Marzia
Fantauzzi
*a,
Fausto
Secci
ac,
Marco
Sanna Angotzi
ab,
Cristiana
Passiu
c,
Carla
Cannas
ab and
Antonella
Rossi
ab
aDipartimento di Scienze Chimiche e Geologiche, Università degli Studi di Cagliari, Campus di Monserrato S.S. 554, Italy. E-mail: fantauzzi@unica.it
bINSTM, UdR Cagliari, Italy
cDepartment of Materials, ETH Zürich, Vladimir-Prelog-Weg 5, Zurich, Switzerland
First published on 18th June 2019
Nanostructured spinel cobalt ferrite samples having crystallite size ranging between 5.6 and 14.1 nm were characterized by X-ray photoelectron spectroscopy and X-ray induced Auger electron spectroscopy in order to determine the chemical state of the elements, the iron/cobalt ratio and the cation distribution within tetrahedral and octahedral sites. The presence of size-dependent trends in the binding energy of the main photoelectron peaks and in the kinetic energy of the X-ray induced O KLL signal was also investigated. The results showed that iron is present as FeIII and cobalt is present as CoII. The iron/cobalt ratio determined by XPS ranges between 1.8 and 1.9 and it is in very good agreement, within experimental uncertainty, with the expected 2:
1 ratio. The percentage of Fe in octahedral sites ranges between 62% and 64% for all samples. The kinetic energy of the O KLL signals increases with crystallite size. These results are explained in terms of changes in the ionicity of the metal–oxygen bonds. The results of this investigation highlight how the XPS technique represents a powerful tool to investigate the composition, the chemical state and inversion degree of cobalt spinel ferrites, contributing to the comprehension of their properties.
The chemical and physical properties are directly correlated to the composition and to the structure. Bulk cobalt ferrite shows an inverse spinel structure: all CoII cations occupy the octahedral sites, and the FeIII cations are equally distributed in Td and Oh sites (γ = 1, γ is the inversion degree or the fraction of divalent cations in octahedral sites). The coupling of the magnetic moments associated with the ions in Td and Oh sites justifies the ferrimagnetic behavior of the cobalt ferrite below 860 K.19 Conversely, when prepared in form of nanostructured material CoII and FeIII are randomly distributed (γ = 0.66). In the literature, different techniques (57Fe-Mössbauer spectroscopy, EXAFS, neutron diffraction) have been applied in order to determine the inversion degree of nanostructured CoFe2O4 synthesized with various approaches and the values fall in the 0.68–0.76 range.10,20–22
It is clear that among all the factors that mainly influence the properties of the ferrites, besides the size and the shape of the particles, the stoichiometry and the cation distribution play a crucial role. Therefore, the knowledge of the chemical composition, of the structure and of the properties is the key to design a material for a specific application.
In this context, X-ray photoelectron spectroscopy is acknowledged to be a state-of-art tool for providing elemental, chemical state and quantitative information also when investigating nanostructured materials.23 Due to the above mentioned features it is often used as a complementary technique in papers dealing with nanostructured ferrites.24–26
The chemical state analysis of XPS data usually starts with the curve fitting of the most intense photoelectron signal. For the transition metals this procedure is not straightforward being the line shapes of the 2p peaks complicated by peak asymmetry, multiplet splitting, by the presence of shake-up satellites and plasmon loss structures. Furthermore, if iron and cobalt are simultaneously present in a sample, as in CoFe2O4, the photoelectron Fe 2p and Co 2p peaks overlap the Auger Co L3M45M45 and Fe L3M45M45 respectively, when the Al Kα X-ray source is used.
Among the scientific literature on ferrites, only few papers provide details on the curve fitting procedure.27,28 Some of the papers report XPS spectra, tables listing the binding energy values and the quantitative composition without details on the spectra processing.29–31 The present work focuses on the use of XPS for the characterization of nanostructured cobalt ferrites samples with crystallite size ranging between 5.6 and 14.1 nm.
In order to determine the iron and cobalt chemical states, their most intense photoelectron signals were resolved in components following two different approaches. The first one is based on the results of the calculation of the multiplet structure expected for 2p core levels of iron and cobalt free ions, proposed by Gupta and Sen32 and the strategy developed in ref. 25 for first row transition metals oxides and hydroxides. This model was successfully applied for curve fitting the iron signals in various applications ranging from the environmental33 to the biological ones.34 According to the curve fitting strategy proposed in literature27 the iron of a nickel ferrite is composed of four components that were chosen as starting point for curve fitting the spectra in this paper.
The magnetic properties of the ferrites are influenced by disorder in the cation distribution between octahedral and tetrahedral sites, which can influence saturation magnetization, exchange couplings, and ferrimagnetic ordering temperatures;35 in order to determine the spinel inversion degree of the cobalt ferrite, a second curve fitting strategy, based on that proposed by Aghavniana and co-workers,36 has been adopted in this paper.
Since it is known that the binding energy of photoelectron lines is a size dependent parameter,37–40 in this paper it is also evaluated if a systematic chemical shift of the binding energy of iron Fe 2p3/2, cobalt Co 2p3/2, oxygen O 1s and of the kinetic energy of the X-ray induced O KLL signal is observable, changing the cobalt ferrite crystallite size. Since initial and final state effects influence the chemical shift of both photoelectron and Auger lines, an explanation of the observed trends is provided.
Table S1† summarises the synthesis conditions adopted in this work. The first column reports the sample notation.
X-ray diffraction, using a PANalytical X'Pert PRO with Cu Kα radiation (1.5418 Å), secondary monochromator, and a PIXcel position-sensitive detector was used for characterizing the crystalline phases. Calibration of the peak position and instrumental width was carried out using powder LaB6 from NIST. The hexane dispersions were dried on a glass plate and measured in the angular range of 10–90° with a step of 0.039°. The XRD patterns aiming to the identification of the crystalline phase, after background subtraction, were processed using PANalytical X'Pert HighScore software. The most intense X-ray peaks [(220), (311), (400), (422), (511), (440)] were fitted with Origin Software by PseudoVoigt function, using a 1:
1 Gaussian/Lorentzian ratio. In reciprocal space for the cubic system, the reciprocal of the interplanar distance q is defined as
In this equation, the value 1.10 takes into account the information on size distribution and shape of the particles and the model derived from Repko et al.41 based on the fitting of the peak of Fourier-transformed solid spheres with volume-weighted log-normal size distribution with σ = 0.18, which are almost Gaussian. For each sample, the crystallite size is given as average of the values calculated on the most intense peak and the associated error is the standard deviation.
AFM analysis was performed on spin-coated silicon wafer-samples to observe the changes occurring on their surfaces when varying the spin-coating conditions.
To obtain a uniform layer, the ideal suspension concentration was 4 mg cm−3. 100 μL of each solution were deposited on each silicon wafer and the spin-coating conditions were 4000 rpm for one minute.
AFM images were acquired using an Icon Dimension (Bruker, Camarillo, California, USA) with a silicon cantilever (BudgetSensors, Bulgaria) with a spring constant of about 26 N m−1 (manufacturer's value).
XPS measurements were carried out using a Sigma2 XPS and a Thetaprobe manufactured by ThermoFisher Scientific Inc., East Grinstead, UK. Sigma 2 is equipped with an Al/Mg Kα twin source. A non-monochromatic Mg Kα source (1253.6 eV) operated at 200 W was used in order to eliminate the overlap between Fe 2p and Co LMM peaks and between Co 2p and Fe LMM peaks, which is observed if an Al Kα source is used. The residual pressure during the analysis was always lower than 10−6 to 10−7 Pa. The emitted electrons are collected by an Alpha 110 hemispherical analyzer, and a multichannel detector consisting of seven channeltrons is used. The analyzer was operated in fixed-analyzer transmission mode setting the pass energy at 25 eV for collecting the high-resolution spectra and at 50 eV for the survey spectra; a step size of 0.05 eV and 1 eV was respectively used. The instrument was used in large-area mode (diameter of the analyzed area: 8 mm). The energy resolution in large-area XPS mode, expressed as the full width at half maximum (FWHM) of the Ag 3d5/2 signal was 1.1 eV; the intensity was 3.7 Mcps using an Al window less than 1 μm thick to limit the contribution of backscattered electrons and to reduce the contribution of Bremsstrahlung radiation.
The Thetaprobe spectrometer is equipped with a monochromatic Al Kα X-ray source; the residual pressure during the analysis was always lower than 10−7 Pa. Spectra were collected using a monochromatic beam (Al Kα1,2E=
1486.6 eV) operated at 4.7 mA and 15 kV (70 W), with a nominal spot size of 300 μm. The analyzer was operated in fixed analyzer transmission mode at 100 eV pass energy for collecting narrow scan spectra and at 200 eV for the survey spectra with a step size of 0.05 eV and 1 eV, respectively. The linearity of the binding energy scale of both spectrometers was monthly checked according to the literature43 and to ISO 15472:2010, using Cu, Au and Ag reference samples. Data were acquired under computer control. Binding energy values were corrected by referencing to the aliphatic carbon at 285.0 eV. Thetaprobe was used to acquire the O 1s and O KLL spectra since the lower FWHM allowed obtaining a better-resolved O KLL signal.
Magnetic nanoparticles dispersed and dried on silicon wafer can be characterized by XPS taking care of checking the shape of the survey spectrum at the low kinetic energy side i.e. below 200 eV. This part of the spectrum is reported to be affected in the presence of stray magnetic fields.44 No changes at the low kinetic energy values were observed during this investigation. Fe/Co ratios were calculated correcting the experimental areas for the relative sensitivity factors. The details are reported in the ESI.†
Fe 2p peaks (Fig. 1a) exhibit a doublet separated by 13.5 eV due to spin–orbit coupling. For the curve-fitting purpose only the Fe 2p3/2 peak was taken into account.
![]() | ||
Fig. 1 (a) Fe 2p and (b) Co 2p peaks of the cobalt ferrite samples following the curve-fitting procedure based on the multiplet splitting approach.27 The crystallite size is reported in the graph. X-ray source: Mg Kα. |
Fe 2p3/2 peaks were resolved in four components due to multiplet splitting phenomena, following what is reported in literature27 for NiFe2O4. The binding energy of the four components are listed in Table 1.
Crystallite size (nm) | Fe/Co ratio | Fe 2p I | Fe 2p II | Fe 2p III | Fe 2p IV | Co 2p I | Co 2p II | Co 2p III | Co 2p IV |
---|---|---|---|---|---|---|---|---|---|
5.6 (2) | 1.8 | 710.3 (1) | 711.3 (1) | 712.5 (1) | 713.9 (1) | 780.5 (1) | 782.6 (1) | 786.0 (1) | 787.0 (1) |
6.7 (1) | 1.9 | 710.2 (1) | 711.2 (1) | 712.4 (1) | 713.8 (1) | 780.5 (1) | 782.6 (1) | 786.0 (1) | 787.0 (1) |
8.8 (2) | 1.9 | 710.1 (1) | 711.1 (1) | 712.4 (1) | 713.8 (1) | 780.5 (1) | 782.6 (1) | 786.0 (1) | 787.0 (1) |
11.2 (6) | 1.9 | 710.1 (1) | 711.1 (1) | 712.4 (1) | 713.8 (1) | 780.5 (1) | 782.6 (1) | 786.0 (1) | 787.0 (1) |
14.1 (8) | 1.9 | 710.0 (1) | 711.0 (1) | 712.2 (1) | 713.6 (1) | 780.4 (1) | 782.5 (1) | 785.9 (1) | 786.9 (1) |
The curve-fitting parameters are shown in the ESI (Table S2†): the full width at half-maximum of the peak height and the area ratios were kept constant during curve synthesis, while the position of the components was let free to change. Despite this degree of freedom, the binding energy difference between the first peak (Fe 2p I) and the other peaks was reproducible, and it is reported in Table S2.† The binding energy of the signals (Table 1) is typical for FeIII27 and no evidence of FeII is observed. Fig. 1a and Table 1 show a small binding energy shift of the Fe 2p3/2 components with the crystallite size: the smaller the cobalt ferrites particles size, the higher the binding energy of the peak.
Co 2p peak is a doublet due to spin–orbit coupling (Fig. 1b); the binding energy separation between Co 2p3/2 and Co 2p1/2 was 16 eV and it allowed us to exclude the presence of CoIII, thus the experimental lines were curve-fitted starting from the fitting parameters of CoO reported in Biesinger et al.27 Also in this case, for curve-fitting purposes, only the Co 2p3/2 was taken into account (Fig. 1b). The shape of the Co 2p3/2 signal, which shows an intense satellite separated by 6.5 eV from the main most intense peak, is typical of CoII.27 No significant shifts were observed for the different samples (Table 1). The curve-fitting parameters are listed in Table S3† of the ESI.†
O 1s is a multicomponent peak (Fig. 2a): the most intense component is found at about 530 eV (Table 2) and it is ascribed to oxygen in the oxide lattice.27 The component at about 531.5 eV can be ascribed to both hydroxides and defective oxygen.27 The peak at the highest binding energy value might be due to water and/or to the oleic acid used as capping agent.27
![]() | ||
Fig. 2 (a) O 1s spectra and (b) O KLL spectra of the cobalt ferrite nanoparticles samples. The particle size is reported in the graph. X-ray source: monochromatic Al Kα. |
Crystallite size (nm) | O1s oxide | O1s hydroxide or defective oxygen | O1s water | O KL23L23 | O KL1L23 | ΔKEOKLL | I KL23L23/IKL1L23 | α′a | Δα′b | ΔReac |
---|---|---|---|---|---|---|---|---|---|---|
a α′ = (BEO 1s + KEO KL23L23). b Δα′= (α′CoFe2O4 NPs − α′Fe2O3 taken at 1042 eV). c ΔRea = Δα′. | ||||||||||
5.6 (2) | 530.0 (1) | 531.5 (1) | 532.4 (1) | 511.2 (1) | 490.5 (1) | 20.7 (1) | 1.75 | 1041.2 (2) | −0.8 | −0.4 |
6.7 (1) | 530.0 (1) | 531.6 (1) | 532.6 (1) | 511.4 (1) | 490.8 (1) | 20.6 (1) | 1.78 | 1041.4 (2) | −0.6 | −0.3 |
8.8 (2) | 530.1 (1) | 531.5 (1) | 532.4 (1) | 511.5 (1) | 491.0 (1) | 20.5 (1) | 1.80 | 1041.6 (2) | −0.4 | −0.2 |
11.2 (6) | 530.1 (1) | 531.6 (1) | 532.3 (1) | 511.7 (1) | 491.3 (1) | 20.4 (1) | 1.89 | 1041.8 (2) | −0.2 | −0.1 |
14.1 (8) | 530.1 (1) | 531.5 (1) | 532.5 (1) | 511.9 (1) | 491.3 (1) | 20.4 (1) | 2.04 | 1042.0 (2) | 0 | 0 |
O KLL spectrum showed the presence of two prominent peaks, assigned to O KL23L23 and O KL1L23 (Fig. 2b),45 together with a component at about 477 eV assigned to the KL1L1 Auger transition45 and a small peak at 501 eV (peak s in Fig. 2b) that may be due to shake-up and shake-off contribution to the Auger peak.45
The kinetic energy values of the O KL23L23 components are reported in Table 2, together with KEO KL23L23 − KEO KL1L23 difference, which is related to the ionicity of the oxides, in agreement with literature.46 The identification of the three different oxygen species (oxide, hydroxides or defective oxygen, and water or oleic acid) by curve-fitting of the O KLL signals was not attempted since it was beyond the scope of this work. The O KLL signals were resolved in their main components in order to calculate the modified Auger parameter (α′) and to evaluate any possible size effect in the kinetic energy values of the peak maxima. The values of the modified Auger parameter47α′ = BEO1s + KEOKL23L23 are also reported in Table 2.
Iron/cobalt ratios of the CoFe2O4 samples are found to be close to 2:
1 (Table 1) for all the samples, in agreement with the ICP-OES (inductively coupled plasma–optical emission spectrometry) results obtained on CoFe2O4 samples synthesized by an identical synthesis route.42 To evaluate the ratio by XPS data, the Tougaard's background-subtracted Fe 2p and Co 2p peaks (both 2p3/2 and 2p1/2 components) were considered. Tougaard's background was chosen since it is less sensitive to the choice of the minimum and maximum of the selected energy range than the Shirley background, especially if large energy ranges are taken into account.48 In the present case the energy range was 35 eV for Co 2p and 25 eV for Fe 2p.
The curve-fitting strategy of Co 2p peaks proposed by other authors36 was thus followed in order to estimate the spinel inversion parameter from XPS data. The estimation strategy starts with distinguishing between CoII in octahedral sites (Oh) and CoII in tetrahedral (Td) sites on Co 2p peaks (Fig. 3). Both Co 2p3/2 and Co 2p1/2 are taken into account after Shirley background subtraction, in agreement with literature.36 Co 2p3/2 was curve-fitted with a peak located at 780.5 eV, which is ascribed to cobalt in Oh sites, with a peak at 782.9 eV due to CoII in Td sites, and with a satellite at 786.3 eV for all the samples (Table 3 and Fig. 3).
Crystallite size (nm) | Co 2p3/2 (Oh) | Co 2p3/2 (Td) | Co 2p3/2 SAT | ΔBE 2p3/2 − 2p1/2 (eV) | γ | % Fe (Oh) | |||
---|---|---|---|---|---|---|---|---|---|
BE (eV) | FWHM (eV) | BE (eV) | FWHM (eV) | BE (eV) | FWHM (eV) | ||||
5.6 (2) | 780.4 (1) | 2.6 (1) | 782.8 (1) | 2.6 (1) | 786.2 (1) | 4.9 (1) | 16.0 (1) | 0.76 | 62 |
6.7 (1) | 780.4 (1) | 2.6 (1) | 782.8 (1) | 2.6 (1) | 786.2 (1) | 4.9 (1) | 15.9 (1) | 0.75 | 62 |
8.8 (2) | 780.5 (1) | 2.6 (1) | 782.9 (1) | 2.6 (1) | 786.4 (1) | 5.0 (1) | 15.9 (1) | 0.76 | 62 |
11.2 (6) | 780.5 (1) | 2.5(1) | 783.0 (1) | 2.5(1) | 786.6 (1) | 4.7 (1) | 15.8 (1) | 0.75 | 63 |
14.1 (8) | 780.5(1) | 2.6 (1) | 782.9(1) | 2.6 (1) | 786.3 (1) | 5.0 (1) | 16.0 (1) | 0.72 | 64 |
Despite the larger number of oxygen atoms near the CoII in octahedral sites, its BE is lower than the BE of CoII in tetrahedral sites. This is explained in terms of effective charge on the cobalt atoms and of oxygen polarizability.51 The BE separation between CoII in Oh and Td sites is fixed.36 The area ratio between CoII in Oh and Td in Co 2p1/2 was constrained to be equal to their ratio in Co 2p3/2. Assuming a general formula: (FeyCo1−y)Td[Fe2−yCoy]OhO4 the percentages of area of the CoII components in Oh and Td sites allow the estimation of y, thus the fraction of cobalt in Oh sites. The percentage of iron in octahedral sites is thus calculated with the formula %Fe (Oh) = (y × 0.5 + (1 − y) × 1) × 100.36 The percentage of Fe in octahedral sites calculated following this approach for all the samples ranges between 62 and 64% (Table 3) and it is in very good agreement with the cation distribution formula obtained by 57Fe Mössbauer spectroscopy and for nanostructured CoFe2O4:10,20,21(FeIII0.7CoII0.3)[FeIII1.3CoII0.7]O4.
On the basis of these results it seems that this curve synthesis based on two symmetric peaks for distinguishing between differently coordinated iron and cobalt is effective to estimate the cation distribution.
Due to the presence of the oleic acid shell, which is estimated to be about 1 nm thick in agreement with Shard,52 the signals from cobalt ferrite are attenuated. The sampling depth of an electron travelling in a ferrite nanoparticle can be estimated by the formula 3λferritecosθ where λferrite is the inelastic mean free path determined for inorganic compounds, according to Seah and Dench,53 and θ is the emission angle. For iron Fe 2p, for example, the sampling depth was found to be 4.6 nm for electrons travelling in pure cobalt ferrite. Due to the presence of the 1 nm-thick oleic acid layer, the sampling depth is reduced and it can be estimated by the formula
, where t = 1 nm is the thickness of the organic layer and λC is the inelastic mean free path of an electron from Fe 2p travelling in an organic layer, determined according to Seah and Dench.53
The sampling depth for an Fe 2p electron starting from the cobalt ferrite nanoparticle and attenuated by the oleic acid layer is estimated to be 4.0 nm.
The agreement between the results obtained by XPS (cation distribution and the Fe/Co ratios) on the surface and those extracted by bulk techniques (57Fe Mössbauer spectroscopy and ICP-OES)42 supports a compositional and structural homogeneity throughout the nanoparticle, also for the larger NPs, for which only 1/3 of their volume is sampled by XPS.
It is known that the binding energy of photoelectron peaks can be strongly influenced by the particle size. In many papers dealing with metal nanoparticles a positive shift of the binding energy values is observed with decreasing particle size.38,54 In other papers a negative shift is observed for metal supported nanoparticles and it might be attributed to the interaction between the metal nanoparticles and the support rather than to an intrinsic particle size effect.37,38 The observed chemical shift, due to both size effects and to metal/support interactions, has been explained taking into account both initial- and final-state effects. The reduced coordination number of surface atoms, which is more significant for small NPs, is an initial-state effect; the electron exchange between the support and the NPs and the screening of the core hole on the photo-emitting atom (relaxation energy) are final-state effects and are influenced by particle size as well.38
Concerning the oxides, a positive shift of Cu 2p3/2 from CuO with decreasing nanoparticles size was observed and it was explained on the basis of the quantum-size effect.39 The shift toward higher BE for smaller NPs was interpreted with the increase of Cu–O bond ionicity. A study on the effect of particle size of ZnO showed an opposite trend in the binding energy of the Zn 2p signal: the smaller the particle the lower the binding energy.40 The authors explained this shift with a decrease of the number of oxygen atoms in nanocrystalline ZnO that reduces the charge transfer from Zn to oxygen and increases the shielding effect of the valence electrons in Zn ions, resulting in the decrease of the Zn 2p binding energy. In this paper only a small positive shift (0.3 eV) of the binding energy of the Fe 2p3/2 peak was observed, when decreasing the particle size from 14.1 to 5.6 nm (Table 1): these results seem to be in agreement with the results obtained on CuO.
If there is a size-dependent effect on the ionicity of the iron-involving bonds (Fe–O) the same effect should be observed on oxygen. For this reason, the effect of the particle size on O 1s and the O KLL signals was also evaluated (Table 2). The binding energy of the photoelectron O 1s signal does not show significant differences with the ferrite particle size, being the chemical shift between the smallest and the largest particles within the experimental uncertainty (0.1 eV). On the contrary, a trend in the kinetic energy of the most intense Auger peak, O KL23L23, is observed. Auger O KLL lines provide information not only on the chemical state of oxygen, but also on the ionicity of the oxygen-involving bonds and on their polarizabilities.46,55 The ΔKEOKLL difference between the kinetic energies of O KL23L23 and O KL1L23 and the ratio between their intensity were calculated (Table 2). Both these parameters are influenced by the ionicity of the M–O bond.46,56 A correlation was found between the ΔKEOKLL value and the charge on the oxygen atom (q) calculated after Pauling's electronegativity scale, following a previous study.55 The lower was q (lower ionicity) the higher was ΔKEOKLL. Similarly, a correlation between the IOKL23L23/IOKL1L23 intensity ratio and the ΔKEOKLL value was found, and it was observed that highly ionic oxides (MgO) showed higher values of the intensity ratios and lower values of ΔKEOKLL.46 They proposed a phenomenological model for oxygen Auger spectra that allows justifying this finding. The authors considered an oxygen ion having a charge q immersed in a dielectric oxide and proposed an equation that relates the ΔKEOKLL shift to the charge on the oxygen atom.46 In this paper ΔKEOKLL decreases with increasing crystallite size (Table 2), thus we can deduce that the charge on oxygen atom is higher and thus the ionicity of the M–O bond. Furthermore, the smallest particles (5.6 ± 0.2 nm) showed the lowest IOKL23L23/IOKL1L23 intensity ratio and the highest ΔKEOKLL value. This result suggests that the ionicity of the metal–oxygen bond decreases with the particle size. Further information can be obtained from the oxygen Auger parameter shifts. It is known that differences in the Auger parameter values (Δα′) reflect differences in the extra-atomic relaxation energies:57
Δα′ = ΔRea |
In this work Δα′ was calculated referring it to bulk Fe2O3. The values are reported in Table 2. Δα′ increases with the decrease of the particle size (Fig. 4) and it reaches the maximum value for the smallest particles (5.6 nm). This results in lower relaxation energies than those of bulk Fe2O3 (similar results are obtained if Co3O4 is taken into account). Lower values of relaxation energy indicate lower polarizabilities of the chemical environment: for the smallest particles, the polarizability of the electronic cloud towards the hole is less effective than in bulk oxides.
![]() | ||
Fig. 4 Oxygen chemical state plot. Data from this work are reported (black dots) together with data from literature55 (colored squares). |
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c9ra03488a |
This journal is © The Royal Society of Chemistry 2019 |