Bulk spin polarization of magnetite from spin-resolved hard x-ray photoelectron spectroscopy

There is broad consensus that magnetite (${\mathrm{Fe}}_3{\mathrm{O}}_4$) is a promising material for spintronics applications due to its high degree of spin polarization at the Fermi level. However, previous attempts to measure the spin polarization by spin-resolved photoemission spectroscopy have been hampered by the use of low photon energies resulting in high surface sensitivity. The surfaces of magnetite, though, tend to reconstruct due to their polar nature, and thus their magnetic and electronic properties may strongly deviate from each other and from the bulk, dependent on their orientation and specific preparation. In this study, we determine the intrinsic—i.e., bulk—spin polarization of magnetite by spin-resolved photoelectron spectroscopy on (111)-oriented thin films, epitaxially grown on ZnO(0001), with hard x-rays, making it a truly bulk-sensitive probe. This becomes possible by using a novel, specially adapted momentum microscope, featuring time-of-flight energy recording and an imaging spin-filter.

Figure 1

(a) LEED pattern (98.5 eV) of a freshly grown magnetite film, showing the reciprocal primitive cells of the oxygen (red) and iron (blue) sublattices. (b) hXPD patterns of the Fe $3p$ ($E_{\text{kin}}=5150$ eV) and O $2p$ ($E_{\text{kin}}=4993$ eV) core levels: experimental data on the left and calculated diffraction patterns on the right are in good agreement, confirming the excellent structural quality of the film and also supporting the claim of bulk sensitivity of the measurements. The hXPD patterns cover an angle range of $\pm 9^{\circ }$.

Figure 2

(a) Fe $2p$ core-level spectrum measured with Al $K\alpha$ radiation, showing ${\mathrm{Fe}}^{2+}$ and ${\mathrm{Fe}}^{3+}$ contributions due to the mixed valency of magnetite as well as charge-transfer (CT) satellites. (b) Hard x-ray (blue) and laboratory-based (red) valence-band spectra of the same magnetite film. The Fe $3d$ states at $E_F$ are clearly distinguishable from the O $2p$ valence band.

Figure 3

X-ray magnetic circular dichroism measured at the Fe $L_{2,3}$ absorption edge. The blue and red curves reflect the absorption for different helicity ${\sigma }_{+}$ and ${\sigma }_{-}$, respectively. Both curves were normalized by the mirror current $I_0$. The resulting XMCD signal is shown as a green dashed line (multiplied by 2).

Figure 4

(a) Valence-band spectra measured with an excitation energy of $h\nu =5.0$ keV for two opposite in-plane magnetizations denoted by $I^{+}$ and $I^{-}$. (b) Energy-dependent asymmetry $A$, calculated with Eq. (1).

Figure 5

Energy-dependent Sherman function $S\left(E\right)$, calibrated using the known 100% spin polarization of the Eu $4f$ states of an EuO thin film. The bottom axis refers to the actual scattering energy at the spin-filter crystal. In the top axis these values are converted to a binding energy scale, as adapted to the spin-resolved magnetite spectra of Fig. 4.

Figure 6

Spin polarization of the magnetite film, calculated with Eq. (2) (the line serves as a guide to the eye). It is negative at $E_F=0$ eV with a value of $-80$%. Between 0.60 and 0.95 eV, there is a sign change of the polarization. The error bars denote the statistical error.

Figure 7

Majority and minority spin-DOS (red and blue triangles, respectively) for ${\mathrm{Fe}}_3{\mathrm{O}}_4$, calculated from the results of Fig. 4 using Eq. (3) with error bars denoting the statistical error. Mostly minority spin states are located at $E_F$. The spin-DOS overall resembles those from DFT calculations (solid lines) in Ref. [42], especially upon convolution with the instrumental resolution (dashed lines).

Figure 8

Design (side view) and principal operation of the newly set up microscope at beamline P22, DESY. Notice that the spin-filter also preserves the energy information encoded in the time-of-flight signal (see the inset), resulting in a 3D ($k_x,k_y,E_{\text{kin}}$) recording scheme.