Kinetic and Thermodynamic Enhancement of Low-Temperature Oxygen Release from Strontium Ferrite Perovskites Modified with Ag and CeO2

The redox behavior of the nonstoichiometric perovskite oxide SrFeO3−δ modified with Ag, CeO2, and Ce was assessed for chemical looping air separation (CLAS) via thermogravimetric analysis and by cyclic release and uptake of O2 in a packed bed reactor. The results demonstrated that the addition of ∼15 wt % Ag at the surface of SrFeO3−δ lowers the temperature of oxygen release in N2 by ∼60 °C (i.e., from 370 °C for bare SrFeO3−δ to 310 °C) and more than triples the amount of oxygen released per CLAS cycle at 500 °C. Impregnation of SrFeO3−δ with Ag increased the concentration of oxygen vacancies at equilibrium, lowering (3 – δ) under all investigated oxygen partial pressures. The addition of CeO2 at the surface or into the bulk of SrFeO3−δ resulted in more modest changes, with a decrease in temperature for O2 release of 20–25 °C as compared to SrFeO3−δ and a moderate increase in oxygen yield per reduction cycle. The apparent kinetic parameters for reduction of SrFeO3−δ, with Ag and CeO2 additives, were determined from the CLAS experiments in a packed bed reactor, giving activation energies and pre-exponential factors of Ea,reduction = 66.3 kJ mol–1 and Areduction = 152 mol s–1 m–3 Pa–1 for SrFeO3−δ impregnated with 10.7 wt % CeO2, 75.7 kJ mol–1 and 623 molO2 s–1 m –3 Pa–1 for SrFeO3−δ mixed with 2.5 wt % CeO2 in the bulk, 29.9 kJ mol–1 and 0.88 molO2 s–1 m–3 Pa–1 for Sr0.95Ce0.05FeO3−δ, and 69.0 kJ mol–1 and 278 molO2 s–1 m–3 Pa–1 for SrFeO3−δ impregnated with 12.7 wt % Ag, respectively. Kinetics for reoxidation were much faster and were assessed for two materials with the slowest oxygen uptake, SrFeO3−δ, giving the activation energy Ea,oxidation = 177.1 kJ mol–1 and pre-exponential factor Aoxidation = 3.40 × 1010 molO2 s–1 m–3 Pa–1, and Sr0.95Ce0.05FeO3−δ, giving the activation energy Ea,oxidation = 64.0 kJ mol–1, and pre-exponential factor Aoxidation = 584 molO2 s–1 m–3 Pa–1.

S2 discernible CeO2 peaks, indicating that the Ce was incorporated into the perovskite structure, rather than forming a separate phase. Instead, a characteristic shift in the SrFeO3 peak at 2θ = 47.0° can be noticed in Fig. S3, in agreement with previous results for SCeFO 2 . The XRD results for Ag/SFO contain additional peaks corresponding to metallic silver, with an estimated Ag loading of 14.5 wt%, in line with the expected value, 15 wt%.    Crystallite size of Ag and CeO2 in the prepared materials was estimated from XRD measurements, using the Scherrer equation: where τ is the mean crystallite size (nm), K is a dimensionless shape factor (taken as K ≈ 0.9), λ is the Cu-Kα X-ray wavelength (0.15406 nm), β is the full width at half maximum (radians), and θ is the Bragg angle (radians). Estimated values are given in Table S3, with no change in Ag crystallite when comparing Ag/SFO before and after CLAS cycling, but with spent CeO2/SFO showing an increase in mean CeO2 crystallite size after 250 CLAS cycles. To confirm estimated loadings of CeO2 on CeO2/SFO and (CeO2)ssSFO, and Ag on Ag/SFO, ICP-AES measurements were performed, with results summarised in Table S4.

S3. Derivation of rate expression
Assuming a pseudo-steady state condition and first-order reaction, the concentration profile along the packed bed can be simplified for the case when the reaction is just starting ( = 0) and is uniform across the bed (i.e. no profile in has been developed yet, so = 0 ), as explained in 3 . Then, the balance for oxygen gives: where 2 is the mole fraction of oxygen at a point along the bed, 2 , is the mole fraction of oxygen that would be in equilibrium with the solid material, v is the superficial gas velocity through the bed (m s -1 ), z is the distance along the bed (m), k is the rate constant (mol s -1 m -3 Pa -1 ), T is the reaction temperature (K), and R is the molar gas constant (kJ mol -1 K -1 ). The superficial velocity can then be related to the constant molar flux of inert N2, 2 (mol m -2 s -1 ): where P is the total pressure in the reactor (taken as ~ 1.01 × 10 5 Pa).
Combining Eqs. S2 and S3, and integrating over the bed length, L: In the case of a switch in feed gas from air to nitrogen, the value of 2 , was taken to be 10 -5 (i.e. using the nominal purity of the cylinder N2 as 99.999 vol%), the value of 2 , was taken to be 0.21, and the value of 2 , was estimated by finding the maximum value of 2 from the UEGO signal, as shown in Fig. 6 of the main manuscript. Then, by solving the integrals in Eq. S4 analytically, the first-order rate constant, k, is given by The apparent reaction constant for re-oxidation of SFO was determined by evaluating Eq. S5 using xin = 0.0505 (i.e. the mole fraction of oxygen in the supplied gas from the cylinder), xeq = 10 -5 , and xout = (0.0505 -|xUEGO|), where xUEGO is the maximum negative deviation measured using the UEGO oxygen sensor with respect to the blank curve recorded over an inert bed. S11 Figure  Initial stoichiometry of the SrFeO3-δ perovskite was estimated from the change in gradient associated with the reduction of SrFeO2.5 4,5 .

S4. Supplementary thermogravimetric analysis (TGA)
where mBM is the relative mass at the change in gradient (as shown on Fig. S11 cooling rate) to room temperature in N2 and characterised via XRD (shown in Fig. S11c), confirming that SrFeO2.5 was the main phase present.

S5. Mass transfer limitations on oxygen release in packed bed experiments
During experiments in a packed bed reactor, the measured rate of release of oxygen might be limited by (1) external mass transport between the surface of the particles of OC material and the flowing gas stream, (2) internal mass transport within the pores of particles of OC materials, (3) thermodynamic equilibrium.
To confirm that the experimental arrangement used to estimate kinetics of oxygen release and uptake would not be limited by external mass transfer from the solids to the gas phase, the maximum rate of mass transport from the surface of the OC particles to the gas stream was estimated. Assuming that the rate of oxygen transport is approximately constant across a thin packed-bed at the start of the particle reduction step, the rate of oxygen transport can be assessed as: where r is the rate of oxygen transport per mass of oxygen carrier (mol s -1 ), kM,g is the external where Sh is the Sherwood number, Sc is the Schmidt number for the gas phase, and Re is the Reynolds number. To evaluate Sh and Re, the characteristic length was used, where ⟨dp⟩ is the geometric mean of the range of particle sizes used in the packed bed (i.e. ⟨dp⟩ = √180 • 300 = 253 µm for the packed beds of OC particles used here).
Evaluating Eq. S7 over the temperature range 500-600°C, the estimated rate of oxygen transport at the start of the reduction experiments was ~ 50 mol s -1 gOC, i.e. 6 orders of magnitude greater than the rates of reduction and 5 orders of magnitude greater than the maximum rate of oxidation measured from experiments. where rv,obs is the observed rate of reaction per particle in the packed bed (mol s -1 m -3 ) (positive values of rv,obs correspond to oxygen release, and negative values of rv,obs correspond to oxygen uptake), and ηeff is an effectiveness factor. The effectiveness factor is taken as for a reversible first order reaction, which was shown by Bohn to give the same dependency on ϕ as an irreversible first order reaction 11 : We note, however, that the oxygen release from a solid oxide involves a zero-order forward reaction and a first order backward reaction (if expressed per mol of O2), possibly affecting the resulting dependency of ( ).
In the case where the influence of external mass transport of gas away from the surface of the particle is negligible, as assumed here, 2 , ≈ 2 , , where 2 , is the overall concentration of oxygen in the bulk gas stream.
Effective diffusivity of oxygen within the pores is then given by where Dm is the molecular diffusivity of oxygen in nitrogen (m 2 s -1 ), estimated using Chapman-Enskog theory 12 , εp is the porosity of the OC materials, and τp is the pore tortuosity (taken as τp = 2). Porosity was estimated from where ρs is the density of particles of SFO (taken as ρs = 2486 kg m -3 ), and 3 is the density of non-porous SrFeO3, taken to be 3 = 5570 kg m -3 from the Springer Materials database 13 . The rate of reaction per particle was estimated from the measured overall rate of reaction measured over the entire bed volume, r (mol s -1 m -3 ) using is satisfied. Hence, for each temperature investigated, the critical value of r such that both sides of Eq. S13 are equal (i.e. the threshold for internal mass transfer contributing significantly to observed rate of reaction) was calculated. As shown in Fig. S14a,   For re-oxidation of SFO, shown in Fig. S14b, the estimated values of ϕ substantially exceed the criterion specified in Eq. S13; hence, the observed rate of reaction was likely limited by internal mass transfer. To account for this, the estimated values of the rate constant from Eq. S5, k', S21 were divided by an effectiveness factor, ηeff, given as a function of ϕ for spherical particles by 15 .
As ϕ is an implicit function of k, Eq. S14 was solved iteratively until values of k, ϕ, and ηeff

S6. Comparison of linear and non-linear kinetic fitting
Kinetic parameters were estimated from gas-cycling experiments by assuming that when the maximum rate of reaction is observed (approximately 0.07 s after the inlet gas is switched from air to N2), the material in the bed is approximately in equilibrium with air (i.e. pO2 = 0.21 atm).
Connecting the measured rates of reaction to in oxygen non-stoichiometry, requires further assumption that can be approximated as 0 . Apparent activation energy, Ea, and preexponential factor, A, were then extracted from the measured values for the maximum rates by using linear and non-linear regression 16,17 . For all samples of interest, the rate of oxygen release was affected by internal mass-transfer at 600°C (as described in Fig. S14). At 475°C in Fig. 16

S7. Oxygen depletion over Ag/SFO
In experiments using dilute (5.05 vol%) O2 to re-oxidise oxygen carrier materials, the high reactivity of Ag/SFO resulted in total depletion of the oxygen in the gas stream. A comparison between the oxygen depletion profiles for SFO, and Ag/SFO, is shown in Figs. S18 and S19, showing the rate of re-oxidation of Ag/SFO was sufficient to fully deplete the available oxygen, for all experiments at or above 525°C. Contrastingly, for unmodified SFO, the full depletion of available oxygen was only observed for measurements at or above 575°C.