The III-nitrides have attracted much attention owing to their wide range of band gap, high thermal conductivity, and high electrical resistivity. On account of the similar lattice constant and coefficient of thermal expansion with Al content AlGaN epilayer, AlN has become the irreplaceable substrate materials for solid and ultraviolet luminescence devices.^[1] Currently, halide vapor phase epitaxy (HVPE) is one of the most promising methods to grow high quality AlN single crystals owing to its relatively high growth rate and ability to obtain high UV transparent films.^[2] On the other hand, the nitrogen source ammonia, which is often used in the HVPE growth of AlN, will react with the chloride of Al to form various complexes. The pre-reaction between AlCl₃ and NH₃ before reaching the substrate consumes a large amount of Al source, decreasing the growth rate of AlN and deteriorating its surface.^[3] To suppress the pre-reaction, many experiment researches and theoretical predictions of using N₂ as precursor gas in the metal-organic chemical vapor deposition (MOCVD) or plasma-enhanced atomic layer deposition (PEALD) growth of III-nitrides have been reported.^[4–6] Our recent research^[ad] also found that using N₂ as nitrogen source in HVPE grown AlN could effectively inhibit the pre-reaction. The growth rate of AlN films grown by using N₂ increased linearly with the increase of the input ratio of N₂ and HCl flux (V/III ratio), rather than decreased exponentially in the growth by using NH₃ as the nitrogen source. This may be an assistant to illustrate that it is more advantageous to replace NH₃ with N₂ in HVPE growth of AlN.

Nevertheless, the low reactivity of N₂ at the conventional growth temperature of HVPE grown AlN introduced extra difficulties compared to the more mature growth process of using NH₃ as nitrogen source. In this paper, we present a comprehensive comparison research on the influence of nitrogen sources on AlN films grown by HVPE.

The 2-in AlN films were grown on sapphire substrates with a miscut of about 1.5° toward the [] axis at 1500 °C using a home-made high-temperature HVPE system.^[7] AlCl₃ was generated by introducing HCl gas over the Al metal at 520 °C in the source zone of the reactor. Before growth, the sapphire substrates were initially heated to 1500 °C in a flowing NH₃ gas and then cleaned in H₂ at 1500 °C for 5 min. After the cleaning of the surface of the substrate, the input flux (N₂ + H₂ or NH₃) rates were changed to the designed values for AlN growth. All samples were divided into four groups (A–D) depending on the nitrogen sources and the growth methods. Samples in groups A and B used N₂ as the nitrogen source, while samples in groups C and D used NH₃ as the nitrogen source. Among them, groups B and D were grown by a two-step method strategy. The detailed growth conditions of these films are listed in Table 1. The thicknesses of these samples were 5–10 μm. The quality of the AlN epilayers was examined by x-ray diffraction (XRD, Bruker D8 Discover), optical microscope (Leica DM400M), atomic force microscope (AFM, NanoScope III tapping-mode), and transmission electron microscopy (200 kV FEI Talos F200X).

The FWHMs of (0002)- and ()-plane x-ray rocking curves (XRCs) for the as-grown AlN films are shown in Fig. 1. The AlN epilayers grown by using N₂ and NH₃ as nitrogen sources are presented with red (N₂) and blue (NH₃) symbols, respectively. Among them, different shapes of markers distinguish one- and two-step growth methods. The FWHMs of the symmetric (0002) and skew-symmetric () ω-scans are used to measure the tilt and twist mosaic of AlN,^[8] respectively. It can be seen from Fig. 1(a) that the samples with FWHM of (0002) reflection smaller than 300 arcsec are AlN (N₂) films. The distribution of the FWHMs in Fig. 1(a) shows that the AlN (N₂) films can obtain smaller FWHM of (0002) reflection than the AlN (NH₃) films. In Fig. 1(b), the samples with FWHM of () reflection below 500 arcsec are grown by the two-step method, and there is no significant difference for AlN (N₂) and AlN (NH₃) films. The XRD measurements reveal that using N₂ as nitrogen source in HVPE growth of AlN tends to obtain homogenous crystal orientation along the c axis. The two-step growth method seems more conducive to improving the crystal quality of AlN but also needs further evidence. According to the FWHMs of XRCs in Table 1, the dislocation densities of the as-grown AlN films are estimated. The results show that the samples with the lowest dislocation density among each group are A2, B1, C1, and D1, and the corresponding dislocation densities are 2.8 × 10⁹ cm⁻², 1.9 × 10⁹ cm⁻², 2.7 × 10⁹ cm⁻², and 1.8 × 10⁹ cm⁻², respectively.

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. The FWHMs of (0002) and () plane XRCs for as-grown AlN films. The samples grown with different nitrogen sources are shown in different colors.

To reveal the influence of N₂ and NH₃ nitrogen sources on the surface morphology of the HVPE grown AlN, films listed in Table 1 were prepared for a more intensive study. Optical micrographs (OMs) in Figs. 2 and 3 show the surface features of the as-grown AlN (N₂) and AlN (NH₃) films, respectively. The surfaces of the AlN (N₂) films grown by one-step method show a rough morphology on the macro scale. Many island structures that are several hundred micrometers in diameter appear on the surfaces. Nevertheless, the surfaces of the AlN (N₂) films grown by two-step method become smoother than the one-step films. The density and size of the islands appearing on the surfaces are greatly reduced. This indicates that an optimized two-step method is effective for the stress control on AlN (N₂) films. The surfaces of the AlN (NH₃) films in Fig. 3 are relatively smoother than the AlN (N₂) films, and have no large islands. The effect of the two-step method on improving the surfaces of the AlN (NH₃) films is not obvious, which is different from that of AlN (N₂) films.

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. Optical micrographs of AlN (N2) samples: (a) A1, (c) A2, (e) A3, (g) A4; (b) B1, (d) B2,(f) B3, (h) B4.

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. Optical micrographs of AlN (NH3) samples: (a) C1, (c) C2, (e) C3, (g) C4; (b) D1, (d) D2,(f) D3, (h) D4.

Table 1.

Table 1.

Table 1. The growth conditions and FWHM values of (0002)- and ()-plane XRCs for AlN epilayers. . Sample HCl flux/sccm N2 flux/sccm Input V/III XRC FWHM/arcsec (0002) () A1 30 200 40 310 630 A2 20 300 90 384 533 A3 50 1000 120 177 686 A4 30 1000 200 138 713 B1 30 50 10 430 470 50 20 2.4 B2 10 1000 600 160 700 50 1000 120 B3 10 1000 600 224 475 20 1000 300 B4 10 1000 600 245 645 30 1000 200 C1 50 15 0.9 507 560 C2 50 50 3 413 608 C3 60 60 3 409 657 C4 30 40 4 429 624 D1 60 60 3 450 470 50 30 1.8 D2 60 60 3 380 540 60 50 2.5 D3 50 50 3 338 522 60 60 3 D4 30 100 10 370 540 50 50 3

Table 1. The growth conditions and FWHM values of (0002)- and ()-plane XRCs for AlN epilayers. .

Figure 4 shows the AFM images of four AlN samples that have the lowest dislocation density in each group as mentioned above. Among them, samples A2 and C1 were grown by one-step method, samples B1 and D1 were grown by two-step method. The normal directions of steps on both AlN epilayers are consistent with the off-cut direction of the sapphire substrates toward the [] axis. The difference is that, the steps on the AlN (N₂) films are undulating, while those on the AlN (NH₃) films are continuous and straight. The RMS roughness of the AlN (N₂) films is a little smaller than that of the AlN (NH₃) films for both one-step and two-step growth methods as shown in Fig. 4. Based on the results of optical microscopy and AFM measurements, it can be found that the two-step growth method can effectively improve the surface morphologies of the AlN (N₂) films. But it is still more difficult to control the surface morphologies of AlN (N₂) films than AlN (NH₃) films, due to the less mature growth technology for the former.

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. The 10 μm × 10 μm AFM images of the as-grown AlN samples: (a) A2, (b) B1, (c) C1, and (d) D1.

To further investigate the threading dislocations evolution in the AlN films, the cross-sectional TEM weak beam dark-field (WBDF) images of samples A2, B1, C1, and D1 were obtained with g vectors of 0002 and , respectively. The vector of g = 0002 is commonly used to observe screw-component dislocations and inversion domains.^[9] As can be seen in Fig. 5, the screw dislocations density of the AlN (N₂) films is smaller than that of the AlN (NH₃) films, which is consistent with the results of XRD. The vector of is used to identify the edge-component threading dislocations (TDs).^[9] It can be found that both AlN films have a distinct edge dislocation merging process near the substrates. The most notable is that many edge (or mixed type) dislocations of sample B1 in Fig. 5(d) have a large angle of bending and then annihilate at about 1 μm above the substrate. It is speculated to be caused by the growth mode change during the two-step growth process.^[10,11] When the V/III ratio is high at the first growth step, the vertical growth rate is higher than the lateral growth rate.^[12] Then, large islands will form on the surface. When the V/III ratio becomes lower at the second growth step, the lateral growth will be dominant.^[12] Ultimately, the mirror force caused by the merging of islands drives the dislocations bending. However, this annihilation mechanism of dislocations is not obvious in the two-step grown AlN (NH₃) film as shown in Fig. 5(h).

Fig. 5.

	Figure Option ViewDownloadNew Window
	Fig. 5. TEM weak beam dark-field images of samples (a), (b) A2, (c), (d) B1, (e), (f) C1, and (g), (h) D1. Each sample was recorded with g vectors of 0002 and , respectively. The white arrows in (a) indicate the location of inversion domains.

It is also found that a high density of columnar inversion domains (IDs) (indicated by white arrows in Figs. 5(a) and 5(c)) formed above the substrates in all AlN films. The origin of these IDs in AlN was reported to be related to the decomposition of the sapphire substrate.^[13] These inversion domains were determined by the comparison of the contrast between the V-shaped columns and the surrounding matrix when switching from g = 0002 and in non-centrosymmetric axis [].^[14–16] Since the columnar inversion domains in our experiment all formed close structures and did not extend to the surface, the overall polarity of the films was not influenced by them. The circles in Fig. 5(c) point out several V-shaped structures composed by inversion domains. These locations correspond exactly to the region with the lowest density of dislocation in Fig. 5(d). We speculate that the process of the island merging in sample B1 may affect the distribution of the inversion domains. The specific relationship between the inversion domains and island merging is difficult to determine from the present experimental results, and further investigations are needed.

The influence of N₂ and NH₃ nitrogen sources on the crystal quality and surface morphology of AlN films grown by HVPE has been comprehensively investigated. In most cases, AlN films grown using N₂ nitrogen sources have smaller FWHMs of (0002) plane XRCs than those grown using NH₃ nitrogen sources, while having no significant difference for the FWHMs of () plane XRCs. An optimized two-step method can improve the surface morphology and reduce the dislocation density of AlN films, and it is especially effective for the AlN films grown using N₂ as the nitrogen source. At present, it is still more difficult to control the crack and surface morphology of AlN films with thicknesses of 5–10 μm grown using N₂ as nitrogen source compared to that grown using NH₃, due to the less mature growth technology for the former.