3.1 Nanocrystalline oxides

Nanomaterials are systems that contain particles with one dimension in the nanometre regime. Currently there is intense interest from biologists, chemists, physicists and engineers in the application of these materials, so-called nanotechnology, which is sometimes referred to as ‘the next industrial revolution’ [26]. The reason for the interest is the unusual properties, very often with useful applications, that are exhibited by these materials when compared to their bulk counterparts [27–35]. Here we will focus on rather simple ionic solids with dimensions predominantly less than 100 nm, materials now known as nanoionics. In these systems the origin of the unusual properties is twofold; (i) the fact that the dimensions of the particles approach, or become smaller than, the critical length for certain phenomena (e.g. the de Broglie wavelength for the electron, the mean free path of excitons, the distance required to form a Frank–Reed dislocation loop, thickness of the space-charge layer, etc.) and (ii) surface effects dominate the thermodynamics and energetics of the particles (e.g. crystal structure, surface morphology, reactivity, etc.). The second factor can lead to nanocrystals adopting different morphologies to bulk crystals with different exposed lattice planes leading to an extraordinary surface chemistry [36,37] and catalytic activity [38,39]. The importance of surfaces and boundaries in nanocrystalline systems is demonstrated in Fig. 4, which shows the fraction of atoms in these regions as a function of grain size.

Unusually rapid diffusion of atoms in nanocrystalline metals was recognised in early studies [34,40,41]. In turn this will affect a wide range of physical and chemical properties of which the enhanced plasticity of metal and ceramic compacts has attracted considerable attention [42]. Fast diffusion in ionic and semi-ionic solids has important technological implications for greatly improved materials in a range of applications such as electrolytes and electrodes for batteries and fuel cells and chemical sensors [43–48]. As a result there has been considerable effort to quantify the extent of the diffusion enhancement and to determine the nature of the fundamental processes involved. However, until relatively recently, surveys of the experimental data of the diffusion in nanoionic materials showed that the results were not consistent, with variations even for the same system [49]. A breakthrough came with the work of Maier and co-workers who provided unambiguous experimental evidence of enhanced diffusion in layered nanocrystalline films of BaF₂ and CaF₂ [50]. Nevertheless, as recent articles have stressed [51,52], the volume of reliable experimental information is still sparse and the detailed understanding of the phenomenon is poor. A major problem is in the preparation of reliable samples and that the properties are very dependent on the synthetic route.

The microstructure is the key to the properties of nanocrystalline materials. It was seen earlier that simple geometric considerations lead to the conclusion that a large fraction of the atoms in a nanocrystal is in the surface (see Fig. 4). However, the crucial questions are what is the nature of the surface, in terms of the level of atomic order, and what is the structure of the interface between grains? In both cases, two extreme possibilities have been considered. One extreme is that there is extensive disorder in an interface that is several atoms in width. In some of the early work on nanocrystals this was intuitively assumed to be the case and the interfaces were referred to as ‘gas-like’ or ‘liquid-like’. This structure would clearly account for rapid diffusion in nanocrystalline samples. The alternative view is that the interface is similar to a grain boundary in normal bulk materials. In this case the interfaces would exhibit usual behaviour, although they would be present in unusually large number. High resolution TEM can provide the microstructural details. TEM measurements on nanocrystalline ceria showed that the grains had a high degree of perfection and were separated by sharp boundaries [53]. Unfortunately the data from HRTEM studies are still relatively sparse and other structural techniques had to be used to explore the microstructure, such as electron diffraction [54], positron annihilation spectroscopy [55] and XAS measurements [56–58].

In principle, for a nanocrystalline sample the XAS signal could be attenuated for two reasons: (i) the particle is so small that the average coordination numbers for the neighbouring shells are reduced or (ii) there is sufficient disorder in the sample (e.g. at the interfaces) that the Debye–Waller factors are increased. At first sight it would appear that XAS has little to offer as a microstructural probe, however, for (i) to be operative the particle size has to be very small and typically less than 5 nm. Thus XAS can offer a probe of disorder in the interfaces of nanocrystals. However, the results have been very confusing and the subject of much argument. The XAS data for ZrO₂ represent a typical example. There have been several XAS studies of this system, which claim evidence for disordered interfaces in nanocrystalline samples, i.e. an attenuation of the XAS for the Zr–Zr correlation [59,60]. However, similar measurements on carefully prepared films, with particle sizes down to 6 nm found that the EXAFS was indistinguishable from the bulk [61] and great care was taken to ensure that all hydroxyl species were removed from the sample. Other studies of nanocrystalline materials have shown that the preparation methods strongly affect the XAS data [62–66].

In recent years we have been using XAS measurements to study the effect of the preparation route of nanocrystalline oxides on their microstructure. We have been using mainly two synthetic routes, namely sol–gel methods and high-energy ball milling. Sol–gel procedures have been used for many years to produce oxides and ceramics offering control over the structure and composition at the molecular level [67,68]. The usual procedure is to subject a metal alkoxide, M(OR)_x to controlled hydrolysis, leading to replacement of the OR group by OH. This leads to the formation of a sol, very small colloidal particles, which then condense to form a gel, an inter-connected network. The gel is then dried and the final product can be either oxide (as in the case of silicon tetraethyl orthosilicate) or hydroxide (zirconium isopropoxide) or a mixed methoxy-hydroxide (as in the case of magnesium methoxide). Thus the final step in the formation of the oxide is calcination at high temperature. This step is difficult to control and presents two major problems, as exemplified by recent work on ZrO₂ [22]. If the calcining temperature is too low then all of the residual OH groups may not be completely removed from the material. If the calcination temperature is too high then the particles will grow and the nanocrystallites will be lost. The surface energy of nanocrystals is such that relatively moderate temperatures (∼400 °C in the case of most oxides) will lead to measurable grain growth over the period of a few minutes [63].

An apparently completely general method of producing all forms of nanocrystals is by mechanical attrition. This involves taking bulk material and reducing the grain size in a high-energy ball mill [69–72]. The advantages of ball milling are the fact that almost every material is accessible, that large amounts can be produced and that the average grain size can easily be varied by choice of the milling time. In addition, it is possible to produce materials in situ in the ball mill by double decomposition reactions [73,74]. This method is therefore useful when many different materials are to be compared. One disadvantage of ball milling is that abrasion of the milling media may occur. This has to be minimized by choosing appropriate materials for the milling vial and balls. A further disadvantage is that the milling can produce amorphous debris, to the extent that recent work on ball-milled Al₂O₃ indicated that the sample consisted of nanocrystalline grains embedded in amorphous material [75].

Zirconium oxide, ZrO₂, is one of the most studied materials due to its applications as an engineering ceramic and as an oxygen ion conductor for solid oxide fuel cells. It has also been the subject of numerous investigations as a nanoionic material due to the apparent ease of preparation. We use the term ‘apparent’ as the most common procedure of heating the hydroxide at modest temperature (500 °C) for a short period (60 min) has been reported as a means of producing 10 nm particle sizes of the oxide [76]. Later work [62] has shown that this procedure does not produce the pure oxide but material that contains a considerable fraction of hydroxyl species, sufficient to invalidate any XAS studies. A polymer precursor route can produce high quality thin films of nanocrystalline ZrO₂. XAS studies [57,58] of these samples have shown that with particle sizes as small as 6 nm there is no attenuation of the XAS signal and the interfaces between crystallites are similar to grain boundaries in bulk materials. We have recently made new XAS measurements on powdered samples of pure nanocrystalline zirconia prepared by sol–gel procedures. We overcame the problem of grain growth during preparation by using silica, 15% by weight, to pin the grains [77]. The preparation involved the addition of TEOS to zirconium alkoxide prior to the hydrolysis stage. The ZrO₂ sample was in the tetragonal phase with a grain size of 10 nm, even after calcining at 1000 °C. XAS spectra for the nanocrystalline ZrO₂ were collected at a series of temperatures down to 17 K. The measurements were undertaken at low temperatures to remove any ambiguity about the possibility of the contributions from dynamic and static disorders to the attenuation of the XAS signal. Parallel data were also collected for a bulk sample of tetragonal zirconia that was prepared by the polymer precursor method with the addition of 8 mol% yttria and calcined at 1000 °C. The data for the two samples were virtually identical at all temperatures and only those at 17 K are shown in Fig. 5. This conclusively proves that the levels of both dynamic and static disorders in sol–gel prepared nanocrystalline ZrO₂ are virtually the same as in the bulk material. Hence the interfaces between the grains are like grain boundaries in normal bulk samples.

Parallel studies were also made of ball-milled ZrO₂ with a particle size of 13 nm. X-ray diffraction showed that the sample was monoclinic [55]. For comparison data were also collected for a bulk, monoclinic sample (particle size 110 nm). The Fourier transforms of the spectra collected at 17 K are shown in Fig. 6. There is a strong attenuation of the EXAFS signal for the ball-milled sample, represented by a dramatic reduction in the magnitude of the peak at ∼3.5 Å, the Zr–Zr correlations. At this measurement temperature the differences can only be due to differences in the level of static disorder in the samples. We interpret this as the formation of significant levels of amorphous material in the ball-milled samples, a feature of this method of sample preparation [75,78].

The report of unusually high lithium-ion diffusion in ball-milled samples of LiNbO₃ [79] was intriguing, as this is not normally regarded as a particularly good ionic conductor. We therefore undertook Nb K-edge XAS studies of nanocrystalline LiNbO₃ prepared by sol–gel methods from alkoxides and ball-milled LiNbO₃ [80]. The results are shown in the form of the Fourier transforms in Fig. 7. The sol–gel sample (particle size 21 nm) shows a transform which is very similar to bulk material, as would be expected as the particles are relatively large and the average CN will not be reduced significantly. In contrast the ball-milled sample (particle size 50 nm) exhibits considerably reduced peaks, even for the first peak, the Nb–O correlation. We have interpreted this as due to the formation of a large fraction of amorphous material in the ball-milled sample, which we roughly estimate to be of the order of 50%. It is worth noting that these EXAFS findings were confirmed in a very recent study of lithium niobate by Heitjans and co-workers [81]. In addition they clearly showed the presence of amorphous material in the ball-milled sample by TEM measurements.

In summary, XAS measurements have clearly demonstrated that the preparation route has a profound effect on the microstructure of nanocrystalline oxides; the softer sol–gel approach produces highly ordered samples, whereas high-energy ball milling produces highly disordered samples with a high fraction of amorphous material. The rapid ionic transport observed in ball-milled samples is due to migration through the amorphous regions. However, it is lost on heating to moderate temperatures as the amorphous regions crystallise.

3.2 Solid-state electrochemical reactions

Recently, it was demonstrated that some nanocrystalline binary oxides, such as CoO, NiO and Co₃O₄ are interesting candidates for the negative electrode of lithium-ion batteries when fully reduced by discharge to ca. 0 V vs. Li [82–84]. The electrochemical reduction of the transition elements to the metallic state takes place simultaneously with the formation of lithium oxide. For cobalt oxides, the partial reversibility of the reduction process was ascribed to the nanosized Co particles dispersed in the Li₂O matrix that were produced during reduction, which possess an enhanced reactivity. In addition, the complex Li₂O + Co system is surrounded by a solid electrolyte interface. Similarly, a good performance was found for NiO and FeO, in which reversible capacities close to 700 mA h g⁻¹ were reported with high cycling efficiency [82]. There is considerable interest in these systems for their technological applications and there has been a focus on the details of the electrochemical processes that occur on charge/discharge cycles. Cycling involves the reduction/oxidation of the transition metal cation and this can be readily monitored in XANES experiments. The measurements can be made on prototype lithium batteries by in situ experiments due to the transparency to hard X-rays of lithium, the organic electrolytes and the plastic containers/windows [85–88]. Thus in mixed metal systems, such as LiNi_1/3Co_1/3Mn_1/3O₂, it is possible to monitor the oxidation state of each transition metal cation in the system during charge/discharge [85].

As part of the ALISTORE project we have studied a number of battery systems using XAS and a good example of this work is the investigation of lithium insertion/extraction in NiCo₂O₄ spinel oxide [89]. The use of the NiCo₂O₄ spinel oxide as the active electrode material in lithium and sodium cells yielded a reversible capacity of ca. 884 mA h g⁻¹ vs. lithium, while reversible cell capacities close to 200 and 300 mA h g⁻¹ were observed for sodium and sodium ion cells, respectively [19]. A complete loss of long-range order of the transition metal compounds takes place during the first discharge. By analogy with Li/NiO and Li/Co₃O₄, a reversible reaction of sodium with the spinel oxide in which Na₂O and the metals are formed was suggested.

The electrodes for the electrochemical experiments were nanocrystalline NiCo₂O₄ as the active material (85%), mixed with PVDF (5%) and SP carbon black (10%). The mixture was dispersed and then the solvent evaporated by heating at 120 °C under vacuum for 2 h. The resulting composite electrode was supported on a copper foil and used as the positive electrode in Li- and Na-anode cells. The electrolyte was 1 M LiPF₆ or 1 M NaClO₄ dissolved in EC:DEC or EC:DMC solvent mixture, respectively. Used electrodes were kept inside a thermo-sealed bag inside a high integrity dry-box and the bag was directly transferred to the XAS station. In addition to the electrode material XAS spectra were collected for Ni metal, NiO, Co metal, CoO, Co₃O₄, NiFe₂O₄ and LiCoO₂ as reference materials for the XANES analysis. The experiments could have been performed in situ, however, the XAS experiments would have taken a lot longer as the system would take time to equilibrate after changes in cell voltage.

Fig. 8 shows the normalised Co K-edge XANES spectra of the as-prepared and used electrodes. For the starting sample, the spectrum is comparable to the literature [90]. A very weak absorption pre-peak at ca. −6 eV from the Co K-edge of cobalt (7710 eV) is visible. This absorption represents the transition of the 1s electron to an unoccupied e_g orbital of cobalt ions with a low-spin electronic configuration. This is an electric dipole forbidden transition in an ideal octahedral symmetry, while the non-centrosymmetric tetrahedral environment allows its observation. Thus the pre-edge can be ascribed to the presence of tetrahedral cobalt in the structure of NiCo₂O₄, which is commonly described as an inverse spinel [91]. The main absorption peak at +12 eV is ascribable to the transition of a 1s core electron to an unoccupied 4p bound state with t_1g symmetry, and thus is electric dipole-allowed for both tetrahedrally and octahedrally co-ordinated cobalt atoms. Interesting conclusions can be extracted by comparison of NiCo₂O₄ and Co₃O₄ XANES. They are virtually identical with a very slight shift. This suggests that the Co oxidation states are the same in both, i.e. 8/3 or 2.667. Hence the charge on Ni is also 2.667, i.e. there is probably some Ni³⁺ present.

The changes in the oxidation state of the absorber atoms during cell operation are particularly well followed by XANES and provide a deeper insight than diffraction techniques. The changes in the Co K-edge XANES spectra of the NiCo₂O₄ electrodes (Fig. 8) evidence the displacement of the edge to lower energies from the very beginning of the discharge process, which is indicative of an immediate cobalt reduction process in NiCo₂O₄, that leads to metallic cobalt on approaching to 0 V. In addition, Fig. 8 (top) reveals a progressive loss of signal intensity. This behaviour can be related to the loss of p character in the 1s → 4p transition, due to d–p mixing, as a consequence of the conversion to metal cobalt particles, absorbing X-rays at the lower energy values.

The XANES spectra obtained after recharge of the lithium cells using NiCo₂O₄ electrodes are shown in Fig. 8 (bottom). The reversibility of the reduction process is evidenced after lithium extraction up to 1.5 V by the fact that the spectra recover their shape partially. However, the maximum oxidation state of cobalt atoms in the spinel oxide is not fully recovered, as shown by the position of the edge. This behaviour agrees with the non-recovered irreversible capacity. Moreover, two steps can be distinguished during the charging process in Fig. 8. From 0 to ca. 0.8 V, the XANES spectra do not alter significantly. Probably in that voltage interval, which coincides with the extraction of ca. 0.5 F mol⁻¹ of NiCo₂O₄, the changes are mostly related to the loss of the polymer-gel film and/or changes in the oxidation state of nickel. After charging to 1.5–3.0 V, some differences from the starting NiCo₂O₄ are observed. Thus, the pre-peak at lower energy is more clearly observed than in NiCo₂O₄. Also, the most intense peak is located at lower energy. Thus, the structure of the oxidised material differs from the initial spinel structure and less symmetric environments of cobalt are expected. Moreover, from the shift of the absorption edge in the XANES spectra, the initial Co^2.667+ average oxidation state is not fully recovered in the re-oxidised material.

The changes in the oxidation state of the nickel atoms during electrochemical reactions were also successfully followed by XANES [89]. The EXAFS results for the fully lithiated spinel, NiCo₂O₄ + 9Li are of particular interest. The results of the best fit show that the electrode has clearly been converted to metal in these samples. This can be seen qualitatively in the Fourier transforms (FT) shown in Fig. 9. The plots of the electrode are considerably attenuated compared to the metal foil. There are two possible causes of this: small and disordered particles. Firstly the nanoparticles were extremely small and there was a much reduced coordination number. However, a detailed fitting showed that the particles would have to be extremely small, ∼1 nm, for this to be true. We note that the first peak in the FT is reduced and this is really unusual. The second possibility is that the particles are very disordered. This would not be surprising as they are being formed in a disordered mixture of lithium and lithium oxide, and it will be a Co/Ni alloy particle and not pure metal. We have subjected the EXAFS for NiCo₂O₄ + 9Li to further analysis, by fitting the first shell only and holding the Debye–Waller factor fixed at the foil value. In other words we are assuming that the reduction in the EXAFS peak is only due to particle size and that there is no disorder present. This gives a coordination number for Co as ∼5 and for Ni ∼4 (for fcc this should be 12). Using simple geometry arguments this puts the size of the particles around 1 nm. This is not unreasonable as it will be ∼1 nm in the smallest dimension. Such an ultrafine particle size, not usually found in the literature, maybe connected with the Li₂O matrix. However, one can speculate that a high level of disorder in metal particles could also contribute to peak attenuation. Lithium oxide exerts a protection against coalescence of the metal particles, which is not present in cobalt or bimetallic samples prepared by other means [92–94]. Also, a thin cobalt oxide (CoO) passivating layer around the metal core, as found in the products using macromolecular surfactants [94], is not seen in the electrode working at ca. 0 V vs. Li, as there is no contact with air, and this could be considered as an indirect method of preparation of metal oxide-free metal nanoparticles.

3.3 Plastically crystalline electrolytes

Plastic crystals are a class of molecular materials [95,96] that consist of roughly spherical or globular molecules; thus the simplest examples are methane (CH₄) and carbon tetrachloride (CCl₄) and more complex systems are the bridged cyclic terpenoids, such as camphor and camphene. The materials are characterized by a solid-state phase transition or transitions. In the lowest temperature non-plastic phase the materials are typical organic solids with low crystallographic symmetries and are mechanically brittle and friable. In the solid phase prior to melting the solids are soft, hence the name plastic crystals, and show relatively rapid translational diffusion [96]. In the plastic phase the molecules are undergoing rapid endospherical reorientation and are often referred to as rotator phase solids. In addition, the plastic phases tend to be of high symmetry, either fcc, hcp or bcc. On melting there are relatively small increases in the entropy as the only gain is translational freedom. Two types of plastically crystalline electrolyte are possible. One type is when the ion or ions of the salt are exhibiting plastic behaviour, i.e. one or both ions are undergoing rapid rotation, the conductivity being improved by doping with a salt [20,97]. The second is when a plastic solid with a high dielectric constant dissolves a salt [21,98]. Both types can have solid-state conductivities at room temperature in excess of 10⁻³ S cm⁻¹, and coupled with appropriate mechanical properties, have potential applications as the electrolytes in solid-state batteries. They offer an alternative to solid polymer electrolytes as the conductivities are higher and the cation transport numbers appear higher.

We have just initiated a programme of XAS studies on plastically crystalline electrolytes and the results presented here are preliminary [99]. We have focused on electrolytes based on ionic salts dissolved in succinonitrile, a readily available material which can be easily purified by zone-refining [100]. In the pure material the plastic phase extends from 238 to 335 K and the dielectric constant is 55, i.e. higher than that for many organic solvents (e.g. acetonitrile has a dielectric constant of 36.6). Succinonitrile will dissolve a wide range of salts and in the non-plastic phase the ionic conductivity is low, typically ∼10⁻⁶ S cm⁻¹. On heating the conductivity of the solutions rapidly rises to 10⁻³–10⁻⁴ S cm⁻¹ at 300 K and then there is a further less pronounced increase up to the melting point [21]. This indicates a link between the molecular motion of the organic molecule, rotation and possibly translational diffusion, with the migration of the ions.

Our aim is to determine the local structure around the ions and monitor changes with temperature. The first system that we investigated was 1 mol% LiBr doped succinonitrile, where we collected Br K-edge XAS data. The measurements made in fluorescence mode and the data were of relatively poor quality. The Fourier transforms of the EXAFS are shown in Fig. 10. At 200 K, i.e. in the non-plastic phase, there is a main peak around 3 Å, corresponding to atoms of the succinonitrile molecule coordinating to the bromide ion. As the temperature was increased the spectra became attenuated and the peak in the Fourier transform is diminished. On melting the EXAFS oscillations were very weak and hardly discernable from the noise. These results are clearly consistent with increasing disorder around the bromide ion with increasing temperature and an increasing mobility of the ions in the system.

High cation conductivity for electrolytes is important for battery applications. Therefore, we examined salts dissolved in succinonitrile that had easily accessible cations, namely Cu(I) and Cu(II) triflate and Cu(I) and Cu(II) bistrifluoromethanesulfonamide (TFSI), and for which there were conductivity measurements [21]. We collected Cu K-edge XAS spectra for all four systems using transmission mode over a range of temperatures and doping concentrations. The general features of the EXAFS and their temperature behaviour were the same for all four systems. The Fourier transforms of the 1 mol% Cu(I) triflate in succinonitrile are shown in Fig. 11. There are three distinct peaks at 2, 3 and 4.5 Å and they show only minor reduction on heating from the non-plastic phase through to the melting point. The peaks were tentatively modelled as shells of N, C and Cu atoms. These data suggest that there is a high degree of order around the cation, even though the succinonitrile molecules are rotating, and appears inconsistent with the observed conductivity. Presently we have no firm explanation of the relation between the EXAFS and the ionic conductivity. One possibility is that the majority of the Cu ions are in locally ordered regions (EXAFS will monitor the average local environment of all the Cu ions in the system) and that a small fraction is mobile and contributes to the conductivity. A second possibility is that the Cu ions are migrating with a tightly bound coordination shell of succinonitrile molecules, similar to a hydrated ion in aqueous solution. However, this seems unlikely as the overall size of the mobile species would be very large in this case. Currently we are devising new experiments to resolve this dilemma of high ionic conduction and highly ordered structure.

3.4 Dopants in a host matrix

The location of the sites occupied by dopants in materials is a long-standing application of XAS, particularly for oxides and ionic crystals [101]. In cases where the dopant concentration is low this is fairly straightforward and involves only a consideration of substitutional and interstitial sites. A comparison of the dopant and host XAS can resolve the problem, as in the case of cation doped tin oxide [63]. At high concentration quite complex defect clusters can form and the XAS measurements require other inputs, such as computer simulations or diffraction data, to resolve the structures. An early example of this approach was the work on defect clusters in rare earth doped calcium fluoride [102]. Recently, we have been using XAS studies to locate dopant sites in nanocrystalline anatase titania ceramics with particle sizes of 12–30 nm [22–24].

Titanium dioxide is a model binary oxide for fundamental studies that presents many of the exciting properties of electroceramics, including variations of oxygen stoichiometry and ranges of ionic or electronic conductivity [103]. It can also be readily doped with acceptor or donor dopants, i.e. solutes with negative or positive effective charge, respectively. TiO₂ has a multitude of applications: it has been a grand-scale white pigment for many years and is currently under investigation for advanced technological and environmental applications, such as solid-state gas sensors [104], photovoltaic cells [105], and photocatalysis [106]. The metastable anatase modification presents better characteristics for photoelectrochemistry.

In our first studies we used XAS to study Nb and Zn doped TiO₂ [22]. The data showed that the Nb⁵⁺ ion substituted for the Ti⁴⁺ ion in the lattice at concentrations of 0.1 and 1 at%. This might have been expected due to the similarity of the ionic radii of Nb⁵⁺ (0.69 Å) and Ti⁴⁺ (0.68 Å). The ionic radius of Zn²⁺ (0.74 Å) is similar to that for Ti⁴⁺ but was found only to be soluble, dissolving substitutionally for Ti⁴⁺, at concentrations of 0.1 at% or less. This emphasises the role of the ion charge (rather than ion size) and that it is the electrostatic effects that determine the segregation of impurities from the host lattice.

Recently we studied the Sn K-edge XAS of 1 at% Sn doped nanoceramic TiO₂ [23]. In this case we observed the expected substitutional doping on the Ti⁴⁺ site. The Fourier transform of the fluorescence XAS spectrum is shown in Fig. 12. This is very similar to the Fourier transform of the Ti K-edge EXAFS of TiO₂ with the peaks at ∼2, 3 and 4 Å representing shells of O, Ti and O atoms, respectively. However, we were able to take the analysis further by comparison with DFT computer simulations of the system using the CASTEP code. The structural parameters are listed in Table 1 for TiO₂ and Sn doped TiO₂ from the various techniques. Firstly we see that the slightly large Sn⁴⁺ ion slightly dilates the lattice. Secondly, there is an excellent agreement for the radial distances from the EXAFS and the simulations, especially for the first two shells.

We have made a similar combined EXAFS and DFT study of Zr⁴⁺ doped TiO₂ [23]. The Fourier transform of the Zr K-edge XAS spectrum is shown in Fig. 13. The transform is very similar to that for Sn²⁺doped TiO₂ and is best fitted by substitutional doping. Also shown in Fig. 13 is the Fourier transform of the Y K-edge of a 1 at% Y doped TiO₂ nanoceramic. In this case there is only one broad peak and is a clear indication that the larger size of the Y³⁺ ion and the different ionic charge when compared to Ti⁴⁺ leads to a segregation of the dopant into the surface regions of the host. Similar effects were found in cation doped nanocrystalline SnO₂ samples when they were heated [63].