Method for estimating ionicities of oxides using O1s photoelectron spectra

Cohen¹ calculated the densities of states (DOS) for valence electrons in the perovskite oxide BaTiO₃ using density functional theory (DFT), and the calculated results indicated that the average valences of Ba, Ti and O are 2, 2.89 and −1.63, respectively. In other words, about 37% of the oxygen ions in BaTiO₃ were calculated to be monovalent, with the result that there must be O2p holes in the outer orbits of the monovalent oxygen ions. These O2p holes are expected to affect the physical properties of the materials, and it is therefore very important to study the average valence of oxygen anions in oxides.

It is well known that there are both ionic and covalent bonds in oxides, and the ionicity has been defined as the fraction of ionic bonds among all the ionic and covalent bonds.² In our opinion, the value of the ionicity (f_i) for the oxides can be understood as half the average valence (V_alO) of the oxygen anions, f_i = |V_alO|/2. Therefore, f_i = |V_alO|/2.000 = 0.815 for BaTiO₃ according to Cohen’s calculation. In 1932 and 1939, Pauling reported his ionicity investigation. In 1970, Phillips² improved the ionicity calculation and reviewed studies of the ionicity of simple compounds. He gave the ionicities of some II-VI compounds, for example, 0.602, 0.841, 0.913 and 0.926 for BeO, MgO, CaO and SrO, respectively. Thomas et al³ reported the ionicities of the dihalides MX₂ (M = Mn, Fe, Co, Ni) obtained using the dielectric theory of chemical bond, and found values ranging from f_i ≈ 0.72 for NiI₂ to f_i ≈ 0.80 for MnCl₂. On the basis of dielectric theory, Chelikowsky et al⁴ examined the relationship between spectroscopic crystal ionicity and the spatial structure of several II-VI and III-V compounds. Somewhat later, using a measure of the asymmetry of the valence charge distribution in A^NB^8−N compounds, García et al⁵ provided a first-principles mapping to Pauling’s ionicity scale for A^NB^8−N solids. Balamurugan et al⁶ examined the relationship between the ionicity of Cu₂O samples and their particle size using Cu2p X-ray photoelectron spectra. Arif et al⁷ reported their investigation of the ionicity factor based on the energy gap of semiconductors with a hexagonal structure, using density functional theory (DFT) with full-potential linear augmented plane wave theory. Ascarelli⁸ investigated ionicity of metallic oxide surfaces on metals using Auger spectroscopy measured by X-ray photoelectron spectra (XPS). We have, however, found that the values of the ionicity given by various authors show significant differences for a given compound, and that a general method for estimating ionicity is therefore required.

In 2000, Dupin et al⁹ proposed that a portion of the O ions in oxides are O¹⁻ ions on the basis of analyses of O1s spectra measured by XPS, and gave the average net charges (q_o) of the oxygen ions in several compounds; namely, −1.15, −1.18, −1.05, −1.78 and −1.85 for TiO₂, ZrO₂, CoO, CaO and SrO, respectively. For these oxides, f_i = |q_o|/2, and thus, for CaO and SrO, the values of f_i (0.890 and 0.925, respectively) are close to those (0.913 and 0.926) from Phillips’s calculation. Unfortunately, there are few other reports on using the O1s spectra to estimate ionicity.

In this paper, we use the O1s spectrum to estimate V_alO for BaTiO₃, and find a value (−1.55) which is close to the calculated value (−1.63) given by Cohen.¹ We then estimate the values of V_alO and f_i for several monoxides using the O1s spectra, and study the relationship between the values of f_i and the ionization energies.

Our samples included the perovskite oxide BaTiO₃, and the monoxides CaO, MnO, CoO, ZnO, NiO and CuO. All of these samples were obtained as commercial chemical powders.

The crystal structures of the samples were identified using an X’Pert Pro X-ray diffractometer with Cu K_α(λ = 1.5406 Å) radiation, at room temperature. The data were collected in the 2θ range 15-120° with a step size of 0.01671°. The X-ray diffraction data for all samples were analyzed using the X’Pert HighScore Plus software, and it was found that all samples had single phase crystal structures. This analysis also provided the crystal lattice parameters and crystallite sizes of the samples, as shown in Table I. Crystallite sizes were estimated using the X-ray diffraction data and the Scherrer Equation with the X’Pert HighScore Plus software, which provides reasonable estimates for sizes less than roughly 100 nm. In Table I, therefore, the sizes of the larger crystallites are listed as >100 nm.

XPS of the samples were measured using an X-ray photoelectron spectroscopy PHI5000 Versa Probe with monochromatic Al K_α radiation (1486.6 eV). The C1s binding energy (284.8 eV) of carbon contamination was used as a calibration to compensate for charging effects. The computer program XPSPEAK Version 4.1 was used to fit the narrow-scan spectra of the O1s peaks after Shirley-type background subtraction.¹⁰ Narrow-scan spectra of O1s peaks were simulated using the symmetric Gaussian–Lorentzian product function.

It can be seen from Table I that the average crystallite size for BaTiO₃ was greater than 100 nm so that surface effects are expected to be very weak. Fig. 1 shows the BaTiO₃ O1s spectrum obtained by fitting with Gaussian–Lorentzian functions. It may be seen that the O1s spectrum can be fitted using three peaks with binding energies (BEs) of approximately 528.5, 530.7 and 532.6 eV. According to the interpretation proposed by Dupin et al,⁹ the lower BE peak is assigned to O²⁻ ions, the middle BE peak to O¹⁻ ions, and the higher BE peak to O^Chem, chemically adsorbed oxygen on the surface. The fitting data for O1s spectrum for BaTiO₃ are shown in the Table II. From the relative peak areas of the main components, it can be deduced that the content ratio between O¹⁻ ions and O²⁻ ions is approximately equal to 0.418/0.509 = 0.45/0.55. Here, we neglect the chemically adsorbed oxygen on the surface. Therefore, we obtain the average valence of the O anion, V_alO, for BaTiO₃ as being −1.55. This value is close to the value (−1.63) calculated by Cohen using density functional theory (DFT).¹ 

In order to show that O1s spectra measured by XPS can be used to estimate V_alO, we also analyzed the O1s spectra for the monoxides CaO, MnO, CoO, ZnO, NiO and CuO, as shown in Fig. 2. Following a procedure similar to the above analysis for BaTiO₃, the values for the O1s peak positions and areas were obtained as shown in Table II. The values for V_alO, f_i and the second ionization energies, V(M²⁺), of the metal ions in these monoxides are shown in Table III. In addition, the values of f_i and V(M²⁺) for the SrO, CaO, MgO and BeO as reported by Phillips² are shown in Table III.

The data in Table III are also presented in Fig. 3. It can be seen that values of f_i decrease linearly with increasing V(M²⁺) and that the two lines are close to each other. The reason for the dependence on V(M²⁺) can be understood as follows: The second electron affinity energy of oxygen is 8.08 eV, while the second ionization energies of the metal cations are significantly higher, between 11.87 and 20.29 eV for these oxides. One would expect that these properties of the free atoms must affect how ions gain and lose electrons in oxides. The result is that for a cation with higher V(M²⁺), it is more difficult to lose a second electron and hence that the ionicity of the material decreases.

The cation XPS spectra for our samples are shown in Fig. 4. It can be seen that there are no obvious satellite peaks indicative of other valence states in these spectra. We therefore can not obtain the ionicities using these metal ionic photoelectron spectra.

Taking into account that there are O¹⁻ ions present and that the ionicity in the oxides is less than unity, many unresolved physical problems can be understood. For example, taking into account the ionicity and the constraints imposed by Hund’s rules, our group proposed an O2p itinerant electron model. Using this model, it was explained why the variations in the magnetic moments as a function of the doping level in Cr, Mn and Ti doped (A)[B]₂O₄ spinel ferrites are different from the changes seen in Fe, Co, Ni and Cu doped spinel ferrites.^11–19 The O2p itinerant electron model is based on three postulates:¹⁹ (i) There are O2p holes in the outer orbits of the O¹⁻ anions. In a given sublattice, an O2p electron of the O²⁻ ion with constant spin direction can hop to the O2p hole of an adjacent O¹⁻ ions via a metal cation. (ii) The two O2p electrons in the outer orbit of an O²⁻ anion, which have opposite spin directions, become itinerant electrons in the two different sublattices (the (A) or [B] sites). (iii) In a given sublattice, constrained by Hund’s rules and by the fact that an itinerant electron has constant spin direction, the magnetic moment directions of cations with 3d electron number, n_d ≤ 4 (such as Cr³⁺, Cr²⁺, and Mn³⁺), are antiparallel to those of the cations with n_d ≥ 5 (such as Mn²⁺ or divalent and trivalent Fe, Co, and Ni cations) whether at the (A) sites or the [B] sites.

Experimental evidence for the existence of O2p itinerant electron in oxides has been reported recently by Suzuki et al.²⁰ They studied the spinel compound, Li_xMn₂O₄, a lithium ion battery cathode material, using high-resolution x-ray Compton scattering. They found that the itinerant electrons migrate through the material mainly via the intermediary of O2p orbitals during battery operation and that manganese 3d states are shown delocalization only involving 0.16 ± 0.05 electrons per Mn sites.

In summary, the content ratios O¹⁻/ O²⁻ for oxygen anions in metal oxides were obtained by fitting O1s photoelectron spectra with Gaussian–Lorentzian functions. We then obtained the average valence (V_alO) of oxygen anions and the ionicity, f_i = |V_alO|/2. It was found that the estimated f_i for BaTiO₃ is close to the value calculated by Cohen and that the dependence of f_i on the second ionization energy of the metal cation obtained in this work for the monoxides is close to that reported by Phillips. We therefore suggest that O1s photoelectron spectrum measurements can be used as a general experimental method for estimating the ionicity of oxides.

This work is supported by the National Natural Science Foundation of China, under Contract No.NSF-11174069, the Natural Science Foundation of Hebei Province (Grant No. A2015205111), the Key item Science Foundation of Hebei Province (Grant No. 10965125D) and the Young scholar Science Foundation of the Education Department of Hebei Province (QN20131008). The authors wish to thank Dr. Norm Davison for helpful discussion.