Atomic and electronic structure of the ${\mathrm{Fe}}_3{\mathrm{O}}_4\left(111\right)∕\mathrm{Mg}\mathrm{O}\left(111\right)$ model polar oxide interface

High-resolution transmission electron microscopy (HRTEM) and density functional calculations are used to study the effect of interface polarity on the atomic and electronic structure of the prototype ${\mathrm{Fe}}_3{\mathrm{O}}_4\left(111\right)∕\mathrm{Mg}\mathrm{O}\left(111\right)$ polar oxide interface. We show that atomically abrupt interfaces exist between the MgO(111) substrate and magnetite (111) film in regions separated by Fe nanocrystals, and propose a solution for this oxide-oxide interface structure. Comparisons of experimental HRTEM images with calculated through-focus and through-thickness images for model interface structures suggest metal-oxygen-metal (i.e., $\mathrm{Mg}\mathrm{O}\mathrm{Fe}$) interface bonding with octahedral (B) coordination of the first Fe monolayer, rather than the combination of tetrahedral-octahedral-tetrahedral (ABA) stacking also found in ${\mathrm{Fe}}_3{\mathrm{O}}_4$. First-principles calculations for all the different models find metal-induced gap states in the interface oxygen layer. Consistent with the HRTEM results, the $\mathrm{Mg}\mathrm{O}{\mathrm{Fe}}_3{\mathrm{O}}_4$ interface stacking $\cdots ,4\mathrm{Mg}∕4\mathrm{O}∕\mathbf{4}\mathbf{Mg}∕\mathbf{4}\mathbf{O}∕\mathbf{3}{\mathbf{Fe}}_{\mathbf{B}}∕4\mathrm{O}∕{\mathrm{Fe}}_{\mathrm{A}}{\mathrm{Fe}}_{\mathrm{B}}{\mathrm{Fe}}_{\mathrm{A}},\cdots$, is calculated to be the energetically most favorable, and effectively screening the MgO(111) substrate surface polarity. The data and calculations exclude mixing of Mg and Fe across the interface, in contrast to the commonly invoked mechanism of cation mixing at compound semiconductor polar interfaces.

Figure 1

Broader-view HRTEM image of the ${\mathrm{Fe}}_3{\mathrm{O}}_4\left(111\right)∕\mathrm{Mg}\mathrm{O}\left(111\right)$ interface showing abrupt oxide-oxide interface regions (identified by the arrows) separated by regions with Fe(110) and ${\mathrm{Fe}}_3{\mathrm{O}}_4\left(001\right)$ nanocrystals in the magnetite film.

Figure 2

Model interfaces for the ${\mathrm{Fe}}_3{\mathrm{O}}_4\left(111\right)∕\mathrm{Mg}\mathrm{O}\left(111\right)$ interface in the $\left[11\overline{2}\right]$ projection: (a) interface B initiated with a monolayer of Fe in the octahedral position $\left(3{\mathrm{Fe}}_{\mathrm{B}}\right)$; (b) interface ABA, initiated with Fe in the tetrahedral position $({\mathrm{Fe}}_{\mathrm{A}}∕{\mathrm{Fe}}_{\mathrm{B}}∕{\mathrm{Fe}}_{\mathrm{A}})$.

Figure 3

Image comparison between experimental and calculated (for interface models B, ABA, BA, and A) HRTEM images in the $\left[11\overline{2}\right]$ zone at a constant thickness of $\sim 6\mathrm{nm}$, and two different defocus settings: (a) $-25\mathrm{nm}$ and (b) $-65\mathrm{nm}$.

Figure 4

HRTEM images of the interface for regions with steps on the MgO(111) surface: (a) For an odd number (one) of bilayers at the step, lattice misalignment is observed in ${\mathrm{Fe}}_3{\mathrm{O}}_4$ film, indicative of a grain boundary. (b) No grain boundaries are present when the step height consists of an even number (two) of $\mathrm{Mg}\mathrm{O}$ bilayers.

Figure 5

(a) HRTEM image from the ${\mathrm{Fe}}_3{\mathrm{O}}_4\left(111\right)∕\mathrm{Mg}\mathrm{O}\left(111\right)$ interface in the $\left[11\overline{2}\right]$ zone. Intensity profiles along the [111] direction across the interface from: (b) the experimental image, (c) a simulated image of model B, and (d) a simulated image with one monolayer of H inserted between MgO and the ${\mathrm{Fe}}_3{\mathrm{O}}_4$ of the interface B model.

Figure 6

(Color online) The planar-averaged Coulomb potentials along the [111] direction for an O-terminated MgO substrate (from a 41 layer MgO film); for an (11-layer) MgO substrate with a $3{\mathrm{Fe}}_{\mathrm{B}}$ overlayer (the initial Fe layer deposition), a $3{\mathrm{Fe}}_{\mathrm{B}}∕4\mathrm{O}∕{\mathrm{Fe}}_{\mathrm{A}}{\mathrm{Fe}}_{\mathrm{B}}{\mathrm{Fe}}_{\mathrm{A}}∕4\mathrm{O}∕3{\mathrm{Fe}}_{\mathrm{B}}$ overlayer $\left(3\mathrm{B}\text{−}{\mathrm{Fe}}_3{\mathrm{O}}_4\right)$, and a ${\mathrm{Fe}}_{\mathrm{A}}{\mathrm{Fe}}_{\mathrm{B}}{\mathrm{Fe}}_{\mathrm{A}}∕4\mathrm{O}∕3{\mathrm{Fe}}_{\mathrm{B}}∕4\mathrm{O}∕{\mathrm{Fe}}_{\mathrm{A}}{\mathrm{Fe}}_{\mathrm{B}}{\mathrm{Fe}}_{\mathrm{A}}$ overlayer $\left(\mathrm{ABA}\text{−}{\mathrm{Fe}}_3{\mathrm{O}}_4\right)$. All results were obtained using symmetric supercells and include structural relaxations. The curves were shifted to approximately align the bulk MgO potentials; the differences in the averaged potentials at large distances are a measure of the differences in the work functions for those surface terminations.

Figure 7

(Color online) Local spin-resolved density of states for different atomic sites around the interface for the interface B model. Majority (minority) spin LDOS are plotted as positive (negative) values. Solid (dashed) lines denote sites with three (one) symmetry equivalent atoms per layer. Note the different vertical scales for the O, Fe, and Mg sites.