Expanding the Perovskite Periodic Table to Include Chalcogenide Alloys with Tunable Band Gap Spanning 1.5–1.9 eV

Optoelectronic technologies are based on families of semiconductor alloys. It is rare that a new semiconductor alloy family is developed to the point where epitaxial growth is possible; since the 1950s, this has happened approximately once per decade. Herein, this work demonstrates epitaxial thin film growth of semiconducting chalcogenide perovskite alloys in the Ba‐Zr‐S‐Se system by gas‐source molecular beam epitaxy (MBE). This work stabilizes the full range y = 0 − 3 of compositions BaZrS(3‐y)Sey in the perovskite structure. The resulting films are environmentally stable and the direct band gap (Eg) varies strongly with Se content, as predicted by theory, with Eg = 1.9 − 1.5 eV for y =  0 − 3. This creates possibilities for visible and near‐infrared (VIS–NIR) optoelectronics, solid‐state lighting, and solar cells using chalcogenide perovskites.


Introduction
Each new family of semiconductors expands opportunities for technological innovation, as new property combinations become available.Alloy families are selected based on their different properties for different applications, from cameras, to solar cells, to telecommunications, and so on.Device engineering and optimization are then achieved through fine control of composition and electronic structure, made possible by epitaxial film growth.

DOI: 10.1002/adfm.202304575
Chalcogenide perovskites have many features suggestive of useful semiconductor functionality, including high dielectric polarizability, chemical and environmental stability, and a wide range of theoretically predicted E g . [1]We recently demonstrated epitaxial growth of the chalcogenide perovskite BaZrS 3 .[4][5] Experimental work to date on chalcogenide perovskite alloys has been limited to solid-state powder synthesis. [5,6]y perovskite, we refer to phases with composition ABX 3 , with structure characterized by corner-sharing BX 6 octahedra (including distorted perovskites, with ∠(B − X − B) ≠ 180°).Most ABX 3 chalcogenides with X = S, Se are thermodynamically unstable in the perovskite structure, and form alternative structures such as the edge-sharing needle-like phase (prototype: NH 4 CdCl 3 -type), and the face-sharing hexagonal phase (BaNiO 3 -type). [7]The moststudied chalcogenide perovskite is BaZrS 3 (distorted perovskite structure, orthorhombic space group Pnma).It can be prepared by high-temperature bulk [8] and thin film growth methods, [8][9][10][11] including MBE, and it is very stable, even against heating in air above 500 °C. [2,12,13]It has a direct band gap of E g = 1.9 eV and absorbs light strongly. [2]For VIS-NIR optoelectronic applications, solid-state lighting, and solar energy conversion, a range of E g is necessary.Theory predicts that Se-alloying rapidly decreases E g , to as low as ≈ 1.3 eV in the pure selenide. [3,14,15]This would make BaZrS (3-y) Se y perovskite alloys particularly attractive for thin-film solar cells.[16][17] Powder synthesis results demonstrate that BaZrS (3-y) Se y is stable in the bulk up to at least y = 1.2 (40% Se). [4]

Results and Discussion
The gas-source MBE method, that we introduced recently for making BaZrS 3 epitaxial thin films, is suited for expanding into alloy development. [2]We use varying quantities of H 2 S and H 2 Se gas flows to make films with varying S:Se ratio, up to and including the selenide perovskite BaZrSe 3 with E g = 1.5 ± 0.2 eV.Our approach represents a revival of methods of gas-source MBE for chalcogenide alloys, developed decades ago for II-VI semiconductors. [18]These methods are newly relevant because of heightened interest in semiconductor fabrication for chalcogenide perovskites, two-dimensional materials, and other materials for computing, optoelectronics, and energy conversion.
In Figure 1, we show the band gap (E g ) and lattice constant measured on five BaZrS (3-y) Se y films spanning y = 0 − 3 and a PC = 4.996 ± 0.004− 5.34 ± 0.09 Å, along with data for other semiconducting materials.For the chalcogenide perovskites, we plot the pseudocubic lattice constant (a PC ), determined from Xray diffraction (XRD) measurements of the (202) reflection, and E g is determined from photoconductivity spectroscopy (PCS).Plots such as these of E g versus lattice constant (called a "Cho plot", after the developer of MBE) are the starting point for designing optoelectronics, solid-state lighting, and solar cells.The advent of epitaxial growth of BaZrS (3-y) Se y alloys with tunable E g means that chalcogenide perovskites may reasonably be placed on a Cho plot, alongside well-established families of semiconductors.Se alloying is effective at reducing E g into the range needed for single-junction solar cells.E g may also be raised above 2 eV, into the green lighting region, in compounds containing Sr and/or Hf. [4,19] To-date such materials have been reported only as powders made by solid-state synthesis; alloys and thin-film epitaxy have not been demonstrated.Nevertheless, we indicate these materials in Figure 1, to suggest that the chalcogenide perovskite alloy family may in the future cover a wide range of E g . [1]e grow films by pseudomorphic epitaxy on BaZrS 3 template layers, leveraging our previously reported process for epitaxial growth of BaZrS 3 on LaAlO 3 (LAO) substrates. [2]Growing BaZrS (3-y) Se y directly on LAO results in films of low crystal quality, as evidenced by an amorphous reflection high-energy electron diffraction (RHEED) pattern.Due to the high growth temperature, S and Se are substantially intermixed between the BaZrS (3-y) Se y film and the underlying BaZrS 3 layer.As a result, the final composition throughout the film (including the original template layer) is closely determined by the ratio of H 2 Se and H 2 S gases in the vapor phase during the alloy growth stage.This facile anion exchange process is why we refer to the initial BaZrS 3 film as a template layer, instead of a buffer layer.The nominal composition parameter y reflects the relative concentrations of We present in Figure 2b reciprocal space maps (RSMs) of the out-of-plane (202) film reflections.The lattice constant of BaZrS (3-y) Se y expands with y, so that the alloy and selenide (202) reflections appear at lower q z than for the sulfide BaZrS 3 .The RSMs show the presence of two (202) reflections, indicative of regions in the film with different d-spacings; out-of-plane, high-resolution X-ray diffraction (HRXRD) measurements also showed two (202) peaks (Figure S2, Supporting Information).In the RSMs, we assign the spot at lower q z to a phase with higher Se content and the higher q z spot to a phase with lower Se content.We see in Figure 2b that both (202) reflections change towards larger d-spacing with increasing Se content.All films were grown on a BaZrS 3 template.If this template layer remained unchanged during alloy layer growth, then one of the (202) reflections would remain at a value of q z consistent with the pure sulfide.We instead conclude that the template layer d-spacing and composition changes due to sulfur and selenium interdiffusion during alloy layer growth.In Figure 2c, we present RSMs of the (121) reflections, containing both in-and out-of-plane contributions.We use these RSMs to compare the differences between in-plane lattice spacing of the two layers.Two overlapping (121) spots can be seen for the y = 1 and 2 samples.These results suggest that the in-plane lattice parameters for the layers (template and alloy) in these samples are similar.For the y = 3 sample, the two spots are clearly separated.In assigning the spots in the y = 3 sample, we use the same logic as in Figure 2b: the spot with higher q z corresponds to a layer with lower Se content, and vice versa.This presents a conundrum in interpreting the q x values-a higher out-of-plane d-spacing seemingly corresponds to a lower in-plane d-spacing.STEM results (discussed below) reveal that these observations in XRD derive from film microstructure.
We use STEM to investigate the local microstructure, crystal structure, and composition for two representative alloy samples.In Figure 3a,b, we present high-angle annular dark-field (HAADF) images, measured along the [010] LAO-PC zone axis, that include the substrate, template layer, and top layer for BaZrS 2 Se (y = 1) and BaZrSSe 2 (y = 2).The substrate/template interface is abrupt.The dominant epitaxial growth mode for BaZrS 3 films on LAO is via a self-assembled, incoherent buffer layer, enabling fully relaxed films despite the large lattice constant mismatch. [2]he interface between the bottom template layer and the top alloy layer is coherent.Therefore, during alloy growth the alloy layer induces tensile strain in the template layer, and the template layer places the alloy layer under compressive strain.This strain gradient may accelerate sulfur/selenium interdiffusion during alloy layer growth.
These strains appear to control the formation of Ruddlesden-Popper (RP) faults or antiphase boundaries (APBs).APBs in perovskites manifest as rock-salt-like layers that disrupt the corner-sharing octahedral connectivity and are related to layered phases such as the RP series.APBs can accommodate cation off-stoichiometry and relieve strain. [20,21]We observe APBs in our chalcogenide perovskite epitaxial films, which we ascribe to the difficulty of maintaining exact Ba:Zr = 1:1 stoichiometry during MBE synthesis, without an accessible parameter regime of growth by self-limited adsorption. [22]Here we observe an additional effect: in the bottom layer we observe mostly vertical APBs, whereas in the top layer we observe a high concentration of horizontal APBs.As the Ba:Zr supply rates are unchanged throughout, we hypothesize that this change in the APB profile is related to strain relaxation.The vertical APBs in the bottom help to relieve the tensile strain imposed by the overlying film.By the same logic, as the top overlayer is under compressive strain from the underlying bottom, there is a preference for APBs to orient horizontally rather than vertically.The APB profile illustrates clearly the location of the bottom/top interface, despite chalcogen intermixing.In Figure 3c, we show STEM EDS maps measured on the y = 2 film.S and Se are dispersed through the film: even in the bottom template layer, that was grown as BaZrS 3 without H 2 Se flow, we measure substantial Se.
In Figure 3d, we show STEM EDS Se/S depth profiles for y = 1 and 2 samples.The Se/S ratios are close to the gas conditions during alloy growth.The template/alloy layer interface is not apparent.The observation of no clear distinction in chalcogen content at the template/alloy interface, despite a sharp distinction in microstructure, is sensible: the diffusivity of two-dimensional extended defects is infinitesimal compared to that of individual chalcogen ions.The APBs nucleate and grow simultaneously with ad-atom attachment, while S 2− and Se 2− migrate throughout film growth without disrupting the crystal structure or film microstructure.
The APBs locally modulate the d-spacing, and explain the reflection pairs observed in XRD.In Figure 4a, we present in-plane and out-of-plane cation-cation distances for sample BaZrSSe 2 (y = 2) determined by STEM. [23]These are the average of Ba-Ba and Zr-Zr distances; data for sample BaZrS 2 Se (y = 1) and additional analysis are presented in Figure S3 (Supporting Information).Horizontal APBs create a tripartite population of out-ofplane cation-cation distances in the top layer, corresponding to expansion in the APB core, compression immediately adjacent to the APB core, and an intermediate value away from APBs.The net effect is that the film has regions of distinct d-spacing, despite nearly uniform composition throughout, due to the presence of APBs.We show in Figure 4b,c that the cation-cation spacing measured by STEM (Figure S3, Supporting Information) is in quantitative agreement with the pairs of out-of-plane (202) reflections observed in XRD (Figure 2b,c, Figure S2, Supporting Information).We compare the average of Ba-Ba and Zr-Zr distances determined from STEM to the d-spacing of the (202) reflection measured by XRD, as they correspond to the same distance in the crystal structure.Nowhere do we see the d-spacing corresponding to the pure sulfide BaZrS 3 .These results further illustrate that S and Se readily intermix in the perovskite structure during film growth.
We use PCS to measure how E g varies with composition.Optical measurement techniques including spectrophotometry and ellipsometry can accurately measure E g for uniform thin films, especially for materials with direct band gap and strong light absorption, such as chalcogenide perovskites. [2,4]However, for films with known variation in composition and E g from top to bottom, data from all-optical methods can be challenging to analyze for the lowest E g in a heterogeneous sample.PCS has a higher dynamic range than all-optical methods, and therefore is more sensitive to the layer with the smallest E g in a multilayer stack.We fabricated photodetectors by depositing interdigitated Ti/Au electrodes on the film surface, which resulted in linear IV behavior as evidenced by Figure S5 (Supporting Information).Using a tunable light source, we measured the responsivity (the ratio of pho-tocurrent to incident light power) spectra of the photodetectors.These spectral data, which we present in Figure 5a, demonstrate a shift in responsivity onset to lower energies as Se content increases, indicating that Se addition lowers E g .24] For the pure sulfide BaZrS 3 we find E g = 1.9 ± 0.1 eV, agreeing with the value of 1.9 eV determined optically. [2]E g decreases rapidly with Se content, reaching 1.5 ± 0.2 eV for BaZrSe 3 , confirming theoretical predictions of the effect of Se alloying of BaZrS 3 . [3,24]The absolute photodetector responsivity decreased monotonically by several orders of magnitude with increasing Se content, going from 100 mA W −1 for the y = 1 detector, to 0.001 mA W −1 for the y = 3 detector (Figure S7, Supporting Information).This reduction in responsivity indicates a decreasing mobility-lifetime product with increasing Se alloying.This could be due to a combination of factors including bulk defect concentrations, surface scattering rates, and surface recombination rates, all of which may depend on composition, processing history, and film thickness.These will be addressed through defect engineering in future work.

Conclusion
Our results are notable for demonstrating an alloy family of perovskite semiconductors that can be grown as epitaxial thin films with continuously variable direct band gap.Chalcogenide perovskites are stable and are made of inexpensive, and low-toxicity elements. [13]][27] The phenomenon of anion (S 2− , Se 2− ) diffusion and mixing observed here suggests that alloy composition and band gap can be controlled by annealing in controlled environments.These reports motivate continued work toward optoelectronic and energy conversion devices based on chalcogenide perovskites, including thin-film solar cells.
The most outstanding obstacle to chalcogenide perovskitebased technology may be lowering the temperature required for synthesis of high-quality films.We also highlight the need for detailed studies of electronic transport and recombination rates, to understand performance-relevant issues such as the effect of anion vacancies on carrier concentration, and the carrier-localizing effects of band gap fluctuations (e.g., at APBs).

Experimental Section
The films were grown using a gas-source MBE system (Mantis Deposition M500).The substrate was heated radiatively from a SiC filament and was rotated at 2 rpm.Ba metal was supplied from an effusion cell (Mantis Comcell 16-500), and Zr metal from an electron beam (e-beam) evaporator (Telemark model 578).Ba and Zr source rates were calibrated using a quartz crystal monitor at the substrate position, and XRR, X-ray photoelectron spectroscopy (XPS), and X-ray fluorescence (XRF) measurements after film growth.Sulfur and selenium were supplied in the form of H 2 S and H 2 Se gases (Matheson).These were supplied from condensed, liquified sources of 99.9% and 99.998% purity, respectively, and pass-through point-of-use purifiers (Matheson Purifilter) before entering the MBE chamber.The gases were injected in close proximity to the substrate using custom-made gas lines and nozzles.
The films were deposited on (001) PC -oriented LaAlO 3 (PC stands for pseudocubic) single-crystal substrates (CrysTec GmbH).The substrates were outgassed in the MBE chamber at 1000 °C in H 2 S gas.The growth temperature measured at the thermocouple was 1000 °C.A template layer of BaZrS 3 , approximately 20 nm thick, was first grown.The deposition was then interrupted, started the H 2 Se flow, adjusted the H 2 S flow, and resumed growth.The H 2 S flow rate was 0.8 sccm during substrate outgassing and buffer layer growth.During BaZrS (3-y) Se y film growth, the H 2 S b) E g versus composition, measured here and as reported elsewhere.Error bars indicate 95% confidence intervals for the extracted E g values, as determined from linear regression standard errors.Nishigaki (2020): optical measurements on powder samples of BaZrS 3 and BaZrS (3-y) Se y , y = 1.2. [4]Sadeghi (2021): optical measurements on BaZrS 3 epitaxial thin films. [2]Meng (2016) and Li (2022) reported theoretical predictions. [3,24]d H 2 Se flow rates were adjusted to achieve the desired chalcogen ratio in the gas phase.The gas flow rates were measured and controlled using mass flow controllers (MFCs, Brooks GF100C).
The chalcogen ratio in the gas phase was determined using a residual gas analyzer (RGA, Inficon Transpector 2.0).The H 2 S and H 2 Se signals appeared in the RGA data at 34 and 81 Da, respectively (both singly ionized).The relative intensity of these signals was used to define the composition variable y, used throughout this work.The relationship between gas flow rates measured by the MFCs and chalcogen ratio measured by the RGA was nonlinear.To achieve H 2 S/H 2 Se = 2 (composition y = 1), the H 2 S and H 2 Se flow rates were 0.6 and 0.1 sccm, respectively (a ratio of 6).To achieve H 2 S/H 2 Se = 0.5 (composition y = 2), the H 2 S and H 2 Se flow rates were 0.3 and 0.36 sccm, respectively (a ratio of 0.83).For BaZrSe 3 (y = 3), this work used an H 2 Se flow of 0.5 sccm and no H 2 S. We note that the gas flow ratios needed to achieve a given film composition change with the gas delivery line conditions, such as line pressure; the numbers provided here are representative of a particular film growth campaign.The cham-ber pressure during film growth was approximately 8 × 10 −5 torr.The film growth rate was 0.1 Å s −1 .H 2 S and H 2 Se gas flows were maintained during cooldown, after growth, to avoid unwanted S and Se desorption.The typical total thickness of the film was approximately 40 nm.
As an alternative to using RGA data, the relative concentrations of H 2 S and H 2 Se in the chamber could be determined by measuring total chamber pressure with different flow rates, using a cold-cathode vacuum gauge (Inficon Gemini).It is found that this approach produces similar results as using the RGA.
RHEED data was measured using a 20 keV, differentially-pumped electron gun (Staib), and a digital acquisition system (k-Space Associates, kSA 400).XRR measurements were performed using a Rigaku Smartlab, with a Cu target, and a tube power of 9 kW (45 kV, 200 mA).Out-of-plane XRD was performed using a Bruker D8 High-Resolution X-ray diffractometer with a Ge (022) four-bounce monochromator in parallel-beam mode, with a Cu target, and a tube power of 1.6 kW (40 kV, 40 mA).RSMs were measured using a Bruker D8 Discover general area detector diffraction system (GADDS) with a Co K  source, μ Eulerian cradle, and Vantec-2000 area detector.RSMs were collected with tilts of 0°and 45°for symmetric and asymmetric scans, respectively.The GADDS data were transformed to reciprocal coordinates using Bruker Leptos 7.3 software.AFM measurements were performed using a Bruker Icon, and the images were processed using Gwyddion software.PCS was performed on photodetector samples fabricated by sputtering 5 nm Ti/200 nm Au interdigitated contacts on the films.The interdigitated contacts had eight individual fingers, each 5 mm long, with finger spacing and width of 100 μm.A tunable light source was used with a 300 W Xe arc lamp (ScienceTech), providing irradiance between 50 and 200 μW cm −2 depending on the selected wavelength.For most samples, a source-meter (Keithley 2400-C) was used to source a bias of 4 V and measure the resulting photocurrent.However, the low responsivity of the y = 3 sample necessitated a more sensitive measurement.For this sample, a lock-in amplifier (Stanford Research Systems SR830) was used to source voltage and measure peak-to-peak photocurrent under mechanically chopped illumination at 13 Hz.Spectroscopic ellipsometry measurements were performed using an XLS-100 (J.A. Woollam) ellipsometer with rotating compensator, in the photon energy range 1.24 to 6.20 eV (1000 to 200 nm) and an angle-of-incidence of 70°.All measurements were performed on mirror-smooth surfaces.
Samples were prepared for STEM using nonaqueous wedge polishing, followed by single-sector Ar-ion milling (Fischione 1051) using decreasing ion-beam energies of 3, 2, and 1 kV.STEM imaging was performed on a Thermo Fisher Scientific Themis Z S/TEM (probe-corrected, accelerating voltage = 200 kV, convergence angle = 17.9 mrad, beam current = 30 pA).Images were analyzed using a Python-based Gaussian fitting script to determine cation-cation distances.STEM EDS was performed using Super X detectors (beam current = 100 pA) and quantified via Thermo Fisher Scientific Velox software.

Figure 1 .
Figure 1.Cho plot representing chalcogenide perovskites as a new family of semiconductor alloys.Deep blue points represent the BaZr(S,Se) 3 alloy series reported here as epitaxial thin films.Light blue points represent other chalcogenide perovskites, reported elsewhere as solid-state synthesis of bulk powders.Other data are established semiconductors and alloy families.Solid (dashed) lines indicate direct (indirect) band gap alloys.The colored band indicates the range of E g appropriate for single-junction solar cells.For the chalcogenide perovskites, the abscissa is the pseudocubic lattice constant (a PC ); horizontal error bars represent composition variation through the film thickness.

H 2
Se and H 2 S in the vapor phase.As shown below, this nominal composition parameter closely predicts the measured film composition, as measured by scanning transmission electron microscopy energy dispersive X-ray spectroscopy (STEM EDS).RHEED data acquired during growth shows evidence of single crystalline, perovskite epitaxial films.In Figure 2a, we present RHEED data measured along the [100] LAO-PC azimuth during growth of BaZrS 3 , BaZrS 2 Se (y = 1) and BaZrSe 3 (y = 3) (LAO-PC indicates pseudocubic indexing of the LAO crystal structure).The RHEED pattern does not change when the growth is switched from BaZrS 3 to BaZrS (3-y) Se y , indicating epitaxial growth of the same crystal structure albeit different composition (i.e., pseudomorphic heteroepitaxy).The pattern remains streaky, without sign of discrete points, indicating atomically smooth film growth.X-ray reflectivity (XRR) and atomic force microscopy (AFM) analysis also show that the top surface and the buried interfaces are smooth (Figure S1, Supporting Information).

Figure 3 .
Figure 3. STEM measurements of structure and composition.HAADF image of a typical film cross section of a) BaZrS 2 Se (y = 1) sample and b) BaZrSSe 2 (y = 2).The interface between the bottom layer and top layer is visible as a change in the distribution of vertical and horizontal antiphase boundaries (APBs).c) EDS maps of elements Ba, Zr, S, Se, and O corresponding to the HAADF STEM image in (b) for BaZrSSe 2 (y = 2).d) S and Se atomic ratios, determined by EDS, through the film thickness for both films.

Figure 4 .
Figure 4. Comparison of local cation-cation distances measured by STEM and d-spacings measured by XRD.a) In-plane (left) and out-of-plane (right) cation-cation distances, corresponding to the HAADF STEM image (Figure 3b) for sample y = 2. b) Histogram of out-of-plane cation-cation distances (mean μ, and standard deviation ), divided into the top and bottom layers.The top layer is defined by the presence of horizontal APBs.c) 1D out-ofplane XRD scan on BaZrSSe 2 (y = 2) for the (202) reflection.The x-axis is converted from 2 to d-spacing for direct comparison to STEM analysis results.The black dotted line corresponds to the d-spacing for pure BaZrS 3 .The XRD peak positions (d top = 531.4± 0.1 pm, d bottom = 516.7 ± 0.1 pm) agree quantitatively with the STEM analysis.