Zr – doped ceria for the redox splitting of water

Zr – doped ceria for the redox splitting of water

international journal of hydrogen energy xxx (xxxx) xxx Available online at www.sciencedirect.com ScienceDirect journal homepage: www.elsevier.com/l...

3MB Sizes 0 Downloads 0 Views

international journal of hydrogen energy xxx (xxxx) xxx

Available online at www.sciencedirect.com

ScienceDirect journal homepage: www.elsevier.com/locate/he

Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water Darwin Arifin a, Andrea Ambrosini b, Steven A. Wilson c, Bennett Mandal c, Christopher L. Muhich c,**, Alan W. Weimer a,* a

Department of Chemical and Biological Engineering JSCBB, University of Colorado Boulder, Boulder, CO 80309-0596, USA b Sandia National Laboratories, PO Box 5800, MS 0734, Albuquerque, NM 87185, USA c Chemical Engineering, School for the Engineering of Matter Transport and Energy, Arizona State University, ERC 257, 551 E. Tyler Mall, Tempe, AZ, 85281, USA

highlights

graphical abstract

 H2O splitting of reduced Zr, Gd/Zr, and Pr/Zr doped ceria is surface reaction limited.  Pr and Gd addition induce nonredox active sites, detrimental to H2 production.  Surface

H2

formation

is

rate

limiting, activation barriers > bulk O2 diffusion.  Steam oxidation is best described by a deceleratory power law model (F-model).

article info

abstract

Article history:

There is a renewed interest in CeO2 for use in solar-driven, two-step thermochemical cy-

Received 15 August 2019

cles for water splitting. However, despite fast reduction/oxidation kinetics and high ther-

Received in revised form

mal stability of ceria, the cycle capacity of CeO2 is low due to thermodynamic limitations.

12 October 2019

In an effort to increase cycle capacity and reduce thermal reduction temperature, we have

Accepted 23 October 2019

studied binary zirconium-substituted ceria (ZrxCe1-xO2, x ¼ 0.1, 0.15, 0.25) and ternary

Available online xxx

praseodymium/gadolinium-doped Zr-ceria (M0.1Zr0.25Ce0.65O2, M ¼ Pr, Gd). We evaluate the oxygen cycle capacity and water splitting performance of crystallographically and

Keywords:

morphologically stable powders that are thermally reduced by laser irradiation in a stag-

Solar thermochemical

nation flow reactor. The addition of zirconium dopant into the ceria lattice improves O2

Water splitting

cycle capacity and H2 production by approximately 30% and 11%, respectively. This

Doped ceria

improvement is independent of the Zr dopant level, up to 25%, suggesting that above 10%

Cerium dioxide

Zr dopant level, Zr might be displaced during the high temperature annealing process. The

Abbreviations: WS, water splitting; CDS, carbon dioxide splitting; 10ZrCE, Zr0.1Ce0.9O2; 15ZrCE, Zr0.15Ce0.85O2; 25ZrCE, Zr0.25Ce0.75O2; 10Pr15ZrCe, Pr0.1Zr0.15Ce0.75O2; 10Pr25ZrCE, Pr0.1Zr0.25Ce0.65O2; 10Gd25ZrCE, Gd0.1Zr0.25Ce0.65O2. * Corresponding author. ** Corresponding author. E-mail addresses: [email protected] (C.L. Muhich), [email protected] (A.W. Weimer). https://doi.org/10.1016/j.ijhydene.2019.10.177 0360-3199/© 2019 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved. Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177

2

international journal of hydrogen energy xxx (xxxx) xxx

CeO2

addition of Pr and Gd to the binary Zr-ceria mixed oxide, on the other hand, is detrimental

Hydrogen

to H2 production. A kinetic analysis is performed using a model-based analytical approach to account for effects of mixing and dispersion, and to identify the rate controlling mechanism of the water splitting process. We find that the water splitting reaction at 1000  C and with 30 vol% H2O, for all doped ceria samples, is surface limited and best described by a deceleratory power law model (F-model), similar to undoped CeO2. Additionally, we used density functional theory (DFT) calculations to examine the role of Zr, Pr, and Gd. We find that the addition of Pr and Gd induce non-redox active sites and, therefore, are detrimental to H2 production, in agreement with experimental work. The calculated surface H2 formation step was found to be rate limiting, having activation barriers greater than bulk O diffusion, for all materials. This agrees with and further explains experimental findings. © 2019 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

Introduction One of the most difficult challenges facing the 21st century is building a sustainable energy economy to reduce our overreliance on fossil fuel. Two-step solar thermochemical processes utilizing non-volatile metal oxide cycling are a promising method to capture and utilize solar energy to produce renewable hydrogen. The process can be carbon neutral and does not require a difficult or expensive separation process. When compared to other processes, such as biomass production and processing, artificial photosynthesis, and photovoltaic-driven electrolysis, two-step solar thermochemical processing can operate at higher thermal efficiencies and requires less land and water to operate [1e5]. A generic two-step cycle based on ceria is described by reactions (1) and (2). Heat from concentrating solar energy thermally reduces the metal oxide (CeO2) to a sub-stoichiometric oxide (CeO2-d, where d represents the extent of oxygen nonstoichiometry in the solid) at temperatures (TH) between 1450 and 1550  C, producing O2. The sub-stoichiometric ceria is then taken off sun and oxidized by exposure to steam at some lower temperature (TL, typically  1100  C), thus producing H2 and completing the cycle. 1 1 þD 1 CeO2 / CeO2d þ O2 d d 2

(1)

1 D 1 CeO2d þ H2 O / CeO2 þ H2 d d

(2)

This same cycle can be used to perform carbon dioxide splitting to produce CO, which can be used as a feedstock for synthetic fuel production [2,4,5]. There are numerous types of metal oxide chemistries for two-step solar thermochemical cycling. Metal oxide of nonvolatile ferrite is the prototypical cycle, in which a solid solution of ferrite spinels (MxFe3-xO4 where M is typically Fe, Mn, Co, Ni) is redoxed [6,7]. While this chemistry also has the potential for large cycle capacity, sintering and formation of a molten phase often lead to irreversible deactivation. To improve cyclability and mechanical integrity, these ferrites are often combined with refractory materials such as ZrO2,

YSZ [8e10], or Al2O3 [11]. The mixing of ferrite spinels with Al2O3 led to the development of the “hercynite cycle” where hercynite (FeAl2O4) can be reduced at lower temperatures, but is hampered by slower kinetics for oxidation at the typical 1000  C oxidation temperature [12], although faster oxidation has been demonstrated at 1350  C, where the redox “hercynite cycle” has been operated isothermally [12e15]. The final type of material for solar thermochemical processing is the non-stoichiometric oxide, such as perovskite [16e18] or ceria. The ceria is the prototypical oxide cycle, and has been extensively investigated [19,20]. It has fast redox kinetics [21] but requires high temperatures (TH ¼ 1450e1550  C) to achieve the high extent of oxygen deficiency for efficient fuel production [22]. This non-stoichiometric ceria cycle is in contrast to the stoichiometric reduction of ceria to Ce2O3 that requires temperatures in excess of 2000  C and is plagued with sublimation of the oxide [23]. Currently, cerium oxide is a favored material for the two-step solar thermochemical water splitting because of its high ion conductivity [24e26], rapid exchange kinetics [27], and excellent thermal stability [2]. In an effort to increase cycle capacity and reduce the thermal reduction temperature, substitution and doping of ceria are often employed to introduce lattice defects and create additional oxygen vacancies [28e36]. Taking a page from three way catalysts (TWCs) and heterogeneous catalysis [37e40], Le Gal et al. were the first to examine the effects of incorporating Zr4þ into the lattice of ceria for two-step thermochemical applications. They have shown that Zr substitution can increase the extent of thermal reduction, leading to higher redox performance [41e43], by significantly decreasing reduction enthalpy [44,45]. In addition to zirconium, the introduction of reducible dopants such as praseodymium (Pr) [34] and trivalent cations such as gadolinium (Gd) has also been reported to increase oxygen storage capacity (OSC) and enhance reducibility of ceria based materials [31,46e48]. These authors have also studied the kinetics of water (WS) and carbon dioxide splitting (CDS) of Zr-substituted ceria. They reported that CDS of Zr0.25Ce0.75O2 is limited by diffusion, with activation energies ranging from 83 to 103 kJ/mol depending on the synthesis method [42], and 51 kJ/mol for WS, where surface reaction followed by diffusion is attributed as

Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177

international journal of hydrogen energy xxx (xxxx) xxx

the rate controlling mechanism [41]. The diffusion model only fits the TGA data collected at later points in the oxidation process, and morphological/crystallographical changes appear to occur in their materials, as evidenced by the cycleto-cycle diminishing fuel production. It is critical to properly assess the effects of heat and mass transfer limitations when performing kinetics studies [49,50]. This study evaluates the thermochemical water splitting (WS) of binary (Zr0.1Ce0.9O2, Zr0.15Ce0.85O2, and Zr0.25Ce0.75O2) and ternary oxides (Pr0.1Zr0.15Ce0.75O2, Pr0.1Zr0.25Ce0.65O2, and Gd0.1Zr0.25Ce0.65O1.95). In contrast to the work by Le Gal et al., we evaluate the redox capacity of thermally equilibrated materials that are compositionally and crystallographically stable [51]. In addition, we use quantum simulations to understand and explain the doping effects on both thermodynamics and kinetics.

Experimental The following materials were made and thermochemically cycled: CeO2, Zr0.1Ce0.9O2 (10ZrCe), Zr0.15Ce0.85O2 (15ZrCe), Zr0.25Ce0.75O2 (25ZrCe), Pr0.1Zr0.15Ce0.75O2 (10Pr15ZrCe), Pr0.1Zr0.25Ce0.65O2 (10Pr25ZrCe), and Gd0.1Zr0.25Ce0.65O1.95 (10Gd25ZrCe). The materials were synthesized by coprecipitation of the requisite metallic nitrates, using a similar procedure as reported by Higashi et al. [52] Appropriate amounts of the metal (Ce, Gd, Pr, and Zr) nitrates (Alfa Aesar, Ward Hill, MA) were dissolved in distilled water with mixing. The pH of the mixture was then adjusted by dropwise addition of 1 M oxalic acid until precipitation of mixed metal oxalate was complete. The supernatant liquid was analyzed by induced coupled plasma-atomic emission spectroscopy (ICP-AES) to ensure complete precipitation of the metal cations. The precipitate was filtered, washed with distilled water, and dried in air overnight. The resulting powder was then calcined in air at 700  C for 12 h, and further calcined in air at 1500  C for 16 h. X-ray diffraction (XRD, Scintag PAD5 Powder Diffractometer, CuKa radiation, l ¼ 1.5406  A) analysis was performed using a scan rate of 2 /minute and step width of 0.02 , over the range of 20e70 2q to determine crystallographic phase purity and to evaluate lattice size changes due to incorporation of dopants with different ionic radii. A JEOL JSM-6480 scanning electron microscope (SEM) operating at 15 and 20 kV was used to examine the microstructure and local chemical composition for both the as-calcined and post heat-cycled materials. Elemental composition was determined via energy dispersive X-ray spectroscopy (EDS). Details are provided in the supplemental information (SI) regarding determining zirconium content within the binary zirconium doped ceria (ZrCe). Reduction and oxidation cycles of the pure and doped ceria were performed under different heating/cooling rates, temperatures, times at temperature, and gas atmosperes using an idealized stagnation flow reactor (SFR) The SFR is described in more detail elsewhere [12,21,50,53], but the important features are: (1) the flow field is well-characterized and easily modeled; (2) during thermal reduction, samples are heated by irradiation from a 500 W continuous wave near-IR laser (Apollo Instruments model F500-NIR600), otherwise the SFR

3

temperature is maintained by a SiC furnace (Carbolite STF16/ 180); (3) gas composition is measured by a modulated molecular beam mass spectrometer (Extrel C50); and (4) pressure and flow within the reactor are feedback controlled to give 75 torr and 500 sccm total flow rate, respectively. Between 150 and 200 mg of material were placed in the SFR as a loosely packed bed where all exposed surfaces experience a well mixed gaseous environment within the stagnation flow. Thermal reductions for water splitting were performed by heating the powder to 1500  C, at a low heating rate of 10  C min1 and holding at this temperature for 240 s under a flow of 500 sccm of ultra high purity helium (PO2 < 103 mbar). The powder was then allowed to cool to the SiC furnace temperature that was set at the oxidation temperature. Water splitting experiments (WS) were performed isothermally by exposing the reduced material to 30 vol% steam in helium flow, at 1100 or 1000  C, over a 600 s time interval. Steam was delivered via a research grade humidifier system (RASIRC Rainmaker Humidification System). The concentration of molecular species, such as H2 (m/e ¼ 2) and O2 (m/e ¼ 32), in the reactor effluent was recorded during both thermal reduction and WS. A liquid nitrogen cooled cryogenic trap was used to condense the H2O before gases in the reactor effluent were sampled by the mass spectrometer. Separate oxygen uptake experiments were carried out using a 0.2 mbar O2 in He gas stream for both reduction and oxidation for a high heating/cooling rate of 17  C s 1 and with a 100s hold time. Analytical standards of O2 and H2 were used to calibrate the mass spectrometer. In the postexperimental data analysis, a rigorous numerical procedure was applied to separate material-specific H2 production from experimental effects such as detector time lag, gas phase mixing and dispersion [50].

Computational methods The heterogeneous nature of the oxidation reaction allows physical processes inherent to the experimental apparatus, such as finite detector time lag and gas phase dispersion/ mixing, to impose their temporal imprint on the rate curves of H2 production. Therefore, before any assessment of the kinetics can be performed, these experimental effects must be separated from the as-recorded H2 rates. This was done by the application of a model-based algorithm that (1) assumes that the WS reaction can be described using an empirical model taken from the solid state (SS) kinetic theory [54e56], and (2) combines the solid state model with a validated mixing model of the SFR in the post experimental data analysis [50]. In both the SFR experiments and simulation, the WS step is initiated by a step-change in H2O concentration, occurring in less than 0.5 s after feeding the SFR. This step change initiates the transient H2 production (as governed by equation (3)) that is then propagated through a series of continuously-stirred tank reactors (CSTRs). The series of CSTRs simulates the downstream gas phase dispersion and mixing inherent to the experimental apparatus which alters the temporal characteristics of the rate curve. The dispersed simulated waveform is then compared to the experimentally observed H2 curve. Kinetic rate constants and the governing solid state model are extracted by minimizing a weighted least squares objective

Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177

4

international journal of hydrogen energy xxx (xxxx) xxx

function that compares simulated rates to experimental values at each time step. This optimization process uses a Differential Evolution (DE) algorithm routine to stochastically determine the best-fit SS model and the accompanying kinetic parameters that are unique to the SS model. The governing rate equation used in this analysis is as follows:   1 ¼ k0 ½Yoxid ðtÞg f ðaÞ; where rate s Zt a¼ 0

H2 ðtÞ dt H2; total

(3)

(4)

where k0 is the rate constant of the WS, Yoxid is the timedependent mole fraction of H2O, g is the reaction order of the H2O mole fraction, a is the normalized extent of reaction (Equation (4)), and f(a) is the functional form taken from SS kinetic theory that describes the transient H2 rates as WS progresses. The algorithm samples one of fourteen possible functional forms f(a), representing different rate-limiting phenomena that allows the identification of the governing mechanism within the context of SS kinetic theory. A more detailed description of the numerical approach can be found elsewhere [21,50].

Quantum calculation method The Vienna Ab initio Simulation Package (VAS) [57,58] was used to perform Density Functional Theory (DFT) periodic boundary condition using a plane wave expansion to represent the wavefunction. The PerdewBurke Ernzerhof [59] with a Hubbard correction (PBE þ U) [60] was used for geometry optimization. Projector augmented wave (PAW) pseudopotentials described the oxygen 2s and 2p, the zirconium 5s and 4d, and the cerium, gadolinium, and praseodymium 6s and 5d orbitals explicitly. A 5 eV Hubbard correction, chosen based on previous work [61], was imposed on Ce-f orbitals to account for the strong on-site correlation interaction of localized electrons, which is not accounted for by generalized gradient approximation (GGA) functionals such as PBE. All calculations use at a minimum 500 eV plane wave cutoff energy based on a convergence study from 300 to 600 eV. A difference of <0.1 eV was noted between an energy cutoff of 500 and 600 eV. Calculations were conducted at the G-point. The G point efficiently spans k-space due to the large super cells used and the commensurate extensive Brillion zone folding. HSE06 [62] energies were calculated at each critical point along the reaction path as determined by PBE þ U geometry optimization, including reactants, meta-stable states, transition states and products. The HSE06 hybrid functional provides a more accurate energy because it explicitly includes exact exchange.

Results Structural characterization A representative SEM image of the as calcined 25ZrCe material is presented in Fig. 1 (panel (a)), where primary particles of

diameter 1e5 mm are easily visible. Additionally, EDS maps of Ce(La1) and Zr(La1) absorption edges, shown in panels (b) and (c), indicate that a homogenous dispersion of the cerium and Zr cations within grain level of the solid is maintained, even after the prolonged air calcination at 1500  C. The SEM image of the cycled material is shown in Fig. 1, panel (d). After more than thirty repeated thermochemical cycles at 1500  C (with both O2 and H2O oxidations at 1000 and 1100  C), the image shows that minimal grain growth/sintering occurs; primary particles of similar sizes are still visible. This suggests that evaluation of O2 and H2 cycle capacity performed in this study is conducted with materials that are equilibrated. Panels (e) and (f) show the Ce(La1) and Zr(La1) EDS maps of the cycled material. Similar to the as-calcined powder, the cycled material still exhibits homogenous distribution of Ce and Zr cations. A similar observation was made on the other binary and ternary materials, but micrographs are not shown here. In addition to SEM analysis, phase analysis of the thermochemically-cycled mixed oxide was performed with Xray diffraction, and the resulting spectra are presented in Fig. 2. Identical patterns of cubic fluorite structure are observed for both the pure and doped ceria materials, with no sign of a secondary phase. However, a small shift in 2q to a higher angle is evident in the doped ceria compounds (Fig. 2 (b)); specifically, the Zr-substituted ceria materials. This is A) by the caused by the substitution of Ce cation (Ce4þ ¼ 0.97  A) in the lattice that leads to smaller Zr cation (Zr4þ ¼ 0.84  lattice contraction. Interestingly, the 2q shift appears to be the same for the 10e15% Zr-substituted ceria, while the 25% material exhibits a further shift. This observation is true for both the as-calcined and post-cycled powders (post cycled spectra not shown here). It is possible that Zr ions precipitate out of the ceria solid solution during the initial calcination and reduce the effective Zr substitution in the cubic lattice. Although the EDS maps show homogenous cation dispersion at the grain, they do not rule out the presence of a secondary phase, either due to the lack of magnification, or presence of secondary structure such as Zr:Ce pyrochlore-like structure; since EDS only shows cation dispersion and not changes in structure [63]. The formation of an ordered phase Zr:Ce that cannot be detected with XRD is also possible [64] although Fig. 2 (b) indicates subtle peak splitting that may indicate ordering, especially on the Gd and Pr samples.

Oxygen activity Evaluation of O2 capacity An important first step in evaluating the redox capacity of metal oxides for solar thermochemical water splitting is the assessment of the materials’ oxygen capacity. This can be done by exposing the materials to rapid heating/cooling under a constant oxygen partial pressure, while monitoring the oxygen evolution and uptake behavior. All materials were heated and cooled between 1000 and 1500  C at a rate of 17  C s1 under constant PO2 of 0.2 mbar, with helium balance. The representative O2 uptake and evolution characteristics for CeO2 and 25ZrCe are presented in Fig. 3 (a). Upon heating, the powder undergoes thermal reduction yielding positive O2 production rates. Upon cooling, it undergoes reoxidation evidenced by decreasing background O2 rates. The O2 uptake and

Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177

international journal of hydrogen energy xxx (xxxx) xxx

5

Fig. 1 e SEM micrographs of co-precipitated 25ZrCe: as-calcined (a) and after thermochemical cycling (d). EDS maps of Ce(La1) absorption edge (light areas) of as-calcined material (b) and after thermochemical cycling (e). EDS maps of Zr(La1) absorption edge (light areas) of as-calcined material (c) and after thermochemical cycling (f).

evolution characteristics are rapid: the reduction peaks and reoxidation valleys are almost symmetrical, which implies that the reoxidation by O2 is as rapid as the thermal reduction under these experimental conditions. The transient uptake and evolution behavior are similarly rapid for all the other substituted/doped materials. The total amount of O2 released by the material represents the total amount of O2 capacity at 0.2 mbar PO2. The averages of ten repeated uptake experiments for each composition are summarized in Fig. 3 (b). For the ZreCe binary oxide, we observe a ca. 30% increase in O2 capacity that decreases with increasing Zr content. On the other hand, the addition of 10% Pr and Gd to form ternary oxides does not improve O2 capacity when compared to 25ZrCe. In fact, the ternary compound 10Pr15ZrCe appears to have diminished O2 capacity compared to the Zrsubstituted binary material. This is unexpected since others have reported that a solid solution of Pr:CeO2 [46,65,66] and Pr:Zr:CeO2 [34] reduces energy for oxygen vacancy formation and increases the oxygen non-stoichiometry of ceria at more accessible PO2 levels. Note that during ten repeated cycles, the materials maintain their rapid kinetics and cycle capacity; O2 evolution and uptake curves remain constant and are similar to Fig. 3 (a).

Onset of O2 evolution Another goal of lattice substitution in ceria is to incorporate additional defects into the lattice to weaken the cation-oxygen bonds to reduce the necessary temperature for thermal reduction. We evaluated the onset of O2 evolution exhibited during thermal reduction of the various doped ceria compositions by heating the materials at a rate of 10  C min1 under helium (PO2 < 103 mbar) up to 1500  C and then held for 240s. For this slow heating condition, the materials are presumed to be in thermal equilibrium. Data were collected at 1s intervals. The rate of O2 evolution as a function of temperature is shown in Fig. 4. Co-precipitated CeO2 begins to evolve O2 at ca. 1000  C

(Fig. 4 (a)). The addition of 10e25% Zr and 10% Gd to 25ZrCe does not affect the onset of O2 evolution (Fig. 4 (b) and (c)). Again, the O2 evolution behavior of the Zr-substituted material is independent of the dopant level. This is in agreement with the aforementioned O2 uptake and evolution results. In contrast, the tertiary PrZrCe materials exhibit low temperature O2 evolution peaks, between 500 and 800  C (Fig. 4 (c)). The Tonset where these low temperature peaks occur appears to depend on Zr concentration; Tonset is shifted to lower temperature with increasing nominal Zr concentration. The evolution of the lower energy oxygen can be attributed to the reduction of Pr4þ to Pr3þ that is accompanied by the formation of oxygen vacancies and evolution of O2 from the bulk. In a PrCe system, Pr4þ cations are more amenable to reduction to form Pr3þ than Ce4þ to Ce3þ reduction [46]. In fact, according to XPS analysis by Reddy et al., mixed Pr4þ/Pr3þ can coexist in a ceria solid solution at temperatures as low as 500  C in air [47]. Therefore, unlike Gd, which is a fixed acceptor dopant, incorporation of reducible Pr cations produces additional oxygen vacancies that, if present, are accessible to H2O or CO2 oxidation, can increase thermochemical fuel production. To evaluate if the praseodymium contributes to thermochemical H2 production, we performed paired thermal reduction experiments on 10Pr15ZrCe. The paired thermal reduction experiments consist of two reductions: The first thermal reduction was performed subsequent to O2 reoxidation at 650  C (therefore fully oxidized material was reduced), and the second thermal reduction was conducted subsequent to WS with 30 vol% H2O at 650  C. In short, we are evaluating if the oxygen vacancies that are produced at the lower temperature can be reoxidized by water. The O2 released during these two thermal reductions is presented in Fig. 5. The first thermal reduction exhibits two O2 evolution peaks, similar to the O2 evolution characteristic shown in Fig. 4, where a low temperature O2 evolution, due to the reduction of Pr4þ to Pr3þ, occurs between 500 and 800  C. This is followed by a high

Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177

6

international journal of hydrogen energy xxx (xxxx) xxx

experiments was conducted using 30 vol% H2O in helium at both 1000 and 1100  C. The same powders used for the O2 uptake/evolution experiments were used for the WS evaluation. The average total H2 produced from three WS experiments is summarized in Fig. 6. The 10ZrCe, 15ZrCe, and 25ZrCe produced approximately 11% more H2 than unsubstituted CeO2 at both 1000 and 1100  C. This is much less than the 48% increase afforded by co-precipitated 25ZrCe that was reported by Le Gal et al. [42] Their 48% gain may not be the final, stable H2 productivity gain, however, since H2 productivity decreases by ~20% from cycle #1 to #3. This is possibly due to morphological/crystallographic changes taking place during repeated WS cycles (note: only three cycles were reported by the author). The ternary oxides PrZrCe and GdZrCe produce less H2 than the binary ZrCe. We have shown in the previous section that the Pr4þ/Pr3þ redox pair likely does not participate in the water splitting reaction, and consequently does not improve H2 production. In fact, the addition of Gd to the binary oxide further suppresses H2 production (Fig. 6). Overall, we observe that the addition of trivalent acceptor dopants, such as Gd3þ and Pr3þ to the CeZr binary system is detrimental to not only the H2 cycle capacity, but also the O2 capacity. This is in agreement with recent work by Kuhn et al., where they reported that due to thermodynamic and crystallographic effects, addition of trivalent dopant into the ceria lattice can be detrimental to the gain afforded by Zr substitution [44].

WS Kinetics

Fig. 2 e XRD Patterns of thermochemically cycled ceria oxides. We note that the diffractograms for the as-calcined material (1500  C for 16 h) are identical to the thermochemically cycled material. (a) Complete spectra of the undoped, binary and ternary ceria material. No secondary phase is observed. (b) A closer look at the 2q shift due to cation incorporation.

temperature O2 evolution, which can be attributed to the reduction of Ce4þ to Ce3þ cations. In contrast, the second thermal reduction only exhibits the high temperature O2 evolution. Therefore, the addition of a Pr4þ/Pr3þ redox pair into the Zr-substituted ceria based oxide will most likely not increase H2 production as the low temperature oxygen vacancies are not thermodynamically accessible to oxidation with H2O.

Water splitting activity of doped ceria Evaluation of total H2 produced High oxygen activity is necessary for thermochemical fuel production, but does not guarantee the ability to split water to produce H2. In order to evaluate the H2 cycle capacity of the various mixed metal oxides, a series of isothermal WS

Observation of the progress of the water splitting reaction is limited by the indirect measurement of produced H2 in the effluent, such as in our SFR experiments, or mass gain of the metal oxide during steam exposure, such as in thermogravimetric studies. Both types of observations are susceptible to experimental effects, such as detector time lag and gaseous phase dispersion that leave a temporal imprint on the rate curve of the reaction [50]. To correctly assess the intrinsic kinetics and identify the rate governing mechanism of the oxidation reaction, these experimental effects must be separated from the material-specific transient fuel production. For example, Le Gal et al., via the “master plot analysis” [56], have identified the rate of CO2 splitting using Zr0.25Ce0.75O2 as a diffusion controlled reaction with an activation energy value ranging from 83 to 103 kJ/mol, depending on the synthesis method [42]. They surmise that, due to increased bulk participation in these ceria zirconia materials, oxygen mobility in the bulk limits the oxidation reaction. However, their thermogravimetric data fit the diffusion model only partially. The intrinsic fuel production rate data were not separated from experimental transport effects. This deconvolution is an important consideration, as has been shown by Le Gal and Abanades and Scheffe et al. [41,50], and the work presented herein incorporates such data analysis, distinguishing it from previous studies. In this study, an idealized flow reactor is coupled with a post experimental data reduction algorithm to separate the material specific transient H2 production from the experimental effects. The WS reaction is a transient process in which oxygen is split from water, via surface and bulk mediated processes [67], and transported into the sub-

Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177

international journal of hydrogen energy xxx (xxxx) xxx

7

Fig. 3 e (a) Rate of O2 uptake and evolution by CeO2 and 25ZrCe with constant background O2 of 0.2 mbar. Samples heated at 17  C s¡1 from 1000 to 1500  C, and held at that temperature for 100 s, before being cooled back to 1000  C at the same rate. Positive rates indicate O2 evolution during heating, negative rates indicate O2 uptake of background O2 during cooling. (b) Average O2 uptake over the last 10 cycles.

Fig. 4 e Rate of O2 evolution for slow heating rate. Samples heated at 10  C min¡1 under helium flow at 75 torr. Ternary materials doped with Pr exhibit low temperature O2 peaks that are dependent on the concentration of Zr. Onset of the second O2 evolution appears to be independent of the dopants.

stoichiometric ceria. Unlike a purely catalytic process, the WS reaction goes to completion as the oxygen vacancies within the solid are depleted. The time-dependent H2 production curves of the ZrCe materials at 1000  C are presented in Fig. 7. Upon exposure to 30 vol% steam, H2 production increases to some maximum peak rate, then decreases back to baseline as vacancies are depleted and reaction progresses to completion. This H2 production rate curve is typical of a WS process observed for this type of experiment [50]. Similar to undoped

ceria, the rate of H2 production is rapid, with more than 90% of the H2 produced within the first 300 s [21]. Applying the analytical model-based analysis yields a single kinetic model for WS that describes the H2 rates over the entire time domain. The high-quality fit is evident from the solid lines shown in Fig. 7 (left). A total of fourteen different SS mechanisms, representing different rate governing processes, was tested against experimental data using a least squares objective function evaluated at every

Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177

8

international journal of hydrogen energy xxx (xxxx) xxx

Fig. 5 e Rapid thermal reduction of 10Pr15ZrCe from 650 to 1500  C, with a heating rate of 10  C s¡1, at 75 torr under helium flow. Grey line: thermal reduction after material was oxidized with 0.2 mbar O2. Black line: thermal reduction after material was oxidized with 30 vol% H2O at 650  C and 75 torr. The total amount of observed O2 is shown in parentheses (mmoles/g of material). Low temperature O2 peak can only be reclaimed if the material is oxidized with O2 and not H2O.

Fig. 6 e Average total H2 (mmoles/g of material) produced during isothermal WS with 30 vol% H2O at 1000 and 1100  C for 600 s. WS was carried out after thermal reduction at 1500  C for 240 s. Each bar is the average total H2 produced from three WS experiments.

time step, and since the majority of H2 is produced in the first 300 s of the reaction, preferential weighting of the objective function is used to emphasize fit within this timeframe. We find that the F-model, a power-law deceleratory model, best describes the H2 production. This suggests that the WS kinetics of Zr-substituted ceria is analogous to undoped ceria [21], and that WS is still governed by a surface mediated process that is comparable to a homogenous reaction, contrary to Le Gal’s et al. aforementioned results [2,21]. This

Fig. 7 e (left) H2 production rates as a function of time measured during oxidation in 30 vol% H2O at 1000  C (open symbols). Oxidation performed after thermal reduction at 1500  C for 240 s. The F-model can adequately describe H2 production over the entire time domain. (right) Comparison of actual H2 production rates by 15ZrCe to three different diffusion models acquired by the modelbased analysis. Solid and dashed lines are the result of diffusion governed kinetic modeling. A complete list of the kinetic parameters associated with the SS models used to generate these fits is listed in Table 1. Note: x-axis is in log scale.

result is consistent with WS experiments performed by Petkovich et al., in which they synthesized a three dimensional ordered macroporous structure of 20% Zr-substituted ceria and showed that H2 production and peak rates improve when surface area of material is maximized [21] A complete list of the kinetic parameters associated with the SS models that are used in conjunction with equation (3) to generate these fits is listed in Table 1. To illustrate that the diffusion model cannot sufficiently describe the H2 production, three diffusion models are plotted against the H2 production by 15ZrCe at 1000  C (Fig. 7, right). The diffusion models, regardless of the characteristic dimensions (D1 e 3), cannot adequately describe the H2 produced by 15ZrCe. (Similar observations are true for the other ZrCe materials, but not shown here.) The D1 and D2 fits violate the mass conservation; the total H2 produced exceeds what is experimentally observed due to over-prediction of H2 at later times (>100 s). This is because the D1 and D2 models have decay characteristics that are slower than the experimentally observed phenomenon. Similarly, the D3 model cannot adequately describe the transient H2 production; it over-predicts the rates at early times, and under-predicts at later times. Under the experimental conditions used in this study, the oxygen non-stoichiometry, d, range that is thermodynamically possible for undoped ceria is very small, and the resulting oxidation process has been shown to be limited by surface mechanisms [21]. Although it is possible that the global model

Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177

international journal of hydrogen energy xxx (xxxx) xxx

9

Table 1 e Oxidation reaction rate constants (k0) extracted from fits to experimental data, shown in Figs. 6 and 7, using a coupled kinetic and dispersion model. COMPOUND

CeO2 10ZrCe 15ZrCe 25ZrCe 10Pr25ZrCe 10Gd25ZrCe

RateH2 f k0 YH2O (1 e a)n log k0 (s1)

n

1.10 1.03 0.94 0.90 1.31 1.30

1.37 1.37 1.48 2.02 1.70 1.72

of oxidation for the Zr-Ceria materials is a diffusion model, or a mixed global model where diffusion is a component of the oxidation process, the following two aspects likely must occur: (1) Extent of reduction, d, is sufficiently large that bulk oxygen vacancies become a significant contribution to the oxidation process. (2) Crystallographic phase change occurs due to addition of Zr dopants. However, under the experimental conditions employed in this study, neither (1) or (2) occurred. At 1000  C, the addition of Zr dopants increases the average H2 production by 11%. This corresponds to a very small change in extent of reduction (Dd will only correspondingly increase by 11%). Thus, it is unlikely that bulk VO species play a significant role in the oxidation. Furthermore, Chueh et al. have experimentally shown that due to large bulk and surface differences in entropic contributions, the distribution of VO is non-uniform. Concentration of surface defects of ceria can be one to two orders of magnitude higher than the bulk, especially at low d, such as in our case [21]. Additionally, the crystallographic phase of the Zr-doped materials is expected to be in the cubic phase, the same to that of undoped CeO2 [21,27] and presence of other crystallographic phases were not observed by XRD (Fig. 2). These, along with observations of surface area dependent of the oxidation kinetics support our empirical models [21]. The same numerical procedure is applied to the H2 production of the ternary metal oxides, specifically 10Pr25ZrCe and 10Gd25ZrCe. The H2 production rates of these oxides are summarized in Fig. 8. Open symbols represent the experimentally recorded H2 production rates, whereas the solid lines represent the results of the SS model. The addition of Pr and Gd to the binary ZrCe system alters the local oxygen environment within the lattice, which affects the O2 and H2 cycle capacity. However, the H2 production is still best described by the F-model. Our kinetic model suggests that these bulk perturbations have no effect on the rate governing mechanism of the WS reaction, which remains the order-based deceleratory model. These results imply that the WS process of the ternary doped ceria material studied here is surface mediated and not bulk diffusion limited, the same as the rate governing mechanism of undoped CeO2.

Fig. 8 e H2 production rates as a function of time measured during 30 vol% WS at 1000  C (open symbols). Solid lines are the result of kinetic modeling. Total H2 produced is shown in parentheses (mmoles/g of material). A complete list of the kinetic parameters associated with the SS models that are used to generate these fits is listed in Table 1.

Zr, Gd, and Pr doping on CeO2. To understand the thermodynamics, we calculated the reduction, water splitting energies, and bulk/surface O-vacancy segregation. We calculated the oxygen diffusion activation barriers and surface reaction activation barriers of each material to understand the kinetics.

Calculated thermodynamics The ceria materials have relative oxygen vacancy formation energies of CeO2 > ZrCe > PrCe > GdCe, as shown in Table 2. This suggests that the doped materials will have larger reduction extents under the same conditions, and that GdCe will be the most reduced. It is critical to note that while CeO2, ZrCe, and PrCe reduction is sufficiently endothermic to drive water splitting (i.e. >2.5 eV), GdCe is not, similar to other trivalent cations [68]. The lowered reduction. The energy suggests that sites neighboring the Gd cations in GdCe are not redox active; this explains the decreased performance of the Gd doped tertiary materials. For all considered materials, there is an energy preference for the O-vacancy to be on the surface or in the subsurface layer as compared to the bulk, except for PrCe. In PrCe, it is energetically favorable for an oxygen vacancy to remain in the bulk. This may explain the reduced kinetic performance of the PrCe and Pr tertiary materials because O-vacancies must be on the surface for the water splitting reaction to occur.

Calculated oxidation kinetics Quantum mechanical simulations of kinetics and thermodynamics Density functional theory calculations were conducted in order to understand the thermodynamic and kinetic effects of

To understand the rate limiting step and relative kinetics of the various materials, we examined bulk diffusion activation barriers and surface reaction pathways. It is endothermic for the O-vacancy to move away from the dopants and the diffusion activation barriers are higher for the doped

Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177

10

international journal of hydrogen energy xxx (xxxx) xxx

Table 2 e Oxygen formation energy, vacancy diffusion energy, H2 release energy difference, and overall reaction energy for TWS of each surface. Formula

CeO2 Ce(1-x)ZrxO2 Ce(1-x)PrxO2 Ce(1-x)GdxO2 a b

Bulk O Vacancy Formation Energy (eV)

Bulk O Vacancy Energy preference (Bulk eSurface) (eV)a

Diffusion Activation Energy (eV)

H2Release Energy Requirement (eV)

Overall oxidation Energy (eV)

Thermodynamically able to split waterb

3.77 3.21 3.04 0.57

0.57 0.20 0.27 0.26

0.39 0.79 0.60 0.50

1.9 2.23 1.86 4.84

0.64 0.85 0.73 2.27

yes yes yes no

A negative value indicates that it is energetically favorable for the oxygen vacancy to remain in the bulk. The oxidation reaction energy must be exothermic to drive water splitting [67].

materials than pure CeO2 (0.39 eV); however, the activation barriers are still relatively low for all materials, <0.8 eV, as shown in Fig. 9. The low activation barriers indicate that O diffusion in the bulk is facile for all materials, and this is unlikely to be the rate limiting step. This agrees with the experimental results which show a surface limited reaction rather than a diffusion limited one. We considered that the surface water splitting reaction occurred along a reaction path with five metastable states, namely: 1) a bulk O vacancy, (2) a surface O-vacancy; (3) water absorbed into the vacancy, (4) deprotonation of the water molecule leaving two surface protons on a vacancy free surface, and finally (5) formation and release of H2, as shown in Fig. 10. In this analysis, the vacancy is assumed to occupy the site next to a dopant atom, i.e. Zr, Pr or Gd, in the doped materials. We calculate that it is energetically preferred for the vacancy to occupy this site. The water-filled surface oxygen vacancy is the most stable state for all materials, while the formation of the H2 is the most endothermic step, as shown in Fig. 10. The finding that the H2 formation step is rate determining on CeO2 is in agreement with the work of Hansen and Wolverton [69]. The H2 formation reaction energy is lowest for CeO2 and PrCe (~1.9 eV), slightly higher for ZrCe (2.23 eV), and more than double for GdCe (4.84 eV). The latter means that the overall water splitting reaction is endothermic for Gd and is,

Fig. 9 e Calculated bulk O vacancy diffusion reaction pathway energies. The O-vacancy is considered to move from a dopant neighboring position (left of graph) to a site way from the dopant.

therefore, not thermodynamically viable [70]. The relative energy between the water filled O-vacancy and released water step is the largest energy change along the oxidation reaction path, and the activation barrier for this step must be at least this large. Since this energy is so much higher than the bulk Ovacancy diffusion activation barrier or the differences in energies between other meta-stable states, we determine that H2 formation is the water splitting rate limiting step, even without a calculated transient state energy. We leave the explicit calculation of surface transition states to future work.

Discussion and conclusions The results presented are consistent with literature reports that the incorporation of Zr4þ into the ceria lattice improves ceria’s oxygen capacity, for both low temperature [71e73] and high temperature solar thermal applications [42,74]. In the O2 evolution and uptake experiment (Fig. 2), we observe an increase in O2 capacity afforded by the Zr substituents. It is theorized that the presence of the smaller Zr4þ cation relieves additional strain associated with the reduction of Ce4þ to Ce3þ [71,75], or that the Zr e O bonds store tensile energy that is released upon O-vacancy formation [68]. Others have hypothesized that Zr4þ prefers the 7-fold coordination, in contrast to the 8-fold coordination in the fluorite system, which can drive formation of additional oxygen vacancies through the reduction of ceria cations [44]. Furthermore, for [Zr]  18%, it is possible that the Ce/Zr system undergoes phase enrichment that forms a mixture of fluorite and pyrochlore structures, where the pyrochlore could be more amenable to reduction [63,76]. Regardless of the reason, calculated values of DHred for the Ce/Zr system are typically lower than those of undoped ceria; up to a 30% decrease can be achieved by 10% Zr addition [75]. Interestingly, we observe the increase in O2 capacity with addition of a Zr substituent that decreases with increasing Zr content, and observed improved H2 production that is independent of Zr content, in contrast to what was observed by Le Gal et al. [41] We believe one primary difference is that the O2 and H2 cycle capacities reported here are for materials that are equilibrated. In their WS and CDS studies, oxidation experiments were only repeated two to three times, and fuel production, within the limited cycles, continues to decline [41,42]. Furthermore, in the preparation of these mixed metal oxides, the authors used a relatively low calcination temperature

Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177

international journal of hydrogen energy xxx (xxxx) xxx

11

Fig. 10 e (a) DFT calculated energies along the CeO2 water splitting path at the HSE06 hybrid functional level. All materials are referenced to the (0) fully oxidized form. States along the reaction path are: (1) bulk oxygen vacancy formation, (2) surface oxygen vacancy, (3) water absorption onto the surface vacancy, (5) surface water deprotonation, and (5) released hydrogen from the surface. (b)e(e) Geometric representations of the surface structure along the reaction path for a representative material showing: (b) the surface oxygen vacancy (step 2), (c)adsorbed water (step 3), (d) deprotonated water (step 4) and (e) released H2 (step 5). Large gray and blue, medium red and small white spheres represent Ce, dopant, O and H atoms respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

(between 600 and 800  C); therefore, the material tested may not be cycle-stable, and morphological and chemical changes can be occurring during thermochemical cycling. In this study, the material is calcined at higher temperature for a longer time (1500  C for 16 h); thus, it is conceivable that after prolonged calcination, some Zr comes out of the ceria cubic solid solution. This in effect reduces the effective zirconium

dopant level within the cubic fluorite ceria lattice, yielding a similar effective dopant level for the nominally 10, 15, and 25ZrCe materials. Therefore, for the materials employed here, the addition of Zr above 10 mol % does not provide additional H2 cycle capacity after prolonged thermochemical cycling. Similar observations regarding the relationship between redox performance and zirconium-ceria composition in high

Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177

12

international journal of hydrogen energy xxx (xxxx) xxx

temperature settings have been made by others. Atribak et al. reported that after 1000  C calcination, catalytic activity of ceria-zirconia mixed oxides have little dependence on Zr molar content (from 0.11 to 0.5) [77]. Additionally, Call et al. published a report on carbon dioxide splitting by Ce1-xZrxO2 (with 0  x  0.4) and found that CO2 splitting performance is increased by ~50% for 0.15  x  0.225. However, after four thermochemical cycles, the authors reported that the redox performance difference at different Zr concentrations is not statistically significant [78]. This is likely due to the collapse of homogeneity of the atomic level arrangement of Ce:Zr:O, due to heating above 1000  C, as has been suggested by Nagai et al. [79] The consequence of this phenomenon is that the overall improved H2 production we observe is less significant than what has been reported by others. The improved yield reported here can be considered marginal, and not as extreme as has been previously demonstrated for non-aged materials, and much less than has been reported for lower temperature catalysis applications. Our numerical analysis of the H2 production rates of the doped ceria materials resulted in the high-quality fits to our kinetic models that are evident in Figs. 7 and 8. Regardless of the dopant, all the mixed cerium oxides evaluated in this work have micron-sized primary particles, and kinetics are governed by an order dependence model that suggests that WS is surface-mediated. This is consistent with our previous work, where we have shown that the WS kinetics of ceria is surface limited [21]. It appears this surface dependency holds true despite crystallographic modifications of the binary (ZrCe) or ternary (PrZrCe and GdZrCe) oxides. This implies that oxidation rates will scale with surface area, as has been shown by others [74,80,81]. Furthermore, we have shown that the diffusion model, regardless of the dimension over which it is applied, cannot describe the H2 rate curve over the entire time domain. We have also performed DFT calculations in order to understand the thermodynamic and kinetics based experimental results. Calculations predict that the overall reaction is surface limited for all doped and undoped cases. As seen in Table 2, the diffusion of vacancies through the bulk is ~0.4e0.8 eV while the energy requirement to remove hydrogen from the surface is ~1.8e4.8 eV. Transition state calculations were not performed because it is clear that the activation energy for H2 removal will be even higher than the thermodynamic requirement and this energy is drastically higher than the bulk diffusion energy as shown in Fig. 10. The predicted surface limited nature of the water splitting reaction is in agreement with the experimental results, and details that the process is limited by H2 formation. Calculations show that Zr enhances the water splitting performance of CeO2 because of its higher H2 production capacity [68]. Zr bulk oxygen formation energy is lower (3.21 eV) than that of CeO2 (3.77 eV). This increases the formation of oxygen vacancies within the material during reduction and thus H2 production capacity, as seen experimentally. Additionally, it is energetically favorable for the oxygen vacancy site to occupy surface sites of Zr doped materials; this then increases the number of sites available for reaction. It is worth noting that Zr has a higher barrier for H2 removal from the surface which would slow down the reaction. In conjunction

with the experimental results, we hypothesize that the higher H2 capacity of Zr overcomes the higher H2 removal barriers, as compared to CeO2. The Gd oxygen formation energy was calculated to be 0.57 eV which is lower than the required 2.5 eV required to spilt water [72]. Because Gd neighboring sites are not water splitting active, the inclusion of Gd in the material acts to suppress the H2 production capacity. This explains the substantial decrease in activity of the 10Gd25ZrCe sample as compared to both CeO2 and Zr doped CeO2. Computationally, we calculated that Pr dopants provided no added benefit for TWS. The Pr doped CeO2 oxygen vacancy formation (3.04 eV) and H2 release (1.86 eV) energies are similar to that of Zr (3.21 eV and 2.23 eV respectively). However, Pr preferentially maintains its oxygen vacancies in the bulk, thus lowering the overall number of sites available for TWS on the surface available for reaction. This may explain the lower productivity and water splitting rates of Pr doped ceria. In summary, we have evaluated the oxygen storage capacity and WS kinetics of equilibrated binary Zr-substituted ceria and ternary Pr/Gd-doped Zr-ceria. We observe an increase of approximately 30% oxygen capacity and 11% H2 production increase with the addition of Zr substituent, independent of the Zr concentration. Furthermore, we observe that addition of Pr and Gd to the binary Zr/Ce oxides offer no improvement in the total H2 produced. In fact, addition of Gd is detrimental to H2 production. We have applied a modelbased analytical methodology to reveal intrinsic kinetic behavior for H2 production in Zr substituted ceria. Analogous to undoped ceria, the water splitting (WS) kinetics of the binary and ternary doped ceria system is governed by a powerlaw model consistent with an activated surface mediated process and not diffusion controlled chemistry, as has been previously suggested. Additionally, DFT calculations suggested that the water splitting reaction is surface limited, which is in agreement with experiment, and provided an explanation for the improved performance of Zr substituted ceria but the negligible or detrimental effects of ternary doped ceria systems.

Author Contributions The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

Acknowledgements This material is based upon work supported by the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy (EERE), specifically the Fuel Cell Technologies Office. Sandia National Laboratories is a multimission laboratory managed and operated by National Technology and Engineering Solutions of Sandia LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525. The authors

Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177

international journal of hydrogen energy xxx (xxxx) xxx

acknowledge experimental and intellectual support from Dr. Anthony McDaniel.

Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.ijhydene.2019.10.177.

references

[1] Blankenship RE, Tiede DM, Barber J, Brudvig GW, Fleming G, Ghirardi M, et al. Comparing photosynthetic and photovoltaic efficiencies and recognizing the potential for improvement. Science 2011;332:805e9. [2] Chueh WC, Falter C, Abbott M, Scipio D, Furler P, Haile SM, et al. High-flux solar-driven thermochemical dissociation of CO2 and H2O using nonstoichiometric ceria. Science 2010;330:1797e801. [3] Chueh WC, Haile SM. Ceria as a thermochemical reaction medium for selectively generating syngas or methane from H2O and CO2. Chemsuschem 2009;2:735e9. [4] Miller JE, Allendorf MD, Diver RB, Evans LR, Siegel NP, Stuecker JN. Metal oxide composites and structures for ultrahigh temperature solar thermochemical cycles. J Mater Sci 2008;43:4714e28. [5] Rosen BA, Salehi-Khojin A, Thorson MR, Zhu W, Whipple DT, Kenis PJA, et al. Ionic liquid-mediated selective conversion of CO2 to CO at low overpotentials. Science 2011;334:643e4. [6] Allendorf MD, Diver RB, Siegel NP, Miller JE. Two-step water splitting using mixed-metal ferrites: thermodynamic analysis and characterization of synthesized materials. Energy Fuel 2008;22:4115e24.  mez-Mancebo MB, Ferna  ndez-Saavedra R, [7] Goikoetxea NB, Go rez F, Jime nez JA, et al. Understanding Borlaf F, Garcı´a-Pe water-splitting thermochemical cycles based on nickel and cobalt ferrites for hydrogen production. Int J Hydrogen Energy 2019;44:17578e85. [8] Kodama T, Nakamuro Y, Mizuno T. A two-step thermochemical water splitting by iron-oxide on stabilized zirconia. J Sol Energy Eng Trans-ASME 2006;128:3e7. [9] Gokon N, Murayama H, Umeda J, Hatamachi T, Kodama T. Monoclinic zirconia-supported Fe(3)O(4) for the two-step water-splitting thermochemical cycle at high thermal reduction temperatures of 1400-1600 degrees C. Int J Hydrogen Energy 2009;34:1208e17. [10] Kodama T, Gokon N, Yamamoto R. Thermochemical twostep water splitting by ZrO2-supported NixFe3-xO4 for solar hydrogen production. Sol Energy 2008;82:73e9. [11] Scheffe JR, Li JH, Weimer AW. A spinel ferrite/hercynite water-splitting redox cycle. Int J Hydrogen Energy 2010;35:3333e40. [12] Arifin D, Aston VJ, Liang XH, McDaniel AH, Weimer AW. CoFe2O4 on a porous Al2O3 nanostructure for solar thermochemical CO2 splitting. Energy Environ Sci 2012;5:9438e43. [13] Hoskins AL, Millican SL, Czernik CE, Alshankiti I, Netter JC, Wendelin TJ, et al. Continuous on-sun solar thermochemical hydrogen production via an isothermal redox cycle. Appl Energy 2019;249:368e76. [14] Muhich CL, Evanko BW, Weston KC, Lichty P, Liang X, Martinek J, et al. Efficient generation of H2 by splitting water with an isothermal redox cycle. Science 2013;341:540e2. [15] Muhich CL, Weston KC, Arifin D, McDaniel AH, Musgrave CB, Weimer AW. Extracting kinetic information from complex gasesolid reaction data. Ind Eng Chem Res 2014;54:4113e22.

13

[16] McDaniel A, Miller E, Arifin D, Ambrosini A, Coker E, O’Hayre R, et al. Sr- and Mn-doped LaAlO3d for solar thermochemical H2 and CO production. Energy and Environmental Science. Royal Society of Chemistry; 2013. p. 2424e68.  ski W, [17] Riaz A, Kreider P, Kremer F, Tabassum H, Yeoh JS, Lipin et al. Electrospun manganese-based perovskites as efficient oxygen exchange redox materials for improved solar thermochemical CO2 splitting. ACS Appl Energy Mater 2019;2:2494e505. [18] Barcellos DR, Coury FG, Emery A, Sanders M, Tong J, McDaniel A, et al. Phase identification of the layered perovskite CexSr2exMnO4 and application for solar thermochemical water splitting. Inorg Chem 2019;58:7705e14. [19] Bhosale RR, Takalkar G, Sutar P, Kumar A, AlMomani F, Khraisheh M. A decade of ceria based solar thermochemical H2O/CO2 splitting cycle. Int J Hydrogen Energy 2019;44:34e60. [20] Lu Y, Zhu L, Agrafiotis C, Vieten J, Roeb M, Sattler C. Solar fuels production: two-step thermochemical cycles with cerium-based oxides. Prog Energy Combust Sci 2019;75:100785. [21] Arifin D, Weimer AW. Kinetics and mechanism of solarthermochemical H-2 and CO production by oxidation of reduced CeO2. Sol Energy 2018;160:178e85. [22] Siegel NP, Miller JE, Ermanoski I, Diver RB, Stechel EB. Factors affecting the efficiency of solar driven metal oxide thermochemical cycles. Ind Eng Chem Res 2013;52:3276e86. [23] Abanades S, Flamant G. Thermochemical hydrogen production from a two-step solar-driven water-splitting cycle based on cerium oxides. Sol Energy 2006;80:1611e23. [24] Atkinson A, Barnett S, Gorte RJ, Irvine JTS, McEvoy AJ, Mogensen M, et al. Advanced anodes for high-temperature fuel cells. Nat Mater 2004;3:17e27. [25] Mogensen M, Sammes NM, Tompsett GA. Physical, chemical and electrochemical properties of pure and doped ceria. Solid State Ion 2000;129:63e94. [26] Zhao ZL, Uddi M, Tsvetkov N, Yildiz B, Ghoniem AF. Enhanced intermediate-temperature CO2 splitting using nonstoichiometric ceria and ceria-zirconia. Phys Chem Chem Phys 2017;19:25774e85. [27] Trovarelli A. Catalytic properties of ceria and CeO2containing materials. Catal Rev Sci Eng 1996;38:439e520. [28] Kaneko H, Miura T, Ishihara H, Taku S, Yokoyama T, Nakajima H, et al. Reactive ceramics of CeO2-MOx (M ¼ Mn, Fe, Ni, Cu) for H-2 generation by two-step water splitting using concentrated solar thermal energy. Energy 2007;32:656e63. [29] Kaneko H, Ishihara H, Taku S, Naganuma Y, Hasegawa N, Tamaura Y. Cerium ion redox system in CeO2-xFe(2)O(3) solid solution at high temperatures (1,273-1,673 K) in the two-step water-splitting reaction for solar H-2 generation. J Mater Sci 2008;43. [30] Kaneko H, Tamaura Y. Reactivity and XAFS study on (1-x) CeO(2)-xNiO (x¼0.025-0.3) system in the two-step watersplitting reaction for solar H(2) production. J Phys Chem Solids 2009;70:1008e14. [31] Le Gal A, Abanades S. Dopant incorporation in ceria for enhanced water-splitting activity during solar thermochemical hydrogen generation. J Phys Chem C 2012;116:13516e23. [32] Muhich C, Hoes M, Steinfeld A. Mimicking tetravalent dopant behavior using paired charge compensating dopants to improve the redox performance of ceria for thermochemically splitting H2O and CO2. Acta Mater 2018;144:728e37.

Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177

14

international journal of hydrogen energy xxx (xxxx) xxx

[33] Bulfin B, Lange M, de Oliveira L, Roeb M, Sattler C. Solar thermochemical hydrogen production using ceria zirconia solid solutions: Efficiency analysis. Int J Hydrogen Energy 2016;41:19320e8. [34] Dasari HP, Ahn K, Park SY, Ji HI, Yoon KJ, Kim BK, et al. Hydrogen production from water-splitting reaction based on RE-doped ceria-zirconia solid-solutions. Int J Hydrogen Energy 2013;38:6097e103. [35] Hao Y, Yang CK, Haile SM. Ceria-zirconia solid solutions (Ce1-xZrxO2-delta, x <¼ 0.2) for solar thermochemical water splitting: a thermodynamic study. Chem Mater 2014;26:6073e82. [36] Zhao ZL, Uddi M, Tsvetkov N, Yildiz B, Ghoniem AF. Redox kinetics and nonstoichiometry of Ce0.5Zr0.5O2-delta for water splitting and hydrogen production. J Phys Chem C 2017;121:11055e68. [37] Di Monte R, Kaspar J. Heterogeneous environmental catalysis - a gentle art: CeO2-ZrO2 mixed oxides as a case history. Catal Today 2005;100:27e35. [38] Di Monte R, Kaspar J. Nanostructured CeO2-ZrO2 mixed oxides. J Mater Chem 2005;15:633e48. [39] Kaspar J, Fornasiero P, Baiducci G, Di Monte R, Hickey N, Sergo V. Effect of ZrO2 content on textural and structural properties of CeO2-ZrO2 solid solutions made by citrate complexation route. Inorg Chim Acta 2003;349:217e26. [40] Boaro M, Desinan S, Abate C, Ferluga M, de Leitenburg C, Trovarelli A. Study on redox, structural and electrical properties of CexZr1-xO2 for applications in SOFC anodes. J Electrochem Soc 2011;158:P22e9. [41] Le Gal A, Abanades S. Catalytic investigation of ceriazirconia solid solutions for solar hydrogen production. Int J Hydrogen Energy 2011;36:4739e48. [42] Le Gal A, Abanades S, Flamant G. CO(2) and H(2)O splitting for thermochemical production of solar fuels using nonstoichiometric ceria and ceria/zirconia solid solutions. Energy Fuels 2011;25:4836e45. [43] Abanades S, Legal A, Cordier A, Peraudeau G, Flamant G, Julbe A. Investigation of reactive cerium-based oxides for H(2) production by thermochemical two-step water-splitting. J Mater Sci 2010;45:4163e73. [44] Kuhn M, Bishop SR, Rupp JLM, Tuller HL. Structural characterization and oxygen nonstoichiometry of ceriazirconia (Ce1-xZrxO2-delta) solid solutions. Acta Mater 2013;61:4277e88. [45] Takacs M, Scheffe JR, Steinfeld A. Oxygen nonstoichiometry and thermodynamic characterization of Zr doped ceria in the 1573e1773 K temperature range. Phys Chem Chem Phys: R Soc Chem 2015:7813e22. [46] Chatzichristodoulou C, Hendriksen PV. Oxygen nonstoichiometry and defect chemistry modeling of Ce(0.8) Pr(0.2)O(2-delta). J Electrochem Soc 2010;157:B481e9. [47] Reddy BM, Thrimurthulu G, Katta L, Yamada Y, Park SE. Structural characteristics and catalytic activity of nanocrystalline ceria-praseodymia solid solutions. J Phys Chem C 2009;113:15882e90. [48] Meng Q-L, Lee C-i, Kaneko H, Tamaura Y. Solar thermochemical process for hydrogen production via twostep water splitting cycle based on Ce1xPrxO2d redox reaction. Thermochim Acta 2012;532:134e8. [49] Ji HI, Davenport TC, Gopal CB, Haile SM. Extreme high temperature redox kinetics in ceria: exploration of the transition from gas-phase to material-kinetic limitations. Phys Chem Chem Phys 2016;18:21554e61. [50] Scheffe JR, McDaniel AH, Allendorf MD, Weimer AW. Kinetics and mechanism of solar-thermochemical H2 production by oxidation of a cobalt ferrite-zirconia composite. Energy Environ Sci 2013;6:963e73.

 V. [51] Le Gal A, Abanades S, Bion N, Le Mercier T, Harle Reactivity of doped ceria-based mixed oxides for solar thermochemical hydrogen generation via two-step watersplitting cycles. Energy Fuel 2013;27:6068e78. [52] Higashi K, Sonoda K, Ono H, Sameshima S, Hirata Y. Synthesis and sintering of rare-earth-doped ceria powder by the oxalate coprecipitation method. J Mater Res 1999;14:957e67. [53] McDaniel AM. Virtually accessible laser heated stagnation flow reactor for characterizing redox chemistry of materials under extreme conditions. Sandia National Laboratories; 2019 [HydroGen Advanced Water Splitting Materials]. [54] Sestak J, Malek J. Diagnostic limits of phenomenological models of heterogeneous reactions and thermal-analysis kinetics. Solid State Ion 1993;63e5:245e54. [55] Gotor FJ, Criado JM, Malek J, Koga N. Kinetic analysis of solidstate reactions the universality of master plots for analyzing isothermal and nonisothermal experiments. J Phys Chem 2000;104:10777e82. [56] Khawam A, Flanagan DR. Solid-state kinetic models: basics and mathematical fundamentals. J Phys Chem B 2006;110:17315e28. [57] Kresse G, Furthmu¨ller J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys Rev B 1996;54:11169. [58] Kresse G, Furthmu¨ller J. Efficiency of ab-initio total energy calculations for metals and semiconductors using a planewave basis set. Comput Mater Sci 1996;6:15e50. [59] Perdew JP, Burke K, Ernzerhof M. Generalized gradient approximation made simple. Phys Rev Lett 1996;77:3865. [60] Hubbard J. Electron correlations in narrow energy bands. In: Proceedings of the royal society of london series amathematical and physical sciences. vol. 276; 1963. p. 238. [61] Lu D, Liu P. Rationalization of the Hubbard U parameter in CeOx from first principles: unveiling the role of local structure in screening. J Chem Phys 2014;140:084101. [62] Heyd J, Scuseria GE, Ernzerhof M. Erratum: “Hybrid functionals based on a screened Coulomb potential” [J. Chem. Phys. [bold 118], 8207 (2003)]. J Chem Phys 2006;124. 219906-1. [63] Shah PR, Kim T, Zhou G, Fornasiero P, Gorte RJ. Evidence for entropy effects in the reduction of ceria-zirconia solutions. Chem Mater 2006;18:5363e9. [64] Abdollahzadeh-Ghom S, Zamani C, Andreu T, Epifani M, Morante JR. Improvement of oxygen storage capacity using mesoporous ceria-zirconia solid solutions. Appl Catal B Environ 2011;108:32e8. [65] Bishop SR, Stefanik TS, Tuller HL. Electrical conductivity and defect equilibria of Pr(0.1)Ce(0.9)O(2-delta). Phys Chem Chem Phys 2011;13:10165e73. [66] Bishop SR, Stefanik TS, Tuller HL. Defects and transport in PrxCe1-xo2-delta: composition trends. J Mater Res 2012;27:2009e16. [67] Davenport TC, Kemei M, Ignatowich MJ, Haile SM. Interplay of material thermodynamics and surface reaction rate on the kinetics of thermochemical hydrogen production. Int J Hydrogen Energy 2017;42:16932e45. [68] Muhich C, Steinfeld A. Principles of doping ceria for the solar thermochemical redox splitting of H2O and CO2. J Mater Chem: R Soc Chem 2017:15578e90. [69] Hansen HA, Wolverton C. Kinetics and thermodynamics of H2O dissociation on reduced CeO2(111). J Phys Chem C 2014;118:27402e14. [70] Muhich CL, Ehrhart BD, Witte VA, Miller SL, Coker EN, Musgrave CB, et al. Predicting the solar thermochemical water splitting ability and reaction mechanism of metal oxides: a case study of the hercynite family of water splitting cycles. Energy Environ Sci 2015;8:3687e99.

Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177

international journal of hydrogen energy xxx (xxxx) xxx

[71] Mamontov E, Egami T, Brezny R, Koranne M, Tyagi S. Lattice defects and oxygen storage capacity of nanocrystalline ceria and ceria-zirconia. J Phys Chem B 2000;104:11110e6. [72] Hori CE, Permana H, Ng KYS, Brenner A, More K, Rahmoeller KM, et al. Thermal stability of oxygen storage properties in a mixed CeO2-ZrO2 system. Appl Catal B Environ 1998;16:105e17. [73] Madier Y, Descorme C, Le Govic AM, Duprez D. Oxygen mobility in CeO2 and CexZr(1-x)O2 compounds: study by CO transient oxidation and O-18/O-16 isotopic exchange. J Phys Chem B 1999;103:10999e1006. [74] Petkovich ND, Rudisill SG, Venstrom LJ, Boman DB, Davidson JH, Stein A. Control of heterogeneity in nanostructured Ce1-xZrxO2 binary oxides for enhanced thermal stability and water splitting activity. J Phys Chem C 2011;115. [75] Balducci G, Kaspar J, Fornasiero P, Graziani M, Islam MS, Gale JD. Computer simulation studies of bulk reduction and oxygen migration in CeO2-ZrO2 solid solutions. J Phys Chem B 1997;101:1750e3. [76] Zhou G, Shah PR, Kim T, Fornasiero P, Gorte RJ. Oxidation entropies and enthalpies of ceria-zirconia solid solutions. Catal Today 2007;123:86e93.

15

[77] Atribak I, Bueno-Lopez A, Garcia-Garcia A. Thermally stable ceria-zirconia catalysts for soot oxidation by O-2. Catal Commun 2008;9:250e5. [78] Call F, Roeb M, Schmu¨cker M, Bru H, Curulla-Ferre D, Sattler C, et al. Thermogravimetric analysis of zirconiadoped ceria for thermochemical production of solar fuel. Am J Anal Chem: Sci Res (N Y) 2013:34e45. [79] Nagai Y, Yamamoto T, Tanaka T, Yoshida S, Nonaka T, Okamoto T, et al. XAFS and XRD analysis of ceria-zirconia oxygen storage promoters for automotive catalysts. Top Catal 2008;47:137e47. [80] Venstrom LJ, Petkovich N, Rudisill S, Stein A, Davidson JH. The effects of morphology on the oxidation of ceria by water and carbon dioxide. J Sol Energy Eng Trans ASME 2012;134:8. [81] Furler P, Scheffe J, Marxer D, Gorbar M, Bonk A, Vogt U, et al. Thermochemical CO2 splitting via redox cycling of ceria reticulated foam structures with dual porosities. Phys Chem Chem Phys: R Soc Chem 2014:10503e11.

Please cite this article as: Arifin D et al., Investigation of Zr, Gd/Zr, and Pr/Zr e doped ceria for the redox splitting of water, International Journal of Hydrogen Energy, https://doi.org/10.1016/j.ijhydene.2019.10.177