Epitaxial growth of atomically thin Ga₂Se₂ films on c-plane sapphire substrates

Van der Waals (vdW) materials provide a promising platform for fabrication of heterostructures capable of multiple functionalities.^1–5 The weak interlayer interactions in vdW crystals result in self-passivated surfaces with no dangling bonds that enable the stacking of different materials with different lattice constants with an ultimate thickness scaling down to the atomic level.^6,7 Following graphene,^6,8,9 a wide range of vdW materials such as topological insulators,^10–12 transition metal dichalcogenides (TMDs),^13,14 and post-transition metal chalcogenides^1,15,16 were explored to expand the family of vdW materials.^6,17–19 The boom in vdW materials has triggered innovations in many fields such as photocatalysis,^18,20 optoelectronics,^19,21 and photovoltaics.²² Among vdW materials, Ga₂Se₂ deserves particular attention due to its exotic electronic and optoelectronic properties.

Compared to TMDs, Ga₂Se₂ has an anomalous bandgap behavior, i.e., it changes from a direct bandgap bulk material to an indirect bandgap monolayer.^1,23,24 The top of the valence band in the Ga₂Se₂ monolayer exhibits a “Mexican hat” dispersion,^24–26 resulting in a high density of states. Thus, Ga₂Se₂ is an attractive two-dimensional (2D) material for semiconductor industry.^27–30 Ga-Se compounds have two common phases: Ga₂Se₂ and Ga₂Se₃.³¹ Ga₂Se₂ is the focus of this study because in the crystal structure of Ga₂Se₃, one-third of the gallium sites is typically occupied by undesired vacancies.³² A primitive Ga₂Se₂ cell consists of two hexagonal gallium planes sandwiched by two hexagonal selenium planes, called a tetralayer (TL), with a thickness of 8 Å.^33,34 There are four common polytypes of Ga₂Se₂ designated as β (2R)-, ɛ (2H)-, γ (3R)-, and δ (4H)-Ga₂Se₂, arising from the stacking sequence of AA′AA′…, ABAB…, ABCABC…, and AA′B′BAA′B′B…, respectively.^35–38 ɛ-Ga₂Se₂ is the focus of this study since it is the most extensively explored polytype.

There are a variety of ways to synthesize vdW materials such as exfoliation from a bulk crystal^39–41 and chemical vapor deposition.^42,43 However, molecular beam epitaxy (MBE) synthesis has the advantages of extremely high sample purity, precise control of doping and layer thickness, in situ monitoring of the growth surface, and a wafer-scale platform commonly used in industry.³¹ Epitaxial growth of Ga₂Se₂ has been studied by the scientific community for decades.^31,44–48 The multiple phases and polytypes of Ga-Se compounds make them challenging to achieve accurate Ga₂Se₂ stoichiometry.^49–51 Moreover, there has been relatively little effort on how to grow atomically thin Ga₂Se₂ films on c-plane sapphire (c-sapphire) substrates. In 1998, Chegwidden et al.⁴⁸ developed a high-temperature annealing treatment to remove their initial Ga₂Se₃ film and obtain polycrystalline Ga₂Se₂ on a subsequent deposition on the same substrate. In 2015, Chia-Hsin et al.⁴⁴ obtained a single crystal Ga₂Se₂ epilayer on c-sapphire by MBE. In 2017, Lee et al.³¹ grew ɛ-Ga₂Se₂, and they found that increasing the Se:Ga flux ratio resulted in a Ga₂Se₃ phase and a spotty reflection high-energy electron diffraction (RHEED) pattern. They obtained Ga₂Se₂ crystalline films with a thickness of 75 nm and an RMS roughness as low as 2 nm. However, they also observed particles on their films using atomic force microscopy (AFM) and speculated that they were Ga droplets.

Compared to growth on 3D substrates, such as GaAs, the advantage of growing Ga₂Se₂ on c-sapphire is the negligible influence of lattice mismatch and thermal expansion coefficient mismatch,⁵² leading to fewer misfits and threading dislocations. C-sapphire wafers are also inexpensive, readily available, and commonly used for the vdW epitaxy of other materials. However, vdW growth may suffer from more 0D defects (e.g., vacancies) due to easier re-evaporation and more 1D defects (e.g., grain boundaries) due to less control over domain orientation.^53–56 

In this paper, we study the growth of atomically thin (about 3 TL) Ga₂Se₂ single crystal films on c-sapphire using MBE. The growth window was optimized by growing thick (about 32 nm) Ga₂Se₂ films. At this stage, the substrate temperature, Se:Ga flux ratio, and growth rate were varied to obtain films with the desired phase, lowest surface roughness, and highest crystallinity, as well as to understand how these parameters impact the quality of resulting films. We achieved Ga₂Se₂ crystals with an RMS roughness as low as 1.82 nm using an optimal set of parameters, i.e., a Se:Ga flux ratio of 20, a growth rate of 0.06 Å/s, and a growth temperature of 425 °C. These parameters were then used to grow atomically thin films. We developed a three-step growth mode to obtain coalesced 3 TL-thick Ga₂Se₂ films with significantly improved surface morphology, in which the RMS roughness was reduced to 0.61 nm. Finally, we will discuss the presence of 3D features on the films, how they are related to vdW epitaxy and how to reduce or eliminate them.

All samples were grown on epi-ready c-plane sapphire substrates using a dedicated Veeco GENxplor MBE system in the University of Delaware Materials Growth Facility. A Knudsen cell for gallium and a valved cracker cell for selenium were used for all growths, maintaining a temperature of 900 °C for the selenium cracking zone while controlling the selenium flux via the valve. Prior to growth, substrates were degassed in a load lock chamber at 200 °C and then transferred to a growth chamber where they were heated to 650 °C to desorb any residual contaminants. The substrates were then cooled to the desired growth temperature. Immediately before growth, the selenium shutter was opened 2 min before the gallium shutter to create a selenium atmosphere.^31,48 Then, the Ga-Se films were grown via co-deposition; details are described in Sec. III. For the cooling process, contrary to our group's experiences growing other selenides like Bi₂Se₃ (Refs. 57–59) and In₂Se₃,⁶⁰ the exposure of Ga₂Se₂ to the selenium flux should be stopped immediately at the end of growth. If it is not, more 3D features will appear, as shown in the AFM images in Fig. S1.⁷¹ After growth, the samples were removed from the MBE and immediately vacuum packed to await the analysis. All substrate temperatures were measured by a noncontact thermocouple. The reported source fluxes are beam equivalent pressures measured at the beginning of each growth day using a beam flux monitor.

In this study, an in situ RHEED system was used to monitor surface morphology and crystallinity during growth. AFM and scanning electron microscopy (SEM) measurements were used to characterize the surface morphology. AFM was performed on a NeaSNOM Microscope at the Advanced Materials Characterization Laboratory, and SEM was done on an Auriga 60 CrossBeam at the Keck Center for Advanced Microscopy and Microanalysis. X-ray reflectivity was conducted on Rigaku Ultima IV XRD to determine the thickness of the deposited films. X-ray diffraction (XRD) 2θ/ω scans were performed on a Bruker D8 XRD using a Cu Kɑ₁ source to analyze the film phase, and ω scans were performed on a PANalytical 4-Circle X'Pert3 MRD to assess crystal quality. The absence of sapphire substrate peaks in our XRD patterns is ascribed to the rough alignment between the sample and the XRD system. Room-temperature Hall effect measurements were conducted in van der Pauw configuration to study the film electrical properties. Raman spectra were collected from a Horiba LabRAM HR Evolution for better identification of phases.

Starting with a thick film allows for easier optimization of following growth conditions: growth temperature, Se:Ga flux ratio (hereinafter referred to as the flux ratio), and growth rate. At this stage, unless otherwise stated, all samples were grown using a gallium flux of approximately 2.6 × 10⁻⁸ Torr equivalent to a 0.06 Å/s growth rate for 90 min, resulting in a thickness of 32 (±3) nm. Later, the optimal conditions were applied to grow ultrathin films approximately 3 TL thick.

We first investigated the effect of flux ratio by fixing the growth temperature at 450 °C and the growth rate at 0.06 Å/s. Figure 1 shows the XRD analysis of samples grown using different flux ratios. Sample 1, grown using a ratio of 100, only presents a peak at 2θ = 28.3° associated with Ga₂Se₃ (111) orientation.⁶¹ A ratio of 60 used for sample 2 results in an amorphous phase. The two samples indicate that an excess of selenium leads to a Ga₂Se₃ or amorphous phase. Samples 3, 4, and 5 grown using the ratios of 37, 22 and 18, respectively, all show similar XRD curves, in which the two peaks at 2θ = 11.1° and 22.2° correspond to Ga₂Se₂ (002) and (004) orientations, respectively.⁵¹ The three Ga₂Se₂ samples behaved significantly differently in AFM scans, as shown in Fig. 2: Sample 3 consists of randomly distributed poorly coalesced domains, while sample 5 has large clusters on the surface. An intermediate flux ratio of 22 (sample 4) produced the most continuous film with an RMS roughness of 2.35 nm.

An unusual and unfortunate finding compared to our experience growing Bi₂Se₃ (Refs. 57–59) and In₂Se₃⁶⁰ as well as others growing Ga₂Se₂^31,45,62 is the extremely narrow flux ratio window within which smooth Ga₂Se₂ crystalline films can be obtained. For example, comparing samples 4 and 5, although both flux ratios of 22 and 18 produced Ga₂Se₂ single crystals, the slightly smaller ratio led to large clusters as shown in Fig. 2(c). These clusters may be caused by gallium aggregation in the early stages of growth due to a slight selenium deficiency. More examples are shown in Figs. S2 and S3.⁷¹ Figure S2 shows that when films are grown at 0.06 Å/s with a lower substrate temperature of 425 °C, the appropriate flux ratio window is even narrower. Figure S3 compares the surface morphology of samples grown using a ratio of 28 and 23, respectively, based on a higher growth rate of 0.11 Å/s. Sample 20 using a ratio of 23 displays a much smoother surface compared to sample 19 using a ratio of 23, again confirming the stringent requirement for a suitable flux ratio. This extremely high sensitivity to flux ratio may be due to multiple phases of Ga-Se compounds and the use of a selenium cracker cell. It is vital that the selenium flux is not in excess to stop the formation of Ga₂Se₃ phases, and insufficient selenium will result in gallium agglomerates, especially on the c-sapphire substrate. Therefore, it is more difficult to achieve the desired stoichiometry of Ga₂Se₂ than to obtain Bi₂Se₃ or In₂Se₃, since both of those materials can be grown with a large selenium excess without formation of unwanted phases. In addition, the cracked selenium molecules we used have a higher reactivity and incorporation efficiency than the uncracked ones, which may place more pressure on the precision of the Se:Ga flux ratio.

Next, we explored the effect of substrate temperature on film morphology. In this section, 37 was chosen as the flux ratio instead of 22 because sample 4, despite showing a more continuous surface, has undesired 3D features. The main goal of this portion of the study was to obtain continuous films without 3D features by optimizing the growth temperature. In Fig. 3, neither sample 6 grown at 480 °C, nor sample 8 grown at 400 °C show visible peaks in XRD curves, indicating that a too high or a too low substrate temperature will lead to amorphous films. Ga₂Se₂ is known to sublimate around 500 °C as Ga₂Se_1/2+ Se₂,⁴⁸ which sets a ceiling for the growth temperature. Furthermore, the hotter the substrate, the less likely adatoms or compounds are to adhere to the growth surface, especially for vdW materials due to a weak interaction between the substrate and the deposited materials. However, low-temperature growth causes the problem of reduced adatom mobility,⁵⁷ resulting in more and smaller nucleation domains, which is a typical source of grain boundary defects. Figure 3 shows that Ga₂Se₂ single crystal films can be obtained in the window of 425–450 °C. From the AFM images in Fig. 4, we see that sample 7 grown at a relatively low temperature of 425 °C has a more continuous surface than sample 3 grown at 450 °C, although it has 3D features. By reducing the flux ratio to 20 at 425 °C, the density and height of 3D features were minimized as shown in Fig. 4(e). Details on optimizing the flux ratio at the growth temperature of 425 °C are described in Fig. S2.

The last parameter we studied to optimize the growth of thicker Ga₂Se₂ films was the growth rate. Theoretically, a slow growth rate is beneficial for the formation of single crystal films with fewer defects, as it allows more time for adatoms reaching the growth surface to diffuse and nucleate domains, thus yielding larger but fewer domains and suppressing grain boundary defects or clusters. The ω scans depicted in Fig. S5 (j)⁷¹ confirm that for the vdW growth of Ga₂Se₂, a slower growth rate leads to a higher crystal quality. Therefore, 0.06 Å/s, the lowest stable growth rate our MBE system can achieve, was chosen for this study.

Another interesting finding is that the growth rate significantly affects the necessary Se:Ga flux ratio, which can be derived from XRD curves in Fig. 5. A flux ratio as high as 60 accompanied by a high growth rate of 0.45 Å/s yields a Ga₂Se₂ phase, while maintaining the high flux ratio but reducing the growth rate to 0.06 Å/s results in an amorphous phase. To regain desired Ga₂Se₂ phases, the corresponding flux ratio must be reduced, such as the ratio of 22 used for sample 4. It can be deduced that higher growth rates broaden the window of flux ratios in which Ga₂Se₂ single crystals can be achieved. This is surprising, as typically the flux ratio required to grow a compound is not strongly dependent on the growth rate. For these materials, the likely reason is that the lower growth rate leaves more time for gallium adatoms to migrate across the growth surface and diffuse over longer lengths to nucleate with further selenium atoms, resulting in a smaller amount of selenium required.

Furthermore, sample 21, which was grown under optimal conditions, was selected for Φ scans (Fig. S6).⁷¹ The Φ scan of Ga₂Se₂ revealed a sixfold pattern at about every 60°, consistent with the expected hexagonal crystal structure, but indicating the presence of twin domains. Although an epitaxial relationship between the Ga₂Se₂ layer and the sapphire substrate is observed, the film displays substantial in-plane rotational disorder, suggesting a weak binding between the Ga₂Se₂ layer and the sapphire. Finally, samples 4 and 9, which exhibit relatively optimal surface morphology, were chosen for Hall effect measurements. Both samples were found to be insulating with infinite resistance, contradicting previous research^5,31,33 that found Ga₂Se₂ to be p-type. This discrepancy could be attributed to defects within samples or a poor connection between the indium contact and the film.

Photoluminescence (PL) measurements were taken with the goal of analyzing the material bandgap, but the results were inconclusive. The PL spectra for all samples were very broad, with a spectral full-width half-max of 115 nm that is blueshifted by 40 nm relative to the expected GaSe PL peak centered on 620 nm with a 12 nm full-width half-max.⁶³ Moreover, there is significant variation in PL intensity across all of the samples, which could be indicative of inhomogeneity in the number of defects, grain boundaries, or structure. We suspect that the blue shift of the center of the PL peak, the extreme broadening of the PL peak, and the inhomogeneity of the PL intensity across the sample may all be due to the existence of 3D features, which may be causing changes in strain, plasmonic enhancement, oxidation, or thin film interference effects. Because the 3D features are much smaller than the diffraction limited spot size of the optical microscope and are measured only by AFM, we cannot directly correlate any variations in PL to the proximity of specific 3D features. Further study will be needed to isolate and understand these effects. For the purposes of optimizing growth conditions, which is the focus here, we considered only the average PL intensity across each sample as an indicator of film quality.

Optimal growth conditions identified by growing thick Ga₂Se₂ films are a Se:Ga flux ratio of 20, a substrate temperature of 425 °C, and a growth rate of 0.06 Å/s. These parameters were next applied to grow Ga₂Se₂ films with a thickness of 3 TL by shortening the growth time to 400 s. However, both the RHEED pattern [Fig. 6(a)] and the AFM scan [Fig. 6(b)] indicate that the resulting sample 12 is only a collection of discrete islands rather than a coalesced layer. It is expected that the weak interaction between the substrate and deposited materials makes it difficult to form a coalesced film with atomic-scale thickness,⁵⁹ and this morphology may be caused by the poor wettability of gallium atoms on sapphire.⁶⁴ The surface chemistry of sapphire does not support gallium adatoms wetting the surface; thus, first arriving gallium adatoms tend to ball up to form discrete islands rather than spread out and nucleate Ga₂Se₂ domains. This interpretation is supported by the observation of more continuous films with fewer visible islands with increasing thickness. This may be because the initially formed Ga_xSe_y changes the surface chemistry, allowing the later formed Ga_xSe_y to be uniformly distributed and gradually cover the initial islands.

Based on these observations, we decided to pursue a three-step growth mode, which was inspired from the MBE growth of other vdW films.^59,65 Figure 7 illustrates the overall process, in which a growth rate of 0.06 Å/s and a flux ratio of 20 were adopted. In the first step, 20 TL Ga₂Se₂ were deposited on a sapphire substrate at 425 °C; then, the substrate was heated to 1000 °C and annealed for 10 min in a selenium atmosphere to completely decompose the initial film. Decomposition was confirmed by the observation of only sapphire Kikuchi lines in the RHEED pattern. After that, the substrate was cooled to the desired growth temperature. Then, we grew the first tetralayer of Ga₂Se₂ at a relatively low temperature of 300 °C. Afterward, the substrate was heated to 450 °C to deposit the rest of the Ga₂Se₂ layers. In this way, at the end of the growth of 3 TL Ga₂Se₂, the RHEED shows a clear streaky pattern [Fig. 6(c)] associated with Ga₂Se₂ crystalline films,^1,32 which was in sharp contrast to the cloudy RHEED pattern [Fig. 6(a)] at the end of direct growth. This suggests a significant improvement in the resulting film quality, which is supported by the AFM image in Fig. 6(d). The morphology of the film grown using this three-step mode is more typical of what is observed for thick Ga₂Se₂ films, and we can see a much-improved surface continuity with an RMS roughness as low as 0.61 nm. Notably, XRD scans [Fig. 6(e)] successfully detected a Ga₂Se₂ (004)-oriented peak from such a thin film. Although the peak of Ga₂Se₂ (002) orientation is not present, the peak at 2θ = 22.2° confirms that this film is primarily Ga₂Se₂.

It is critical to understand the potential mechanism of this three-step mode. Based on experiences in similar systems (e.g., Ga₂Se₂ on silicon,³² Ga₂Se₂ on sapphire,⁴⁸ Bi₂Se₃ on sapphire,⁵⁹ and WSe₂ on silicon⁶⁶), it is suspected that the substrate pretreatment in step 1 can alter the surface chemistry of sapphire to overcome the poor wettability of gallium atoms. Scott et al.⁴⁸ used x-ray photoelectron spectroscopy to verify that a reacted interface layer containing traces of gallium and selenium was left on top after the evaporation of the initial film at high temperature. It is hypothesized that an exchange occurs between selenium and oxygen, as well as between gallium and aluminum. The weaker ionicity of selenium compared to oxygen results in a less polar, less reactive surface. Such passivated surfaces with fewer dangling bonds are favorable for vdW growth and may promote gallium atoms to wet the surface. Although the actual mechanism remains elusive, Fig. S7⁷¹ demonstrates that the substrate-pretreatment (step 1) plays a role in the growth of coalesced ultrathin Ga₂Se₂ films by comparing AFM and RHEED images from sample 23 (grown on a pristine substrate) and sample 24 (grown on a pretreated substrate). Because these films are otherwise identical, these data show that the pretreatment step is critical. Next, a combination growth of low temperature (step 2) followed by high temperature (step 3) was inspired by the growth of Bi₂Se₃ on sapphire^{57–59,67–69} since they both grow by vdW epitaxy. In general, the epitaxial growth of covalently bonded materials occurring at higher substrate temperatures results in larger but fewer domains and, thus, fewer defects because the adatom mobility increases with the substrate temperature. However, the story is opposite in vdW growth, in which higher substrate temperatures not only makes it more likely to re-evaporate the deposited film, but also causes gallium adatoms to diffuse faster and ball up with adjacent gallium adatoms, thereby yielding islands with lower substrate coverage. We believe that nucleating the film at a low temperature provides a platform for further growth and coalescence of Ga₂Se₂ domains. Furthermore, in-plane covalent bonds formed in the Ga-Se seed layer are stable up to the next high temperature step, resulting in a further improvement in surface morphology by increasing the adatom mobility.

One remaining question concerns dotlike 3D features shown in AFM images of Ga₂Se₂ samples, as exhibited in Figs. 2(b), 4(c), 4(e), 6(d), 8(a), and 9(b). Within the growth window where Ga₂Se₂ forms, no matter how we tuned the flux ratio, growth temperature, or growth rate, the 3D features could not be eliminated or significantly affected. AFM and SEM images (Fig. S8)⁷¹ only tell us that these 3D features are protrusions with spherical surfaces, ranging in diameter from 30 to 100 nm. The height of 3D features on samples with a thickness around 32 nm ranges from 29 to 48 nm. A common perception is that a streaky RHEED pattern only appears when the surface of the grown film is smooth. However, taking samples 4 and 9 shown in Fig. S9⁷¹ as examples, we saw clear streaky RHEED patterns throughout the growth process, whereas the resulting film surface has many 3D features, which is contrary to our expectation. Based on these observations, two possible origins of 3D features were proposed.

The first is that they may be impurities. If Ga₂Se₂ tends to adsorb impurities onto its surface, during the process of removing it from the MBE and performing AFM scanning, dust from the external environment may be absorbed and accumulated on the sample surface, which can explain why the RHEED pattern did not reflect those surface features. Based on this conjecture, we ultrasonically cleaned sample 14 with N-methyl-2-pyrrolidone (NMP), a common solvent for removing organics. Figure 8 compares surface morphology before and after cleaning. There are visible features on the pristine sample as shown in Fig. 8(a). After sonication in NMP for 40 s, most of the 3D features disappeared, but ring-shape features were left in Fig. 8(b), and the RMS roughness was reduced to 1.14 nm. An amazing result was yielded by another 300 s of sonication. Figure 8(c) shows no visible features and, thus, a smooth surface with an RMS roughness as low as 0.39 nm. In addition, we can see triangular Ga₂Se₂ nucleation domains. However, the SEM image in Fig. 8(d) demonstrates that a few 3D features persisted, and that this cleaning method left traces that appear as shadows in the SEM image and ring-shape features in the AFM image. While the 3D features can be ultrasonically removed to some extent, surface impurities should not be that difficult to remove. Moreover, the traces left after cleaning point imply that these features did not just adhere to the sample surface.

The evidence that 3D features remain after cleaning support our second conjecture, which is that the features formed at the bottom of the film, likely during initial nucleation. To test this conjecture, we need to understand the composition of the features. One challenge is their small size: the largest feature is only about 0.1 μm in diameter. An investigation of the feature composition/phase was attempted on sample 15, which, like other Ga₂Se₂ crystal samples, exhibited a streaky RHEED pattern during growth, pure XRD peaks of Ga₂Se₂ (002) and (004) orientations in Fig. 9(d), and the AFM image [Fig. 9(b)] with undesired 3D features. Raman tests were performed on a small area (1 μm diameter spot) and a large area (20 × 20 μm² square), respectively. Figure 9(a) depicts the spectrum from a single spot in which the three peaks at 132, 210, and 310 cm⁻¹ correspond to A¹_1g, E¹_2g, and A²_1g vibration modes of Ga₂Se₂, respectively.^1,31 The two peaks at 206 and 213 cm⁻¹ are thought to be a splitting of the 210 cm⁻¹ peak, which may be caused by surface strain or stress. The final two peaks at 155 and 249 cm⁻¹ can be attributed to Ga₂Se₃.^69,70 However, in a 20 × 20 μm² scan area, Fig. 9(c) shows that the three characteristic peaks of Ga₂Se₂ remain significant, but Ga₂Se₃ peaks are almost invisible (only a weak signal at 155 cm⁻¹ can be seen). The relative weak intensity at 155 cm⁻¹ compared to 132 cm⁻¹ in Fig. 9(a) as well as diminishing of the Ga₂Se₃ signal over a large area both suggest that the film consists of a majority of Ga₂Se₂ and a minority of Ga₂Se₃. The Ga₂Se₃ phase is, therefore, likely to be too small to be detected by XRD or large-area Raman scans. From these data, we conclude that 3D features are most likely Ga₂Se₃ inclusions. The formation mechanism may be that gallium adatoms ball up on the sapphire surface to form bumps at the initial stage of growth, providing more surface area for selenium atoms to incorporate, thus making the stoichiometry in these small regions closer to Ga₂Se₃ rather than Ga₂Se₂. This also explains the appearance of streaky RHEED patterns with 3D features on the growth surface. We hypothesize that the selenium cracker may also exacerbate the problem of 3D features by supplying more reactive selenium that may more easily form Ga₂Se₃ as noted previously.

In this work, we studied the MBE growth of atomically thin Ga₂Se₂ single-crystal films on c-plane sapphire. C-sapphire substrates are appealing due to their few dangling bonds, epi-ready availability, and relatively inexpensive prices. However, growth of Ga₂Se₂ on sapphire has rarely been studied. By examining thick films grown at varying temperatures, Se:Ga flux ratios, and growth rates, we found that the optimized growth window is: a substrate temperature of 425 °C, a Se:Ga flux ratio around 20, and a growth rate of 0.06 Å/s. The ceiling of the Se:Ga flux ratio at which Ga₂Se₂ single crystals can form increases with the growth rate, possibly because the slow growth allows gallium adatoms to diffuse farther to nucleate with selenium atoms, thereby reducing the necessary selenium flux. We also explored the growth of atomically thin Ga₂Se₂ films. We found that direct growth of 3 TL Ga₂Se₂ using this window produced only uncoalesced films with many islands. By developing a three-step growth mode, a coalesced 3 TL Ga₂Se₂ crystalline film with an RMS roughness as low as 0.61 nm was achieved. Unfortunately, the resulting films have undesired 3D features that are most likely caused by Ga₂Se₃ inclusions. Ultrasonic cleaning was used to remove most of the 3D features, after which we saw triangular features that are typical of the growth of Ga₂Se₂. Overall, this work describes the challenges associated with the growth of Ga₂Se₂ on sapphire and presents some general solutions, enabling the growth of thin Ga₂Se₂ films with low surface roughness.

The authors acknowledge funding from the Coherent/II-VI Foundation. The authors acknowledge the use of the Materials Growth Facility (MGF) at the University of Delaware, which was partially supported by UD-CHARM, a National Science Foundation MRSEC under Award No. DMR2011824. The co-authors acknowledge the Penn State Materials Characterization Lab for use of the PANalytical 4-Circle X'Pert3 MRD and A. Richardella for helpful discussions on data interpretation.

The authors have no conflicts to disclose.

Mingyu Yu: Conceptualization (supporting); Data curation (equal); Formal analysis (lead); Investigation (lead); Methodology (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (lead). Lottie Murray: Formal analysis (supporting); Investigation (supporting); Methodology (supporting); Writing – original draft (supporting). Matthew Doty: Conceptualization (equal); Funding acquisition (equal); Methodology (supporting); Project administration (supporting); Writing – review & editing (supporting). Stephanie Law: Conceptualization (equal); Data curation (equal); Formal analysis (supporting); Funding acquisition (equal); Methodology (supporting); Project administration (lead); Resources (lead); Supervision (lead); Writing – review & editing (supporting).

The data that support the findings of this study are openly available in Zenodo at http://doi.org/10.5281/zenodo.7719247, Ref. 72.