ARTICLE IN PRESS Ultramicroscopy 108 (2008) 1432– 1440
Contents lists available at ScienceDirect
Ultramicroscopy journal homepage: www.elsevier.com/locate/ultramic
In situ environmental TEM studies of dynamic changes in cerium-based oxides nanoparticles during redox processes Peter A. Crozier , Ruigang Wang 1, Renu Sharma LeRoy Eyring Center for Solid State Science, School of Materials, Arizona State University, Tempe, AZ 85287-1704, USA
a r t i c l e in f o
PACS: 61.46.Hk 61.72.jd 64.70.K 68.37.Og Keywords: Ceria Ceria zirconia In situ environmental TEM Catalysts
a b s t r a c t We apply in situ environmental transmission electron microscopy (ETEM) to study the dynamic changes taking place during redox reactions in ceria and ceria–zirconia nanoparticles in a hydrogen atmosphere. For pure ceria, we find that a reversible phase transformation takes place at 730 1C in which oxygen vacancies introduced during reduction order to give a cubic superstructure with a periodicity of roughly twice the basic fluorite lattice. We also observe the structural transformations taking place on the surface during reduction in hydrogen. The (11 0) ceria surface is initially constructed with a series of low-energy (111) nanofacets. Under strong reduction, the surface slowly transforms to a smooth (11 0) surface which was not observed to change upon re-oxidation. The surface transformation allows the reduced surface to accommodate a high concentration of oxygen vacancies without creating a strong perpendicular dipole moment. In the ceria–zirconia system, we are able to use ETEM to follow the redox activity of individual nanoparticles and correlate this property with structure and composition. We find considerable variation in the redox activity and interpret this in terms of structural differences between the nanoparticles. & 2008 Elsevier B.V. All rights reserved.
1. Introduction Cerium-based oxides continue to attract considerable attention because of their current and potential use in catalytic applications. Cerium is especially interesting among rare earths because of its ability to easily change its valence state from Ce3+ to Ce4+. In cerium-based oxides, this electronic re-configuration is associated with the loss (for Ce4+) or addition (for Ce3+) of oxygen and can be triggered at relatively low temperatures (300–800 1C) by changes in the oxygen partial pressure of the gas in contact with the solid [1]. The ability of this material to reversibly accept or contribute oxygen to its surroundings is the basis for its use in catalytic applications where redox processes are important. The exact nature of the activity of cerium-based oxides depends primarily on the nature and concentration of defects in the material (e.g. oxygen vacancies, Ce3+ centers and dopants). When oxygen vacancies are present, oxygen ions can move through the lattice relatively easily giving rise to high-oxygen ion conductivity [2,3]. This also plays an important role in catalytic applications because oxygen vacancies can move rapidly from surface to bulk and vice versa. Corresponding author. Tel.: +480 965 2934; fax: +480 965 9004.
E-mail address:
[email protected] (P.A. Crozier). Present address: Materials Sciences Division, Lawrence Berkeley National Lab, Berkeley, CA 94720, USA. 1
0304-3991/$ - see front matter & 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.ultramic.2008.05.015
The stability and low-temperature redox activity are dramatically improved by the addition of zirconia and mixed ceria– zirconia oxides are now extensively used in three-way catalyst applications [4–7]. Doping ceria with zirconium can substantially lower the temperature at which the phase transformations take place resulting in a more active catalyst component. The activity of ceria–zirconia depends not only on the composition but also on the previous redox history. In particular, high-temperature reduction treatments are found to favor the formation of highly active materials with very desirable low-temperature redox properties. In spite of the tremendous importance of reversible lowtemperature redox in ceria–zirconia, the origin of the behavior is not yet resolved, although considerable progress has been made [8–15]. These applications rely on our ability to understand and control the complex defect structures in cerium-based oxides. However, an ongoing challenge for many catalysts is the difficulty associated with correlating catalytic activity with underlying nanoparticle parameters such as particle size, shape, structure, composition and surface structure. Most methods of characterizing catalyst activity determine the average properties of the entire ensemble of particles. This is a vital step in identifying useful catalyst formulations but it averages over the rich nanoscale variations that must be explored if a deep understanding of the structure–property relations is to be elucidated. This limitation is particularly important in many reducible-oxide catalysts where
ARTICLE IN PRESS P.A. Crozier et al. / Ultramicroscopy 108 (2008) 1432–1440
subtle and dynamic variations in nanoparticle structure and stoichiometry under reducing conditions may play a vital role in catalytic functionality. For example, partially reduced cerium oxide is unstable at low temperatures and/or in high oxygen partial pressure and must be characterized in situ under strong reducing atmospheres. Moreover, under reaction conditions, the active components of the catalyst may not be static and may continually undergo phase transformations as a result of variations in temperature, gas composition and pressure. In situ environmental transmission electron microscopy (ETEM) can be used to explore the dynamic interplay between ambient conditions, catalytic activity and nanoparticle structure for the cerium-based oxides. ETEM has been used to understand the dynamic behavior of catalysts for many years [16,17] and atomic resolution was first demonstrated on a ceria catalysts in 1989 [18]. Gai and co-workers have conducted many studies on the structural transformation mechanisms in catalytic oxides (e.g. [19,20]). Hansen et al. [21] have studied the effect of promoters on Ru catalysts and reduction mechanisms in CuO/ZnO–methanol catalyst by high-resolution imaging and electron energy-loss spectroscopy (EELS) with ETEM [22]. The Topsoe group has also made other significant contributions using in situ electron microscopy of catalysts [23,24]. At Arizona State University, we have experience in developing and applying ETEM techniques to problems in nanomaterials [25–33] and catalysts [34–47]. By varying the oxygen chemical potential in the microscope, we can measure the redox activity of individual nanoparticles of cerium-based oxides. Moreover, using a combination of in situ nanospectroscopy and imaging, we are able to correlate nanoscale changes in activity with nanoscale variations in composition and structure. This approach allows us not only to explore the phase transformations taking place within the nanoparticles, but also allows us to monitor surface transformations which may also impact catalytic properties. An added advantage of undertaking surface studies on nanoparticles is that we are able to observe both high and low-energy surfaces within the same particles. These high-energy surfaces may not be easy to prepare for singlecrystal surface science studies but may play a critical role in many catalytic processes. Here, we demonstrate the power of in situ ETEM by exploring the dynamic changes taking place in nanoparticles of ceria and ceria–zirconia in hydrogen. Under the conditions assumed by us, catalytic oxidation of hydrogen occurs via the extraction of lattice oxygen leading to reduction of the catalyst, i.e. 2mH2 þ nCeO2 ! 2mH2 O þ Cen O2ðnmÞ where n and m are integers with mpn. Similarly, for the mixedoxide system in an H2 atmosphere we have the reaction 2mH2 þ n Ce1x Zrx O2 ! 2mH2 O þ ðCe1x Zrx Þn O2ðnmÞ These reactions cause the catalyst to change composition and structure to accommodate the oxygen vacancies being introduced to the crystal lattice. We employ a combination of high-resolution electron imaging (HREM), electron diffraction and EELS to monitor the evolution of the structure and chemistry in situ during the redox processes from room temperature up to 800 1C in a hydrogen atmosphere. Section 2 describes sample preparation and practical aspects of running-ETEM experiments. Section 3.1 gives results where we observe reversible ordering of oxygen vacancies during redox processes in pure ceria. In Section 3.2, we show how re-constructions taking place under reducing conditions on ceria surfaces can be interpreted in terms of the contribution of charged sheets to the surface dipole moment. In Section 3.3, we determine the activity of individual nanoparticles of ceria–zirconia and show that the sample heterogeneity leads to the formation of active and inactive nanoparticles.
1433
2. Experimental approach 2.1. Sample synthesis Ceria nanopowders were synthesizing using a precipitation technique. Cerium (III) nitrate (Ce(NO3)3 6H2O, 99.999% from Aldrich Chemical Company, USA) and ammonium carbonate ((NH4)2CO3, 99.999% from Alfa Aesar, USA) were chosen as titrant and precipitant, respectively. A 0.15 M Ce(NO3)3 6H2O solution was slowly (5 cm3/min) dropped to the solution of 1.5 M (NH4)2CO3 under vigorous stirring and homogenized at 70 1C for 1 h. The resulting precipitant was then filtered, washed and dried at 70 1C. The material was calcined in flowing oxygen (100 cc/min) for 2 h at the 700 1C. Ceria–zirconia powders of nominal composition Ce0.5Zr0.5O2 were synthesized using a spray-freezing method. The procedure is described in detail elsewhere (Wang et al., 2006). Cerium and zirconium salt solutions of appropriate concentration were mixed and sprayed with an air brush on the wall of a glass bowl held at liquid nitrogen temperature. The liquid drops freeze instantaneously on contact with the wall and stays frozen in the glass bowl. The glass bowl with frozen granules was quickly transferred to a vacuum chamber, which is connected to a multipump system. When water was removed, the gel-like precursor was dissolved in anhydrous alcohol, dried and calcined at 500 1C for 5 h in a furnace. The ceria–zirconia sample was then subjected to a hightemperature heat treatment in hydrogen. Details of the treatment are described elsewhere [47,48], but it consisted of heating the sample slowly up to 1000 1C in a 5%H2/95% He atmosphere. This processing was conducted in a Setaram TG92—thermal gravimetric analysis system, so that the weight change of the sample could be monitored during heating. We observe 100% reduction of the cerium at a temperature of about 490 1C. The sample was then allowed to cool down to room temperature and re-oxidize.
2.2. Instrumentation and experimental methodology The powders were gently pressed between two glass-microscope slides and then dry-dispersed onto clean Pt grids. The Pt grids were made by punching 3 mm grids out of Pt mesh (99.9% pure and 100 mesh woven from 0.003 in diameter wires). It is worth considering the effect of using Pt grids on in situ measurements on catalytic oxides. It is true that Pt will rapidly dissociate hydrogen and this atomic hydrogen may spill over onto the oxide which can cause reduction to take place at much lower temperatures. However, this atomic hydrogen is strongly associated with the Pt metal and the oxide in the immediate neighborhood of the metal. In our experiments, we deliberately choose oxide particles that are not in direct contact with the Pt grid but are located at the end of a cluster of oxide particles (see, for example, Fig. 6a). In this case, Pt is not in contact with the grain under study. We are operating at temperatures significantly lower than the Tamman temperature and the oxide crystal of interest is separated from the Pt grid by many grain boundaries. This makes the likelihood of diffusion of Pt atoms to the oxide particles of interest extremely small. The gas volume inside the environmental chambers is very large in comparison to the amount of catalyst and even the TEM grid and it is a flow system. Any hydrogen conversion that may occur on the Pt metal surface should have minimal influence on the overall gas composition. The final confirmation of this interpretation is that the reduction temperatures measured in the TEM are similar to or higher than the ex situ measurements with TGA [47,48]. This is the ultimate proof that there is no enhanced reduction caused by the use of Pt-TEM grids.
ARTICLE IN PRESS P.A. Crozier et al. / Ultramicroscopy 108 (2008) 1432–1440
3. Results and discussion 3.1. Ceria nanocrystals Fig. 1 shows a typical data set consisting of HREM images, electron diffraction patterns and energy-loss spectra recorded in 0.5 Torr of hydrogen from the same ceria nanoparticle at 600 and 730 1C. At 730 1C, we observe a reversal in the intensity of the Ce M45-white lines indicating that Ce has undergone a transition
(-200) (-11-1)
Ce+4 Intensity
The sample was immediately loaded into a Gatan-Inconel heating holder, introduced in the ETEM and heated to 150 1C to discourage hydrocarbon adsorbates from settling on the surface and blocking access to active sites on the ceria. In situ measurements were performed in an FEI Tecnai F20 field-emission ETEM operating at 200 kV with a point resolution of 0.24 nm. The gas reaction cell is a differentially pumped system and we have demonstrated an information limit of 0.13 nm in 4 Torr of H2 proving that the atomic resolution capability can be easily attained [49]. The ETEM is also equipped with a Gatan Imaging Filter; thus, EELS can be employed for in situ high-spatial resolution nano-chemical analysis. In situ HREM images and electron diffraction patterns were recorded with a Gatan-Orius CCD camera which permits digital recording up to 15 frames per second. The redox behavior of ceria-based oxides is very sensitive to any residual oxygen sources that may be present within the TEM column. For this reason, it is important to take steps to reduce the sources of oxygen inside the column in order to facilitate the reduction reactions of interest. The TEM and gas handling system were baked at around 70 1C for 24 h and then flushed, while hot, with dry nitrogen and dry hydrogen, to remove residual moisture and hydrocarbons. All tools and grids were subjected to ultrasonic cleaning in isopropanol followed by baking at 150 1C and plasma cleaning. To further reduce the water content of the system, ultrapure hydrogen was passed through liquid nitrogen to freeze out any residual water in the gas lines and liquid nitrogen was also put into the cold finger. Nanoparticles were chosen for the analysis that were in stable configurations on the Pt grid and were close to suitably convenient zone-axis orientations. Hydrogen was admitted into the environmental cell and the temperature stabilized close to 150 1C. An initial data set consisting of high-resolution images, electron diffraction patterns and electron energy-loss spectra were recorded from nearly 10 different nanoparticles. The temperature was then increased in variable steps and, after a waiting period of nearly 20 min for drift stabilization, another set of data were recorded for each nanoparticle after adjusting the particle tilt if necessary. This procedure was repeated up to about 750 1C. The samples were then allowed to cool in variable steps and additional data sets recorded. In some cases, near-phase transition points, the temperature was varied by relatively small amounts to study the reversibility of a phase transformation. Digital video data were also recorded to obtain information with improved time resolution. Data were processed using the Gatan digital micrograph software and homemade video scripts. In order to avoid electron beam irradiation effects, the electron beam was blanked during most of the experiments and was turned on to record data only for short periods of time during the in situ experiments. Furthermore, comparison between the nonirradiated and irradiated areas of the sample demonstrated that the use of relatively low-dose electron irradiation did not cause observable effects on the in situ observations. (In a previous work, we determined the temperature dependence of the critical dose in ceria [36].)
(02-2)
880900920940 900 880 920 Energy lossloss (eV) Energy (eV)
(11-1)
940
1 nm aa
Intensity
1434
Ce+3
880
900 920 Energy loss (eV)
940
Fig. 1. In situ HREM, electron diffraction pattern and energy-loss spectrum from a ceria crystal recorded at (a) 600 1C and (b) 730 1C in 0.5 Torr of H2. Images and diffraction patterns are recorded from (11 0) fluorite projection.
from the 4+ to 3+ oxidation state [36]. This indicates that oxygen has been removed from the ceria lattice as hydrogen gets converted to water in the gas phase. This reduction in the ceria crystal is associated with the appearance of superlattice reflections in both high-resolution images and electron diffraction patterns. The superlattice results in a quadrupling of the (2 2 0) plane-spacing associated with the cubic fluorite structure. The spectroscopic and crystallographic data show that oxygen vacancies have ordered in this structure. For the hydrogen pressure employed here, the structure is only stable at temperatures above 730 1C. Fig. 2 shows the effect of a small drop in temperature and
ARTICLE IN PRESS P.A. Crozier et al. / Ultramicroscopy 108 (2008) 1432–1440
1 nm
1435
conditions. First, experimental and theoretical studies on non-stoichiometric ceria suggest that the fluorite structure does not favor high oxygen vacancy concentrations. Experiments on thermal expansion measurements of CeO2y suggest that the fluorite structure is only stable for y values up to 0.15 (i.e. oxygen vacancy concentration exceeds 7.5%) after which a phase transformation occurs [50]. Moreover, molecular mechanics calculations suggest that, in partially reduced ceria, it is energetically more favorable to form domains of fully reduced Ce2O3 than to form dilute vacancies in CeO2 [51]. Fast-transformation kinetics will be facilitated by the large oxygen ion conductivity of ceria at 700 1C [52]. A likely scenario is that, in the presence of low oxygen pressure, when the oxygen vacancy concentration in the fluorite crystal exceeds 7.5%, the system rapidly transforms to Ce2O3. Raising the oxygen pressure causes the system to immediately revert back to fluorite. We are still studying the structural re-arrangement that takes place during the transformation process and a complete treatment will be published elsewhere. Our preliminary analysis indicates that the new structure is also cubic, but with a unit cell that is double the parent superlattice cell. In the refined structure, we expect there to be some cation relaxation away from their fluorite positions to accommodate the vacancy ordering. However, the rapid and reversible nature of the phase transformation suggests that there is no large displacement in the cation positions. Our observations do not appear to match any of the reduced ceria structures observed earlier [53]. However, these structures were created by keeping ceria under static reducing conditions for a period of 3 days and the structures were stable upon exposure to air. The structures we observe appears to be an active intermediate phase because it immediately transforms back to fluorite upon cooling. The observations illustrate the importance of making in situ observation in order to observe dynamic intermediate phases that may only be observable under changing ambient conditions.
3.2. Ceria surfaces
Fig. 2. In situ HREM from same ceria crystal as Fig. 1 recorded at (a) 735 1C in 0.53 Torr of H2 and (b) 722 1C in 0.46 Torr of H2.
hydrogen pressure on the superlattice. A reduction of only 13 1C along with a 15% drop in the H2 pressure causes the superstructure to vanish almost immediately. The energy-loss spectrum shows that the Ce oxidation state immediately reverts to 4+ indicating that the sample has rapidly re-oxidized. The vacancy ordering is completely reversible as the ceria particle reduces and re-oxidizes with changing temperature and hydrogen pressure. There is a very strong driving force for Ce to exist in the 4+ oxidation state [1] and even though the ambient is predominantly hydrogen, there is sufficient oxygen in the background gas of the column to re-oxidize the very small volume of ceria on the Pt grid. In this experiment, the presence of hydrogen is critical to lowering the oxygen chemical potential and allowing the ceria reduction to take place. It may be possible to trigger a similar reaction simply by heating in vacuum but much higher temperatures would be required. The entire phase transformation seems to occur within a few seconds. Our EELS analysis shows that this relatively small change in the oxygen chemical potential causes an exchange of large quantities of oxygen between the crystal interior and the gas phase. Both thermodynamic and kinetic considerations may contribute to the rapid phase transformation observed under reducing
Surfaces also undergo transformations in reducing atmospheres [54]. As the extent of reduction increases, reconstructions may occur because the low-energy surface of CeO2 is composed of charged sheets which are not stable when high concentrations of oxygen vacancies are introduced. Fig. 3 shows a series of profile images recorded from a ceria (11 0) surface in (11 0) projection during heating in 0.5 Torr of H2. At 266 1C (Fig. 3a), the surface is rough and displays a saw-tooth profile made up of (111) nanofacets approximately 1 nm in length. At this temperature, there should be no significant reduction of the ceria and the crystal should be fully oxidized with a composition close to CeO2. For pure CeO2, density functional calculations show that the bulkterminated (11 0) surface has an energy of 0.86 J/m2 whereas the (111) surface has an energy of 0.55 J/m2 and is the low-energy surface for ceria [55]. Even though sharp edges are generally associated with high energy because of the increase in the number of dangling bonds, our observations suggest that it is energetically favorable for the (11 0) surface to form with a series of lower energy (111) nanofacets. Presumably, the lower energy associated with the (111) facets more than offsets the highenergy cost associated with the sharp apex of the resulting sawtooth profile. Fig. 3b shows the same surface after heating for 30 min at 730 1C. The surface shows a gradual reduction in the roughness with the elimination of saw-tooth points and the formation of a smooth profile. The sample was held at this temperature for a total of 2 h and then cooled to 600 1C. Fig. 3c shows the surface
ARTICLE IN PRESS 1436
P.A. Crozier et al. / Ultramicroscopy 108 (2008) 1432–1440
O 2Ce4+ O 2-
1 nm
O 2Ce3+ O 2-
-
-
-
-
-
-
2 2- 2 2 2 2 2 + + + + + + + 4 4 4 4 4 4 4 - - - - - - 2 2 2 2 2 2 2 -
2
-
-
-
2 2 2 + + + + + 3 3 3 3 3 3 3 - - - - - - 2 2 2 2 2 2 2 +
+
Fig. 4. Schematic diagram showing charged layers making up the (111) ceria surface when (a) fully oxidized bulk-terminated and (b) with high concentration of oxygen vacancies.
Fig. 3. In situ profile images showing evolution of an identical region of a (11 0) surface of ceria during heating in 0.5 Torr of H2 recorded at (a) 270 1C, (b) 730 1C and (c) 600 1C.
after heating for 1 h at 600 1C in hydrogen and it is clear that the surface has continued to flatten and is now very smooth and is predominantly composed of (11 0) terraces with very little (111) component remaining. In the case of ceria, it is energetically favorable to form oxygen vacancies on the surface relative bulk. However, theoretical calculations show that the energy advantage is very small for the formation of single vacancies on the (111) surface and slightly better for vacancy associations [51,56]. Oxygen vacancies have been observed on the (111) surface of ceria heated to 500 1C in ultra-high vacuum conditions with scanning tunneling microscopy [57]. However, the vacancies are often clustered into triangular-type defects and eventually form line defects. The vacancy structures occupy only 1–2% of the surface and the rest of the surface is defect free. The tunneling currents also suggest that the electrons left behind when the oxygen leaves may not completely transfer to the nearby cerium atoms which is presumably an attempt to restore the symmetry of the charge in the oxygen sheet and minimize the surface dipole moment. In our experiments, the surface is exposed to a strong reducing atmosphere and we know from the previous section that ceria undergoes a phase transformation as a result of ‘‘bulk’’ reduction when it is heated in hydrogen above 700 1C. A large number of oxygen vacancies are introduced which should imply that a very large number of oxygen vacancies should be present on the surface. However, we observe a significant reconstruction of the surface atoms in response to the oxygen-vacancy formation process that takes place during strong reduction. An insight into the reconstruction process can be obtained by considering surface stability in terms of the surface dipole moment associated with planes of charge at the surface of ionic crystals [58]. A surface will
be stable if it consists of charged planes in a symmetrical configuration resulting in no dipole moment perpendicular to surface. Fig. 4a shows a schematic diagram of a likely configuration of a bulk-terminated (111) surface of ceria. In this case, the surface layer is composed of a sheet of Ce4+ cations sandwiched between two sheets of O2 anions. The O sheets are symmetrically placed around the Ce sheet and the net perpendicular dipole moment will be zero giving rise to a stable surface configuration in agreement with experimental observation and theoretical calculations. This simple symmetrical configuration is disrupted when vacancies are introduced to the surface during redox processes. Fig. 4b shows a possible configuration of oxygen vacancies on the surface during the initial stages of reduction. During reduction, oxygen from the top layer reacts with hydrogen to form water leaving two electrons behind, thus ensuring the overall surface charge remains the same. The electrons transfer to adjacent Ce cations causes the oxidation state of the Ce to change to 3+ (as observed in the EELS measurements). Even though the average surface charge remains constant, the presence of oxygen vacancies on the outer O2 sheet disrupts the symmetry of the surface configuration and gives rise to a net surface dipole moment. Thus, this configuration of the non-stoichiometric (111) surface in ceria will be unstable. This argument against the formation of oxygen vacancies is clearly too strong, since the theoretical calculation and STM observations show that oxygen vacancies can exist on the (111) surface. However, during bulk-reduction of ceria, the oxygenvacancy concentration on the surface will be very high (approaching 25%). It is unlikely that vacancy clustering can prevent the creation of a large perpendicular dipole moment in the bulkterminated surface with a high-vacancy concentration. As the concentration of defects on the surface increases, a strong driving force will develop to immediately fill the vacancies with oxygen diffusing from the interior of the crystal. Indeed, the surface instability in the presence of high-oxygen vacancy concentrations may provide an active driving force for removing oxygen from the crystal interior. Eventually, the easy supply of oxygen from the interior of the crystal will be depleted when all the Ce4+ is converted to Ce3+. At this point, the surface has to undergo a significant reconfiguration of the surface atoms to avoid creating a net dipole moment. Our experimental observations show that the surface loses its (111) saw-tooth character and becomes a flat (11 0) surface. Fig. 5 shows the nature of the charge sheet making up the bulk-terminated (11 0) surface. In this case, the surface is composed of only one sheet with one Ce4+ cation for every two O2 anions. The sheet is neutral and presumably has a higher surface energy compared to the (111) surface because of the closer proximity of adjacent oxygen anions. In this case, when oxygen vacancies are introduced, the sheet remains neutral and the symmetry about a plane parallel to the surface is preserved,
ARTICLE IN PRESS P.A. Crozier et al. / Ultramicroscopy 108 (2008) 1432–1440
2- 4+ 2-
2-
3+
2-
2-
4+
2-
2-
3+ 2-
2-
4+
3+
2-
2-
4+
2-
2- 3+
2-
Fig. 5. Schematic diagram showing charged layers making up the (11 0) ceria surface when (a) fully oxidized bulk-terminated and (b) with high concentration of oxygen vacancies.
thus avoiding the creation of a net dipole moment perpendicular to the surface. Consideration of surface kinetics suggests that the proposed transformation can occur at the temperatures and observation times for our experiment. For the case considered here, it would take only a relatively small re-arrangement of atoms for the surface to transform from the nanofacetted (111) configuration to the flat bulk-terminated (11 0) configuration. We know that at 700 1C, some sintering of ceria does take place so there is certainly enough thermal energy to drive the necessary surface diffusion of cations in the space of a few hours. Our observations show that it is both energetically and kinetically favorable for the surface to transform to the bulk-terminated (11 0) surface under reducing conditions. When the sample was cooled to 600 1C, the crystal reoxidized and our EELS analysis shows that most of the oxygen vacancies in the bulk were refilled with Ce returning to the 4+ oxidation state. Fig. 3b shows that the surface does not transform back to the nanofacetted (111) configuration. At this point, we do not know if this is because the surface remains partially reduced or if it simply re-oxidizes in the (11 0) bulk termination. The driving force for surface reconfiguration may be very weak because of the (11 0) surface energy is only slightly higher than the (111) surface.
3.3. Ceria– zirconia solid solutions The ability to detect redox activity and the associated phase changes in individual nanoparticles becomes important in more complex systems where there are possibilities for multiple phases to form. In the ceria–zirconia systems, the many possible phase transformations present an opportunity for nanoscale compositional and structural heterogeneity to exist [48]. Identifying the most active phases and understanding the phase changes taking place under redox conditions is critical to developing improved catalysts. The phase transformations taking place in the high-surface area samples are difficult to detect with XRD analysis even though TGA measurements show substantial changes in the redox activity of the materials. These observations imply that subtle nanoscale structural and chemical changes within individual nanoparticles might be the key to understanding the origin of low-temperature reduction behavior. Our TGA measurements suggest that one component in the nanopowder is activated during the high-temperature redox cycling giving rise to a material with lowered reduction temperature, while a second component of the powder loses its activity resulting in a decrease in reduction percentage of the entire powder. In situ measurements provide us with a unique opportunity to explore the redox activity of individual nanoparticles and correlated the activity with nanoscale structure and composition measurements. The degree of reduction in individual nanoparticles can be determined by quantification of the ceria M45 white lines. The
1437
white lines are caused by electron transitions from the 3d5/2 (M5890 eV) and the 3d3/2 (M4910 eV)) to the unoccupied states in the 4f band. It has been shown that the relative intensity of the M5 and M4 peaks depends on the occupancy in the 4f band, which in turn depends on the oxidation state of Ce [59]. The occupancy of the 4f band changes from approximately 0 to 1 as the Ce transforms from a 4+ to a 3+ state. There are several different methods of extracting the variation of the white-line intensity in order to determine the average oxidation state of cerium [59–61]. Some of these methods require significant spectral processing and impose restrictive conditions on the data acquisition. Such methods are not always practical under the dynamic conditions present during ETEM work where images, diffraction patterns and spectra must be acquired rapidly often with significant sample drift while phase transformations are in progress. Fortunately, in the Ce system, the change in the white-line intensity is pronounced and thus we can use a simplified procedure to quantify changes in the Ce oxidation state [35,36]. Here, we remove the background beneath the M45 edge, integrate the M5 and M4 intensities over 17 eV windows and determine the M5/M4 intensity ratio. This simple approach is rather robust and forgiving and seems to work well for the data sets acquired in this and previous work. It is not easy to calibrate the white-line ratios simply by directly measuring the O/Ce concentration because of difficulties associated with variations in cation concentrations, adsorbates and oxygen non-stochiometry. Instead, we obtain a simple calibration by examining the spread of white-line ratios obtained over the entire temperature range and assume that the low value obtained from a fresh fully oxidized sample at room temperature corresponds to Ce4+ and the value obtained under severe reducing conditions (800 1C in 2 Torr of H2) corresponds to Ce3+. We further assume a linear relationship between the white-line ratio and the Ce oxidation state. These assumptions seem reasonable and based on the statistical spread in data points should give oxidation states that are accurate within 10%. Using the above approach, we are able to monitor the oxidation state of individual nanoparticles as a function of temperature and hydrogen pressure and determine the degree of redox activity. Fig. 6 shows a typical ceria–zirconia nanoparticle of size approximately 20 nm and the variation in the Ce oxidation state during heating in hydrogen. The particle has an initial Ce oxidation state of close to 3.7 at 480 1C in hydrogen. The oxidation state drops to about 3.1 when the temperature reaches 590 1C. On cooling to 470 1C, the oxidation state returns to the initial value of 3.7. This nanoparticle demonstrates a high degree of reversible redox activity and would be ideal for the redox reactions of relevance to three-way catalysts and fuel-cell anodes. Fig. 7 shows a similar curve for a different nanoparticle which was also about 20 nm in size. This particle also shows an initial Ce oxidation state of close to 3.7 at 480 1C in hydrogen. However, this time the oxidation state drops only to about 3.5 when the temperature reaches 590 1C. Moreover, on cooling to 470 1C, the particle does not re-oxidize and the oxidation state stays at a value of 3.5. This particle demonstrates a poor degree of reversible redox activity and may be in the process of transforming into a less active phase. We have observed this variation in many nanoparticles and have correlated the redox activity measured in situ on individual nanoparticles with the average redox activity of the entire sample measured through thermal gravimetric analysis [47]. We find good agreement between the nanoscale in situ measurement and the macroscopic measurement. Thus, the in situ EELS approach can be employed to differentiate between active and less active particles. In order to obtain more complete information about the nanostructure and dynamic changes taking place in the material
ARTICLE IN PRESS 1438
P.A. Crozier et al. / Ultramicroscopy 108 (2008) 1432–1440
20 nm 600 500 400 300
4.0
200 100
Ce oxidation state
3.8
0 880
900
920
940
920
940
3.6 3.4 1 nm
3.2 3.0
heating up cooling down
2.8 460
480
500
520 540 560 Temperature (°C)
580
600
Fig. 6. (a) Typical ceria–zirconia nanoparticles and (b) redox activity of active nanoparticle determined by measuring the cerium oxidation state as a function of temperature in 0.5 Torr of H2 using in situ EELS.
4.0
Type II
Ce oxidation state
3.8 3.6 880
3.4
Fig. 8. In situ HREM, electron diffraction pattern and energy-loss spectrum of an active ceria–zirconia particles recorded in 0.4 Torr of H2 at (a) 450 1C and (b) 650 1C. The EELS shows that the particle undergoes reduction at 650 1C. Images and diffraction patterns recorded from (11 0) projection.
3.2 heating up cooling down
3.0
900
2.8 460
480
500
520 540 Temperature (°C)
560
580
600
Fig. 7. Redox activity of less active nanoparticle determined by measuring the cerium oxidation state as a function of temperature in 0.5 Torr of H2 using in situ EELS.
with different redox activities, we combine our in situ EELS measurement with in situ HREM and electron diffraction. Fig. 8 shows a data set recorded from an active nanoparticle at 450 1C and 650 1C in 0.5 Torr of H2. The energy-loss spectrum shows that at 450 1C, the cerium is in the +4 oxidation state
whereas at 650 1C it has been reduced to the 3+ oxidation state. Clearly a large number of oxygen vacancies have been introduced and the ceria–zirconia nanocrystal is in a strongly reduced state. In contrast to the result for pure ceria, there is no apparent difference in the high-resolution images recorded at 450 1C and 650 1C and no evidence of an superstructure. This particle maintains a predominantly cubic fluorite structure as shown in the electron diffraction pattern and is characteristic of active nanoparticles for ceria–zirconia. Fig. 9 shows a data set recorded from an inactive nanoparticle at room temperature in 0.5 Torr of hydrogen. The energy-loss spectrum shows that, even at room temperature, cerium is
ARTICLE IN PRESS P.A. Crozier et al. / Ultramicroscopy 108 (2008) 1432–1440
already in a 3+ oxidation state. In contrast to the result for pure ceria and the active ceria–zirconia particle, the HREM image and electron diffraction pattern show the presence of a superstructure at room temperature. Heating the particle to 700 1C in hydrogen causes no observable change in the in situ EELS, HREM image or electron diffraction pattern showing that the particles are not redox-active over this large temperature range. In this case, the observed superstructure is related to ordering between the cerium and zirconium cations giving rise to a pyrochlore structure. In pyrochlore, Ce and Zr cations order along the (11 0) direction giving rise to individual columns of Ce and Zr along the (11 0) direction (Fig. 10a). Pyrochlore is known to form in these materials when they are heated to high temperature in a strong reducing environment. In this case, the sample was heated to 1000 1C during the TGA analysis and clearly some pyrochlore has formed. We carried out a series of image and diffraction simulations to confirm our identification. The image simulations were calculated using the multislice method with the Molecular Simulations CERIUS applications. The parameters used in the calculations were: incident electron energy Eo ¼ 200 keV, spherical aberration coefficient Cs ¼ 1.2 mm, focal spread ¼ 30 nm, convergence angle ¼ 1.0 mrad, atomic vibration ¼ 0.035 nm.
880 900 920 940 Fig. 9. In situ HREM, electron diffraction pattern and energy-loss spectrum of an inactive ceria–zirconia particles recorded in 0.4 Torr of H2 at room temperature. The EELS spectra show that the particle is already in the 3+ oxidation state. Dotted insert is simulated image of pyrochlore structure in (11 0) projection.
Ce
1439
Images were calculated using 320 320 beams out to cut-off radius of 14 nm1. The simulated image is shown Fig. 9 gives a good match to the experimental data confirming the identification as pyrochlore.
4. Conclusion In situ environmental transmission electron microscopy is a powerful technique for characterizing gas–solid reactions on nanoparticles. The ability to perform atomic resolution imaging and subnanometer chemical analysis in the presence of reactive gases at elevated temperature makes this approach ideal for studying heterogeneous catalyst under near-reactor conditions. We have employed ETEM to study the dynamic changes taking place during redox reactions in cerium-based oxide nanoparticles in a hydrogen atmosphere. For pure ceria, we find a reversible phase transformation taking place at 730 1C in which oxygen vacancies introduced during reduction order to give a cubic superstructure with a periodicity of roughly twice the basic fluorite lattice. The transformation takes place rather rapidly with the cerium cation switching from an initial oxidation state of 4+ to 3+ in the superstructure. With profile imaging, we are also able to observe the structural transformations taking place on the surface during reduction in hydrogen. The (11 0) ceria surface initially is constructed with a series of low-energy (111) nanofacets. Under strong reduction, the surface slowly transforms to a smooth (11 0) surface which was not observed to change upon re-oxidation. The surface transformation allows the reduced surface to accommodate a high concentration of oxygen vacancies without creating a strong perpendicular dipole moment. When ceria is doped with zirconia the reduction temperature is substantially reduced and high-temperature processing in hydrogen enhances the low-temperature reducibility. We are able to use ETEM to follow the redox activity of individual nanoparticles and correlate this property with structure and composition. This provides us with an exciting opportunity to isolate and identify the phases with different redox activity in catalytic nanoparticles of cerium-based oxides. We find considerable variation in the redox activity suggesting the presence of subtle compositional and structural differences between the nanoparticles. Some the nanoparticles have a pyrochlore structure showing strong cation ordering at room temperature and are not redox active. The redox-active particles have a predominant fluorite structure.
Acknowledgments This work was supported by National Science Foundation through Grant NSF-CTS-0306688. We thank the use of TEM facilities at the John M. Cowley Center for High-Resolution Electron Microscopy at Arizona State University. References
Zr
Fig. 10. Schematic diagram of pyrochlore structure (Ce2Zr2O7).
[1] D.J.M. Bevan, J. Inorg. Nucl. Chem. 1 (1955) 49. [2] M. Mogensen, N.M. Sammes, G.A. Tompsett, Solid State Ionics 129 (2000) 63. [3] A. Trovarelli, in: A. Trovarelli (Ed.), Catalysis by Ceria and Related Materials, Imperial College Press, London, 2002, pp. 15–50. [4] P. Fornasiero, G. Balducci, R. Di Monte, J. Kaspar, V. Sergo, G. Gubitosa, A. Ferrero, M. Graziani, J. Catal. 164 (1996) 173. [5] H.W. Jen, G.W. Graham, W. Chun, R.W. McCabe, J.P. Cuif, S.E. Deutsch, O. Touret, Cataly. Today 50 (1999) 309. [6] M. Shelef, R.W. McCabe, Cataly. Today 62 (2000) 35. [7] M. Shelef, G.W. Graham, R.W. McCabe, Ceria and other oxygen storage components in automotive catalysts, in: A. Trovarelli (Ed.), Catalysis by Ceria and Related Materials, Imperial College Press, London, 2002, pp. 343–375.
ARTICLE IN PRESS 1440
P.A. Crozier et al. / Ultramicroscopy 108 (2008) 1432–1440
[8] E. Mamontov, R. Brezny, M. Koranne, T. Egami, J. Phys. Chem. B 107 (2003) 13007. [9] E.S. Punta, J. Stubenrauch, J.M. Vohs, R.J. Gorte, Langmuir 85 (1995) 4832. [10] O. Costa-Nunes, R.M. Ferrizz, R.J. Gorte, J.M. Vohs, Surf. Sci. 592 (2005) 8. [11] O. Costa-Nunes, R.J. Gorte, J.M. Vohs, J. Mater. Chem. 15 (2005) 1520. [12] T. Montini, N. Hickey, P. Fornasiero, M. Graziani, M.A. Banares, M.V. MartinezHuerta, I. Alessandri, L.E. Depero, Chem. Mater. 17 (2005) 1157. [13] Z.C. Kang, J. Alloys Compds. 408 (2006) 1103. [14] E. Aneggi, M. Boaro, C. de Leitenburg, G. Dolcetti, A. Trovarelli, J. Alloys Compds. 408–412 (2006) 1096. [15] S. Bernal, G. Blanco, J.J. Calvino, J.C. Hernandez, J.A. Perez-Omil, J.M. Pintado, M.P. Yeste, J. Alloys Compds. 451 (2008) 521. [16] R.T.K. Baker, M.A. Barber, P.S. Harris, F.S. Feates, R.J. Waite, J. Catal. 26 (1972) 51. [17] R.T.K. Baker, P.S. Harris, R.B. Thomas, R.J. Waite, J. Catal. 30 (1973) 86. [18] G.M. Parkinson, Cataly. Lett. 2 (1989) 303. [19] P.L. Gai, Acta Crystallogr. B53 (1997) 346. [20] P.L. Gai, Topics Catal. 8 (1999) 97. [21] T.W. Hansen, J.B. Wagner, P.L. Hansen, S. Dahl, H. Topsoe, C.J.H. Jacobsen, Science 294 (2001) 1508. [22] P.L. Hansen, J.B. Wagner, S. Helveg, J.R. Rostrup-Nielsen, B.S. Clausen, H. Topsoe, Science 295 (2002) 2053. [23] J.B. Wagner, P.L. Hansen, A.M. Molenbroek, H. Topsoe, B.S. Clausen, S. Helveg, J. Phys. Chem. B 107 (2003) 7753. [24] S. Helveg, C. Lopez-Cartes, J. Sehested, P.L. Hansen, B.S. Clausen, J.R. RostrupNielsen, F. Abild-Pedersen, J.K. Norskov, Nature 427 (2004) 426. [25] R. Sharma, K. Weiss, Microsc. Res. Tech. 42 (1998) 270. [26] M.J. McKelvy, R. Sharma, A.V.G. Chizmeshya, R.W. Carpenter, K. Streib, Chem. Mater. 13 (2001) 921. [27] P.A. Crozier, J. Tolle, J. Kouvetakis, C. Ritter, Appl. Phys. Lett. 84 (2004) 3441. [28] R. Sharma, P. Rez, M.M.J. Treacy, S.J. Stuart, J. Electron Microsc. 54 (2005) 231. [29] W.F. van Dorp, B. Someren, C.W. Hagen, P. Kruit, P.A. Crozier, Nanoletters 5 (2005) 1303. [30] R. Sharma, P. Rez, M. Brown, G. Du, M.M.J. Treacy, Nanotechnology 18 (12) (2007) 125602, (8pp). [31] P.A. Crozier, Nanoletters 7 (8) (2007) 2395. [32] S. Hofmann, R. Sharma, C.T. Wirth, F. Cervantes-Sodi, C. Ducati, T. Kasama, R. Dunin-Borkowski, J. Drucker, P. Bennett, J. Robertson, Nat. Mater. (2008) (web issue). [33] P.A. Crozier, J. Vac. Sci. B 26 (1) (2008) 249.
[34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52]
[53] [54]
[55] [56] [57] [58] [59] [60] [61]
P.A. Crozier, R. Sharma, A.K. Datye, Microsc. Microanal. 4 (1998) 278. R. Sharma, P.A. Crozier, Inst. Phys. Conf. Ser. No 161: Section 10 (1999) 569. R. Sharma, P.A. Crozier, Z.C. Kang, L. Eyring, Philo. Mag. 84 (2004) 2731. P.A. Crozier, A.K. Datye, Stud. Surf. Sci. Catal. 130 (2000) 3119. P.A. Crozier, V.P. Oleshko, A.D. Westwood, R.D. Cantrell, Inst. Phys. Conf. Ser. 168 (2001) 393. V. Oleshko, P.A. Crozier, R. Cantrell, A. Westwood, Stud. Surf. Sci. Catal. 130 (2000) 935. V.P. Oleshko, P.A. Crozier, R.D. Cantrell, A.D. Westwood, Macromol. Rapid Commun. 22 (1) (2001) 34. V. Oleshko, P.A. Crozier, R. Cantrell, A. Westwood, J. Electron Micros. 51 (Suppl.) (2002) S27. Rou-Jane Liu, P.A. Crozier, C.M. Smith, D.A. Hucul, J. Blackson, G. Salaita, Microsc. Microanal. 10:01 (2004) 77. Rou-Jane Liu, P.A. Crozier, C.M. Smith, D.A. Hucul, J. Blackson, G. Salaita, Appl. Catal. A: Gen. 282 (2005) 111. P. Li, J. Liu, N. Nag, P.A. Crozier, J. Chem. Phys. B 109 (2005) 13883. P. Li, J. Liu, N. Nag, P.A. Crozier, Surf. Sci. 600 (2006) 693. P. Li, J. Liu, N. Nag, P.A. Crozier, Appl. Catal. A 307 (2) (2006) 212. R. Wang, P.A. Crozier, R. Sharma, J.B. Adams, Nanoletters 8 (3) (2008) 962. R. Wang, P.A. Crozier, R. Sharma, J.B. Adams, J. Chem. Phys. B 110 (37) (2006) 18278. R. Sharma, P.A. Crozier, R. Marx, K. Weiss, Microsc. Microanal. (2003) 912 CD. I. Riess, R. Koerner, M. Ricken, J. Noelting, Solid State Ionics 28–30 (1988) 539. J.C. Conesa, Surf. Sci. 339 (1995) 337. M. Mogensen, Ceria-based electrodes, in: A. Trovarelli (Ed.), Catalysis by Ceria and Related Materials, Imperial College Press, London, 2002, pp. 453–481. P. Knappe, L. Eyring, J. Solid State Chem. 58 (1985) 312. S.H. Overbury, D.R. Mullins, Ceria surface and films for model catalytic studies using surface analysis techniques, in: A. Trovarelli (Ed.), Catalysis by Ceria and Related Materials, Imperial College Press, London, 2002, pp. 311–341. Y. Jiang, J.B. Adams, M. van Schilfgaarde, J. Chem. Phys. 123 (2005) 064701. T.X.T. Sayle, S.C. Parker, C.R.A. Catlow, Surf. Sci. 316 (1994) 329. H. Norenberg, G.A.D. Briggs, Phys. Rev. Lett. 79 (1997) 4222. P.W. Tasker, J. Phys. C 12 (1979) 4977. T. Manoubi, C. Colliex, J. Electron Spectrosc. Relat. Phenom. 50 (1990) 1. D.H. Pearson, C.C. Ahn, B. Fultz, Phys. Rev. B47 (1993) 8471. L.A.J. Garvie, P.R. Buseck, J. Phys. Chem. Solids 60 (1999) 1943.
