Europium monochalcogenides (EuX; X = O, S, Se, and Te) are known as ideal Heisenberg ferromagnets exhibiting a high magnetic moment originating from localized Eu²⁺ 4f states and a large exchange splitting energy (ΔE _EX) of the Eu 5 conduction state below the Curie temperature (T _C) (Mauger and Godart, 1986). In particular, EuO has shown the highest T _C (about 70 K), the largest ΔE _EX, and the lowest work function of all these compounds (Eastman et al., 1969; Mauger and Godart, 1986). Because of recent observations of fully polarized electrons in electron-doped EuO and an increase of T _C to about 200 K upon La doping (Steeneken et al., 2002; Schmehl et al., 2007; Miyazaki et al., 2010), EuO has attracted considerable attention for the development of next-generation spintronics applications, such as spin injectors and spin-polarized photocathodes. However, EuO is an indirect energy gap ferromagnetic semiconductor that shows the lowest energy transition from the valence to conduction bands which involves a change in wave vector. This typically creates inefficient and complicated events in interband transitions, making these materials unsuitable for spintronics applications (Zhang et al., 2001).

The use of ferromagnetic semiconductors in spintronics devices depends on reduction of their thickness to a few atomic layers by creating low-dimensional magnetic ultrathin films and controlling their electronic band structure by changing their indirect energy gap into their indirect counterpart. A thickness-dependent calculation of the electronic structure of EuO thin films (Schiller, Müller, and Nolting) has shown that T _C and ΔE _EX decreased with decreasing thickness. Experimental T _C values were qualitatively consistent with theoretical predictions for ultrathin films with several atomic layers (Santos and Moodera, 2004; Müller et al., 2009; Negusse et al., 2009). However, the thickness-dependent ΔE _EX and detailed band structure have not been investigated experimentally despite their great importance in the design of spintronics devices. The ΔE _EX value in the EuO ultrathin films of fixed thickness has been determined using the Andreev reflection (Schmehl et al., 2007) and spin-resolved soft X-ray absorption measurements (Steeneken et al., 2002). These studies indicate that ultrathin films exhibit a smaller ΔE _EX than bulk films, consistent with the theoretical predictions.

Three-dimensional angle-resolved photoemission spectroscopy (3D-ARPES) studies have pointed out that ΔE _EX appears in the energy shift of Eu 4f bands, and the indirect band gap forms according to the theoretically predicted band structures (Miyazaki et al., 2009; Miyazaki et al., 2013). Thickness-dependent 3D-ARPES is expected to provide information about the thickness-dependent T _C, ΔE _EX, and detailed electronic band structure changes occurring with decreasing thickness. Therefore, in order to obtain information on the detailed electronic structure and magnetic properties, such as the T _C and ΔE _EX, we performed the thickness- and temperature-dependent 3D-ARPES measurements for EuO ultrathin films with various thicknesses.

Single-crystalline 2–8 and 20-nm EuO (100) ultrathin films were grown using a molecular beam epitaxy method. To grow high-quality ultrathin films, a 5-nm SrO thin film acting as a buffer layer was deposited between the EuO ultrathin films and SrTiO₃ substrate (Figure 1A) because SrO forms a rock-salt type crystal structure with a lattice constant of 0.5144 nm, similar to EuO (the lattice mismatch between SrO and EuO is 0.3%) (Miyazaki et al., 2012). The substrates were heated at 600°C for 1 h under an ultrahigh vacuum (2.0⁻⁷ Pa × 10⁻⁷ Pa) to clean their surfaces. The SrO buffer layer and EuO ultrathin films were grown by the sequential evaporations of Sr and Eu metals under oxygen pressures of 1.0 × 10⁻⁴ and 8.0⁻⁶ Pa × 10⁻⁶ Pa, respectively, at a substrate temperature of 400°C.

Three-dimensional ARPES measurements along the surface normal were performed at the beamline 5U of the UVSOR-III synchrotron radiation facility at the Institute for Molecular Science (Ito et al., 2007). Γ and X points in the normal emission geometry corresponded to photon energies hν of 78 and 38 eV, respectively (Miyazaki et al., 2013). Total energy and momentum resolutions were set to 123 meV (45 meV) and 0.020 Å⁻¹ (0.014 Å⁻¹) at Γ (X) point, respectively. There is no significant change in the photoelectron spectrum due to the difference in energy resolution between the Γ and X points. Temperature-dependent 3D-ARPES spectra were recorded during the temperature decrease.

The electronic structure of EuO was calculated by a local spin density approximation plus Coulomb repulsion U (LSDA + U) using the Wien2k code (Blaha et al., 1990). The electron correlation energy U of the Eu 4f state was fixed at 7.0 eV (Ghosh et al., 2004). To calculate the electronic structure of EuO ultrathin films, we used the slab approach, in which periodic boundary conditions are applied to the surface supercell including a slab of atomic layers and a vacuum region as shown in Figures 2(a1) and 2 (a2).

To negate the existence of impurities, such as Eu₂O₃ and the Eu metal in the fabricated EuO ultrathin films, their compositions and the Eu valence were investigated by angle-integrated PES measurement. Figure 1B shows the PES spectra of 2 nm EuO/5 nm SrO/SrTiO₃ [Figure 1B(1)], 5 nm EuO/SrTiO₃ [Figure 1B(2)], and 5 nm SrO/SrTiO₃ films [Figure 1B(3)] measured at 10 K for hν = 38 eV. The 2-nm EuO/5-nm SrO/SrTiO₃ film only presented Eu²⁺ 4f and O 2p [Figure 1(b1)] although the 5-nm EuO/SrTiO₃ displayed Eu³⁺ 4f and metallic states originating from Eu₂O₃ and Eu metal [Figure 1B(2)], respectively (Lee et al., 2007). These results suggest that the SrO buffer layer prevents oxygen diffusion from the SrTiO₃ substrate to the EuO ultrathin film. The PES peaks attributable to the Sr 5s and O 2p states of the SrO buffer layer appeared at the same energy [Figure 1B(3)] as that of the O 2p state of EuO [Figure 1B(1)]. Because the O 2p state of EuO was indistinguishable from that of SrO, their corresponding spectra are not discussed in this study. Figure 1C shows an ARPES image of a 2-nm EuO ultrathin film at 10 K for hν = 38 eV, and the band structure of bulk EuO near the X point. A large dispersion curve was observed at the top of the Eu 4f state, consistent with the band calculation. This also agrees with previous results obtained for 50-nm EuO films (Miyazaki et al., 2009), implying that both EuO films exhibit equally high crystallinity. We succeeded in fabricating single-crystalline EuO ultrathin films with a continuously controlled few-atomic layer thickness.

Before reporting the detailed results of 3D-ARPES measurements, we show the expected electronic structure of EuO ultrathin film calculated using the LSDA + U method. Figures 2 (a1) and 2 (a2) show the spin-dependent density of states (DOS) of bulk and 1-monolayer EuO, respectively. The splitting energy between the minority and majority spin states of both Eu 5d and O 2p states decreases and the band gap increases in the thin-film limit. It indicates that the ΔE _EX is decreasing from bulk to 1-monolayer EuO. Figures 2 (b1) and 2 (b2) show the spin-dependent band structure of bulk and 1-monolayer EuO, respectively. In the bulk EuO, the band structure shows the characteristics of a ferromagnetic semiconductor with an indirect band gap. On the other hand, in the 1-monolayer EuO, a direct band gap is realized. With decreasing thickness, the hybridization between the Γ and Z directions is expected to decrease. Since the strong hybridization between the constituent elements is the origin of the ferromagnetic phase in EuO, decreasing hybridization may cause a decrease in the T _C and ΔE _EX.

Temperature-dependent 3D-ARPES spectra were recorded for EuO ultrathin films with various thicknesses to determine the influence of film thickness on T _C and ΔE _EX. Figures 3A,B show the temperature-dependent energy distribution curves (EDCs) of the Eu 4f state in 8- and 2-nm EuO ultrathin films, respectively, at the X point. Because the X and Z points along the perpendicular line of the Brillouin zone, as shown in Figures 2(b1) and 2 (b2), were equivalent, we hereafter examine the X point. Peak and shoulder positions of the EDCs were evaluated from the local minima on the second-derivative EDCs. At the X point, EDCs consisted of two components. Based on previous band calculations (Miyazaki et al., 2009), higher and lower binding-energy components were attributed to the non-bonding and bonding Eu 4f states hybridized with the O 2p and Eu 5d states, respectively.

The temperature-dependent energy shift of the non-bonding Eu 4f state stems from the exchange splitting of Eu 5d majority and minority bands during the ferromagnetic phase transition because the Fermi level (E _F) is pinned at the bottom of the Eu 5d band at X point (Miyazaki et al., 2009). Because the exchange splitting occurred below T _C, T _C was determined using the temperature-dependent energy position of the non-bonding Eu 4f state. Figure 3C shows the temperature-dependent relative energy position of the non-bonding Eu 4f state at the X point in EuO ultrathin films for thicknesses of 2, 3, 5, and 8 nm. The onset temperature corresponds to T _C as shown in Figure 3D. T _C values calculated by the mean field theory (Jensen et al., 1992) and band calculation (Schiller, Müller, and Nolting) are also plotted as a function of temperature. T _C gradually decreased with decreasing thickness, qualitatively consistent with previous experimental and theoretical results, (Santos and Moodera, 2004; Müller et al., 2009; Negusse et al., 2009), albeit with slightly lower T _C values for films thinner than 5 nm.

The energy difference between the Eu 4f non-bonding state and the center of gravity of the Eu 5d majority and minority spin states must be constant across T _C, and E _F is located at the bottom of the Eu 5d majority band (Miyazaki et al., 2009). Consequently, the ΔE _EX of the energy difference between these majority and minority spin states can be attributed to the doubling of the energy shift of the non-bonding Eu 4f state below T _C. Figure 3E shows the thickness-dependent ΔE _EX at 10 K derived from Figure 3C, along with the theoretical ΔE _EX evaluated from band calculations in Figures 2 (a1) and 2 (a2). The experimental ΔE _EX gradually decreased with decreasing thickness, exhibiting the same behavior as T _C. At a thickness of 20 nm, the experimental ΔE _EX was approximately one-third of the calculated value. On the other hand, the interpolated experimental ΔE _EX amounted to one-tenth or less of the calculated value at 1 nm. Band calculations assumed a perfect hybridization and a crystal without interaction at its interface. Thus, this disagreement may originate from an overestimated hybridization and/or an underestimated interaction at its interface.

We now seek a reasonable explanation for the possible origin of the changing band structure in EuO ultrathin films leading to decreasing the thickness. Figure 4 shows the thickness-dependent EDCs of the Eu 4f state at Γ (a) and X (b) points in the ferromagnetic phase at 10 K. The hybridized Eu 4f state is denoted by the letter A, and non-bonding states are denoted by the letters B and B'. For the 8-nm EuO ultrathin film, the hybridized Eu 4f state at the Γ point was located at a lower binding energy side than that at the X point. Whereas for the 2-nm EuO ultrathin film, the hybridized Eu 4f state at the X point was observed at a lower binding energy side than that at the Γ point. Band calculations indicated that the bottom of the Eu 5d conduction band was located at the X point for EuO. Therefore, these results suggest that the electronic structure of EuO ultrathin film changes from indirect to direct energy gap at the X point when the thickness decreases. However, the previous band calculation results for 64-atom EuO thin films reported a direct energy gap (Monteiro et al., 2015), which is inconsistent with the present results. In the future, a systematic study of the band structure of EuO at various film thicknesses is needed to discuss the mechanism of the band gap change from direct to indirect with decreasing film thickness in more detail.

To investigate the origin of the changes in the electronic structure with decreasing thickness, the zero relative binding energies in Figures 4 (g1) and 4 (x1) were set to the energy of peak A for the 2-nm EuO ultrathin film at 100 K at the Γ and X points, respectively. Because peak B corresponded to the non-bonding Eu 4f state (Miyazaki et al., 2009), the relative binding energies of peak B in the other EDCs were fixed to the energy of the 2-nm EuO ultrathin films at 100 K (1.0 and 0.9 eV at the Γ and X points, respectively). At the Γ point, as shown in Figure 4 (g1), peak A at 100 K split into two (A and B’) below T _C for the 2-nm EuO ultrathin films. On the other hand, at the X point, as shown in Figure 4 (x1), shoulder A only shifted toward the lower relative binding energy side when the temperature decreased from 100 to 10 K, consistent with the shift of shoulder A observed at the Γ point. Only the shallowest shoulders A at the Γ and X points shifted to higher and lower binding energies, respectively, upon decrease in thickness at 10 K. Because E _F was pinned at the bottom of the Eu 5d bands, the band gap increased with decreasing thickness, in agreement with the thickness-dependent band calculations shown in Figures 2 (a1) and 2 (a2). In addition, the intensity of shoulder A at 10 K at the Γ and X points decreased gradually with decreasing thickness. The shoulder A mainly originates from the hybridization of the Eu 4f state with the O 2p and Eu 5d states (Miyazaki et al., 2009). The photoemission intensity of shoulder A decreased with decreasing thickness, suggesting that thinner films may suppress the hybridization of the Eu 4f state with O 2p and Eu 5d states.

The thickness-dependent energy shifts of shoulder A at 10 K are shown at the Γ and X points in Figures 4 (g2) and 4 (x2), respectively. When the thickness decreased, the energy shift gradually decreased at the Γ point, but progressively increased at the X point and band gap increased. This opposite trend can be explained by the change in the electronic structure across T _C as shown in Figure 5. Since the 3D-ARPES measurements only observe the valence electronic structure, the conduction Eu 5d states were assumed by the results of the band structure calculation. The Eu 4f–O 2p hybridization occurs in the same manner at both high symmetry points. At the Γ point, the Eu 5d state is located far from the Eu 4f state, producing a weak hybridization between these states. Because the Eu 4f state is mainly hybridized with the O 2p state, the resulting state moved to the lower binding energy side. The bottom of the Eu 5d conduction band was located at X point. The Eu 5d state is nearer to the Eu 4f state at the X point than at the Γ point. The Eu 4f state is strongly hybridized with the Eu 5d and Eu 4f states at the X point, displacing the hybridized Eu 4f state to the higher binding energy side. In the atomic layer regime, the hybridization between the Eu 4f and O 2p states is suppressed because of a decrease in the number of neighboring O atoms around the Eu atom. As a result, at the Γ point, the energy shift also diminished with decreasing thickness. At the X point, the hybridization strength between the Eu 4f and O 2p states and between the Eu 4f and 5d states was reduced. Therefore, the Eu 4f state shifts to the lower binding energy side with decreasing thickness if the hybridization between the Eu 4f and 5d states prevails over that between the Eu 4f and O 2p states, consistent with previous experimentally and theoretically evaluated indirect-exchange and super-exchange interaction parameters J ₁ and J ₂ (Kasuya and Yanase, 1968; Kasuya, 1970; Dietrich et al., 1975; Kocharyan and Khomskii, 1975; Passell et al., 1976; Mauger and Godart, 1986; Khomskii and Sawatzky, 1997; Larson and Lambrecht, 2006). Therefore, the suppressed hybridization between the Eu 4f–O 2p and Eu 4f–5d states explains the decreases in T _C and ΔE _EX with decreasing thickness in EuO ultrathin films. Because of the balance of J ₁ and J ₂, the energy gap changes from indirect to direct energy gap with decreasing thickness. A direct energy gap is highly favorable for spintronics applications, suggesting that EuO ultrathin films contribute to the development of these future spintronics applications.

In conclusion, we measured the temperature- and thickness-dependent 3D-ARPES spectra of single-crystalline EuO ultrathin films. The electronic structure of EuO ultrathin films changed from indirect to direct band gap at the X point. Both T _C and ΔE _EX gradually decreased with decreasing thickness. The decrease in neighboring atoms caused a suppressed hybridization between the Eu 4f and O 2p states and between the Eu 4f and 5d states, explaining the changes in the electronic structure and magnetic properties of EuO ultrathin films. The thickness dependences of the top of the Eu 4f state at the Γ and X points differ from each other because of the balance of the hybridization intensities between the Eu 4f–O 2p states and between the Eu 4f–Eu 5d states. A direct observation of the thickness-dependent variations in the electronic structure related to indirect- and super-exchange interactions is expected to provide a deeper understanding of low-dimensional magnetic ultrathin films. In the future, new applications such as photoluminescence using the direct band gap in EuO ultrathin films are expected.