Understanding and controlling the growth of chalcogenide perovskite thin films through interface design is important for tailoring film properties. Here, the film and interface structure of BaZr thin films grown on by molecular beam epitaxy and postgrowth anion exchange is resolved using aberration-corrected scanning transmission electron microscopy. Epitaxial films are achieved from self-assembly of an interface “buffer” layer, which accommodates the large film/substrate lattice mismatch of nearly 40% for the alloy film studied here. The self-assembled buffer layer, occurring for both the as-grown sulfide and post-selenization alloy films, is shown to have rock-salt-like atomic stacking akin to a Ruddlesden–Popper phase. These results provide insights into oxide-chalcogenide heteroepitaxial film growth, illustrating a process that yields relaxed, crystalline, epitaxial chalcogenide perovskite films that support ongoing studies of optoelectronic and device properties.
I. INTRODUCTION
Growth of high-quality semiconductor thin films, needed for applications in microelectronics and optoelectronics, requires the availability of substrates that are well-matched structurally and chemically. This presents problems for many emerging materials for which suitable substrates are not readily available. For instance, many chalcogenide semiconductors have chemistry, crystal structure, and lattice constants that are substantially different than available oxide and compound semiconductor substrates. Further, many proposed device concepts—including two-dimensional materials for transistors and chalcogenide perovskites for solar cells—may depend on better control over oxide-chalcogenide hetero-interfaces. Better control hinges on a better understanding of interfaces with many elements (here, La, Al, O, Ba, Zr, S, and Se) in atomic structures that may deviate substantially from nominal bulk phases and that are difficult to predict a priori. Advancing semiconductor heteroepitaxial growth of new materials, therefore, depends on overcoming substantial challenges in advanced characterization. Here, we take on this challenge in the context of the growth of chalcogenide perovskite thin films on complex oxide substrates by molecular beam epitaxy (MBE).
Chalcogenide perovskites of type (A = Ca, Sr, Ba, B = Ti, Zr, Hf, X = S,Se), with a direct bandgap that is tunable within visible to near-infrared (VIS-NIR), strong optical absorption, low toxicity, and excellent environmental stability, are of interest for future solar cell technologies.1–4 The most-studied chalcogenide perovskite, , features strong dielectric response, strong optical absorption, exceptional thermal stability, and has been made as powders, crystals, nanocrystals, and thin films.2,4–12 Epitaxial thin films in the perovskite structure have been grown by the authors across the full alloy range, with direct bandgap spanning the range 1.5–1.9 eV.12–14 Notably, despite a very large lattice constant mismatch of with the substrates, the formation of relaxed, epitaxial films was observed to be accommodated by a self-assembled buffer layer at the interface.12
By comparing the bulk pseudocubic lattice parameters for the film (497.5 pm) and substrate (381.1 pm), a coincident site lattice (CSL) was proposed to explain the observed epitaxial relationship.12 The appearance of structural perturbations at the interface and the potential for surface passivation and reconstruction in the presence of at high temperature, however, indicate that the interface is too complex to be described only by a CSL.12,15–17 Difficulties in imaging and quantifying sulfur at atomic resolution, however, precluded direct determination of the interface structure. More recently, the authors realized thin film sulfide-selenide perovskite alloys by both direct growth and postgrowth selenization methods.13,14 Beyond modifying the bandgap, the higher atomic mass of selenium compared to sulfur along with the anion exchange method enables a more detailed atomic-scale study of the interface, in which selenium serves as an effective “tracer.”
Here, the epitaxial interface between and BaZr films with oxide substrates is characterized using aberration-corrected scanning transmission electron microscopy (STEM). Postgrowth selenization of films through annealing in gas, used to make the BaZr films, exchanges a fraction of the chalcogen ions without changing the interface structure or film microstructure. The self-assembled buffer layer is shown to have a rock-salt-like structure, similar to stacking faults often found in perovskites with imperfect cation stoichiometry and in Ruddlesden–Popper (RP) layered structures. This buffer layer accommodates the film-substrate mismatch and explains the structural deviations at and near the interface, which quickly diminish away from the interface. By elucidating the interface structure in these films, opportunities for the controlled epitaxial growth of other chalcogenide perovskites on oxide substrates using surface passivation and buffer layers can be explored.
II. MATERIALS AND METHODS
Thin films were grown and annealed using a chalcogenide MBE system (Mantis Deposition M500). Further details on the deposition system and protocols are described in Refs. 12 and 14. Thin films were grown on (001) -oriented substrates (MTI). PC stands for pseudo-cubic; corresponds to the (012) family of reflections in a rhombohedral basis. Prior to growth, the substrates were outgassed in the MBE chamber at 1000 C under gas. Epitaxial films were grown at 900–1000 C with elemental Zr and Ba sources under 0.8 SCCM flow. Alloying with selenium through anion exchange was performed by annealing epitaxial films under a mixture of 0.2 SCCM and 0.2 SCCM at 800 C for 60 min in the MBE chamber. Elemental quantification with energy dispersive x-ray spectroscopy (EDS) indicates a uniform distribution of Ba, Zr, S, and Se throughout the film cross section, with an overall composition of .14
Cross-sectional samples for electron microscopy were prepared by nonaqueous mechanical wedge polishing followed by single sector ion milling (Fischione 1051 TEM Mill).18 A probe aberration-corrected Thermo Fisher Scientific Themis Z S/TEM operated at 200 kV was used for STEM imaging and spectroscopy. Four-dimensional (4D) nano-beam electron diffraction (NBED) data were acquired with a convergence angle of 0.70 mrad using an electron microscopy pixel array detector (EMPAD).19 Strain mapping from these data was performed using the exit-wave power-cepstrum transform.20 Imaging was performed using a probe semi-convergence angle of 18 mrad and a probe current of 30 pA. High-angle annular dark-field (HAADF), annular dark-field (ADF), and integrated differential phase contrast (iDPC) images were acquired using collection angles of 65–200, 25–153, and 6–24 mrad, respectively, and a dwell time of 1 s. The revolving STEM (revSTEM) method was used to correct for image distortions from sample drift during imaging for the 12-frame image series.21 STEM image simulations were performed using a custom implementation of the multislice method using parameters matching experiment.22 STEM EDS maps were acquired using SuperX detectors and quantified using the Thermo Fisher Scientific Velox software. A 5 pixel averaging pre-filter and 3-sigma Gaussian postfilter were used for noise reduction in the EDS maps.
III. RESULTS AND DISCUSSION
An HAADF STEM image of the 40-nm-thick selenized film is shown in Fig. 1(a), with varying contrast arising from overlap and boundaries between film rotation variants. These rotation variants originate due to the nearly equivalent in-plane pseudocubic lattice parameters along the orthorhombic [0 1 0] and [1 0 1] projections, which is similar to observations in pure films.12 While maintaining an overall epitaxial relationship, the presence of rotation variants as well as substrate step edges and other planar defects introduces slight lattice distortions, as evidenced in cepstral strain analysis (supplementary material). Further, the atomic scale structure of these two domain types or rotation variants is shown in STEM HAADF images in Figs. 1(b) and 1(c) and compared to STEM image simulations in Figs. 1(d) and 1(e), assuming the orthorhombic structure with randomly distributed anion occupancy according to the average composition. Experiment and simulation are in agreement, where tilting of the S/Se octahedra and alternating Ba–Ba displacements are visible in the [0 1 0] and [1 0 1] projections, respectively [Figs. 1(f) and 1(g)].
The atomic structure of the film/substrate interface is highlighted in the representative simultaneously acquired ADF and iDPC images [Figs. 2(a) and 2(b)]. Approaching the interface, the anion octahedral tilt angle [Fig. 2(c)] and Ba–Ba displacement angle [Fig. 2(d)] are reduced compared to their values measured from the film bulk ( and , respectively). These changes near the interface are reminiscent of previous reports of perovskite oxide heteroepitaxy, where oxygen octahedra tilting is either dampened or propagated in the thin film through boundary constraints with the substrate.23–27 At the boundary between the rotation variants within the film in-plane direction, the magnitude of these tilts and displacement angles diminishes, likely as a result of structure projection along the electron beam direction. Throughout the thin film, both structurally overlapped and sharp rotation variant boundaries are observed (supplementary material). Further, the first two atomic planes of the selenized film—immediately at the interface with —exhibit structure contrasting with the rest of the film [Figs. 2(a) and 2(b)]. This observation mirrors the CSL hypothesized for the pure sulfide , suggesting that selenization changes the chalcogen composition without affecting the interface structure. Note that this structure is observed across multiple selenized and pure sulfide samples and represents the predominant film substrate interface type.
To confirm the CSL relationship, 4D-STEM NBED is performed. Averaged patterns from the BaZr film [yellow, Fig. 3(a)] and substrate [magenta, Fig. 3(b)] are color mixed and overlaid to show the film–substrate relationship, with coincident sites circled [Fig. 3(c)]. The CSL structure is further elucidated from atomic resolution images of the interface. ADF, iDPC, and inverse Fourier filtered images [Figs. 3(d)–3(f)] cropped from [Figs. 2(a) and 2(b)] illustrate an approximate coincidence of every five unit cells with every four BaZr unit cells, in agreement with the overall periodicity observed from NBED. At the coincident sites [connected by white lines in Figs. 3(d) and 3(e)], the Ba and La atom columns exhibit shifted positions and contrast in comparison to those of the rest of the interface. In particular, the La coincident site atom columns are less intense, suggesting the presence of vacancies or static displacements. Based on this interface model, the top and cross-sectional views for the two rotation variants are shown in Figs. 3(g)–3(j), with in-plane lattice parameters determined from experiment. The corners of the projected square La sub-lattice are close to the near-square Ba sub-lattice, exhibiting a mismatch of 0.5% along [1 0 1] and 0.4% along [0 1 0].
Despite this, the asymmetry of Ba positions in the orthorhombic structure yields imperfect coincidence for both of the rotation variants. This is visible by the misalignment of Ba and La sub-lattices and planes, as well as the absence of lattice distortions and tilt angles in experiment [Figs. 3(g)–3(j), 2(c), and 2(d)]. From the structural model, the displacements of Ba are symmetric and opposite with respect to the La sub-lattice of the substrate, resulting in an additional mismatch of Å, yet in experiment, these displacements, in addition to anion octahedral tilting, are absent near the interface. From comparing images from experiment with the structural models [Figs. 3(d) and 3(e) vs Figs. 2(i) and 2(j)], additional atom columns are also found within the interface gap. These atom column positions exhibit significant displacements compared to the bulk structures of both the BaZr film and substrate, representing another deviation from a purely CSL interface structure.
From atomic number-sensitive HAADF images of the interface [Fig. 4(a)], the contrast of the atom columns found within the interface gap suggests occupation by S ( ) or Se ( ). This is evidenced by a significantly brighter contrast than that of Al ( ) columns in the substrate (both marked by the white arrows) and lower contrast than those of Ba ( ), La ( ), or Zr ( ) columns. Further, the substrate surface is terminated by this plane, suggesting that these atom columns and the rest of the buffer layer self-assemble during exposure of the substrate immediately before and at the start of film growth.
The self-assembled buffer layer is further characterized using EDS elemental mapping at the interface [Figs. 4(a) and 4(b) and supplementary material). The profile across the interface, shown in Fig. 4(b), is integrated along the plane of the interface to improve the signal-to-noise ratio with a comparatively reduced dose. A well-defined film is observed, with limited inter-diffusion beyond 1–2 nm across the film/substrate interface, visible from the sharp gradients and low relative intensity of oscillations in the EDS profiles of the film and substrate elements (supplementary material). Further, the HAADF image intensity is decreased within a 0.5 nm wide region at the interface [shaded region in Fig. 4(b)] compared to the substrate and the rest of the film. Likewise, reduced HAADF and EDS signal intensity is observed at the terminating plane of the substrate immediately below the film, which is reminiscent of periodic La vacancy formation and surface reconstruction of (supplementary material).15 Immediately adjacent to this plane of La atoms is a notable peak in the sulfur signal [triangle marker, Figs. 4(a) and 4(b)], distinct from the selenium signal, whose location coincides with the unidentified atom columns extending into the interface gap from the substrate La plane [indicated with horizontal arrows in Fig. 4(a)].
These observations suggest that exposure at high temperature produces a surface reconstruction of the substrate that includes La vacancy and La–S bond formation. At the coincidence sites at the interface, the La vacancies appear to be replaced by S/Se, given by their intermediate contrast, the absence of neighboring S/Se [vertical arrows in middle of inset in Fig. 4(a)], and displacement toward the film. Combined with the Ba displacement toward the substrate at these coincident sites, this would suggest Ba–S bond formation enabled by surface reconstruction and treatment (supplementary material).
Beyond the passivated surface, the anion octahedral distortions and cation displacements in the self-assembled buffer layer are similar to the structure of the RP phase (space group ), which is shown in Figs. 4(c) and 4(d) overlaid on HAADF images acquired on selenized and pure-sulfide films.28,29 To highlight similarities in structure, interplanar distances, calculated from the in-plane averaged atom column positions, are measured from the acquired images of the grown films and simulated images of the RP phase [Fig. 4(e)]. These spacings are numbered with the film/substrate gap starting at #0. Beginning in the substrate, the average separation of 380 pm in the substrate is consistent with the pseudocubic lattice parameter of . Across the interface (spacing #0), the distances between La and Ba planes in the selenized and pure sulfide films are 331 and 325 pm, respectively, close to the ideal rock-salt layer separation of 350 pm in [Fig. 4(c)]. In the first plane of the films, the anion atom columns are also displaced toward the interface gap [visible in Figs. 4(c) and 4(d)]. This is characteristic of the rock-salt layers in the RP phase, in which anion planes terminate each perovskite slab and extend into the gap, connecting the structure of the perovskite film to that of the passivated substrate.
Moving further into the film, the next interplanar distances (spacing #1) are 250 pm in the selenized and 205 pm in the pure sulfide films. This decrease is again consistent with the RP structure, in which the planar spacing immediately adjacent to the rock-salt layer is contracted (to 200 pm) due to the polar displacements of Ba and the octahedra.30 A subsequent increase in interplanar distance at the Ba- octahedra (spacing #2) follows. Beyond spacing #2, there is a reduction in the oscillation of planar separations, converging to an average out-of-plane interplanar distance of 251 pm for and 266 pm for the selenized film, matching that of the expected perovskite structure (Pnma).14,29 This transition from the buffer layer to the bulk film, apparent in the variation of anion octahedral tilts and Ba–Ba displacement angles [Figs. 2(c) and 2(d)], emerges and becomes correlated within 1–3 unit cells of the buffer layer, similar to observations at perovskite oxide interfaces.27,31 Finally, even with anion exchange of S with Se, the structure of the RP buffer is preserved.
IV. CONCLUSIONS
The formation of epitaxial and relaxed and BaZr chalcogenide perovskite films on substrates with large lattice mismatch is made possible by a self-assembled buffer layer. This layer likely forms during the initial stages of film growth, including annealing and gas exposure of the substrate. The buffer layer resembles the rock-salt-like layers found in RP phases, and the interface structure is quantitatively comparable to . Notably, the selenization process, used to convert the pure sulfide to an alloy film, substitutes Se for S throughout the film, including in the buffer layer, without affecting the buffer layer structure or its strain relief functionality. The epitaxial growth and facile selenization process associated with this buffer layer enables the study of the optoelectronic properties of chalcogenide perovskite alloys with a wide range of composition, bandgap, and lattice constant, for which matching substrates may not be available. Finally, the gas-source film growth method used and self-assembled buffer layer formed here introduces broader opportunities for investigating heteroepitaxial growth of chalcogenide semiconductors on non-matched, non-chalcogenide substrates.
SUPPLEMENTARY MATERIAL
See the supplementary material for additional discussion and data regarding rotation variant boundaries, film epitaxy and cepstral strain analysis, and extended elemental mapping results.
ACKNOWLEDGMENTS
M.X. and J.M.L. acknowledge support for this work from the Air Force Office of Scientific Research (No. FA9550-20-0066) and the MIT Research Support Committee. We acknowledge support from the National Science Foundation (No. DMR-1751736). K.Y. acknowledges support from the National Science Foundation Graduate Research Fellowship, Grant No. 1745302. This research was partly supported by a grant from the United States-Israel Binational Science Foundation (BSF), under Grant No. 2020270. This research was supported in part by the Sagol Weizmann-MIT Bridge Program. This research was supported in part by the Skolkovo Institute of Science and Technology as part of the MIT-Skoltech Next Generation Program. This work made use of the MIT.nano Characterization Facilities.
AUTHOR DECLARATIONS
Conflict of Interest
The authors have no conflicts to disclose.
Author Contributions
Michael Xu: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Kevin Ye: Conceptualization (equal); Investigation (equal); Writing – review & editing (equal). Ida Sadeghi: Investigation (equal); Resources (equal); Writing – review & editing (equal). R. Jaramillo: Supervision (equal); Writing – review & editing (equal). James M. LeBeau: Conceptualization (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal).
DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author upon reasonable request.