Itinerant ferromagnetism mediated by giant spin polarization of the metallic ligand band in the van der Waals magnet ${\mathrm{Fe}}_5\mathrm{Ge}{\mathrm{Te}}_2$

We investigate near-Fermi-energy ($E_F$) element-specific electronic and spin states of ferromagnetic van der Waals (vdW) metal ${\mathrm{Fe}}_5\mathrm{Ge}{\mathrm{Te}}_2$. The soft x-ray angle-resolved photoemission spectroscopy (SX-ARPES) measurement provides spectroscopic evidence of localized Fe $3d$ band. We also find prominent hybridization between the localized Fe $3d$ band and the delocalized Ge/Te $p$ bands. This picture is strongly supported from direct observation of the remarkable spin polarization of the ligand $p$ bands near $E_F$, using x-ray magnetic circular dichroism (XMCD) measurements. The strength of XMCD signal from ligand element Te shows the highest value, as far as we recognize, among literature reporting finite XMCD signal for nonmagnetic element in any systems. Combining SX-ARPES and elemental selective XMCD measurements, we collectively point to an important role of giant spin polarization of the delocalized ligand Te states for realizing itinerant long-range ferromagnetism in ${\mathrm{Fe}}_5\mathrm{Ge}{\mathrm{Te}}_2$. Our finding provides a fundamental elemental selective viewpoint for understanding mechanism of itinerant ferromagnetism in low-dimensional compounds, which also leads to insight for designing exotic magnetic states by interfacial band engineering in heterostructures.

Figure 1

(a) Crystal structure of ${\mathrm{Fe}}_5\mathrm{Ge}{\mathrm{Te}}_2$ depicted by VESTA [28]. 50% occupation for Fe(1) and Ge sites is known [27]. (b) The valence-band PES spectra at the incident photon energy ($h\nu$) of 1486.6 and 450 eV. See main body for details of the spectral features indicated by A–C. (c) The DFT-calculated partial density of states, multiplied by Fermi Dirac function. (d) The bulk and projected Brillouin zones of ${\mathrm{Fe}}_5\mathrm{Ge}{\mathrm{Te}}_2$. (e) The DFT calculation for relative spectral weight between $d$ and $p$ orbitals. (f) Photoemission intensity plot along $\overline{\mathrm{\Gamma }}\text{−}\overline{\mathrm{M}}$ at 50 K and $h\nu =450\mathrm{eV}$.

Figure 2

(a) The experimental geometry of XAS/XMCD measurements in total electron yield (TEY) mode. See method part for details. (b) The Fe $L_{2,3}$-edge XAS and XMCD spectra at 20 K under 6 T. The ${\mu }^{+}\left({\mu }^{-}\right)$ denotes the XAS absorption intensity for parallel (antiparallel) alignment of the photon helicity and the sample magnetization direction. (c) The magnetic field dependence of the XMCD intensity of Fe $L_3$ (707.1 eV) at the photon energy of the XMCD peak.

Figure 3

(a), (b) The Te $M_{4,5}$-edge and Ge $L_{2,3}$-edge XAS and XMCD spectra of ${\mathrm{Fe}}_5\mathrm{Ge}{\mathrm{Te}}_2$ at 20 K under 6 T. In Te $M_{4,5}$ edge, the background spectrum including in the Cr $L_{2,3}$ edge ($\approx 576$ and 585 eV) absorption from the focusing mirrors are shown as the black solid line. (c), (d) The element-specific magnetization curves of Te and Ge with the photon energy set at the XMCD peaks; X (572.4 eV), Y (1215.1 eV), and Z (1219.3 eV).

Figure 4

(a) The magnetic dependence of XMCD signals normalized at the saturated value of each absorption edges. (b) The sketch of the relationship about magnetic moments between Fe $3d$ and seemingly nonmagnetic elements revealed by SX-ARPES and XMCD measurements.