Floating Particles in the Melt during the Growth of β-Ga₂O₃ Single Crystals Using the Czochralski Method

Abstract

Floating particles often appear during the Czochralski (CZ) growth of β-Ga₂O₃ in the Ir crucible, thereby impeding the seeding process. Identifying the floating nanoparticles and then inhibiting or removing them is critical for growing high-quality β-Ga₂O₃ single crystals. We grew β-Ga₂O₃ crystals containing floating particles using the CZ method. It is indicated that the floating particles were composed of Ir with a face-centered cubic (fcc) structure. In addition, the β-Ga₂O₃/Ir interface was comprehensively characterized, showing sharp and straight configuration on the whole with small fluctuations at the nanoscale. Combined with density functional theory (DFT) calculation, we found that Ir-O bonding was responsible for stabilizing the interface. Accordingly, the atomic configuration of the interface with the stablest structure, including the relaxed one, was determined. Based on the formation mechanism of the floating particles, we propose three effective strategies, including blowing sufficient oxygen into the bottom of the Ir crucible, coating a protective layer on its inwall and equipping a mechanical arm for inhibiting or removing them.

1. Introduction

β-Ga₂O₃ has emerged as an important next-generation semiconductor for power devices due to its ultra-wide bandgap [1,2,3], large breakdown electric field [4] and high Baliga’s figure of merit [5,6,7,8,9]. These attractive features have spurred researchers to grow β-Ga₂O₃ single crystals with large diameters. Thanks to its capability for formation of a relatively stable melt, large β-Ga₂O₃ single crystals can be grown from a melt source with several well-known advantages, including a rapid growth rate, promising potential for a large diameter and large-scale production [10]. Since the first report of achieving a β-Ga₂O₃ single crystal using the Verneuil technique in 1964 [11], many approaches, such as Czochralski (CZ) [12,13,14,15,16,17,18], floating zone (FZ) [19,20,21,22], edge-defined film-fed growth (EFG) [23,24,25,26,27,28,29], vertical Bridgman (VB) [30], etc., have been developed in the last two decades. Among these, EFG has received much more attention and has recently become a major method for commercial production of β-Ga₂O₃ wafers due to the accessibility of a relatively stable temperature field, especially at the liquid–solid interface. However, during the growth of a β-Ga₂O₃ single crystal, some defects are usually inevitable in the bulk [31,32,33,34,35]. CZ has been regarded as another promising method for growing large high-quality β-Ga₂O₃ single crystals. Unfortunately, spiral growth often takes place due to the concave interface arising from the severe free-carrier adsorption and poor thermal conductivity of β-Ga₂O₃ [14,15,16,36].

In a melt growth, the crucible is of great importance for achieving high-quality single crystal. The high melt point (ca. 1800 °C) [37] of β-Ga₂O₃ and the prerequisite of an oxidizing environment for growth severely limit the choice of crucible. Pt-Rh crucibles have been used in the VB method in ambient air [30]. However, the low melting temperature of Pt-Rh alloy (Rh < 30 wt%), high price of Rh (three times the price of Ir) and significant Rh contamination in crystal limit its wide application. Up to now, Ir has proved to be the best choice for fabricating the crucible, especially for the growth of β-Ga₂O₃ single crystals using the CZ or EFG process. In the CZ growth of β-Ga₂O₃ single crystals, floating particles are often formed and prevent the seeding process [38]. Galazka et al. [16] and Mu et al. [39] reported that floating particles generally appeared near at a temperature around the melt point and existed stably on the surface of the solution for CZ growth of β-Ga₂O₃ single crystals. Although the previous literature [16,39] has speculated on the formation mechanism and composition of these floating particles of β-Ga₂O₃, more direct experimental evidence is still lacking, especially on their detailed structural characterization. In addition, since the existence of the floating particles is deleterious to the production of high-quality β-Ga₂O₃ single crystals, revealing their interface stability mechanism at the nanoscale will provide a structural basis for exploring its destabilization methods. As such, it is highly desirable to determine the composition and structure of floating particles for a better understanding of the formation mechanism and to figure out a way to inhibit or remove them.

Here, we report the comprehensive characterizations of the floating particles formed during the CZ growth of a β-Ga₂O₃ single crystal. It is indicated that the floating particles were definitely composed of metallic Ir with a face-centered cubic (fcc) structure. In addition, the interface between β-Ga₂O₃ and Ir was characterized using the atomic-resolution microscopy technique, showing a generally sharp and locally fluctuant configuration. This demonstration is further supported by theory calculation.

2. Materials and Methods

2.1. Growth of β-Ga₂O₃ Samples with Floating Particles

The samples containing floating particles were obtained from the process of growing β-Ga₂O₃ single crystals via the CZ method. The CZ growth was conducted in an Ir crucible in a 98% Ar and 2% O₂ atmosphere. After the formation of a melt, a β-Ga₂O₃ seed rod was used to dip into the melt, touch and pull up the floating particles. The orientation of seed crystal was [010]. The pulling rate was 1–2 mm/h, and the rotation rate was 3–5 rpm. With a higher melting temperature than Ga₂O₃, the floating particles on the melt served as a real nucleation site for the crystal sample. The samples with the floating particles were cut into small pieces by a wire-cutting machine. Owing to the strong cleavage feature of the (100) plane, it could be easily split and cracked by itself. As such, the samples were properly split from the (100) plane layer by layer manually until the floating particles were exposed on the surface.

2.2. Morphological, Compositional and Structural Characterizations

X-ray diffraction (XRD) patterns were achieved by using the X-ray diffractometer (Rigaku D/max-ga, Tokyo, Japan) with Cu Kα radiation (graphite-monochromatized, λ = 1.541 Å). X-ray photoelectron spectrometer (XPS) was conducted on Axis supra (KRATOS, Manchester, UK)). Raman spectra were obtained by using a NT-MDT Raman spectrometer at room temperature. Field-emission scanning electron microscope (FESEM, ZEISS Sigma 300, Oberkochen, Germany) equipped with energy dispersive X-ray spectroscopy (EDX, Oxford Aztec Live One Xplore 30 mm², Oxford, UK) was used to characterize the surface morphology and composition of the floating particles. A focused ion beam (FIB, Thermo scientific Helios, Waltham, MA, USA) was employed to prepare the specimen for transmission electron microscope (TEM) analysis. The FIB current was decreased gradually in order to reduce possible damage to the specimen surface, because the floating particles were harder to mill than the β-Ga₂O₃ crystal. High-resolution TEM (HRTEM) was performed using a FEI Tecnai F30 G2(Thermo scientific, Waltham, MA, USA) microscope equipped with a high-angle angular dark-field (HAADF) detector and EDX system (Oxford). The atomic-resolution structure of the floating particles was investigated with an aberration-corrected TEM (FEI Titan 80–300 operated at 300 kV, Thermo scientific, Waltham, MA, USA).

2.3. DFT Calculation of Interface Structure between Floating Particles and β-Ga₂O₃ Matrix

First-principles calculations were performed using density functional theory (DFT) with the projector-augmented wave (PAW) method, as implemented in the VASP code [40,41,42]. The exchange–correlation potential was described within the generalized gradient approximation formulated by Perdew–Burke–Ernzerhof (PBE) [43]. The O’s 2s²2p⁴, Ga’s 3d¹⁰4s²4p¹ and Ir’s 5d⁸6s¹ electrons were treated as valence electrons. Based on the experimentally determined atomic stacking sequence, a series of β-Ga₂O₃/Ir interface models were built and labeled as I–V. An energy cutoff of 500 eV and a Gaussian smearing of 0.05 eV were used for all computations. The Monkhorst-Pack k-point meshes were chosen as 8 × 8 × 8, 4 × 16 × 9 and 3 × 2 × 1 for the bulk β-Ga₂O₃, the bulk Ir and the β-Ga₂O₃/Ir interface models, respectively [44]. The ionic relaxation was considered as convergence when the force on every atom was less than 5 meV/Å.

3. Results and Discussion

Figure 1 shows a photograph of the β-Ga₂O₃ crystal grown from the melt containing the floating particles using the CZ method. We observed some black areas existing in the β-Ga₂O₃ crystal, indicating that the floating particles were embedded into the crystal. Figure S1 shows the XRD pattern of this sample. All the diffraction peaks can be attributed to β-Ga₂O₃ (JCPDS Card No. 43-1012), and no other impurity related with Ir was found. The Raman (Figure S2) and XPS (Figure S3) spectra also show the existence of a β-Ga₂O₃ crystal. No signal associated with the floating particles was detected, possibly because they were wrapped and merged in the β-Ga₂O₃ crystal, together with their extremely low concentration. In order to ensure the existence of the floating particles, the crystals were cleaved layer by layer along (100) planes until the particles appeared the surface, as shown in Figure 2. From the FESEM image in Figure 2a, it can be seen that there were some irregular particles with bright contrast (marked by blue arrows) outcropping the surface of the matrix. The bright contrast indicates the excellent conductivity of the particles. The size of the particles was measured at about 2~15 μm. Figure 2c–e show the EDX mapping images of the area marked by a red square in Figure 2b. Clearly, the particle (e.g., bright region) was only composed of Ir, and no other element was detected (Figure 2d). The dark area consisted of O and Ga (Figure 2c,e), indicating that the matrix was β-Ga₂O₃. As such, the FESEM analysis indicated that the floating particles were made of pure Ir embedded in the β-Ga₂O₃ crystal.

To further clarify the composition and structure of the floating particles, a specimen was prepared using an FIB for TEM observation. Figure 3a shows a cross-sectional TEM image of the specimen at a relatively low magnification. It is clear that the interior of the sample exhibited a much darker contrast relative to the exterior. The interface between these two areas with completely different contrasts was sharp and straight. Combined with SEM-EDX data and the FIB process, the dark interior corresponds to the floating particles, while the bright area is the β-Ga₂O₃ matrix. Figure 3e shows the HAADF-STEM image of the magnified region marked by a red square in Figure 3a, confirming the sharp and straight interface. Since the contrast of the HAADF-STEM image is positively related to the atomic number of an element (Z-contrast image), the part with brighter contrast was made of elements with a larger atomic number. As determined from EDX mapping images in Figure 3f–h, the lower right part was mainly composed of Ir, and the upper left one consisted of Ga and O, indicating that the composition of the floating particles was Ir. In addition, selected-area electron diffraction (SAED) was employed to determine the structure of the floating particles. Figure 3b–d shows a series of SAED patterns of the region marked by blue square in Figure 3a along three different zone axes. The diffraction patterns along the $\left[10\overline{3}\right]$, $\left[\overline{1}12\right]$ and $\left[01\overline{1}\right]$ zone axes can be attributed to Ir, which had a face-centered cubic (fcc) structure. This demonstration was confirmed by the characterization of another sample (Figure S4). As such, we can conclude that these floating particles only consisted of fcc-structured Ir.

After cooling, the floating particles were embedded into the matrix of β-Ga₂O₃ to form the β-Ga₂O₃/Ir interface, which was first studied using the aberration-corrected HAADF-STEM technique (Figure 4). In Figure 4a, the sharp and straight interface can clearly be observed. Due to the different crystal structures of β-Ga₂O₃ and Ir, only lattice fringes associated with Ir appeared along the [211] zone axis of fcc-structured Ir. Careful observation of the atomic-scale image (Figure 4b) indicates that the β-Ga₂O₃/Ir interface had small continuous fluctuations at the nanoscale, but it could still maintain the overall flatness in a wider range. The atomic driving force of such interface relaxation will be discussed later. In addition to the interface structure, the crystallographic orientation of the β-Ga₂O₃/Ir interface was then determined using the HAADF-STEM image and diffraction pattern analysis. Figure 4c,d present the typical HAADF-STEM images of β-Ga₂O₃ and Ir away from the interface position observed from the [010]_Ga2O3 and [211]_Ir zone axes, respectively. It is evident that the plane in the β-Ga₂O₃ phase parallel to the interface was (001), as clarified by the fast Fourier transformation (FFT) pattern in the inset of Figure 4c. Combined with the diffraction pattern seen in the FFT image in Figure 4d and atomic position analysis, we could also determine that the plane in the Ir phase parallel to the interface was a high-index ($\overline{4}$17) facet.

To gain deeper insights into the atomic feature of the β-Ga₂O₃/Ir interface, we also performed first-principles calculations to determine its specific atomic configuration and reveal the mechanism of interface fluctuation observed experimentally (see calculational method for the details). In this study, we built five possible interface models between Ir and β-Ga₂O₃ with different interfacial terminations and stacking sequences (namely model I–V in Figure 5). Then, we calculated the binding energy according to the following formula:

$$W_{\mathrm{bind}}=(E_{\beta -{\mathrm{Ga}}_2{\mathrm{O}}_3}+E_{\mathrm{Ir}} -E_{\mathrm{int}})/A,$$

(1)

where A is the interface area, $E_{\mathrm{int}}$, $E_{\beta -{\mathrm{Ga}}_2{\mathrm{O}}_3}$ and $E_{\mathrm{Ir}}$ are the total energies of the interface model and the two slab models containing the isolated β-Ga₂O₃ and Ir parts, respectively. As shown in Figure 5, the binding energy for model Ⅰ was 35.87 J/m², which was the largest one among the five models. This result indicates that model Ⅰ was the most stable atomic configuration. This is understandable since in model Ⅰ, at the interface, β-Ga₂O₃ was O-terminated, which could provide more opportunities for Ir-O bonding. This demonstration was further confirmed by the relaxed atomic structure of model Ⅰ, as shown in Figure 6. It was found that various Ir-O bonds were formed between β-Ga₂O₃ and Ir layers. This strong interaction forces Ir atoms at the interface to deviate from their original atomic position, resulting in the fluctuation in the β-Ga₂O₃/Ir interface observed in the experiment. As such, first-principles calculations revealed that the Ir-O covalent bonding at the interface played a key role in inducing the local interface fluctuation.

As is well-known, β-Ga₂O₃ is inclined to decompose into different gallium oxides and eventually metallic Ga around the melting point, especially in an oxygen-deficient atmosphere. In an Ir crucible, the bottom often lacks oxygen, facilitating the formation of Ga in this region. Subsequently, in situ-generated Ga corrodes the Ir crucible to form eutectic or intermetallic GaIr. Some GaIr particles float on the surface of the melt driven by the flow. Due to the sufficient oxygen at the top of the crucible, the GaIr particles are believed to be oxidized, eventually leading to Ir floating particles. This demonstration is definitely supported by the characterization results for the floating particles. In addition, the fact that no floating particle was generated in the absence of the Ga-based compound implies the critical role of Ga in the formation of floating particles. This result also excludes the possibility of the formation of floating particles directly through the diffusion of Ir from the crucible. Based on the formation mechanism of the floating particles, we can propose several strategies to inhibit or remove them. First, a sufficient amount of oxygen can be blown into the bottom of the crucible to inhibit the formation of Ga. Second, a protective layer can be coated on the inwall of the Ir crucible to block the reaction between Ga and Ir. Third, the floating particles can be removed by a mechanical arm. It should also be noted that the still increasing price of Ir has caused some problems in the mass production and application of Ga₂O₃ wafers. However, the exciting thing is, researchers are developing more novel growth methods for Ga₂O₃ single crystals using less iridium or no iridium to solve this problem.

4. Conclusions

In summary, we employed the HAADF-STEM-EDX mapping technique to identify the floating particles formed during the CZ growth of β-Ga₂O₃ crystals. Distinguished from the general perspective, we demonstrate that the floating particles were composed of Ir with an fcc structure. The β-Ga₂O₃/Ir interface showed a sharp and straight characteristic on the whole with small fluctuations locally. The stablest atomic structure of the interface was built through DFT calculation, which was stabilized by Ir-O bonding. In addition, several strategies to inhibit or remove the floating particles were proposed. This work clarifies the argument on the composition and structure of the floating particles, which is of great importance for the growth of high-quality β-Ga₂O₃ single crystals.