Influence of sputtered AlN buffer on GaN epilayer grown by MOCVD

The ex situ sputtered AlN buffer and GaN epilayer grown on top of it by metalorganic chemical vapor deposition were studied comprehensively by a variety of techniques including atomic force microscope, high resolution x-ray diffraction, Raman and x-ray photoelectron spectroscopy characterizations. It exhibited that the AlN buffer deposited by using sputtering technique could be oxidized with exposure in atmosphere. Such oxidation phenomenon significantly influences the characteristics of GaN epilayer, for example leading to poor surface morphology, high dislocation density, and large compressive stress. This study demonstrated the effect of oxygen impurities on GaN growth and has an important guiding significance for the growth of high-quality III-nitride related materials.


Introduction
III-nitrides are wide band-gap semiconductors suitable for light emitting diodes (LEDs) [1], power electronics and highfrequency devices [2] due to their excellent physical properties, such as direct bandgap, high breakdown voltage, high electron saturation and electron mobility [3]. The progress of III-nitrides is significantly remarkable since the crystalline quality of GaN epilayer can be greatly improved with two-step growth technology by metal-organic chemical vapor deposition (MOCVD) [4]. However, with maturing IIInitride technology in recent years, some limitations of the conventional approach, albeit very successful, have emerged. * Author to whom any correspondence should be addressed.
Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. Therefore, further reduction of dislocation density and residual strain is necessary for improvement of device efficiency and reliability [5]. Recently, sputtering technique, a mature deposition method widely employed in semiconductor industry, has been considered a viable alternative to MOCVD growth of IIInitrides, or at an aid [6]. It has been demonstrated that a thin ex situ sputtered AlN buffer layer greatly improves crystalline quality of upper MOCVD epitaxial GaN layer, particularly, it has been confirmed the validity in LEDs, electronic devices and photo detectors [7][8][9][10][11]. In practical, such technique has been widely employed in LED industries [12], as it could greatly reduce the growth time and cost. However, to the best of our knowledge, the characteristics of such a composite template, especially the influence of sputtered AlN buffer layer on GaN grown on top of it has not been systematically studied.
In this paper, we found that the crystal quality and surface morphology of GaN epilayer could significantly deteriorate with the increased the exposure time of the sputtered AlN/sapphire template in atmosphere. It can be mainly attributed to the oxidation effect on the sputtered AlN, confirmed by detailed x-ray photoelectron spectroscopy (XPS) characterization. Importantly, we also confirmed that oxygen impurities obviously affect GaN nucleation and screw dislocation generation. This result is of great significance for using ex situ sputtered AlN as a nucleation layer for epitaxial growth of GaN, especially for high-quality material epitaxial preparation and other related applications.

Experiment
In this study, AlN buffer layers with a thickness of 25 nm are sputtered on 2 inch c-plane sapphire substrates by using a physical vapor deposition system. The whole preparation process is carried out in a high-vacuum chamber. The tray with the substrate is transferred into the vacuum chamber by robot arm, later the tray is heated up to 650 • C, and a mixture gas of argon and nitrogen in a ratio of 1:8 is introduced, which has the flow of 30 sccm and 240 sccm, respectively. The pressure during the sputtering process is kept at 3.72 × 10 −3 Torr, and the effective power is fixed to 3000 W. In addition, at the beginning of the sputtering process, a trace amount of oxygen was introduced to control the lattice mismatch between the AlN buffer and the sapphire substrate. An electric field is added between the target and the tray to ionize the Ar ions. Under the action of the electric field, they hit the surface of the target at a high speed, then the sputtered Al ions react with nitrogen and deposit on the surface of the sapphire substrate [11,13]. In this study, as-deposited fresh AlN buffer sample was denoted as sample AlN-A. In order to investigate the evolution process of the sputtered AlN buffer layer, as-deposited sample was intentionally exposed in atmosphere for 5 days, 10 days and 20 days, respectively. Accordingly, these samples are denoted as sample AlN-B, AlN-C and AlN-D, respectively. After exposure, GaN epilayer was grown on these sputtered AlN buffer samples by a MOCVD system with identical growth condition. During this growth process, trimethylgallium (TMGa) and ammonia (NH 3 ) were used as the precursors for gallium (Ga), and nitrogen (N) sources respectively. The hydrogen was used as a carrier gas with a total flow of 80 standard cubic centimeter per minute (SLM). First, the AlN films was baked with high-temperature cleaning (1050 • C) for 300 S to remove the surface contamination. Second, a low temperature (1020 • C) with low V/III (800) and high reactor pressure (450 mbar) to grow 20 nm u-GaN layer. Third, the temperature was increased to 1050 C for growing GaN layer with high V/III (1600) and low reactor pressure (150 mbar). The total thickness of GaN epilayer is 2 µm. Accordingly, GaN epilayers grown on sample AlN-A, AlN-B, AlN-C and AlN-D were denoted as sample GaN-A, GaN-B, GaN-C and GaN-D, respectively.
The crystal quality of samples was measured by the highresolution x-ray diffraction (HR-XRD). The surface morphology of samples was characterized by optical microscope (OM) and atomic force microscopy (AFM). The in-plane residual stress of GaN epilayer was characterized by Raman spectroscopy. Finally, the surface state of AlN buffer layer was investigated by x-ray photoelectron spectroscopy (XPS), which were recorded by ESCALAB 250Xi XPS system (ThermoFisher Scientific) equipped with a focused monochromatized Al-Kα radiation(1486.6 eV) Figure 1 shows the OM of GaN epilayers grown on sputtered AlN buffer layers with different exposure time in atmosphere. It demonstrates that the surface of GaN epilayers is obviously changed with the increase of exposure time of AlN buffer. As shown in figure 1(a), for sample GaN-A grown on fresh AlN buffer layer, the surface of GaN is fairly smooth and clean. However, some surface fluctuation can be observed on the top of sample GaN-B with underlying AlN buffer layers being exposed in the atmosphere (shown in figure 1(b)). With further increase of the exposure time, stripe shape surface morphology is clearly presented in GaN-C shown in figure 1(c). Finally, there are a large number of hexagonal hillocks covering almost the entire surface for the sample GaN-D in figure 1(d). The hexagonal base of the hillocks on sample GaN-D is estimated about 20-50 µm in size. Based on previous study, the origin of the formation of the hexagonal hillocks in GaN films is mainly attributed to that inversion domains in GaN could nucleate at thin platelets of oxygen containing amorphous material [14][15][16]. Therefore, the significant morphology changes of GaN epilayers could be related to the oxidation effect on AlN buffer with the increase of exposure time. The more undulation of the GaN surface, the higher degree of oxidation of the AlN buffer, Then, its morphology gradually changed from smooth surface to stripe shape, finally tending to form hillock morphology.

Results and discussion
As AFM images of GaN epilayers shown in figure 2, it can be observed obviously step flow on the surface, indicating that the growth model of GaN on all the sputtered AlN buffer are step flow growth mode. However, it should be noticed that with the AlN buffer exposure time increase, the roughness for growth GaN become larger, i.e. 0.176 nm, 0.197 nm, 0.249 nm and 0.258 nm, respectively. Moreover, we can also see that the step flow's width and height of gradually increases. This is probably associated to the modification of the surface state of AlN buffer after being exposed to the atmosphere, which may affect the free path of diffusion on the surface of the source molecule for the GaN growth at the initial stage of growth. Figure 3 shows the x-ray rocking curve and Raman scattering spectra of GaN-A, GaN-B, GaN-C and GaN-D, respectively. The measured full width at half maximum (FWHM) of the GaN epilayer along (002) plane is 69.4, 96.9, 106.2 and 169.1 arcsec for GaN-A, GaN-B, GaN-C and GaN-D samples, respectively, as shown in figure 3(a). According to the Mosaic model [17], the screw dislocation densities of GaN are calculated as 1.05 × 10 7 cm −2 , 2.04 × 10 7 cm −2 , 2.45 × 10 7 cm −2 and 5.93 × 10 7 cm −2 , respectively. We can see a five-fold increase in screw dislocations in GaN-D than that of GaN-A. Figure 3  2.67 × 10 8 cm −2 , 2.58 × 10 8 cm −2 , 3.07 × 10 8 cm −2 and 3.09 × 10 8 cm −2 , respectively. As demonstrated more clearly in figure 3(c), it can be seen that the quality of the GaN (002) plane crystals will significantly deteriorate, if growing on the AlN buffer with longer exposure time. But the quality of the crystal for the (102) plane is not obviously variety. In order to show the difference between different type of dislocations, we used the method of the tails of the rocking curves to analyze the as prepare samples, this is the nondestructive method for the threading dislocation analysis [18,19], it shows the same result for the variation of crystal quality of GaN in GaN epilayers. The XRD data shows that the influence of AlN buffer have a much larger impact on the formation of screw dislocations during the epitaxial growth of GaN, but has a relatively small impact on the formation of the edge dislocations. It can be seen that oxygen segregation to screw dislocations in GaN epitaxial layer from XRD results. It may be due to the substitution of oxygen for nitrogen, which is to extend over many monolayers for the open core dislocation It has been reported that the major origin of screw dislocations in GaN is related to appearance of oxygen impurity [20]. Therefore, such significant increased screw dislocations in GaN epitaxial layer strongly associated with the oxidation effect of the AlN buffer with the increase of exposure time.
Raman scattering measurements are performed in backreflection geometry using a Raman microscope with nonresonant 532 nm (2.3 eV) excitation. It can be seen the peak of GaN high-E 2 plasmon-longitudinal optical (LO) phonons at 570.8 cm −1 (GaN-A) shift to 571.4 cm −1 (GaN-D), as shown in figure 3(d). Note that the observation of the peaks for the E 2 (LO) mode drift in the direction of higher wave number. As we know, the E 2 (LO) phonon peak is used to evaluate the strain/stress present in the GaN films and the phonon frequencies of unstrained GaN E 2 (high) is equal to 567.6 cm −1 . The E 2 (LO) phonon mode of Raman spectrum is sensitive to the amount of strain and has been widely used in the characterization of GaN. A quantify stress by the following equation [21,22]: where δ is the residual stress and ∆ω is the E 2 (LO) phonon peak shift. It can be calculated that the stress existing in the GaN/AlN/sapphire is approximately drift from 0.761 GPa to 0.904 Gpa, indicating that the top GaN were under compressive stress due to higher thermal expansion coefficient of sapphire than that of GaN. More accurate stress values in GaN epilayers can be determined as from 0.574 GPa to 0.957 GPa for all four samples based on a reliable approach [23]. It can be observed that such biaxial compressive stress of GaN epilayer become larger with AlN buffer exposure time longer, which is also confirmed by XRD characterization by using 2theta/omega scan. Such results suggested that the tensile stress during the GaN epitaxy process is released through the formation of dislocations. Generally speaking, such tensile stress functions as the compensation of the compressive stress, caused by the sapphire to GaN epilayer during the cooling process. And it ultimately leads to a larger compressive stress in GaN epilayer after cooling. Furthermore, to study the evolution process of AlN buffer by sputtering deposition with the change of exposure time, the as prepared AlN buffer samples were also studied comprehensively by a variety of techniques including AFM, HR-XRD, XPS. Firstly, AFM surface images of the sample AlN-A, AlN-B, AlN-C and AlN-D were recorded in tapping mode within 2 µm × 2 µm scanning area and microscopic images are shown in figure 4. AFM image shows there is no essential change in morphology, basically columnar, which is the classic morphology of low temperature AlN growth prepared by sputtering method. The roughness is 0.339 nm, 0.554 nm, 0.362 nm and 0.407 nm for AlN-A, AlN-B, AlN-C and AlN-D, respectively. Figure 5 shows the (002) plane x-ray rocking curve of the AlN buffer on the sapphire. It can be seen that the AlN buffer has a wurtzite crystalline structure, the FWHM of sample AlN-A, AlN-B, AlN-C and AlN-D are 265.7, 274.6, 297.1 and 281.1 arcsec, respectively. While the (102) plane is too thin to obtain a signal. The relatively stable FWHM of (002) of all four samples proved that the overall crystalline quality of AlN buffer did not change significantly after exposure in atmosphere. Such mild fluctuation in FWHM of (002) rocking curve could probably be attributed to the severe oxidation effect related modification of surface state of sputtered AlN, which is associated to the polycrystalline nature of buffer deposited by sputtering method at relatively low temperature. High density of grain boundary in polycrystalline AlN serve as a fast diffusion path for reactive oxygen atoms. As a result, a severe oxidation of AlN buffer as the time of exposure increases. Such conclusion can be strongly confirmed by detailed XPS characterizations as follow. Figures 6(a) and (b) show the XPS survey scans of AlN-A, AlN-B, AlN-C and AlN-D series buffers and magnify view of peak oxygen levels O 1S . The spectra confirmed that these AlN buffer were composed of the elements aluminum, nitrogen, oxygen and carbon [24,25]. Due to the high chemical  activity of Al, the remnant oxygen is likely to interact with the Al atoms. The most noteworthy feature in the spectra is that the prominent oxygen peaks are taller than those of aluminum. In addition, a considerable amount of carbon is observed, which is mainly from adventitious contamination at atmosphere. In order to analysis the surface chemical surface state, high-resolution spectra of Al2p and N1s were performed on the AlN buffer layers. Figure 7(a) shows the comparison of the Al2p peaks for all four AlN buffer samples [26,27]. The binding energy scale was calibrated by measuring adventitious C1s peak at 284.8 eV and A Shirley type background subtraction It can be observed with the increase of exposure time, the center of the peak shifts from 73.47 to 73.70 eV, indicating the drift of binding energy in the direction of high bond energy, what's more, Al2p peak has closely spaced spin-orbit components, it shows only one peak. Similarly, the center of N1s peaks shift from 396.60 to 396.87 eV, as shown in figure 7(b). Moreover, the intensity of the peak decreases with the increasing of exposure time. Normally, such intrinsic peak value of the material moves in the direction of high bond energy after oxidation as the bond energy of oxygen is relatively high electronegative [28]. Therefore, this result strongly suggested that AlN buffer samples were gradually oxidated with the exposure time increasing. Moreover, as shown in figure 7(a), it can also be noticed that the high-resolution Al2p scans curve is not symmetrical due to Al-O bonding on the right side of the curve. In order to analyze such oxidation effect more clearly, the Al2p photoelectron spectrum can be deconvoluted into two components assigned to Al-N and Al-O bonding. For example, as presented in figure 7(a), the spectrum of GaN-D can be deconvoluted by two dash curves, Al-N (73.90 eV) and    In all, the XPS results indicate that the Al2p binding energy of AlN buffer grown at low temperature fabricated by sputtering method is unstable. When the surface of AlN buffer is oxidized in atmosphere, the binding energy of aluminum and oxygen atoms is stronger than that between aluminum and nitrogen atoms. The oxidized area is easier to obtain gallium atoms molecules for preferential nucleation growth. Therefore, the enhanced oxidation effect with the increase of exposure time will generate higher density of the nucleation point for GaN growth. It will lead the formation of screw dislocations in GaN epilayer and finally result in the larger FWHM of the (002) plane of the GaN epilayer, as shown in figure 3(a). Moreover, such oxidation effect of the AlN buffer with the increase of exposure time will also induce the significant morphology changes of GaN epilayers as presented in figure 1.

Conclusion
In this study, we found that the surface morphology, crystalline quality and strain state of GaN epilayer grown on sputtered AlN buffer is significantly influenced by the oxidation effect of AlN buffer after exposure in atmosphere. It indicated that the oxygen atom segregation to screw dislocations in GaN epitaxial growth. It may have similar influencing factors for other nitrides epitaxy. According to our experiments, the sputtered AlN/sapphire template should be used for epitaxy growth promptly in order to achieve high crystal quality of the subsequent material and high reliability device applications.
Moreover, AlN/sapphire template should be stored in a nitrogen cabinet to delay its oxidation effect, although such oxidation process cannot be completely avoided entirely, there is still a trace of oxygen in the cabinet in the actual operation process. Such conclusions/suggestions should be applicable to other materials, which grow at low temperature condition and are prone to oxidation.
This work was partially supported by National Key

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.