Surface-to-bulk core level shift in CoFe₂O₄ thin films

CoFe₂O₄ favors an inverse spinel structure in which tetrahedral (T_d) sites are occupied by Fe³⁺ and octahedral (O_h) sites are occupied by both Fe³⁺ and Co²⁺ cations.^1–11 CoFe₂O₄ is stable, hard, and readily fabricated with magnetic properties that include high magnetic anisotropy, high coercivity, high Curie point, and moderate saturation magnetization.^12–19 The nominally single-component ferrimagnetic CoFe₂O₄/Al₂O₃ thin films can exhibit exchange bias.²⁰ 

CoFe₂O₄ has attracted attention for a wide range of potential applications such as in magnetic recording media, gas sensors, and catalysts.^12,21–23 An understanding of the structure and surface properties of the cobalt ferrite thin films is also important in corrosion control in boiling water reactors. In boiling water reactors, cobalt ions, released as a corrosion product, become Co-60 after neutron activation. The cobalt species may become deposited on out-of-core surfaces.^24–28 It is believed that the soluble Co-60 is absorbed into iron oxides resulting in the formation of a tenacious and more thermodynamically stable cobalt ferrite layer, under the boiling water reactor conditions.^{24,25,29–32} The result is a possible corrosion layer and embrittlement of the stainless steel pipe and valve surfaces in boiling water reactors.^24,25,27,31

It is typical that the surface of a thin film material, especially oxides, is chemically different from the bulk because of the different chemical environment of the surface atoms compared to that of the bulk.^32,33 The surface of an oxide can differ from bulk as a result of surface segregation, a different oxide at the surface (including oxygen loss), hydroxy adsorption, and a surface enthalpy and lower coordination that differs from the bulk, effects that are all intertwined. The dissimilar chemical environment at the surface as well as differences in the screening of the photohole during the photoemission process may result³² in a significant surface-to-bulk core level shift in core level photoemission.^33–35 The dielectric properties of the surface of an oxide may also differ significantly from the bulk leading to a different screening of the photohole in the photoemission process perturbing the photoemission final state.³² The characterization of the surface of CoFe₂O₄, by x-ray photoemission, presents an even more significant challenge because the tetrahedral (T_d) sites are occupied by Fe³⁺ and the octahedral (O_h) sites are occupied by both Fe³⁺ and Co²⁺ cations.

There have been multiple efforts to characterize the inverse spinel CoFe₂O₄, using x-ray photoemission spectroscopy (XPS), and the most common interpretation of the XPS spectra is to ascribe the T_d and O_h sites to different Fe and Co 2p_3/2 core level features.^1–11 In addition to the assignment of the T_d and O_h interstitial sites to different Co and Fe 2p_3/2 core level XPS features, there is also a satellite feature in the core level XPS spectra typically ascribed to a photoemission two hole bound state.

This analysis where the different Fe and Co 2p_3/2 core level features are assigned to the T_d and O_h sites ignores the obvious complexities that, as noted above, can occur at the surface of an oxide. These complications at the surface of CoFe₂O₄ may be more significant than simply noting that the XPS spectra are affected by tetrahedral (T_d) sites occupied by Fe³⁺ and octahedral (O_h) sites occupied by both Fe³⁺ and Co²⁺ cations.

Here, we investigate the possibility that there are both surface and bulk weighted components in the Co and Fe 2p_3/2 core level spectra of CoFe₂O₄ thin films, of various thicknesses, grown on Al₂O₃.

Epitaxial CoFe₂O₄ (111) thin films of thicknesses 5.5, 1.7, and 1.2 nm were grown on Al₂O₃ (0001) substrates using pulsed laser deposition, as described elsewhere.²⁰ X-ray photoemission spectra were acquired using a SPECS x-ray Mg Kα anode (hv = 1253.6 eV) source and VG100AX hemispherical analyzer. All the angle-resolved x-ray photoemission spectroscopy (ARXPS) measurements were carried out at room temperature in an ultrahigh vacuum, as a function of emission angle, with respect to the surface normal (0°). The limited electron mean free path ensures that by changing the photoelectron emission angle, the angle-resolved XPS can be used to distinguish components in the XPS spectra that have greater surface weight, i.e., larger emission angles with respect to the surface normal of the sample result in core level photoemission spectra with greater surface weights. The XPS measurement at higher emission angles provides a more surface sensitive characterization of a sample.^33–38 The kinetic energy of the photoelectrons coming out of Fe and Co core levels in this work is below 600 eV, and the photoelectrons with this kinetic energy have an inelastic mean free path of about 1 nm.^39,40

The XPS hemispherical analyzer has an acceptance angle of ±10°. At each emission angle, the binding energies of the various XPS core level components are referenced to the adventitious C 1s contamination position to eliminate variations in binding energy due to surface charging.

The ARXPS spectra of the Co 2p_3/2 core level spectra for CoFe₂O₄ thin films of thicknesses 5.5 and 1.7 nm grown on Al₂O₃ are shown in Figs. 1(a) and 1(b), respectively. For the 5.5 nm thick CoFe₂O₄ thin film, the Co 2p_3/2 XPS core level spectra contain three components: P₁ at a binding energy of 779.4 eV, P₂ at a binding energy of 781.4 eV, and S (satellite) at a binding energy of 786.0 eV. For the 1.7 nm thick CoFe₂O₄ thin film, these three Co 2p_3/2 core level features, P₁, P₂, and S, have binding energies of 779.6, 781.6, and 785.8 eV, respectively. For both the 5.5 and the 1.7 nm thick CoFe₂O₄ thin films, the binding energies of the Co 2p_3/2 P₁, P₂, and S core level components measured here are somewhat smaller than the CoFe₂O₄ film Co 2p_3/2 core level component binding energies of 779.8, 781.9, and 785.9 eV reported elsewhere.² Given that the measured binding energies depend somewhat on film thickness, the binding energies of the Co 2p_3/2 P₁, P₂, and S core level components for both the 5.5 and the 1.7 nm thick CoFe₂O₄(111) thin films are in general agreement with the CoFe₂O₄(111) Co 2p_3/2 core level component binding energies reported previously.^1,2,20

Similarly, the ARXPS spectra of Fe 2p_3/2 core level spectra for the CoFe₂O₄(111) films of thicknesses 5.5 nm and 1.7 nm are depicted in Figs. 1(c) and 1(d), respectively. Unlike cobalt, the Fe 2p_3/2 core level spectra have two components P₁ and P₂. For the 5.5 nm thick CoFe₂O₄ film, the Fe 2p_3/2 XPS core level spectra contain two components: P₁ at 709.8 eV and P₂ at 711.8 eV. For the 1.7 nm CoFe₂O₄ film, these two Fe 2p_3/2 core level features, P₁ and P₂, are at the binding energies of 710.0 and 712.0 eV, respectively. The binding energies of the components of Fe core levels in both 5.5 and 1.7 nm CoFe₂O₄(111) films are significantly smaller than Fe core level binding energies of 710.7 and 713.7 eV reported previously.² The binding energies of the component peaks of both the Co 2p_3/2 and Fe 2p_3/2 core levels slightly increased for thinner samples, indicating that the thinner films may be more dielectric or the dielectric substrate affects the measured core level photoemission binding energies.

The principal issue here has less to do with the actual core level binding energies but rather whether the P₁ and P₂ Fe and Co 2p_3/2 core level components in the x-ray core level photoemission are the result of the different T_d and O_h sites in CoFe₂O₄, as ascribed elsewhere^1–11 or if these different 2p_3/2 core level components are the result of the surface or surface oxide that is distinct from the bulk in terms, as has been suggested elsewhere.²⁰ The ARXPS spectra of 2p_3/2 core levels for both Co and Fe, from both the 5.5 and 1.7 nm thick CoFe₂O₄(111) films, were carried out at seven different emission angles, as shown in Fig. 1.

The surface sensitivity of the various cobalt and iron 2p_3/2 core level components is established by plotting the ratio of the P₁ and P₂ component intensities (shown in Fig. 1), as a function of emission angle, for 1.7 and 5.5 nm thick CoFe₂O₄(111) films, as shown in Fig. 2(a) (for Co) and Fig. 2(b) (for Fe). As illustrated in Fig. 2(a), the value of the P₁/P₂ intensity ratio for cobalt decreases at higher angles, indicating that the Co 2p_3/2 P₂ component is more representative of the surface than the P₁ component. This is consistent with the slight shift in the Co 2p_3/2 core level to high binding energies with increasing emission angle [Fig. 1(b)]. As noted above, the concept that the surface oxidation and surface properties of CoFe₂O₄ (111) thin films may result in a surface oxide of Co that differs from the bulk has been previously proposed;²⁰ nonetheless, here the analysis has the P₂ component of the Co 2p_3/2 core level photoemission spectra more representative of the cobalt oxide at the surface and P₁ component more representative of the bulk.

For Fe 2p_3/2 core levels, from the P₁/P₂ component intensity ratio as illustrated in Fig. 2(b), the P₁ component, of the Fe 2p_3/2 core level photoemission spectra, is found to be more representative of the surface. The assignment of the P₁ component as having more surface weight is possible because the intensity ratio P₁/P₂ increases with increasing emission angle [Fig. 2(b)]. Obviously, then the Fe 2p_3/2 core level photoemission P₂ component is more indicative of the bulk iron oxide environment. We note that the greater surface of the P₁ component of the Fe 2p_3/2 core level is consistent with a shift of the Fe 2p_3/2 envelope to smaller binding energies with increasing emission angle. This small core level shift with emission angle is more clear than is the case for Co and is independent of any fitting routine. The deconvolution of the Fe 2p_3/2 XPS core level feature into P₁ and P₂ components, as shown here, is very similar to the 2p_3/2 XPS core level component breakdown done for x-ray photoemission spectra from CoFe₂O₄(111) thin films reported elsewhere.^1–11,20 Rather than simply assigning these 2p_3/2 core level components to the T_d and O_h sites of CoFe₂O₄, here we find that one component is favored at the surface.

The surface-to-bulk core level shifts in both 1.7 and 5.5 nm thick CoFe₂O₄ samples, for both Fe and Co 2p_3/2 core levels, are found to be 2 eV but we recognize the surface oxides may differ from the bulk. Although here we ascribe the individual P₂ component of Co and P₁ component of Fe as surface weighted components, this does not exclude the prior assignment of the Fe 2p_3/2 XPS core level components to the T_d and O_h sites in CoFe₂O₄. If the components of the Fe 2p_3/2 XPS core level are the result of the T_d and O_h sites in CoFe₂O₄, then the T_d and O_h sites cannot be uniformly occupied at the surface and near surface region. Although such an explanation is difficult to reconcile with the ideal structure of CoFe₂O₄, it is not entirely excluded by the data. Assigning the 2p_3/2 XPS core level components to T_d and O_h sites means that the surface is rich in iron in the octahedral sites and cobalt in the tetrahedral sites. Just the same, it is important to note that Co³⁺ and Co²⁺ generally have similar Co 2p_3/2 binding energies. Neither the Co 2p_3/2 binding energy of Co³⁺ [binding energies 779.9 eV (Ref. 41) and 780.1 eV (Ref. 42)] or Co²⁺ [binding energies 780.0 eV (Ref. 41) and 780.1 eV (Ref. 43)] resembles the P₂ Co 2p_3/2 component with a binding energy of 781.4–781.9 eV measured here but rather the binding energy of the P₁ Co 2p_3/2 component of CoFe₂O₄ (779.4–779.8 eV as noted above). This makes it difficult to assign the Co 2p_3/2 XPS core level components to the T_d and O_h sites.

Surface segregation also does not explain these data. When plotting the ratio of the Co and Fe 2p_3/2 core level XPS intensities, there is a very little variation with changes in the emission angle. The Co to Fe ratio for the 5.5 nm film of CoFe₂O₄ is found to be 0.59, which is close to the expected value of 0.50. Figure 3 shows the ratio of the sum of surface and bulk 2p_3/2 core level components for Co relative to Fe, for 1.2, 1.7, and 5.5 nm thick CoFe₂O₄(111) thin films. Surface segregation of either cobalt or iron oxide does not appear to be significant, as this should result in a variation in the 2p_3/2 core intensities for Co relative to Fe, with increasing emission angle.³² 

In ultrathin CoFe₂O₄(111) thin films, it has been observed that reduction of the cobalt to a suboxide is possible and leads to a low binding energy Co 2p_3/2 core level photoemission spectra component.²⁰ The evolution of this low binding energy Co 2p_3/2 core level photoemission spectra component through annealing of a 1.2 nm thick CoFe₂O₄(111) thin film is shown in Fig. 4(a). The Co 2p_3/2 core level photoemission spectra from 5.5 and 1.7 nm thick CoFe₂O₄(111) thin films contain three components, which clearly differs from the four Co 2p_3/2 core level components observed for the 1.2 nm thick film. The component peaks P₁, P₂, and S of the 2p_3/2 core level, from 1.2 nm thick CoFe₂O₄, were observed at binding energies of 780.1, 782.1, and 786.6 eV, respectively. An additional Co 2p_3/2 core level photoemission component, at a binding energy of 778 eV, was also observed, on the lower binding energy side of the main Co 2p_3/2 core level feature, after extensive annealing [Fig. 4(a)]. The x-ray photoemission spectra taken after three different annealing stages of the 1.2 nm thick CoFe₂O₄(111) thin film in vacuum are plotted in Fig. 4(a). The CoFe₂O₄ (111) thin film was annealed at 200 and 285 °C, respectively, for about 5 h, then annealed at 303 °C for about 3.5 h. The evolution of Co 2p_3/2 core level is evident in the appearance of a small shoulder on the lower binding energy side and a slight broadening of the main 2p_3/2 peak. This low binding energy Co 2p_3/2 component implies that there exists another cobalt species represented by P₀ in Fig. 4(b).

The result is that the shape of the Co 2p_3/2 core level photoemission envelope changes, as shown in Fig. 4(a), with annealing of the 1.2 nm thick CoFe₂O₄(111) thin film. In contrast, there exists no additional feature appears in the Fe 2p_3/2 core level photoemission spectra, with the annealing of a 1.2 nm thick CoFe₂O₄(111) thin film. The absence of a corresponding additional component at lower binding energies in the Fe 2p_3/2 core level photoemission spectra for 1.2 nm thick CoFe₂O₄(111) thin films may be associated with the observations that 1.2 nm thick CoFe₂O₄(111) thin films are cobalt rich compared to thicker CoFe₂O₄(111) thin films, as seen in Fig. 3.

The observed core level binding energy position of the P₀ Co 2p_3/2 core level photoemission spectral component at a very low binding energy (778.0 eV) is at a binding energy closely observed to be the binding energy of Co 2p_3/2 of Co metal [778.0 (Ref. 41) to 778.2 eV (Ref. 43)] so represents a highly reduced oxide of cobalt. This suggests that extensive annealing of a 1.2 nm thick CoFe₂O₄(111) thin film, at the higher temperature of 285–303 °C, may cause loss of some oxygen from the surface. Some oxygen loss from the surface, after the annealing at 303 °C, is expected and a 1.2 nm thick CoFe₂O₄(111) thin film does not have a large volume to replace lost surface oxygen. The observation of P₀ segregation/enrichment at the surface (see the supplementary material⁴⁴ for surface enrichment of reduced cobalt, as indicated in Fig. S5) suggests that there may be some relation between the P₀ component and oxygen vacancies at the surface.

The surface-to-bulk core level shifts in the binding energy of Co and Fe core levels are observed in 1.2 nm thick CoFe₂O₄(111) film, illustrated in Fig. 4(b), which are similar to those observed in 5.5 and 1.7 nm thick CoFe₂O₄(111) films, which means that the P₁ component remains representative of a surface (bulk) weighted species and while the P₂ component is bulk (surface) of Fe (Co) 2p_3/2 core levels. This surface-to-bulk core level shift appears to be little disturbed by the presence of the P₀ component.

A large intrinsic exchange bias was reported²⁰ for these nominally single-component CoFe₂O₄ thin films, and an interfacial mechanism, i.e., the emergence of a third layer containing antiferromagnetic CoO between the CoFe₂O₄ thin film and the sapphire substrate, was proposed²⁰ to explain the pinning of the magnetic moments of CoFe₂O₄ to produce the observed exchange bias. However, the mechanisms for such an interfacial reconstruction remain elusive. The observation of P₀ component in the 1.2 nm sample after annealing but also indicates that the assumed CoO could actually exist by drawing oxygen to the surface while creating oxygen vacancies at the interface between the CoFe₂O₄ thin film and the sapphire substrate. This could also occur in those CoFe₂O₄ thin film samples grown with a relatively low oxygen pressure (10 m Torr).²⁰ 

We have found no compelling evidence of either Co or Fe surface segregation in CoFe₂O₄(111), yet the surface differs substantially from the bulk, perhaps as a result of preferential surface oxide formation. The traditional assignment of the different XPS Co 2p_3/2 and Fe 2p_3/2 core level components of CoFe₂O₄ to the T_d and O_h sites cannot be reconciled to the ARXPS results. This is particularly true of the XPS Co 2p_3/2 core level. Cobalt oxide at the interfaces of CoFe₂O₄(111) appears to be subject to reduction.

This research was supported by the National Science Foundation (NSF) through the EPSCoR RII Track-1: Emergent Quantum Materials and Technologies (EQUATE), Award No. OIA-2044049. The authors would like to acknowledge the seminal role Scott Chambers has played in our understanding of oxides through XPS, and P.A.D. acknowledges many fruitful and educational discussions with Scott Chambers.

The data that support the findings of this study are available from the corresponding author upon reasonable request.