Incorporation of square-planar Pd 2+ in ﬂ uorite CeO 2 : hydrothermal preparation, local structure, redox properties and stability †

The direct hydrothermal crystallisation at 240 (cid:1) C of Pd 2+ -containing ceria is investigated to study the extent to which precious metal dopants may be introduced into the cubic ﬂ uorite lattice. Samples of composition Ce 1 (cid:3) x Pd x O 2 (cid:3) d , where 0 # x # 0.15 can be produced in which Pd is included within the CeO 2 structure to give a linear lattice expansion. Attempts to produce higher Pd 2+ -substitution result in the formation of PdO as a secondary phase. Ce and Pd were determined to be in the +4 and +2 oxidation states, respectively, by X-ray absorption near edge structure, suggesting oxide de ﬁ ciency as the mechanism of charge balance. Extended X-ray absorption ﬁ ne structure (EXAFS) analysis at the Pd K-edge reveals that Pd 2+ has local square-planar coordination, as expected, and that a structural model can ﬁ tted in which the average ﬂ uorite structure is maintained, but with Pd 2+ sitting in the square faces of oxide ions present in the local cubic geometry of Ce. This model, consistent with previous modelling studies, gives an excellent ﬁ t to the EXAFS spectra, and explains the observed lattice expansion. Transmission electron microscopy analysis shows that Pd is well dispersed in the nanocrystalline ceria particles, and in situ powder XRD shows that upon heating in air the samples remain stable up to 800 (cid:1) C. H 2 -TPR shows that Pd-substitution leads to low temperature (<200 (cid:1) C) reduction of the oxide, which increases in magnitude with increasing Pd-substitution. On prolonged heating, however, the Pd is lost from the ceria lattice to give dispersed Pd metal, suggesting an inherent instability of Pd-doped CeO 2 .


Introduction
Cerium dioxide (ceria) and doped variants are well known to be an essential component of a range of important heterogeneous catalyst systems of relevance to energy and the environment. 1,2 The ability of ceria to release oxygen whilst maintaining the average uorite structure (consisting of a cubic close packed array of Ce 4+ ions with oxide ions occupying the tetrahedral sites) and readily re-oxidise in oxidising conditions makes it an ideal oxygen buffer. This is facilitated by high oxide mobility in the uorite structure and the Ce 4+ /Ce 3+ redox couple at moderate temperatures. 3 A small amount of precious metal, such as Rh, Pd or Pt, highly dispersed on the surface of ceria greatly improves its low temperature reducibility, 4,5 a feature much exploited for applications such as the water gas shi reactions 2,6,7 and automobile three-way catalysts. 5,6,8,9 Recent work has also shown other catalysis applications for preciousmetals/ceria, such as in the methane oxidation reaction. 10 Since this enhancement of properties, involving intimate, atomicscale interaction, is widely considered to be a cooperative effect between the cerium oxide and the precious metal, 11 incorporation of precious metal ions into the ceria lattice may be considered as an atomic efficient alternative to surface immobilisation of precious metals. 12 Substitution of isovalent 9,13 or aliovalent 14 metal ions into ceria is well known to improve oxide mobility as a result of lattice distortion and/or the formation of charge-compensating oxide vacancies. 15 The group of Hegde has been very successful in incorporating small amounts (#5 mol% replacement of Ce) of precious metal into ceria, principally by a solution combustion method. 16 The materials Ce 1Àx M x O 2Àd , (M ¼ Pt, Pd, Rh, Ru) prepared via this route have shown favourable CO oxidation and NO reduction activity compared to the precious metal dispersed on the surface of CeO 2 . [17][18][19] More recently, milder synthetic routes to preparing phase-pure Ce 1Àx Pd x O 2Àd with a higher degree of Pd substitution have been explored, such as ultrasonication, 20 ultrasonic spray pyrolysis 21 (x up to 0.10 in both cases) and coprecipitation of a microemulsion (x up to 0.21). 22 One report of hydrothermal preparation of precious metalsubstituted ceria (Ce 0.95 Ru 0.05 O 2Àd and Ce 0.90 Ru 0.10 O 2Àd ) has emerged, 23 despite the wide range of other metals ions substituted into ceria by this route. 24 The oxidation state of Pd when included in ceria has been demonstrated to be +2 by X-ray absorption near edge spectroscopy (XANES) 18 and X-ray photoelectron spectroscopy (XPS). [18][19][20]25 Although a square-planar geometry is expected for Pd 2+ with its 4d 8 electron conguration, existing extended X-ray absorption ne structure (EXAFS) studies 18,19 have failed to explain how a square-planar ion is accommodated into the uorite structure. Scanlon et al. used density functional theory analysis 26 to show computationally that a Pd 2+ dopant ion moved off the Ce site by 1/4 of a unit cell in one direction ($1.2Å) to sit on the face of an oxide cube rather than the centre, is an energetically stable conguration (Fig. 1). In this model Ce 4+ is replaced by Pd 2+ so to balance charge oxide vacancies must also be present, and one of the oxygens becomes under-coordinated. To our knowledge, such a model has not been tested with any experimental data. In this paper we explore the hydrothermal synthesis of nanocrystalline Ce 1Àx Pd x O 2Àd (0 # x # 0.25) and examine the local chemical environment of Pd using EXAFS analysis before studying the stability of these materials in air and their redox properties in a reducing gas stream. Our aim was to explore the level of Pd incorporation possible in the CeO 2 structure by this so chemical route and to examine the structural chemistry of the resulting phases, something not resolved satisfactorily in the literature at present. At the same time we have examined redox properties and stability.

Structure and solubility limit
Powder XRD patterns of Ce 1Àx Pd x O 2Àd (Fig. 2a) show a single nanocrystalline uorite phase is present in each case up to x ¼ 0. 25. When x $ 0.30, Bragg peaks attributed to crystalline PdO were seen in the diffraction pattern. Whilst there is no indication of PdO in Ce 0.75 Pd 0.25 O 2Àd in the diffraction data, the lattice parameter is almost identical to that of Ce 0.80 Pd 0.20 O 2Àd , suggesting that no more Pd has been substituted into the lattice by this point. In fact a linear lattice expansion can be seen only up to x ¼ 0.15 (Fig. 2b), suggesting the upper limit of Pd inclusion is actually between 0.15 and 0.20 (the EXAFS analysis described below claries this point). Other Pd-containing ceria samples reported in the literature have also been shown to exhibit lattice expansion with increasing Pd content: for example, Kurnatowska et al. produced materials by a water-in-oil microemulsion method followed by calcination in air at 500 C and found a similar lattice expansion with homogeneous materials up to x ¼ 0.20, 22 while Misch et al. produced materials with up to 10 mol% Pd substitution by ultrasonic spray pyrolysis at 500 C, and also found a lattice parameter expansion, 21 although somewhat smaller than seen for our materials. Materials prepared by ultrasonication also show a maximum value of x of 0.10, 20 and the lattice parameters reported are   Other published papers on Pd doping of ceria use much lower Pd content and lattice parameters are very similar to that of undoped CeO 2 , 18,27 or are co-doped with other metals. 25,28 The ionic radius of Pd 2+ is signicantly smaller (0.64-0.86 A depending on geometry 29 ) than eight-coordinate Ce 4+ (0.97Å (ref. 29)), and a lattice contraction would thus be expected if Pd were simply occupying the Ce 4a crystallographic site in the uorite structure. The lattice expansion observed is, however, consistent with the Pd 2+ ion residing in a square-planar site at the centre of one of the faces belonging to the CeO 8 cubes, Fig. 1   bond lengths. The powder XRD also shows that there is signicant broadening of the Bragg peaks as x is increased, suggesting that the crystallite size is decreasing. This effect has also been seen with other dopants in CeO 2 . 32 It is also possible that strain associated with the distorted metal environment is a factor in peak broadening. Rietveld analysis of the structure is complicated by the broad diffraction prole and the likely high thermal parameters of oxygen, but using a model based on Fig. 1 the rened oxygen content is consistent with increasing oxide deciency with increased Pd content and the general chemical formula Ce 1Àx Pd x O 2Àd (see ESI †).
Elemental analysis showed that the Ce : Pd ratio in the asmade oxides is the same as expected in the reaction mixtures (ESI, Table S1 †). High resolution transmission electron microscopy of hydrothermally prepared CeO 2 , Ce 0.90 Pd 0.10 -O 2Àd and Ce 0.80 Pd 0.20 O 2Àd (Fig. 3) shows that the particle size decreases as x increases from around 10-20 nm of CeO 2 to sub-10 nm particles of Ce 0.80 Pd 0.20 O 2Àd . Energy-dispersive X-ray spectroscopy (EDX) elemental analysis of Ce 0.90 Pd 0.10 -O 2Àd and Ce 0.80 Pd 0.20 O 2Àd at multiple points found that the Ce : Pd ratio is both consistent with the intended values and the metals are homogeneously distributed throughout the oxides (ESI, Table S2 †). Ce L III -edge XANES spectra of all the samples (ESI, Fig. S3 †) shows that the average Ce oxidation state and local environment remains as in CeO 2 . While the presence of Ce(III) could also lead to lattice expansion, and indeed in small particles of CeO 2 some reduction of the cerium is oen found, 33 for our materials there is no evolution of the XANES signal with Pd doping, so the increased Pd content must be largely be responsible for the lattice expansion. The Pd K-edge XANES suggests that the Pd is found primarily in the +2 oxidation state, although there is possible evidence for some local oxide deciency giving partially oxidised Pd, since the local structure is not identical to as seen in PdO (ESI, Fig. S4 †).

EXAFS analysis of local structure of Pd
Pd K-edge EXAFS spectra of Ce 1Àx Pd x O 2Àd (x ¼ 0.05, 0.10, 0.20) show a marked difference to the spectrum of PdO (Fig. 4). The calculated c(k) of PdO based on its crystal structure with distances and thermal parameters allowed to rene, ts closely the measured data (Fig. 4a, and Table 1), and this same model yields a poor t to the Pd-substituted ceria uorites with incorrect rened correlation distances and unphysical thermal parameters, eliminating the possibility of Pd existing solely as amorphous PdO in the ceria materials. We note here that previous EXAFS studies of Pd-doped ceria materials by Hegde and co-workers, erroneously assigned an O-O distance of 2.67Å in PdO as due to Pd-Pd and used this in their EXAFS modelling, 18,34 so we are unsure as to the validity of their tted parameters in the lightly doped samples they prepared by solution combustion. Instead we have used the structural model shown in Fig. 1 to describe the local structure, where Pd is substituted into ceria, but rather than occupying the Ce site it is shied to sit at the centre of a plane of four equidistant O atoms. An initial model with no expansion of the unit cell was used (i.e. with Pd-O distances of 1.9Å) and on renement of structural parameters against the Pd K-edge EXAFS spectra of   (Fig. 2b), it is conceivable that a small amount of PdO, undetectable by powder XRD, is present as a second phase at x ¼ 0.20. A two-phase model is difficult to t to the EXAFS signal, however, since the Pd-Pd distance in PdO and the Pd-Ce distance in the doped model are both close to 3Å. Therefore, some parameters had to be xed in order to obtain a satisfactory t: S 0 2 was xed at 0.826, the value found for PdO (and very close to the value for the doped model), and the Debye-Waller factor for the Ce shell was xed at 0.0044Å 2 (found for the doped model) and the value for the second O shell at 0.001Å 2 . The proportion of Pd present as PdO was estimated by tting the lattice parameter of the uorite to the extrapolated linear t (tted over the region 0 # x # 0.15) to give value of x ¼ 0.17. This means that approximately 15% of the Pd is present as PdO and 85% is found in Ce 0.83 Pd 0.17 O 2Àd . Fitting this model to the EXAFS spectrum gives a good t (R-factor ¼ 0.0295, Table 1). The addition PdO in this material must be of too small a crystal size to be resolved in the powder XRD. This is then consistent with the maximum amount of Pd in Ce 1Àx Pd x O 2Àd being x ¼ 0.15 from our hydrothermal synthesis method.

Thermal stability and redox properties
In situ powder diffraction on heating in air of Ce 1Àx Pd x O 2Àd (x ¼ 0.05, 0.10) shows the materials are stable up to at least 800 C (Fig. 5), with no trace of Pd or PdO observed. However, although Ce 0.85 Pd 0.15 O 2Àd was observed to have only uorite reections at 750 C, at 800 C Pd metal peaks are also present, suggesting that at high Pd-content the oxide is less stable (Fig. 5). H 2 -TPR proles of Ce 1Àx Pd x O 2Àd (0 # x # 0.25) up to 250 C (Fig. 6a) show in each case the presence of a large, sharp, low temperature reduction peak, which cannot be solely attributed to the reduction of Pd 2+ to Pd metal on the basis of quantication of the amount of hydrogen consumed. The temperature of the maximum hydrogen uptake decreases as x increases, and the total amount of hydrogen oxidised tends to increase linearly with increasing x up to x ¼ 0.20 (Fig. 6b). The TPR prole of hydrothermally prepared CeO 2 over the same range (not shown) showed no visible hydrogen oxidation. The increasing hydrogen oxidation in the Pd-containing materials cannot be solely attributed to the increasing amount of Pd 2+ present being reduced: a signicant, increasing proportion of Ce 4+ reduction must be occurring (up to x ¼ 0.20). Assuming complete  reduction of the Pd, then the percentage of cerium that is reduced can be calculated: this is plotted in Fig. 6b and the results shown in Table S3 of the ESI. † When x $ 0.20 the reduction peak begins to split into two peaks, possibly due to the presence of a PdO impurity (as observed in the EXAFS analysis). The increasing proportion of Ce 4+ reduction (up to x ¼ 0.20) suggests that the presence of ionic Pd 2+ distributed throughout the structure leads to greater reducibility of the Ce 4+ . The origin of the enhanced reducibility of the materials can thus be ascribed to be due to a combination of the expansion of the lattice and the higher level of oxide defects as more Ce 4+ is substituted by Pd 2+ , both of which can be envisaged to enhance oxide-ion migration, in turn leading to more easy removal of oxygen with increased Pd content. Since the analysis above showed that further addition of Pd (i.e. intended x ¼ 0.25) leads to the formation of amorphous PdO in an inhomogeneous mixture this explains the fact that this material has a lower than expected hydrogen uptake.
For the lowest concentration of Pd (5 mol%), the effect of recycling the TPR experiment was investigated, Fig. 6c. This shows that aer re-oxidation to 700 C the hydrogen reduction occurs at markedly lower temperature ($75 C). This is reproducible (see TPR4 on Fig. 6c) but re-oxidation at the lower temperature of 500 C is not sufficient to regenerate the low temperature reduction (TPR3). The material thus formed aer the rst TPR-TPO cycle resembles Pd dispersed on CeO 2 : for example Luo et al. studied ceria onto which Pd had been added by wet impregnation and found the same sharp single reduction peak at 75 C in the TPR. 35 The TPR trace of our Ce 0.95 Pd 0.05 O 1.95 material aer a cycle of reduction-oxidation (TPR2 on Fig. 6c) is remarkably similar to that reported for Luo et al. for 2% and 5% Pd loaded on ceria. This would be consistent with the extrusion of Pd from the Pd-doped CeO 2 in the rst TPR cycle in our materials to give surface Pd, suggesting an inherent instability of Pd when incorporated into the ceria lattice.
TEM before and aer heat treatment, Fig. 7, conrms that Pd metal is easily released from the materials. The EELS maps clearly show how Pd is evenly dispersed in the same region as cerium in the as-made materials (Ce 0.95 Pd 0.05 O 2Àd in this case), but aer heating in air the Pd is phase separated and agglomerates in separate particles, >10 nm in dimension. This would explain the results of TPR-TPO cycling where aer reoxidising in air to 700 C the materials resemble Pd dispersed on a CeO 2 support.

Experimental
Samples of Ce 1Àx Pd x O 2Àd (nominally 0 # x # 0.25 in increments of x ¼ 0.05) were prepared by hydrothermal reactions at 240 C. CeCl 3 $7H 2 O (Aldrich, 99.9%) and PdCl 2 (Johnson Matthey, 99.8%) were mixed in a (1 À x) : x molar ratio; respectively; to give a total of 6 mmol of metal ions. To this 2 ml cold H 2 O was added, yielding a solution, to which 2 ml of H 2 O 2 (Aldrich, 30% aqueous solution) was added. This mixture was stirred for 15 minutes before 8 ml of 7.5 M NaOH was added dropwise with effervescence. The red-brown gel was stirred for a further 10-15 minutes before being sealed in a Teon-lined steel autoclave and placed in a pre-heated fan oven to 240 C for 24 hours. The autoclave was cooled to room temperature before ltration and washing with copious amounts of boiling water. The resulting brown solids were dried in an oven at 80 C overnight and ground in a pestle and mortar to give ne powders.
Powder X-ray diffraction (XRD) patterns for phase identication were measured using a Siemens D5000 diffractometer operating with Cu Ka 1/2 radiation. High resolution powder XRD data were collected using a Panalytical X'Pert Pro MPD diffractometer with monochromatic Cu Ka 1 radiation and a PIXcel solid-state detector. Non-ambient powder XRD measurements were made using a Bruker D8 Advance diffractometer with Cu Ka 1/2 radiation and tted with a VÅNTEC-1 high-speed detector. Powders were heated in situ using an Anton Paar XRK 900 reaction chamber controlled either in static air or under a ow of 5% hydrogen in nitrogen. Powder XRD patterns were tted using a Le Bail prole renement with the GSAS soware, 36 which allowed renement of lattice parameters and peak broadening parameters.
Thermogravimetric analysis (TGA) was performed using a Mettler Toledo Systems TGA/DSC 1 instrument under an air ow of 50 ml min À1 from room temperature to 1000 C at a heating rate of 10 C min À1 .
Inductively coupled plasma optical emission spectroscopy (ICP-OES) was used to determine the amounts of the metals present in the samples by the company MEDAC Ltd (UK).
XANES and EXAFS spectra were measured using B18 at Diamond Light Source. 37 Data were collected in transmission mode at the Ce L III -edge and the Pd K-edge from samples diluted with appropriate amounts of polyethylene powder and pressed in self-supporting discs around 1 mm thick. The incident energy was selected using a water-cooled, xed-exit, double-crystal monochromator with Si (311) crystals. On B18 a double toroidal mirror, coated with Cr and Pt, 25 m from the source is used to focus the beam horizontally and vertically onto the sample and a pair of smaller plane mirrors are used for harmonic rejection. The raw data were treated using the programme Athena 38 to produce XANES spectra, normalised to the post-edge background, and normalised EXAFS spectra. The Pd K-edge EXAFS spectra were analysed using the soware Artemis, 38 which implements the FEFF code, over the k-range 3-12Å À1 . For the crystalline reference material PdO the coordination numbers were set at values expected from the known structure and shell distances (R) and thermal parameters (s 2 ) along with threshold energy (E 0 ) and amplitude reduction factor (S 0 2 ) were allowed to vary in least-squares renements to obtain a satisfactory t to the k 3 -weighted EXAFS. For the Pd-containing ceria samples, various structural models were considered, as described below, and the goodness of t assessed, while it was checked that the rened amplitude reduction factor (S 0 2 ) gave a value similar to that obtained for PdO. High resolution transmission electron microscopy (TEM) and electron energy loss spectrum imaging (EELS-SI) studies were performed on a probe-corrected JEOL ARM 200F microscope operating at 200 kV and equipped with a Gatan GIF Quantum ER spectrometer. High angle annular dark eld (HAADF) images were produced and for EELS-SI a dispersion of 0.5 eV per channel was set and spectra with 2048 channels were recorded with an energy range of 1024 eV in each spectrum. The Gatan spectrum imaging plugin was used for EELS-SI acquisitions. This yielded 3D data sets with each pixel containing an individual EELS spectrum. Elemental maps of Pd M 4,5 -edge, O K-edge, and Ce M 4,5 -edge were produced aer a suitable preedge inverse power law background tting with energy windows of 30 eV, 35 eV, and 45 eV respectively. Temperature programmed reduction (TPR) was measured by thermal conductivity of the 10% H 2 in N 2 (at a ow rate of 30 ml min À1 ) before and aer contact with the 0.05-0.10 g sample. A H 2 O trap aer the sample was used to absorb water created by H 2 oxidation.
Temperature was increased linearly as a function of time. In order to quantify accurately the hydrogen consumption a known quantity (1 ml) of N 2 was injected into the H 2 /N 2 gas stream before the experiment began to create a calibration peak. Quantication of H 2 consumption was carried by the integration of TPR prole as a function of time. Results are presented as the amount of H atoms oxidised per gram of sample, mmol (H) g À1 . The error is estimated to be AE5% for the total H 2 consumption in these measurements. Temperature programme oxidation (TPO) in the same apparatus was performed with 10% O 2 in He at the same ow rate to examine recyclability of the materials.

Conclusions
In conclusion, a series of nanocrystalline Pd-substituted ceria-type uorites of the form Ce 1Àx Pd x O 2Àd (0.05 # x # 0.15) have been prepared by low-temperature hydrothermal synthesis. Attempts to achieve a higher value of Pd substitution led to the formation of PdO as a secondary phase. Analysis of the Pd K-edge EXAFS and the powder XRD have been used to infer a square-planar coordination, brought about by Pd offset from the Ce sites within the uorite structure and also giving under-coordinated oxide ions and oxide vacancies. This supports a previously proposed model from computer simulation. The oxides have been demonstrated by in situ powder XRD to be stable in air up to at least 750-800 C, but that Pd is extruded from the materials on heating beyond this. H 2 -TPR shows that increasing the Pd substitution leads to improved low-temperature reducibility of the oxides, but that on cycling the materials resemble Pd dispersed on ceria, consistent with an inherent instability of Pd 2+ within the ceria lattice. Despite this, the material formed has a reduction prole that resembles Pd dispersed onto pre-made ceria by we impregnation, and so a material with potential use in catalysis is formed. Further work would be required to understand the structural distortions introduced by Pd 2+ in ceria, such as by pair distribution function analysis, and also it would be informative to perform an in operando spectroscopic study to analyse the migration of Pd from the lattice to the surface of ceria, as this would allow the mechanism of redox chemistry to be understood.