle 8 were separated using a 90 μm mesh. The resulting particle size distribution had a median diameter of 152 μm, and thus heat transfer limitations were not expected. This batch of particles exhibited no signs of the cohesive nature of the original powder. Again, we subjected this batch to a total of 15 consecutive redox cycles for various ṁCeO. Figure shows δ as a function of ṁCeO for V̇Ar = 0.25, 0.5, 1, and 1.5 L min–1 at Tt = 1873, and for Tt = 1773, 1823, and 1873 K at V̇Ar = 1 L min–1. Symbols indicate measured values of δ, whereas lines correspond to the modeled δ at z = 0.35 m (see Figure b). We observed very good reproducibility for δ between different cycles which is likely a direct result of the stable particle size distribution and morphology. During this study, the experimental error was ±5.5%, whereas the experimental error prior to sieving was ±12.5% (see Figure ).

Figure 7

Experimentally measured reduction extents as a function of the ceria feeding rate for varying V̇Ar of 0.25 L min–1 (black ▶), 0.5 L min–1 (red ●), 1 L min–1 (green ■) and 1.5 L min–1 (blue ◆) at Tt = 1873 K and for varying Tt of 1823 K (purple ▲), 1773 K (teal ◀) at V̇Ar of 1 L min–1. Solid lines indicate corresponding results of the heat and mass transfer model.

Independent of the purge flow rate and temperature, δ was found to be inversely proportional to ṁCeO. Furthermore, increasing either Tt or V̇Ar resulted in higher δ, consistent with predictions by the heat and mass transfer model except for high ceria and Ar mass flow rates (ṁCeO > 175 mg s–1, V̇Ar = 1.5 L min–1). The dependency of δ on Tt was mainly due to thermodynamic equilibrium,[7] while the dependency on ṁCeO and V̇Ar was mainly a result of the gas phase advection. At constant Tt and ṁCeO, an increase in V̇Ar resulted in a decrease of the local pO and, consequently, in higher δ according to thermodynamic equilibrium.[7] At constant Tt and V̇Ar, the oxygen evolution rate increased with ṁCeO, resulting in a higher local pO. To some extent the gas phase advection limitation can be overcome by operating under vacuum pressures, with the additional benefit of shifting the equilibrium to lower temperatures and eliminating the use of inert gas, while reducing the particle residence time.

At high ṁCeO (>175 mg s–1), heat transfer is the rate limiting mechanism in the optically thick particle flow. Under such conditions, the modeled Ts becomes less accurate as the near-transparent media assumption is not fulfilled. The trend that δ increases with increasing V̇Ar, observed both experimentally and in the model, suggests that V̇Ar should be maximized for a given reactor geometry and for the largest particle size while avoiding heat transfer limitations and particle entrainment.

Conclusion

We have experimentally demonstrated the continuous operation of a solar tubular aerosol reactor for the thermochemical reduction of ceria as part of a two-step redox cycle for splitting H2O and CO2. The solar reactor featured an alumina tube positioned inside an insulated cavity-receiver and containing countercurrent flows of ceria particles and inert gas to accomplish effective product separation. Solar tests under high-flux irradiation yielded a maximum solar-to-fuel energy conversion efficiency of 0.56% for a single tube, without attempting the recovery of sensible heat of the hot particle and gas streams. Overall we observed that efficiency increased with both ceria mass flow rate and inert gas flow rate, which affected the residence time, gas phase advection, and heat transfer mainly through the optical thickness. To further probe the physics of the system and cyclability effects, experiments were also performed under well-controlled conditions using an electrically heated tubular reactor. Particle cycling runs revealed a growth of the primary particles from a volume averaged median of 2.44 μm to ∼150 μm as a result of sintering, causing a transition from Geldart class C to class A particles and resulting in a gradual decease of the agglomeration behavior. Specifically, the noncycled powder formed agglomerates with sizes larger than 1 mm, while particles of 150 μm that did not form agglomerates were found after eight cycles, which correlated well with an increase in the reduction extent. A coupled 1-D 2-phase heat and mass transfer model indicated that reduction extents were primarily limited by heat transfer for large particles/agglomerates (>200 μm) and by the gas phase advection of product O2 for smaller particles. Experiments using ceria particles with a median diameter of 152 μm showed that δ decreases with increasing ṁCeO. In a limited range (V̇Ar < 1.5 L min–1 and ṁCeO < 175 mg s–1) we additionally observed an increase in δ with increasing V̇Ar, which supports the prediction of the model. We conclude that this tubular aerosol reactor concept requires the selection of appropriate particle sizes that avoid radiative heat transfer limitations and inert gas flows that enhance gas phase advection but avoid particle entrainment.