Improving the ceria-mediated water and carbon dioxide splitting through the addition of chromium

Improving the ceria-mediated water and carbon dioxide splitting through the addition of chromium

Applied Catalysis A: General 537 (2017) 40–49 Contents lists available at ScienceDirect Applied Catalysis A: General journal homepage: www.elsevier...

2MB Sizes 9 Downloads 33 Views

Applied Catalysis A: General 537 (2017) 40–49

Contents lists available at ScienceDirect

Applied Catalysis A: General journal homepage:

Feature Article

Improving the ceria-mediated water and carbon dioxide splitting through the addition of chromium Sotiria Mostrou a,b , Robert Büchel a,∗ , Sotiris E. Pratsinis a , Jeroen A. van Bokhoven b,c,∗∗ a b c

Particle Technology Laboratory, Department of Mechanical and Process Engineering, ETH Zurich, Sonneggstrasse 3, 8092 Zurich, Switzerland Institute for Chemical and Bioengineering, Department of Chemistry and Applied Biosciences, ETH Zurich, Vladimir-Prelog-Weg 1, 8093 Zurich, Switzerland Laboratory for Catalysis and Sustainable Chemistry, Paul Scherrer Institute, WLGA 135, 5232 Villigen, Switzerland

a r t i c l e

i n f o

Article history: Received 27 September 2016 Received in revised form 24 February 2017 Accepted 1 March 2017 Available online 1 March 2017 Keywords: Solar thermochemical splitting Chromium-doped CeO2 Solar fuels Polymerized complex method Flame spray pyrolysis

a b s t r a c t The solar thermochemical water and carbon dioxide splitting, mediated by ceria, has a great potential to produce “green” syngas. Chromium was added to ceria to improve the syngas production. Three preparation methods were applied, resulting in different morphologies allowing to investigate the role of chromium. The samples were characterized by X-ray diffraction, Raman and X-ray spectroscopy, and electron microscopy. Materials made by polymerized-complex-method and dry-impregnation consisted of two crystal phases: ceria and chromia. In contrast, materials made by flame-spray pyrolysis exhibited a homogeneous Cr-doped ceria phase, and chromia was found only at a chromium-content higher than 25 mol%. The chromium-doped ceria released additional oxygen during the formation of CeCrO3 perovskite, which did not enhance hydrogen or carbon monoxide production. All chromia-containing samples exhibited improved oxygen exchange capacity, possibly due to a redox cycle of chromia itself, and significantly improved the activity of water and carbon dioxide splitting. Hydrogen production increased from 3.2 to 6.7 mL/g and the time to reach redox equilibrium was shorten from 41 to 3 min. The best hydrogen and carbon dioxide production rates were up to 20 and 500 times higher than pure ceria, respectively. The presence of chromium is therefore crucial as a catalyst, promoter, and oxygen storage enhancer. This work emphasises the importance of a catalysed re-oxidation reaction and demonstrates that a metal oxide, becoming active in situ, can catalyse water and carbon dioxide splitting. © 2017 Elsevier B.V. All rights reserved.

1. Introduction Liquid fuels, from fossil-free sources, are very important in the attempt to meet the increasing energy demands, while restraining emissions of anthropogenic carbon dioxide [1]. Solar thermochemical water splitting (WS) and carbon dioxide splitting (CDS) utilize solar radiation [2] to produce hydrogen and carbon monoxide from carbon dioxide and water [3,4]. The produced hydrogen can be used as fuel without further treatment, or together with carbon monoxide (syngas) to yield liquid solar fuels and other chemicals via Fischer-Tropsch synthesis [5]. The WS and CDS reactions are realized by means of a reduction-oxidation (redox) cycle of materials with a high oxygen exchange capacity, such as ceria. The latter was used for WS and CDS, separately [6,7] or simulta-

neously [8]. The WS and CDS mechanisms are similar, in that they include a redox cycle as shown in Eqs. (1)–(3). The reduction step (Eq. (1)) is an exothermic reduction of ceria at low oxygen partial pressure and high temperature; between 750–1500 ◦ C, the nonstoichiometry ı is in the region of 0.001–0.3 [9]. Above 2000 ◦ C, ceria reduces stoichiometric to Ce2 O3 [6]. After the reduction step, ceria is cooled and reacts with water and carbon dioxide (Eqs. (2) and (3)) to produce hydrogen and carbon monoxide, respectively. The re-oxidation potential depends on the non-stoichiometry, and it is thermodynamically favourable below 927 ◦ C [9]. Oxygen evolution step (reduction): Tred ı CeO2 (s) → CeO2-ı (s)+ O2 (g) 2 lowpO



Hydrogen production step (re-oxidation): ∗ Corresponding author. ∗∗ Corresponding author at: Institute for Chemical and Bioengineering, Department of Chemistry and Applied Biosciences, ETH Zurich, Vladimir-Prelog-Weg 1, 8093, Zurich, Switzerland. E-mail addresses: [email protected] (R. Büchel), [email protected] (J.A. van Bokhoven). 0926-860X/© 2017 Elsevier B.V. All rights reserved.


OX CeO2-ı (s) + (ı + ı∗ )H2 O(g) → CeO2-ı∗ (s)+(ı + ı∗ )H2 (g)


Carbon monoxide production step (re-oxidation): T

OX CeO2-ı (s) + (ı + ı∗ )CO2 (g) → CeO2-ı∗ (s)+(ı + ı∗ )CO(g)


S. Mostrou et al. / Applied Catalysis A: General 537 (2017) 40–49

The solar-to-fuel energy efficiency of this cycle is defined as the heating value of the product over the operation energy (solar radiative energy input and the energy to operate the system) [10]. Based on thermodynamic calculations, ceria cycle has the potential to reach up to 30% efficiency, depending on the process design [10,11]. Thus far, low amounts of hydrogen or carbon monoxide were produced experimentally, which has led to a fuel-to-solar energy efficiency of only 4%, mainly due to a partial re-oxidation to CeO2-␦* (Eqs. (2) and (3), for ı* > 0) [12]. Under certain reaction conditions, an optimization of one or both steps can lead to greater efficiency. One approach to improve the fuel-to-solar energy efficiency of WS and CDS is to increase the number of vacancies formed during the reduction step (increase ␦). The addition of dopants in the ceria lattice can decrease the energy of vacancy formation from 3.00 eV (pure CeO2 ) to 1.06 eV (Y-doped CeO2 ) [13]. Dopants, such as Ti4+ , Sn4+ , La3+ , Y3+ , Sm3+ [14], Hf4+ [14,15], Zr4+ [14,16,17], and Cr3+ [18], were tested in WS and CDS. Jiang et al. [14] showed that the addition of trivalent cations has a negative effect on the oxygen release, which dropped from 5.7 mL/g (CeO2 ) to 3.8 mL/g (Ce0.85 Sm0.15 O2 ), because the stable intrinsic oxygen vacancies reduce the amount of oxygen available for exchange. Cr-doped CeO2 released oxygen at 465 ◦ C, 1000 ◦ C lower than any other cerium-based material [14,18]; however, only a low amount of hydrogen was produced. Since chromium is the only dopant with such behaviour, we explored herein its potential at 1500 ◦ C, to produce hydrogen and carbon monoxide. The fuel-to-solar energy efficiency increases also if the products are formed faster at a constant energy input. Often, extended times are required (>45 min) to reach equilibrium [15]. Formation of hydrogen poisoned the surface of Sm-doped CeO2 , making CDS kinetically advantageous compared to WS [19]. The opposite was found by Chueh et al., who reported that CDS has a higher activation energy than WS, hence, the rate of CO production was lower than the rate of hydrogen production [8]. The WS rate was significantly increases in the presence of noble metals, for example, the reaction rate over Pt/CeO2 was up to 1000 times higher than over pure CeO2 [20]. Metal oxides, such as NiO and CuO, which reduce to lower valence species, also increased the hydrogen production rate from 0.26·10−8 (CeO2 ) to 175·10−8 (NiO) and 10.7·10−8 (CuO) mol/(s·g); the additives on the surface of ceria have a catalytic effect on the WS [20]. Metal oxides, such as chromia, on the ceria surface also promoted the oxygen storage capacity of ceria [21]. Hence, chromia shows potential for improving both the oxygen storage capacity of ceria and the hydrogen production rate and to the best of our knowledge has never been tested carbon dioxide splitting. This study focuses on the effect of chromium addition on the activity of ceria for WS and CDS. Chromium is present in the tested materials in two forms: as a dopant introduced into the ceria lattice, or as a separate chromia phase, supported on ceria. Three methods of preparation were employed to obtain materials with phases of varying homogeneity: polymerized complex method (PCM), dry impregnation (DI), and flame spray pyrolysis (FSP). The materials were tested for WS and CDS. The addition of a chromia phase increased both hydrogen and carbon monoxide production, indicating the importance of a catalysed re-oxidation step.


solution was added to a 30 mL ethylene glycol (Sigma Aldrich, >99.5%)/24 g citric acid (Acros Organics, 99.35%) solution, and the resulting mixture was heated at 180 ◦ C until all the water had evaporated. Thereafter, the residue was heat-treated in air at 350 ◦ C for 6 h and 800 ◦ C for 4 h, to remove organic residues. Dry impregnation (DI) were synthesized by dissolving chromium (III) nitrate nonahydrate (Sigma-Aldrich, 99%) in water and added to pure ceria powder, prepared by PCM. The volume of the solution was calculated so that it dry-wets the support. The mixture was dried at 100 ◦ C, under vacuum, for 12 h and calcined at 500 ◦ C in air for 5 h. The FSP materials were prepared as described by Mädler et al. [23]. Cerium (III) 2-ethylhexanoate (Strem, 49% in 2ethylehexanoic acid) and chromium (III) acetylacetonate (Fluka, 97%) were mixed and dissolved in a 2-ethylhexanoic acid (Aldrich, 99%)/toluene (Sigma-Aldrich, 99.8%) mixture (1/1 by volume) to give a 0.3 M metal concentration. The precursor solutions were sprayed at 5 mL/min and dispersed by 5 L/min oxygen (Pangas, 99%). The produced particles were collected on a glass fibre filter (Whatman GF/D, diameter 25.7 cm). The nomenclature of the materials indicates the method of preparation (P = PCM, I = DI, and F = FSP) and the Cr content (mol%). For instance, the pure ceria prepared by PCM, is referred to as P-Ce, while FSP ceria with 10 mol% chromium is referred to as F-CeCr10. The nominal composition of the materials can be expressed as Ce1-x Cr x O2(1-x)+1.5x , where x is the Cr stoichiometry so that 0 ≤ x ≤ 1. 2.2. Characterization of the materials

2. Experimental

The specific surface area (SBET ) of the as-prepared materials was measured by nitrogen physisorption at 77 K in a Micromeritics TriStar unit and determined by the BET method. Before measurement, the samples were degassed at 150 ◦ C in flowing nitrogen. X-ray diffraction (XRD) patterns were acquired by a Bruker AXS B8 Advance diffractometer, operating with Cu(K␣ ) radiation. The crystal size (dXRD ) was determined by the XRD patterns, based on the fundamental parameters approach incorporated in TOPAS software. Raman spectra were measured by a Renishaw InVia Raman microscope with 514 nm at 6 mW power and an exposure time of 10 (PCM- and DI-made) and 30 s (FSP-made). The X-ray absorption near edge spectroscopy (XANES) spectra were recorded at the PHOENIX beamline from the Swiss Light Source in Villigen, Switzerland. The samples were pelletized and mounted in a vacuum chamber (10−6 torr) under 45◦ relative to the incident beam. The XANES spectra were measured at room temperature in fluorescence mode, using a four element Si solid state detector (Vortex) with 160 eV energy resolution. The X-ray beam size was 0.5 × 0.5 mm and the detector dead-time was kept under 20% to exclude artifacts due to high count rates originating from the Ce L edge. The X-ray fluorescence (XRF) spectra were measured at 6.5 keV for 30 s on each pellet. The XRF spectra have been quantitatively fitted using PyMca [24], and the XANES spectra have been normalized to a post-edge value of 1 with Athena [25]. Scanning transmission electron microscopy (STEM) with elemental mapping using energy-dispersive X-ray spectroscopy (EDXS) was conducted with a FEI Talos microscope at 200 kV. An FEI ESEM XL30 was used to obtain the scanning electron microscopy (SEM) images.

2.1. Preparation of the materials

2.3. Catalytic testing

The PCM-made materials were made according to an adapted method of Yashima et al. [22]. Cerium (III) acetate hydrate (Alfa Aesar, 99.9%) and chromium (III) nitrate nonahydrate (SigmaAldrich, 99%) were mixed and dissolved in water (100 mL); the

Pellets (48 ± 2 mg and 5 mm in diameter) were made by pressing 50 ± 0.5 mg of the material, at 130 bar (PCM- and DI-made) and 13 bar (FSP-made). The pellets were tested for WS or CDS in three redox cycles. An alumina reactor (99.7 Alsint, Haldenwanger


S. Mostrou et al. / Applied Catalysis A: General 537 (2017) 40–49

Ceramics, 9.5 × 12.7 × 400 mm) was heated in an electric infrared furnace (Ulvak VHT E44) as described elsewhere [26]. The reduction step at 1500 ◦ C lasted 10 min; the total flow was 100 mL/min Ar (Air Liquide, 99.999%). The oxidation step at 800 ◦ C lasted 10 min; the total flow was 100 mL/min Ar + 8.4 mL/min H2 O vapour (at 120 ◦ C) for WS, and 100 mL/min Ar + 8.4 mL/min CO2 (Air Liquide, 99.999%) for CDS. The redox cycles were performed under ambient pressure; the heating and cooling rates were 50 and 100 ◦ C/min, respectively. The product stream passed through a water trap immersed in an ice bath to condense water and was analysed online with an OmniStar mass spectrometer (MS) from Pfeiffer Vacuum. After each test, the MS signals of oxygen and hydrogen were calibrated with two standards (250.0 ppm O2 in Ar and 0.50% H2 in Ar (Air Liquide)) to quantify the oxygen and hydrogen produced during testing. The produced carbon monoxide was quantified by a Bruker Vector 22 FT-IR Spectrometer, recording from 4000 to 400 cm−1 , connected in parallel to the MS. Each material was tested up to 5 cycles, without undergoing any prior treatment. For assessment of the activity of each material, only the hydrogen and carbon monoxide produced after the third cycle was considered. The rate of oxygen, hydrogen, and carbon monoxide evolution are determined as the total amount of gas evolved during one cycle, normalized by the sample mass and the time of evolution.

3. Results and discussion 3.1. Structural properties Fig. 1a shows the specific surface area (SBET ) of the as-prepared materials as a function of chromium content. The wet-made materials (P-CeCr% and I-CeCr%) had a ten times lower surface area (<15 m2 /g) than the flame-made ones (F-CeCr%), which ranged from 116 to 147 m2 /g. FSP-made cerium-based nanopowders exhibit between 100–180 m2 /g [23,27]. The addition of chromium resulted a lower surface area, independent of preparation method. Fig. 1b presents the crystal size determined by XRD peaks, of the materials prior (solid lines) and after WS (dash lines). The as-prepared FSP-made materials had a much smaller (10 nm) crystal size than the PCM- and DI-made materials (40 nm), explaining the difference in their surface area. After testing, the crystal size of ceria had increased by a factor of 2 (wet-made) and 7 (flame-made). The FSP-made materials grew from 10 to 50 nm in the absence of chromium, while its presence promoted sintering up to 75 nm. All the wet-made materials grew from 40 to 80 nm, independent of chromium content. From the observed crystal size, chromium does not improve the sintering resistance. However, SEM images (Fig. 1c) show that chromium affects the growth of the average particle size during reaction. In the presence of chromium, sintered particles are between 1.5 and 2.1 ␮m, in contrast to 53.8 ␮m particles observed in its absence. Fig. 2a depicts the XRD pattern of pure and 15 mol% Crcontaining ceria materials. The XRD patterns of all the tested materials are shown in Fig. S1. All patterns have the characteristic fluorite structure of ceria (PDF 75-0076). The wet-made materials contain a chromia phase (PDF 76-0147), as seen in the magnification of the 2 = 33–40◦ region. PCM and impregnation favoured the formation of a distinct chromia phase, about 12 wt.% according to peak intensity, more than of any other chromium species. Similarly, Moriceau et al. demonstrated that impregnating ceria with high content of chromium will lead to Cr3+ [28]. Chromia was not detected in the FSP-made material. Fig. 2b depicts the XRD pattern of the materials after three redox cycles. Ceria remains the main phase. All Cr-containing samples have the characteristic peaks of the perovskite-like CeCrO3 struc-

Fig. 1. a) Specific Surface Area (SBET ) of the as-prepared materials and b) crystal size (dXRD ) of the as-prepared (solid lines) and spent (dash lines) PCM- (䊏), DI- (), and FSP-made (䊉) materials. c) SEM images of the spent P-Ce, P-CeCr15, I-CeCr15, and F-CeCr15.

S. Mostrou et al. / Applied Catalysis A: General 537 (2017) 40–49

Fig. 2. XRD patterns of a) as-prepared and b) spent materials. XRD patterns of P-CeCr15, I-CeCr15, and F-CeCr15 are magnified from 33 to 40◦ 2. The peaks correspond to CeO2 (, PDF 75-0076), Cr2 O3 (, PDF 76-0147), CeCrO3 (, PDF 75-0289) and Al2 O3 (PDF 50–1496).

ture (PDF 075-0289), quantified as 7 wt.% (F-CeCr15) and 10 wt.% (P-CeCr15 and I-CeCr15) based on pick intensity. Formation of CeCrO3 was observed also in spent Cr-doped ceria after WS [29]. During the formation of CeCrO3 , additional oxygen was released (Eq. (4)), e.g. up to 4.6 mL/g for a 10 wt% CeCrO3 /CeO2 . Typically, CeCrO3 re-oxidizes at 500 ◦ C in air to form CeO2 and Cr2 O3 [30]. The high content of CeCrO3 in the spent materials indicates that re-oxidation was not possible in the presence of water or carbon dioxide even at 800 ◦ C. Therefore, CeCrO3 was not unsuitable for WS or CDS. The remaining chromium species, less than 5 wt.%, whether in the form of Cr(III) or lower valence, were not observed by XRD. An alumina phase (Al2 O3 ) was present in the pure ceria samples, which originated from contamination with the reactor and sample holder material; such contaminations are unavoidable at 1500 ◦ C. The addition of chromium seams to suppress the contamination. 1500o C

2CeO2-ı (s) + Cr2 O3 (s) → 2CeCrO3 (s)+1/2O2 (g) low pO



The state of the chromium on the surface of the as-prepared materials was investigated with Raman spectroscopy (Fig. 3a). The main peak of ceria at 460 cm−1 is visible in all spectra, with a distinctive shift to the right in the FSP-made materials, which could


Fig. 3. Raman spectra of a) as-prepared and b) spent materials. Commercial Cr2 O3 is presented as reference. The peak at 553 cm−1 corresponds to the Cr2 O3 phase and the peak at 842 cm−1 to chromate ion inside the CeO2 lattice.

imply incorporation of chromium in the ceria lattice. The peak at 842 cm−1 , observed only in the FSP-made powders and identified as lattice CrO4 2− , concludes that ceria was successfully doped by chromium [18]. The peak at 552 cm−1 in the P-CeCr15, I-CeCr15, and F-CeCr25 spectra was assigned to chromia, based on the Raman spectrum of pure chromia (commercial powder from Strem Chem.). The interpretation of the Raman spectra and the XRD patterns confirm that PCM and DI produce a mixture of chromium and cerium oxides and only FSP lead successfully to the addition of chromium into ceria lattice. Herein, a thermodynamic calculation could identify the optimal operation conditions for maximizing the non-stoichiometry (␦) [9]. PCM, which is a reliable method to synthesize doped ceria [22], did not work with chromium as the dopant. Chromia was probably formed during the calcination at 800 ◦ C: calcined Cr-doped ceria (F-CeCr15) exhibited the characteristic chromia peak in the Raman spectra, replacing the lattice CrO4 2− (Fig. S2). Calcination probably caused decomposition of CrO4 2− and leaching of chromium to the surface where chromia was formed.


S. Mostrou et al. / Applied Catalysis A: General 537 (2017) 40–49

observed for LaCrO3 perovskite, where the chromium retain a 3+ oxidation state [34]. Due to the similarities of elements of the same group, the spectra were attributed to the perovskite CeCrO3 , which was identified as the main chromium species after reaction by XRD (Fig. 2b). Large difference of photoelectron scattering by chromium and cerium atoms lead to significant difference in the XANES of Cr2 O3 and CeCrO3 . Fig. 4b exhibits the chromium K␣ spectra, deconvoluted by the XRF spectra of the samples at 6.5 keV X-ray energy for 30 s exposure (Fig. S5). The peak location denotes the electron transition and is characteristic of the elements. The intensity of the peaks is proportional to the number of atoms in the samples. A decrease of chromium atoms was observed after the reaction. The flamemade sample was more unstable, exhibiting a ca. 47% decrease in chromium atoms, whereas the impregnated only 13%. STEM combined with EDXS was used to visualize the chromium and cerium in the as-prepared materials (Fig. 5a) and in the spent I-CeCr15 (Fig. 5b). An inhomogeneous distribution of chromium (green) and cerium (red) domains was observed in the P-CeCr15 and I-CeCr15. Due to the partial overlap of the chromium signal and the second cerium line (Fig. S3), Cr seems to be present all over the map but is only present at the bright green sites. The size of the chromium-rich domains varies from 200 to 500 nm (P-CeCr15) and from 5 to 100 nm (I-CeCr15). The F-CeCr15 and F-CeCr25 powders exhibited a homogeneous distribution of both elements, with small (5–10 nm) chromium areas present only in the F-CeCr25. These findings agree with the conclusions from the XRD and Raman measurements (Figs. 2a, 3a): PCM-, DI- made, and the F-CeCr25 contain a distinct chromia phase on the surface of ceria. Fig. 5b illustrates the spent I-CeCr15, where the chromium domains are present and have increased in size; the maps of the rest spent materials were similar (Fig. S4). This chromium is attributed to the CeCrO3 , which is the dominant chromium phase on the used material.

3.2. O2 evolution Fig. 4. X-ray spectroscopy measurements of the as-prepared (solid lines) and spend (dashed lines) F-CeCr15 and I-CeCr15. Cr2 O3 sample is presented as reference. a) XANES spectra near the chromium K-edge. The characteristic features a1 and a2 correspond to the Cr6+ pre-edge and to Cr3+ in chromium perovskite, respectively. The XRF spectra (Fig. S5) are deconvoluted to obtain the chromium K␣ spectra of the samples (b).

Fig. 3b depicts the Raman spectra of spent materials. The peaks of chromia and CrO4 2− are absent, and only ceria is observed. Exception was the spent F-CeCr15 and F-CeCr25, which exhibited many peaks. The peaks between 200 and 438 cm−1 are similar to those of the perovskite structure LCrO3 (L = lanthanides) [31] and may correspond to CeCrO3 , also observed in the bulk by XRD (Fig. 2b). Hence, in the FSP-made materials, CeCrO3 is present in the bulk and on the surface. In the PCM- and DI-made materials CeCrO3 was only observed in the bulk. X-ray spectroscopic techniques were conducted on the samples I-CeCr15 (green) and F-CeCr15 (red) before and after the reaction, which provide information of the chromium state [32]. Fig. 4a presents the chromium K-edge XANES. The as-prepared I-CeCr15 spectra associates to the spectra from the reference chromia sample (magenta). The F-CeCr15 is the only sample exhibiting a pre-edge, at ca. 5.99 keV (feature a1). This pre-edge feature was encountered in chromate species [33], such as the CrO4 2− observed by Raman spectroscopy (Fig. 3a). The spectra of the spent samples (dashed lines) are identical for both I-CeCr15 and F-CeCr15. The edge position suggests a Cr3+ , but the features do not correspond to the chromate. Similar spectra and edge peak position (feature a2) was

Fig. 6a displays the MS signal of oxygen, during the first heating step (25–1500 ◦ C), over pure (dashed lines) and 15 mol% chromium-containing ceria materials (solid lines) and Fig. 6b the corresponding amount of total oxygen released. All tested samples reached maximum oxygen evolution at ca. 1500 ◦ C (maximum tested temperature). The FSP-made samples formed 2 times less oxygen than the PCM-made ones (Fig. 6b). The chromium containing materials released more oxygen than pure ceria, independent of the preparation method. Chromia-ceria mixed oxides (P-CeCr15 and I-CeCr15) produced an oxygen excess of ca. 5.2 mL/g: 4.6 mL was due to the CeCrO3 formation and the remaining could be attributed to the reduction of the remaining chromia on the surface. An increase of oxygen storage capacity of ceria is highly possible and was evaluated by the amount of hydrogen produced (see Section 3.3) [21]. During temperature-programed reduction, bulk chromia reduced at ca. 750 ◦ C [35] and supported chromia, on titania or alumina, already at 300 ◦ C [36]. For ceria, the release of stored oxygen was observed above 800 ◦ C [37]. Hence, chromia reduction in the studied system is expected before the reduction of ceria, at 1500 ◦ C. In that case, Cr3+ reduces to Cr2+ (Eq. (5)) at high temperature and low oxygen partial pressure, forming CrO, which decomposes above 550 ◦ C to metallic Cr and Cr2 O3 (Eq. (6)) [38]. The expected composition of the material, when re-oxidation starts, is metallic chromium supported on mixed CeO2 /CeCrO3 . The actual composition during the cycling can only be observed by in situ measurements [26]. The trace amount of surface chromium and its rapid passivation, formation of a thin chromia layer on the surface [38], might explain

S. Mostrou et al. / Applied Catalysis A: General 537 (2017) 40–49


Fig. 5. Distribution of Ce (red) and Cr (green) elements of a) the as-prepared P-CeCr15, I-CeCe15, F-CeCr15, and F-CeCr25, and b) the spent I-CeCr15. In each row the first image is of both Ce and Cr, the second of only the Ce species and the third of only the Cr species. The images were obtained by STEM coupled with EDXS mapping. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

the absence of its peaks in both the XRD patterns and the Raman spectra of the spent materials (Figs. 2b, 3b). 1500o C

Cr2 O3 (s) → 2CrO(s)+1/2O2 (g) low pO

≥500o C

3CrO(s) → Cr(s)+Cr2 O3 (s) vacuum




Flame-made Cr-doped ceria (F-CeCr15) exhibited a second oxygen evolution peak at 423 ◦ C. Singh et al. also observed this oxygen release [18,29]: lattice oxygen formation started at 423 ◦ C, from the reduction of both cerium and chromium species, leading to a

mixture of a perovskite-like CeCrO3 and fluorite-like CeO2-␦ . Since CeCrO3 species were formed over all the chromium containing material (doped and undoped), one can expect that undoped materials have also two oxygen release peaks, which is not the case. Hence, the presence of CrO4 2− in the ceria lattice (Fig. 3a) induced the CeCrO3 formation at low temperature (423 ◦ C). The excess oxygen was ca. 2.9 mL/g, the oxygen required to form the 7 wt.% CeCrO3 observed in the XRD (Fig. 2b); additional ceria vacancies were not induced by chromium dopant.


S. Mostrou et al. / Applied Catalysis A: General 537 (2017) 40–49

Fig. 7. Time-resolved H2 production during the oxidation step (tox = 10 min) of PCM(top) and FSP-made (bottom) materials with different amounts of Cr.

Fig. 8. Rates of H2 (solid lines) and O2 (dashed lines) production of PCM- (䊏), DI(), and FSP-made (䊉) as a function of chromium content. The rates correspond to the 3rd cycle.

Fig. 6. a) Oxygen formation during heating of pure CeO2 and 15 mol% Cr/CeO2 . Maximum O2 production is observed at 423 and 1450 ◦ C. b) Total O2 evolution during heating up from 25 to 1500 ◦ C, as a function of Cr content. The volume of O2 release was normalized by the mass of CeO2 present in each Cr/CeO2 material.

3.3. H2 and CO production Hydrogen is produced during the re-oxidation of CeO2-␦ (Eq. (2)). Fig. 7 shows the MS signal of hydrogen, during the re-oxidation step (3rd cycle) of PCM- and FSP-made materials. In the absence of surface chromia (P-Ce, F-Ce, and F-CeCr10 and 15), the signal maximum was 1.3·10−10 mA, which decreased very slowly and did not return to the baseline (8·10−11 mA) after 10 min. This indicates that CeO2-ı was not fully re-oxidized and had the potential to produce more hydrogen. Fig. S6 shows that over pure ceria (F-Ce) equilibrium was not reached, even after 30 min on water stream. Similarly, equilibrium was not achieved in a stream of carbon dioxide during 45 min [15]. The slow WS kinetics over pure ceria limits hydrogen production in technical applications. In contrast, samples containing a distinctive chromia phase (P-CeCr10, 15, 30, and F-CeCr25) exhibited an accelerated re-oxidation step, with a very high max-

imum of 4.6·10−9 mA, which decreased sharply and returned to the baseline in less than 5 min. After that time, no more hydrogen was produced, since the oxygen vacancies formed during reduction step were consumed. These signals show clearly that the surface chromia kinetically affects hydrogen production in WS. The addition of chromium can increase the oxygen exchange capacity of ceria, especially at temperatures above 450 ◦ C [21]. That would lead to an increase of the re-oxidation potential and it can be evaluated by the maximum amount of hydrogen produced until equilibrium is reached (equilibrium hydrogen). The equilibrium hydrogen of the tested samples is presented in Table 1; the equilibrium hydrogen of pure ceria samples was approximated using a linear extrapolation. In the presence of chromium, the equilibrium hydrogen was 6.2 (P-CeCr15) and 3.9 mL (I-CeCr15), while only 2.3 mL over P-Ce. The samples made by FSP also improved the vacancy formation of pure ceria, with F-Ce expected to produce 3.7 mL of hydrogen until equilibrium, 60% more than P-Ce. The addition of chromium also reduced the time to reach equilibrium, from 41 min (P-Ce) down to 2 min (I-CeCr15). Fig. 8 presents the rate of hydrogen (solid lines) and oxygen (dashed-lines) production (3rd cycle), normalized by the sample mass and time; the production of the repeated cycles was sim-

S. Mostrou et al. / Applied Catalysis A: General 537 (2017) 40–49


Table 1 Produced and equilibrium H2 produced during the 3rd cycle. Equilibrium H2 refers to the maximum H2 produced until the sample reaches redox equilibrium. Materials

Produced H2 (mL/gs. )

Time to reach equilibrium (min)

Equilibrium H2 (mL/gs. )

P-Ce P-CeCr15 I-CeCr15 F-Ce F-CeCr15

1.0 6.2 3.9 1.8 2.7

41a 5 2 34* 22*

2.3* 6.2 3.9 3.7* 4.9*


Approximation from linear extrapolation.

Table 2 Rates of O2 , H2 , and CO production of the 3rd cycle. Materials

O2 rate (mL/(gs. min))

H2 rate (mL/(gs. min))

CO rate (mL/(gs. min))

P-Ce I-CeCr15 F-Ce F-CeCr15

0.05 0.16 0.07 0.14

0.09 1.94 0.17 0.27

0.005 2.50 0.21 0.66

ilar (Fig. S8). The time of oxygen production was set to 10 min (reduction step), constant over all materials. The time of hydrogen production differed over different samples and was defined as the duration where hydrogen was produced: i.e., 10 min for P-Ce, 2 min for I-CeCr15 (Fig. 7). Herein, the importance of a catalysed re-oxidation reaction is clear: the rate of hydrogen production in the presence of chromia increased from 0.09 mL/(g min) (P-Ce) to 1.94 mL/(g min) (I-CeCr15), exhibiting an improvement of 21 times. This effect was also observed over F-CeCr25, the only FSP-made material containing chromia, which produced 1.65 mL/(g min) of hydrogen, and over all the Cr-containing PCM-made samples, with 1.19–1.41 mL/(g min) hydrogen production. I-CeCr15 was the samples with the best hydrogen production rate (Table 2), despite releasing in total less H2 than P-CeCr15 (Table 1). Thus, a fast reoxidation step was more important than a high oxygen exchange capacity. The improvement seems to depend less on the chromia loading and more on the size of the chromium domains (Fig. 4). The most superior materials were those with smaller chromium domains, I-CeCr15 and F-CeCr25 (Fig. 5). Chromia on the surface behaves as a catalyst, reducing the required reaction time, while improving also the oxygen vacancy formation. Other effects are less likely: CeCrO3 was formed over both active and inactive samples (Fig. 2b), and an independent redox cycle of the chromia surface should not affect the time that ceria’s re-oxidation reaches equilibrium. Nonetheless, a synergetic redox cycle of chromia, which assists to oxidize ceria faster, is possible. No correlation between the kinetic effect and the sintering resistance was established (Fig. S7): both active and inactive samples retain a comparable particle size after reaction. Density function theory (DFT) calculations over reduced ceria showed that in WS, the most energy-demanding step is the evolution of hydrogen from the surface, with an energy barrier of 3 eV [39]. Hence, hydrogen is more likely to desorb from low coordinated defect sites, or from spill-over to noble metal pits [40]. FSP-made ceria contains more low coordinated sites than the wet-made ceria, due to the smaller crystal size, 40 and 10 nm, respectively. That explains the better hydrogen production rates compared to P-Ce. The spill-over effect was observed for additives such as the noble metals Pd and Pt and the metal oxides NiO and CuO, which were reduced during the reaction [20]. Herein, a spill-over effect may explain the significant improvement observed for the samples with chromia on the surface: Cr3+ reduces to Cr0 , followed by hydrogen atoms migrating from ceria to Cr0 sites, from where it desorbs as diatomic hydrogen. Although the hydrogen spill-over mechanism in a Pt/CeO2 system is well studied by DFT calculations [41], the way, in which metal oxides influence the WS, is unidentified.

Fig. 9. Time-resolved CO production during the oxidation step (tox = 10 min) of DI(top) and FSP-made (bottom) materials with 0 and 15 mol% Cr.

Pure and 15 mol% Cr-containing ceria were tested also for CDS. Fig. 9 presents the time resolved carbon monoxide release (FTIR calibrated signals) during one re-oxidation step (3rd cycle). The rate of carbon monoxide production exhibited similar patterns as that of hydrogen (Fig. 7). Pure and Cr-doped ceria reached an insignificant maximum production (<1·10−3 vol%), which decreased slowly but had still not reached the baseline even after 10 min in carbon dioxide stream. However, over I-CeCr15, the peak carbon monoxide production was up to 8.4·10−3 vol% and equilibrium was reached within 3.5 min. The kinetic effect of chromia in the CDS translates to a 3 orders of magnitude (2.50 mL/(g min)) improvement compared to pure ceria (0.005 mL/(g min)) (Table 2) and was twice as high as the best reticulated porous ceramic ceria [42]. The uneven improvement of hydrogen and carbon monoxide production (Table 2) designate the different mechanism of WS and CDS. Over pure ceria carbon dioxide dissociation has a higher activation energy (0.8 eV) than that of water (0.5 eV) [8], but the presence of a dopant switched the kinetics, making CDS faster than WS [19]. Hence, the chromia on the surface may aid in the dissociation of carbon dioxide, after which the oxygen atom spill-over to fill the nearest oxygen vacancy in ceria.


S. Mostrou et al. / Applied Catalysis A: General 537 (2017) 40–49

For the above reasons, the most plausible explanation is that supported chromia has a catalytic effect that influences significantly both the oxygen vacancy formation and the kinetics of WS and CDS, leading to fast equilibrium and increased the production rate of both hydrogen and carbon monoxide by 20 and 500 times, respectively. Until now, mostly noble metals were considered as catalysts towards WS [20], as they are known to dissociate hydrogen. However, they are expensive for commercialized solar thermochemical water and carbon dioxide splitting processes. Metal oxides, although cheaper and easily accessible, have not been considered as catalysts, since the O H bond forming on their surface prohibits H2 desorption, i.e., slow reaction kinetics over pure ceria. However, CuO, NiO, and Cr2 O3 are highly active for WS, because they probably reduce to their metallic form during the reduction step [20]. To the best of our knowledge, this is the first time that an oxide performs as catalyst in CDS. Although CDS mechanism is not as well exploited as WS, the presence of chromia has an enormously positive effect on the re-oxidation rate. These finding introduces a new approach towards the improvement of WS and CDS that is enhanced by the in situ formation of the catalyst. At this point, operando and in-situ techniques would help identify the electronic state of chromium during cycling and define how actually chromium species assist ceria in reaching equilibrium much faster. 4. Conclusions Chromium containing ceria materials were prepared by three different synthesis methods and tested for solar thermochemical water and carbon dioxide splitting. Flame spray pyrolysis synthesis produced Cr-doped ceria materials with high surface area, up to 100 m2 /g. Polymerized complex method and dry impregnation lead to chromia supported on ceria samples, with chromia cluster size between 5 and 500 nm. During the reaction, the presence of chromium (both as dopant and chromium oxide) induced the formation of CeCrO3 perovskite, which increased the oxygen released but appeared to remain stable during cycling and not to participate in the redox cycle. Part of the chromia on the surface is expected to reduce into lower valences, even to metallic chromium, as in a flow of hydrogen, supported chromia reduces 500 ◦ C earlier than ceria. During re-oxidation, only the samples containing a chromia phase on the surface lead to an increase in hydrogen and carbon monoxide production. Cr2 O3 /CeO2 produced 1.94 and 2.5 mL/(g·min) hydrogen and carbon monoxide respectively, while pure ceria only 0.09 and 0.005 mL/(g min). Chromia improved both the oxygen uptake, which increased the total hydrogen production from 3.2 to 6.7 mL/g, and the re-oxidation kinetics, which shortened the time to reach redox equilibrium from 41 down to 3 min. The chromium species, formed during the reduction step, have a catalytic effect on the re-oxidation reaction, possibly by decreasing the high energy barriers of hydrogen desorption and carbon dioxide dissociation. Therefore, the presence of chromium is essential in the tested reaction, since it catalyses the re-oxidation step and enhances the oxygen exchange capacity. The significantly improved production rates are analogous to those observed in the presence of expensive noble metals catalysts, such as platinum, emphasizing the importance of a catalysed re-oxidation step and rendering the easily reducible metal oxides attractive catalysts for the ceria-mediated water and carbon dioxide splitting. Acknowledgements We are grateful to Matthäus Rothensteiner for providing the experimental set-up and for his contribution to the project. We thank the Phoenix beam line scientists and especially Dr. Camelia

Borca and Dr. Thomas Huthwelker, for the X-ray spectroscopy measurements and data evaluation. We thank Dr. Frank Krumeich (ETH) for the STEM/EDXS investigation and the Electron Microscopy Centre of ETH Zurich (EMEZ) for providing the infrastructure. We also thank Simone Sala for the PCM-made materials and his contribution in developing the data processing method, Alexander Bonk (EMPA) for the SEM images and Jan Kovacovic (ETH) for valuable technical support. This research was funded by the Swiss Competence Centre Energy & Mobility and by the European Research Council within the framework of the European Union (FP7/2007-2013)/ERC grant agreement n◦ 247283. Ms. Mostrou thanks Hellenic Petroleum S.A. for funding her postgraduate studies, during which this research was conducted. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at 001. References [1] A.F. Ghoniem, Prog. Energy Combust. Sci. 37 (2011) 15–51, 10.1016/j.pecs.2010.02.006. [2] O. Ellabban, H. Abu-Rub, F. Blaabjerg, Renew. Sustainable Energy Rev. 39 (2014) 748–764, [3] J.A. Trainham, J. Newman, C.A. Bonino, P.G. Hoertz, N. Akunuri, Curr. Opin. Chem.Eng. 1 (2012) 204–210, [4] C. Agrafiotis, M. Roeb, C. Sattler, Renew. Sustainable Energy Rev. 42 (2015) 254–285, [5] T. Riedel, G. Schaub, Top. Catal. 26 (2003) 145–156, [6] S. Abanades, G. Flamant, Sol. Energy 80 (2006) 1611–1623, 10.1016/j.solener.2005.12.005. [7] J.E. Miller, Initial case for splitting carbon dioxide to carbon monoxide and oxygen, in Sandia Report, 2007, pp. SAND2007–8012. [8] W.C. Chueh, S.M. Haile, ChemSusChem 2 (2009) 735–739, 10.1002/cssc.200900138. [9] R.J. Panlener, R.N. Blumenthal, J.E. Garnier, J. Phys. Chem. Solids 36 (1975) 1213–1222, [10] P. Furler, J. Scheffe, M. Gorbar, L. Moes, U. Vogt, A. Steinfeld, Energy Fuels 26 (2012) 7051–7059, ´ [11] J. Lapp, J.H. Davidson, W. Lipinski, Energy 37 (2012) 591–600, http://dx.doi. org/10.1016/ [12] L.J. Venstrom, R.M. De Smith, Y. Hao, S.M. Haile, J.H. Davidson, Energy Fuels 28 (2014) 2732–2742, [13] Z. Hu, H. Metiu, J. Phys. Chem. C 115 (2011) 17898–17909, 10.1021/jp205432r. [14] Q. Jiang, G. Zhou, Z. Jiang, C. Li, Sol. Energy 99 (2014) 55–66, 10.1016/j.solener.2013.10.021. [15] J.R. Scheffe, R. Jacot, G.R. Patzke, A. Steinfeld, J. Phys. Chem. C 117 (2013) 24104–24114, [16] Y. Hao, C.-K. Yang, S.M. Haile, Chem. Mater. 26 (2014) 6073–6082, http://dx. [17] A. Le Gal, S. Abanades, Int. J. Hydrogen Energy 36 (2011) 4739–4748, http:// [18] P. Singh, M.S. Hegde, Chem. Mater. 22 (2010) 762–768, 1021/cm9013305. [19] C.B. Gopal, S.M. Haile, J. Mater. Chem. A 2 (2014) 2405–2417, http://dx.doi. org/10.1039/c3ta13404k. [20] K. Otsuka, M. Hatano, A. Morikawa, J. Catal. 79 (1983) 493–496, http://dx.doi. org/10.1016/0021-9517(83)90346-9. [21] S. Kacimi, J. Barbier, R. Taha, D. Duprez, Catal. Lett. 22 (1993) 343–350, http:// [22] M. Yashima, K. Ohtake, M. Kakihana, M. Yoshimura, J. Am. Ceram. Soc. 77 (1994) 2773–2776, [23] L. Mädler, W.J. Stark, S.E. Pratsinis, J. Mater. Res. 17 (2002) 1356–1362, http:// [24] V.A. Solé, E. Papillon, M. Cotte, P. Walter, J. Susini, Spectrochim. Acta Part B 62 (2007) 63–68, [25] B. Ravel, M. Newville, J. Synchrotron Radiat. 12 (2005) 537–541, http://dx.doi. org/10.1107/s0909049505012719. [26] M. Rothensteiner, S. Sala, A. Bonk, U. Vogt, H. Emerich, J.A. van Bokhoven, Phys. Chem. Chem. Phys. 17 (2015) 26988–26996, c5cp03179f. [27] H. Schulz, W.J. Stark, M. Maciejewski, S.E. Pratsinis, A. Baiker, J. Mater. Chem. 13 (2003) 2979–2984, [28] P. Moriceau, B. Grzybowska, L. Gengembre, Y. Barbaux, Appl. Catal. A 199 (2000) 73–82,

S. Mostrou et al. / Applied Catalysis A: General 537 (2017) 40–49 [29] P. Singh, M.S. Hegde, J. Gopalakrishnan, Chem. Mater. 20 (2008) 7268–7273, [30] R. Shukla, A.K. Bera, S.M. Yusuf, S.K. Deshpande, A.K. Tyagi, W. Hermes, M. Eul, R. Pottgen, J. Phys. Chem. C 113 (2009) 12663–12668, 1021/jp903013u. [31] M.C. Weber, J. Kreisel, P.A. Thomas, M. Newton, K. Sardar, R.I. Walton, Phys. Rev. B 85 (2012) 054303. [32] C. Lamberti, J.A. van Bokhoven, X-Ray Absorption and X-Ray Emission Spectroscopy, John Wiley & Sons, Ltd, 2016. [33] A. Bianconi, J. Garcia, M. Benfatto, A. Marcelli, C.R. Natoli, M.F. Ruizlopez, Phys. Rev. B 43 (1991) 6885–6892, [34] Z. Zeng, I.D. Fawcett, M. Greenblatt, M. Croft, Mater. Res. Bull. 36 (2001) 705–715, [35] M.I. Zaki, N.E. Fouad, G.C. Bond, S.F. Tahir, Thermochim. Acta 285 (1996) 167–179, [36] S.D. Yim, I.S. Nam, J. Catal. 221 (2004) 601–611, jcat.2003.09.026.


[37] A. Trovarelli, Catal. Rev. 38 (1996) 439–520, 01614949608006464. [38] N. Wiberg, A.F. Holleman, E. Wiberg, Holleman-Wiberg; Lehrbuch Der Anorganischen Chemie, 91–100 ed., De Gruyter, Berlin, New York, 1985. [39] H.A. Hansen, C. Wolverton, J. Phys. Chem. C 118 (2014) 27402–27414, http:// [40] B. Chen, Y. Ma, L. Ding, L. Xu, Z. Wu, Q. Yuan, W. Huang, J. Phys. Chem. C 117 (2013) 5800–5810, [41] S. Aranifard, S.C. Ammal, A. Heyden, J. Catal. 309 (2014) 314–324, http://dx. [42] D. Marxer, P. Furler, J. Scheffe, H. Geerlings, C. Falter, V. Batteiger, A. Sizmann, A. Steinfeld, Energy Fuels 29 (2015) 3241–3250, acs.energyfuels.5b00351.