The design and synthesis of functional oxide interfaces with desirable properties is currently limited by gaps in our understanding of growth pathways at the atomic level, as well as a lack of precise methods to control growth dynamics. LaFeO₃/SrTiO₃ heterojunctions, which are of great interest for photocatalytic water splitting, provide a case in point; groups using different growth methods and characterization tools have reported conflicting results. Now, a research team at Pacific Northwest National Laboratory (PNNL) in Richland, Wash., has taken a step forward with its use of shuttered molecular beam epitaxy, electron energy-loss spectroscopy, and ab initio simulations to explore the structure and dynamics of interface synthesis.

“Our work illustrates the importance of carefully controlling the quantity and arrival time of the atoms introduced to a growing film. By using a shuttered growth mode, we can coax the system into new, previously untapped structural configurations. These results will have widespread application to the growth of all oxide thin films,” says lead author Steven Spurgeon. The work was reported in a recent issue of Physical Review Materials (doi:10.1103/PhysRevMaterials.1.063401).

As compounds with perovskite ABO₃ structures, the surfaces of the LaFeO₃ and SrTiO₃ compounds can take on more than one atomic configuration. For example, the (001) SrTiO₃ surface can present either a SrO or TiO₂ layer, which can be controlled by pretreating the substrate. At the heterojunction, the matching LaFeO₃ terminations would be either FeO₂ or LaO, resulting in FeO₂/SrO and LaO/TiO₂ interfaces, respectively. But nothing is ever perfect, so defects such as misfit strain, oxygen vacancies, cation intermixing, and even the migration of entire lattice planes have been identified as potential influences on the structure and properties of oxide interfaces. The PNNL team launched its investigation with an eye toward examining the factors that influence the stability of heterojunctions. They hoped to gain insight into how growth modes can lead to different synthesis outcomes, as well as reconcile the conflicting results of previous LaFeO₃/SrTiO₃ heterojunction experiments.

As their growth technique, the researchers chose shuttered oxygen molecular beam epitaxy (MBE) in preference to other techniques like pulsed laser deposition (PLD). In PLD, all species are ablated simultaneously from a stoichiometric sintered target, whereas MBE permits control over the flux of each species independently by opening and closing a shutter on pure elemental sources, yielding more granular control of the deposition process. In their experiments, SrTiO₃ served as the substrate with a TiO₂ termination. Depositing a single layer of SrO produced the alternative termination. Layers of LaFeO₃ were then grown on these substrates in a shuttered growth sequence. For example, LaO was deposited on the TiO₂ surface followed by a layer of FeO₂ and the sequence repeated until a nine-unit-cells-thick LaFeO₃ layer was obtained, and FeO₂ was deposited on the SrO surface followed by a layer of LaO and repeated for nine unit cells.

The PNNL team used aberration-corrected scanning transmission electron microscopy and electron energy-loss spectroscopy (STEM-EELS) as its primary characterization tool. “STEM-EELS allows us to simultaneously investigate structure, chemistry, and composition at near unit-cell level resolution,” Spurgeon says. What they found was that the interface was predominantly LaO/TiO₂ with a stacking sequence FeO₂/LaO/TiO₂/SrO, regardless of the termination of the SrTiO₃ substrate before LaFeO₃ growth. While their composition maps revealed some intermixing on each side of the interface, there were no obvious chemical state changes that could suggest a mechanism for interfacial reconstruction.

To explain these findings, the team turned to ab initio density functional theory simulations. “In our simulations, we are able to quickly survey a variety of growth parameters to identify mechanisms that operate in different regimes,” Spurgeon explains. “In this regard, ab initio lends strong support to our experimental observations.” The first step was to examine the stability of the two interfacial configurations expected when LaFeO₃ is grown on the two types of SrTiO₃ terminations. Consistent with their experimental observations and a wide body of literature showing that the TiO₂ termination is easier to stabilize, the simulations revealed that the FeO₂/SrO interface is less stable than the LaO/TiO₂ interface for the chosen growth conditions.

Oxide interface reconfiguration. Scanning transmission electron microscopy–electron energy-loss spectroscopy (STEM-EELS) composition maps of LaFeO₃ grown on SrTiO₃ with a SrO termination (left) and that of LaFeO₃ grown on a SrTiO₃ substrate with a TiO₂ termination (right) are nearly identical and show that both have LaO/TiO₂ interfaces. (Center) Illustration showing how the exchange of lanthanum and strontium atoms across the interface (dashed box) could induce the observed transformation of the FeO₂/SrO interface. Green = lanthanum, magenta = strontium, blue = iron, cyan = titanium, and red = oxygen atoms. Credit: Steven Spurgeon, Pacific Northwest National Laboratory.

Why is the LaO/TiO₂ interface more stable than the FeO₂/SrO interface? According to the simulations, the key lies in the first film layer deposited on the substrate. Deposition of lanthanum leads to an ordered, stoichiometric configuration. However, under the experimental conditions, iron deposition results in a Fe₂O₃ plane with oxygen vacancies that can enable cation movement. This instability can ultimately convert the interface to an entirely different structure through a process called dynamic layer rearrangement. The PNNL team’s simulations showed that this rearrangement can occur through several pathways, most likely the exchange of lanthanum and strontium atoms.

Spurgeon sums up, “Understanding growth dynamics will help us more precisely control and engineer the band offset in the LaFeO₃/SrTiO₃ system, which in turn will give us better control over electron–hole pair behavior in photoactive devices. Our group also plans to explore dynamic control of flux in more detail to see what other interesting interface configurations and structures we can unlock.”