Thermogravimetric Analysis of Zirconia-Doped Ceria for Thermochemical Production of Solar Fuel

Developing an efficient redox material is crucial for thermochemical cycles that produce solar fuels (e.g. H2 and CO), enabling a sustainable energy supply. In this study, zirconia-doped cerium oxide (Ce1−xZrxO2) was tested in CO2-splitting cycles for the production of CO. The impact of the Zr-content on the splitting performance was investigated within the range 0 ≤ x < 0.4. The materials were synthesized via a citrate nitrate auto combustion route and subjected to thermogravimetric experiments. The results indicate that there is an optimal zirconium content, x = 0.15, improving the specific CO2-splitting performance by 50% compared to pure ceria. Significantly enhanced performance is observed for 0.15 ≤ x ≤ 0.225. Outside this range, the performance decreases to values of pure ceria. These results agree with theoretical studies attributing the improvements to lattice modification. Introducing Zr into the fluorite structure of ceria compensates for the expansion of the crystal lattice caused by the reduction of Ce to Ce. Regarding the reaction conditions, the most efficient composition Ce0.85Zr0.15O2 enhances the required conditions by a temperature of 60 K or one order of magnitude of the partial pressure of oxygen p(O2) compared to pure ceria. The optimal composition was tested in long-term experiments of one hundred cycles, which revealed declining splitting kinetics.


Introduction
Synthesis gas (or syngas)-primarily a mixture of H 2 and CO-is one of the most promising sustainable energy carriers when produced from renewable resources.It offers an exceptional energy density and is a universal precursor for the production of a very broad variety of chemical substances like polymers, or methanol and especially gaseous and liquid synthetic fuels via the Fischer-Tropsch synthesis, and related processes [1][2][3].Syngas can also be combusted for electricity generation in highly efficient combined Brayton-Rankine cycles [4].
Today, syngas is mainly produced by steam or dry (CO 2 ) reforming or partial oxidation of fossil resources, mainly natural gas, accompanied by a substantial emission of greenhouse gases [5][6][7].For a transition away from fossil energy, several processes have been devel-oped that produce syngas from renewable sources (CO 2 , water) employing solar energy to cover the reaction heat [8,9].These processes meet the demands of a secure, clean and sustainable energy supply converting solar energy into transportable and dispatchable fuels [10] referred to as solar fuels.
The direct method to produce syngas from solar thermal energy is the thermolysis of water and CO 2 molecules in a single step.This prevents energy losses during material handling and exhibits the closest match between the theoretically required solar energy and the energy released by utilizing the produced fuel.On the downside, the equilibrium constant for thermolysis of water and CO 2 is not unity at temperatures less than approximately 4000˚C [11,12].Reasonable H 2 or CO production yields via thermolysis require at least temperatures of about 2000˚C [12,13] and/or a significantly reduced partial pressure of oxygen.Besides the challenges for the reactor construction caused by these impractical temperatures, direct thermolysis produces a mixture of H 2 and O 2 or CO and O 2 , requiring high-temperature gas separation [8].
Thermochemical two-step cycles based on metal oxides that operate at significantly lower temperatures have drawn great attention in the last decades [14][15][16][17][18].The general process concept of two-step CO 2 -splitting cycles is depicted in Figure 1.MO denotes a metal-based redox material, which is either reduced (MO red ) or oxidized (MO ox ).The first step, referred to as thermal reduction (red), is the solar-driven endothermic dissociation of the metal oxide either to the elemental metal or the lowervalence metal oxide.The second step, the CO 2 splitting (ox), is the exothermic oxidation of the reduced material to form CO. The overall reaction of the cycle is as follows: Injecting water steam instead of (or with) carbon dioxide enables the production of hydrogen (or syngas).
While experimental campaigns such as HYDROSOL 2 proved the operability of this process on a solar tower [19,20], identifying that an efficient metal oxide is crucial for the commercialization of this technology.Many materials have been investigated such as various types of ferrites that suffer from sintering and slow kinetics [21][22][23][24][25][26][27][28] as well as cycles based on zinc or tin oxide requiring rapid quenching because of volatilization [29][30][31][32].Recent material studies focused on ceria as the active material [33][34][35][36].Non-stoichiometric ceria combines excellent reactivity due to high oxygen ion conductivity with good cyclability thanks to high-temperature stability.
A test campaign at the High-Flux Solar Simulator of ETH Zurich confirmed the feasibility of a ceria-based cavity reactor [37,38] and innovative reactor designs are promising in regard to the overall process efficiency [39].eria with zirconia or lanthanides enhances the impact of zirconia doping in ce via the citrate nitrate auto by X-ray diffraction (X hesized materials was inve cycle performance [40][41][42].
In the present study, the ria on the CO 2 -splitting performance was investigated by means of thermogravimetric analyses.Particularly, the Zr-content featuring the highest specific yield has been identified and analyzed in terms of reaction conditions and long-term stability.

Experimental Section
The materials were synthesized combustion route similar to ones reported elsewhere [43].Desired amounts of cerium (III) nitrate (99.9% purity, Merck) and zirconoium (IV) oxynitrate hexahydrate (99.99% purity, Sigma Aldrich) were dissolved in deionized water using a reaction vessel made of quartz.Citric acid (99% purity, Merck) also dissolved in deionized water was added to the nitrates in a molar ratio of 1:2 (cations:citric acid).Water evaporation at 95˚C on a heating plate under continuous stirring yielded a yellowcolored gel.Heating this gel to 200˚C for 20 minutes resulted in a swollen foam exhibiting a very low density.During slow heating to 500˚C, the auto combustion took place leaving a fine oxide powder in the reaction vessel.Subsequent calcination in the reaction vessel in a muffle furnace at 800˚C for 1h under air ensured the removal of remaining carbonaceous species.Further calcination at 1400˚C for 1 h in a Pt crucible completed the synthesis route.For each composition two batches were synthesized to guarantee reproducibility.
Phase analyses were performed RD) using a computer-controlled diffractometer (D-5000, Siemens, Germany) with CuK α radiation.As a result, all materials showed a cubic fluorite structure as observed for pure ceria with a small peak shift to high 2θ angle due to the lattice contraction caused by the smaller Zr 4+ ions.Microstructures were observed by scanning electron microscopy (SEM) Ultra 55 FEG (Carl Zeiss, Germany) equipped with an energy-dispersive X-ray spectroscopy (EDS) system.
The performance of the synt stigated by thermogravimetric analysis (TGA) with the aid of a thermo balance STA 449 F3 Jupiter (Netzsch, Germany).The powder was placed on an Al 2 O 3 plate (13 mm in diameter covered with a Pt-foil).During reduction, the measuring cell was purged with Ar (5.0).Due to oxygen impurities of less than 2 ppm, the partial pressure was calculated to be approximately 5 × 10 −6 bar.Comparison of maximum reducibilities obtained in calibration runs of pure ceria with literature values confirmed this presumed partial pressure [44].The calibration runs comprised long reduction steps of more than 4 h at different temperatures providing equilibrium conditions.CO 2 splitting was performed under a mixture of 6 vol.%CO 2 (4.8) in Ar (5.0).The flow rate was set to 85 ml•min −1 for all experiments.A mass spectrometer connected to the gas outlet valve of the TGA qualitatively confirmed the redox reaction, detecting O 2 during reduction and CO during splitting.
The mass loss during thermal reduction corresp were estimated based on the regulatio ISO .These values allow the calculation of the mole ratio of reduced/oxidized cerium atoms to the total amount of cerium atoms (redox extent: X red and X ox in at.% 4 3 Ce   and at.%

Thermo Balance
The reducibility and CO compositions with varying x in the range of 0 to 0.375 were assessed via successive cycling.The data of a typical TGA run is shown in Figure 2.After a pre-heating step to 1300˚C with subsequent cooling to 900˚C (not shown here), four similar cycles were performed consisting of a reduction step (heating to 1400˚C with 20 Kmin −1 ; isothermal for 20 minutes) and a splitting step (cooling to 900˚C with 50 Kmin −1 ; isothermal for 20 minutes under 6.3 vol.%CO 2 in Ar and 20 minutes under pure Ar).Three to five individual samples per composition with masses of approximately 30 mg were subjected to cycling runs.Each of these sample runs were corrected with five independent blank runs that were conducted periodically during the test campaign.Therewith, we achieved reasonable standard uncertainties of the calcu- lated redox ratios and specific yields.
The mass gain and loss alternately reaction along with the temperature variation as seen in Figure 2.During the heating process from 900˚C to 1400˚C, the reduction starts at about 1200˚C corresponding to a sharp mass loss.The reaction rate markedly increases with temperature.As the temperature plateau begins, the m  curve exhibits an inflection point representing a gradual reaction deceleration.This behavior was also observed from other groups and attributed to the rate-limiting transition between the surface reaction and the bulk reaction [45].At the end of the isothermal, the reduction is close to completion.Upon cooling to 900˚C, CO 2 was injected into the measuring cell causing a sharp mass increase.The oxidation reaction is significantly faster than the reduction and does not slow down before 0.1% m    is reached.After 20 minutes injecting CO 2 , the sample mass approximates its initial value equaling full reoxidation.The redox extent augments with increasing Zr content from X ≈ 7% observed for pure ceria to X ≈ 12 red/ox red/ox % r compositions in the range of 15% to 22.5% Zr content.These results are in good agreement with results found in the literature [45,46].Further increasing the Zr content diminishes the reducibility of the material to X red/ox ≈ 10%.Accordingly, the optimal Zr content concerning the redox extent is in the range of 0.15 ≤ x ≤ 0.225.Specific yields are necessary to assess the performance of the fo material towards a technical realization of the process because they convey information important for the reactor design.The specific yields of O 2 and CO averaged over the third and fourth cycles are depicted in Figure 4.

Splitting Performance Depending on the
Increasing the Zr-content inherently influences the specific yields in two ways.On the one hand, the molar weight decreases due to the lower molar weight of zirconium compared to cerium.Hence, higher Zr-contents enhance the specific yields compared to the redox extents.On the other han th Taking these opposed effects into account modifies the ranking based on the redox extents and determines the optimal Zr-content to be 15%.Ce 0.85 Zr 0.15 O 2 releases 0.155 ± 0.016 mmol O 2 per gram material during reduction and produces 0.305 ± 0.026 mmol CO per gram material, respectively.This represents an increase of approximately 50% with respect to pure ceria.
The results clearly indicate that the substitution of Ce 4+ by isovalent Zr 4+ enhances the reducibility, w ith earlier studies [40,[46][47][48].For some authors, the enhancement is attributed to modifications in the crystal structure.The reduction from Ce 4+ to Ce 3+ causes the lattice to expand, since Ce 4+ exhibits a smaller ionic radius resulting in a stress that suppresses further reduction [49].Introducing Zr 4+ into the fluorite structure of ceria compensates for the expansion of the crystal lattice, since Zr 4+ ions are smaller than Ce 4+ and Ce 3+ ions [46,50].ever, for higher Zr-contents the reduction energy remained approximately constant.
Recently, Kuhn et al. fitted a point defect model to TGA data indicating a decline in the reduction enthalpy with increasing Zr-content up to 20%, consistent with the findings in the present study [52].Kuhn et al. also suggested that the smaller Zr 4+ drives t n vacancies caused by the reduction of Ce 4+ to Ce 3+ .This is due to the fact, that Zr 4+ prefers a lower coordination with oxygen (e.g.Zr [7] in monoclinic ZrO 2 ) in contrast to Ce [8].

Impact on the Reaction Conditions
The reaction conditions required to reduce ceria-based materials are one of the major barriers to technical success of the pro and/or the low partial pressure of oxygen p(O significant decrease on the process efficiency.Ermanoski et al. [39] exemplary estimated the process efficiency depending on the reduction extent δ of the following reaction: Thereby, they introduced a routine that fit thermogravimetric data of pure ceria for a wide range of temperatures and oxygen partial pressures p(O 2 ) published by Panlener et al. [44]

Durability of Zirconia-Doped Ceria
In the past, the long-term stability of the redox material was a major barrier to technical success of solar-driven fuel-production processes.Especially, ferriteterials suffered from long-term cycling at temp close to the melting point.Presumably, ceria-based rials exhibit an improved long-term behavior, because they feature much higher melting points (>2000˚C).To evaluate the long-term stability, several 16-cycle experiments were performed consecutively with the same powder sample of the most efficient composition (Ce 0.85 Zr 0.15 O 2 ).The same temperature program was employed for all cycles: a reduction step (heating to 1400˚C with 20 Kmin −1 ; isothermal for 20 minutes) followed by a splitting step (cooling to 900˚C with 50 Kmin −1 ; isothermal for 20 minutes under 6.3 vol.%CO 2 in Ar followed by 20 minutes under pure Ar).The specific yields n m (O 2 ) and n m (CO) vs. the cycle number are depicted in Figure 6 as well as the corresponding ratio r of CO:O 2 release (reduction with its following oxidation).Due to the stoichiometry of the reaction, r should equal two (see Equation 1).
The yields of the sample slightly but continuously decrease with increasing cycle number.After 100 cycles, Hence, with increasing cycle number, the twenty minutes under CO 2 did not suffice to ensure complete reoxidation.
In turn, only a smaller amount of cerium atoms are reduced in the following cycle.
Figure 7 shows the ox m  data of three oxidation steps (beginning, middle and end of long-term cycling), as well as the impact of long-term cycling on the microstructure.As CO 2 is injected (minute 5), the oxidation immediately starts corresponding to a mass gain.In the be datio ginning of all three oxi n steps, the mass change exhibits an almost linear increase that smoothly segues into a logarithmical increase.The lower the cycle number, the longer and steeper is the linear regime.The logarithmical slope, however, is independent of the cycle number and is approximately constant yielding to decreasing oxidation extents with increasing cycle number.As mentioned before, the linear regime is associated with the surface reaction in contrast to the following bulk reaction [45].Hence, difference in the linear regimes should correspond to changes of the specific surface.
SEM imaging of the material before and after longterm cycling reveals that particle and grain sizes of the material significantly increase.The grains grow from sizes less than 1 μm after four cycles to sizes of 1 -3 μm after one hundred cycles and agglomerate to particles of more than 50 μm.This sintering takes place gradually during long-term cycling of the material.Accordingly, the specific surface continuously decreases causing deceleration of the oxidation.Since the reduction kinetics (not shown here) feature no changes with increasing cycle number, we conclude that in the case of powder ma- t denotes the time for the investigated process.Assuming t = 600 s, which is half of the time of the oxidation step, the diffusion length scale l d is more than 1.5 mm.Since only particle sizes of approximately 50 μm are observed, the oxygen diffusion is unlikely to be rate- 1400˚C, the sintering ceria-based materials cannot be entirely prevented.Hence, the observed dependency of the oxidation kinetics on the specific surface emphasizes the importance of adjusting the reduction and oxidation durations during long-term operation.Particularly, this is important when powder m netics was discussed by Le Gal et al. [41].They observed an enhanced reactivity by doping ceria-zirconia solid solutions with small amounts of trivalent lanthanides.The doping caused a less pronounced sintering activity.

Conclusions
The production of solar fuels by means of thermochemical redox cycles that split H 2 O and CO 2 and produce synthesis gas has been considered to enable a secure, clean and sustainable energy supply.The active ceriabased ma regarding the impa formance.In this st 1−x x 2 sized via the citrate nitrate auto combustion route at different Ce:Zr molar ratios (0 ≤ x < 0.4).In order to evaluate their applicability for solar fuel production, they were subjected to thermogravimetric experiments, simulating CO 2 -splitting cycles.The Zr-content featuring the highest specific yield was identified and analyzed in terms of reaction conditions and long-term stability.
The results indicate that a certain Zr-content (0.15 ≤ x ≤ 0.225) enhances the reducibility and therefore the splitting performance.Increasing the Zr-content to x = 0.15 improved the specific CO 2 -splitting performance by 50% compared to pure ceria.Further increasing the Zr-content to x = 0.38 diminished the specific yields to values of pure ceria.This finding agrees with theoretical studies tributing the improvements to lattice modification caused by the introduction of Zr 4+ .Compared to pure ceria, the most efficient composition Ce 0.85 Zr 0.15 O 2 enhances the required reaction conditions by a temperature of 60 K or one order of magnitude of the partial pressure of oxygen p(O 2 ).Long-term cycling of one hundred cycles was performed revealing declining oxidation kinetics.
The future interest is to understand how the microstructure influences the splitting as well as the reduction performance.Particularly, the splitting and reduction kinetics will be investigated in upcoming studies.Doping ceria-zirconia solutions with trivalent lanthanides may improve long-term stability, further enhancing the overall proces

Acknowledgements
Part of the work was co-funded by the Initiative and Networking Fund of the Helmholtz Association of German Research Centers.

Figure 1 .
Figure 1.General schematic of the two-step thermochemical cycle for CO 2 splitting.MO denotes a metal-based redox

2 3 .
Results and Discussions2 -splitting ability of Ce 1−x Zr x O

3 4 Ce
  , respectively) and the speand CO yiel fic yield: n m (O 2 ) and n m (CO) in mmol O 2 and mmol CO per gram oxidized material,

Figu alculated redox extents X red n Zirconia Content x re 3
Figu alculated redox extents X red n Zirconia Content x re 3 summarizes the c and X ox depending on the Zirconia content x, based o the obtained red m  and

Figure 4 .
Figure 4. Specific yields n m (O 2 ) and n m (CO) calculated from TGA runs of Ce 1−x Zr x O 2 with 0 ≤ x < 0.4 (average yield

2 , the specific re ceria are m 2 2 sessable and depicted in Figure 5 .
. Based on this routine yields n (O ) depending on T and p(O ) of pu as 0 mmol O At a reduction temperature of T = 1400˚C (green dashed line) and an oxygen partial pressure of p(O 2 ) = 5 × 10 −6 bar (vertical solid line), pure ceria releases approximately 0.1 mmol O 2 1 mat g  (blue circle).In our experiments, ceria evolved 0.101 ± 0the literature data.The most efficient composition Ce 0.85 Zr 0.15 O 2 , releases 0.156 ± 0.016 mmol O 2 per gram material (red square).Pure ceria does not release this amount until a temperature of 1460˚C is reached or p(O 2 ) is further decreased by e order of magnitude.In other words, Zr-doping saves 60 K or one order of magnitude of p(O 2 ).Hence, Zr-doping significantly enhances the process efficiency.

Figure 5 .-Figure 6 .
Figure 5.The specific oxygen yield n m (O 2 ) versus the partial pressure of oxygen p(O 2 ).The dashed lines were calculated based on a fitting routine published by Ermanoski et al. [39] who fitted data of Panlener et al. [44].The vertical solid line marks p(O 2 ) = 5 × 0 −6 bar, which was achieved in the p thermo balance.The blue circle re resents the result ob tained at 1400˚C by TGA for pure ceria; the red square for Ce 0.85 Zr 0.15 O 2 .-

Figure 7 .
Figure 7.Long-term test campaign of Ce 0.85 Zr 0.15 O 2 : (a) course of m ox during cycles; (c) SEM image after one hundred cycles.terials the microstructure rather influences the kin tics of the splitting step than the reduction kinetics. d limiting.When cycled at temperatures of of aterials are employed.One way to improve the ki terials have been proposed and investigated ct of Zr-doping on the splitting perudy, Ce Zr O oxides were syntheat s efficiency.