1. Introduction
Nanocolumns (NCs) have proven effective for lateral stress relaxation; thus, they exhibit low defect densities in III-nitride heterostructures. This property is particularly beneficial for systems with large lattice mismatch, such as indium nitride (InN), which is typically grown on foreign substrates. InN is an interesting and potentially important III-nitride semiconductor material with superior electronic transport properties [
1]. InN has attracted considerable research attention concerning the development of optoelectronic devices (e.g., full-color light-emitting diodes, solar cells, and optochemical sensors [
2]) because of its room temperature direct optical band gap in the near-infrared region (
Eg = 0.7 eV) [
3]. Moreover, the electron mobility of InN can exceed 2100 cm
2/(V·s) with a residual carrier concentration approaching 8 × 10
17 cm
−3 at room temperature [
4]. However, the relatively low decomposition temperature (<600 °C) in the III-nitride materials [
5], large lattice mismatch with universal substrates, and impurity-prone surface of InN hinder the development of InN-related devices. Furthermore, the narrow growth window of InN (between 500 °C and 550 °C) limits device processing. InN nanocolumns have been grown on Si(111) substrates with buffer layers such as Si
3N
4 [
6], GaN, low-temperature InN [
7], AlN/-Si
3N
4 double-layers [
8], InN/-Si
3N
4 double-layers [
9], and InGaN [
10].
Al
2O
3 has been widely used as a substrate for heteroepitaxial InN growth despite its insulating property and large
a-axis lattice mismatch with InN (~25%). Si(111) is a more suitable substrate for InN growth because it has a smaller lattice mismatch with InN (~8%), better doping properties, and greater thermal conductivity compared to Al
2O
3. β-Si
3N
4 has a lattice constant of 7.602 Å along its
a-axis [
11], which is only slightly larger than twice the lattice constant of InN (
a = 3.548 Å). If twice the lattice constant of InN is considered a nominal lattice constant, the lattice mismatch between Si and InN will be small.
InN nanocolumns have been grown using various methods, including radio-frequency (RF) molecular beam epitaxy (MBE) [
12], RF sputtering [
13], and metal–organic chemical vapor deposition (MOCVD) [
14]. However, because of the low dissociation efficiency of ammonia, these methods require large amounts of ammonia to obtain the flux of active nitrogen required for producing high-quality nitride. Although substantial progress has been made in the development of plasma sources to mitigate this nitrogen problem, the InN growth rates of the above methods are still lower than that achieved via metal–organic molecular beam epitaxy (MOCVD). RF-MOMBE, which is performed in a growth chamber with a load lock and a base pressure of ~10
−9 Torr, combines the characteristics of both MBE and MOCVD and is suitable for the mass production of InN. In the RF-MOMBE growth of InN, a precursor such as trimethylaluminum or trimethylindium (TMIn) serves as the group-III precursor, and atomic nitrogen generated by RF plasma is the source of N. Due to low dissociation temperature and high N
2 equilibrium pressure over InN, plasma-assisted systems are thought to be suitable for InN growth. Compared to the MOCVD growth of epitaxial nitride films, RF-MOMBE generally allows for a lower growth temperature [
15]. In our previous study, the RF-MOMBE growth temperatures of InN-related alloys were all lower than the corresponding MOCVD growth temperatures [
16]. Despite the promising characteristics of RF-MOMBE, few studies have reported the RF-MOMBE growth of hexagonal pyramid-like InN nanocolumns. Therefore, in this study, we focused on nitride growth at low temperature, including AlN growth at 800–900 °C, InN growth at 450–550 °C, and GaN growth at 750–850 °C.
According to previous studies, InN materials grown via RF-MOMBE have often been deposited often sapphire substrates. Silicon (Si) is a promising material for the growth of group-III–V materials. The good thermal conductivity of Si is especially interesting for electronic applications, including low-cost, high-brightness light-emitting diodes, and high-power chips. However, few studies have grown InN nanostructures on Si(1111) substrates via RF-MOMBE. Nitride thin films could be suitable buffer layers for the growth of InN on Si wafers and other lattice-mismatched group-III–V heteroepitaxial systems.
Previous studies have shown that InN grown on nitrided Si(111) substrates exhibits good crystallinity. Kumar et al. reported that InN films deposited onto Si substrates with different buffer layers exhibited high crystallinity and wurtzite lattice structures; furthermore, the InN/-Si
3N
4 double buffer layer exhibited a minimum full-width at half-maximum (FWHM) of the E
2(high) mode [
9]. Segura-Ruiz et al. [
17] reported the growth of misoriented and tilted InN nanorods on Si(111) substrates with amorphous Si
xN
y buffer layers. However, Park et al. [
18] the deposition of porous Si
xN
y buffer layers with numerous nanometer-sized holes under certain experimental conditions; such a porous buffer layer could be a good template for the low-temperature deposition of nitride layers. Nishikawa et al. [
19] reported that both substrate nitridation and growth initiation via In droplet formation are essential for the growth of InN nanocolumns with excellent
c-axis orientation. InN nanostructures such as nanowires, nanotubes, and nanocolumns have attracted considerable attention because the quantum confinement effect may yield novel functions and improved performance [
20]. Lee et al. [
21] reported the predeposition of In before the growth of InN nanocolumns; InN nanocolumns were obtained by forming droplets of Au + In solid solution on Au-coated substrates following by nitriding these droplets.
In this study, hexagonal pyramid-like InN nanocolumns were grown on nitrided Si(111) substrates via catalyst-free RF-MOMBE in conjunction with substrate nitridation. The substrate was nitrided at 900 °C for 60 min. InN nanocolumns were subsequently grown on the nitrided Si(111) substrate at 500–540 °C. The effects of growth temperature on the structural, optical, and electrical properties of the InN nanocolumns were then investigated.
2. Materials and Methods
InN nanocolumns were grown on nitrided mirror-like Si(111) wafers as substrates. The growth chamber was evacuated to a base pressure of 10−9 Torr using a turbomolecular pump. TMIn, which served as the group-III precursor, was delivered to the growth chamber by heating the metal–organic source in the absence of a carrier gas. Active nitrogen radicals were supplied by a RF plasma source (13.56 MHz). The Si(111) wafers were cleaned on a wet bench using the standard Radio Corporation of America process. The wafers were then dipped in buffered oxide etching solution for 5 min to remove native surface oxides. Subsequently, the substrates were placed in the growth chamber and thermally cleaned at 900 °C for 90 min under ultrahigh vacuum to remove any oxide residues in preparation for nitridation and InN growth.
The Si(111) substrates were nitrided at 900 °C for 60 min and then cooled to 500–540 °C for InN growth. The N2 (99.9999%) flow rate was 1 sccm, and the RF plasma power was 300 W. The dynamic pressure of the growth chamber during deposition was approximately 2 × 10−5 Torr.
The structural properties of the InN nanocolumns were characterized by X-ray diffraction (XRD; Bruker D8 ADVANCE, AXS GmbH, Karlsruhe, Germany) with a Cu-Kα X-ray source. The surface morphologies of the InN nanocolumns were examined by field-emission scanning electron microscopy (FE-SEM; Hitachi S-4300 FE-SEM, Hitachi, Ltd, Chiyoda-ku, Japan), while the InN microstructure was analyzed by field-emission transmission electron microscopy (FE-TEM; JEOL ARM200F, JEOL Ltd, Toyoko, Japan) with an electron voltage of 200 kV. Cross-sectional TEM specimens were prepared using a dual-beam focused ion beam (Seiko SII-3050; manufactured by FEI, Hillsboro, OR, USA). Raman spectroscopy (HR800; Horiba Jobin Yvon Labram, Bensheim, Germany) was performed at room temperature in backscattering geometry using the 514.5-nm Ar+ laser line for excitation. X-ray photoelectron spectroscopy (VG ESCA/XPS Theta Probe; Thermo Fisher Scientific Inc., Waltham, MA, USA) with Al Kα (hν = 1486.6 eV) radiation was used to characterize the bonding characteristics of the elements in the films.
3. Results and Discussion
Figure 1 shows the XRD patterns of the InN nanocolumns grown at various temperatures on the nitrided Si(111) substrates. All patterns show the peaks of Si(111) (2
θ = 28.44°) and Si(222) (2
θ = 58.86°), a strong InN(0002) peak (2
θ = 31.36°), a weak InN(10–13) peak (2
θ = 56.9°), and the InN(0004) peak (2
θ = 65.4°). The XRD results indicate the formation of highly <0001>-oriented hexagonal InN nanocolumns on the nitrided Si(111) substrates, similar to the results of Feng et al. [
6]. The peaks at 2
θ = 44.1° and 64.4° are attributed to the XRD sample stage.
The
c-axis lattice parameter of the InN nanocolumns was calculated to be approximately 0.57 nm, in agreement with the value reported for InN bulk crystal [
22]. The results indicate stress relaxation in the InN nanocolumns after growth. The FWHM values of the XRD peaks were determined from the
ω scans of the InN(0002) reflection (results not shown). The FWHM values of the samples grown at 500 °C, 520 °C, 530 °C, and 540 °C were approximately 366, 360, 389, and 480 arcmin, respectively, indicating that the broad peaks can be attributed to the high defect densities and amorphous In–O compounds in the InN nanocolumns. The crystalline quality was slightly better in the InN nanocolumns grown at 520 °C compared to those grown at 500 °C, implying that the growth temperature was within the reactive regime.
Figure 2 presents SEM micrographs showing the surface morphologies and cross-sections of the InN nanocolumns. All of the InN nanocolumns grew preferentially along the <0001> direction. Growth began from a thin polycrystalline InN nucleation layer on top of the nitrided Si
3N
4 ultrathin buffer layer. These results reflect the fact that the InN nanocolumns were grown via chemical reaction from TMIn and RF N
2 plasma. The average growth rate of the InN nanocolumns was approximately 0.8 μm/h. The InN growth rate achieved by RF-MOMBE is higher than that of conventional MBE. In this study, the InN nanocolumns exhibited three-dimensional (3D) features with varying nanometer-scale structures and shapes ranging from pyramid-like columns to pyramids depending on growth temperature. The plane-view SEM images show that aligned, self-assembled InN nanocolumns with diameters of 60 to 120 nm formed uniformly over the entire substrate at a high density of approximately 10
10 cm
2 at growth temperatures of 500 to 540 °C. The densities of InN nanocolumns grown at 500 °C, 520 °C, 530 °C, and 540 °C were approximately 5.3 × 10
10 cm
2, 4.8 × 10
10 cm
2, 3.7 × 10
10 cm
2, and 1.3 × 10
10 cm
2, respectively. Thus, the density decreased with increasing growth temperature. The density, size, and shape of the InN nanocolumns were mainly determined by the diffusion of In adatoms. During the initial stages of the growth process, the formation of TMIn and/or In droplets enhanced nucleation at specific locations, resulting in columnar growth [
23]. Thus, since the diffusion of In increases with temperature, the density of InN nuclei was insufficient for columnar growth at high temperature. The large column diameter can likely be attributed to the large seed size and low density of nuclei during the initial growth stage at high growth temperature, which gave rise to a high lateral growth rate and cracking of the TMIn precursor (i.e., reaction of TMIn with nitrogen), leading to large grain size. The typical nanocolumn height was approximately 240 nm. The nanocolumn diameter increased from 60 to 130 nm as the growth temperature was increased from 500 °C to 540 °C. The InN nanocolumns grown at 500 °C and 520 °C were similar in size. Some of the nanocolumns were oriented along the
c-axis without tapering, whereas others were tilted with respect to the substrate normal. When the growth temperature was further increased, nanocolumns with diameters and heights of approximately 80 and 220 nm (530 °C) or 130 nm and 240 nm (540 °C), respectively, were observed. The deviation angles of the InN nanocolumns grown at 500 °C, 520 °C, and 530 °C were determined to be 75°, 75°, and 80°, respectively. The height:diameter (
h/
d) ratio of the nanocolumns decreased as the growth temperature increased to 540 °C. The growth temperature was shown to affect the diffusion of In adatoms along nanorod sidewalls [
16]. Since the In diffusion length increases with temperature, high temperature leads to the decomposition of In–N. In our previous study, continuous InN films were grown via RF-MOMBE at various temperatures in the range of 500–540 °C. Therefore, the above results demonstrate that the size, morphology, and optimal growth temperature of InN nanocolumns were influenced by the nitrided Si(111) substrate. The results are similar to those observed for InN nanorods grown via MBE [
24].
TEM was used to characterize the microstructures of the InN nanocolumns and nitrided Si
xN
y. Cross-sectional TEM specimens were prepared from the grown layers. The dark-field TEM image of a nitrided Si
xN
y ultra layer (
Figure 3a) clearly shows a rough surface, which was likely caused by plasma-beam damage to the Si
xN
y/Si(111) surface. The observed V-like holes are consistent with the N
2 plasma-beam damaging a nonstoichiometric silicon nitride layer [
25].
Figure 3b shows a cross-sectional bright-field TEM (XTEM) image of InN nanocolumns grown on a nitrided Si(111) substrate at 520 °C. The electron beam incident direction in
Figure 3b is parallel to the [
]
InN direction, while the vertical axis is parallel to the [0001]
InN direction. Similar to the SEM results, the XTEM images indicate a rough InN surface and V-like holes in the interface between InN and the Si substrate. For InN nanocolumns grown on the V-like holes, the crystals likely grew along the sidewalls of the holes, resulting in a deviation in growth angle. As a result, the initial growth of adjacent InN nanocolumns occurred in the voids between the substrate and InN nanocolumns. A previous study found that amorphous Si
xN
y can develop gradually on bare Si(111) before the growth of InN [
26]. Thus, the epitaxial relation is lost on the nitrided area, resulting in misoriented nanorods or a coarse-grained structure on Si(111).
Figure 3c–e shows a high-resolution TEM (HRTEM) image of the top of an InN nanocolumn along the [
] direction of InN along with the corresponding FFT pattern and lattice image. The results indicate that no metallic In was present on the top of the InN nanocolumn, and the InN nanocolumns were single-phase hexagonal crystals with preferred orientations along the
c-axis. As shown in
Figure 3e, the InN surface is the (0002) plane of InN, and direct measurement based on the lattice image yielded a lattice parameter of
c = 5.7 Å. These results are in good agreement with previously reported XRD results [
27].
The compositional depth profile and binding energy of the InN nanocolumns grown at 520 °C were evaluated by XPS (
Figure 4).
Figure 4a shows the variation in the atomic ratios of silicon, carbon, nitrogen, indium, and oxygen along the etching direction from the atmosphere to the InN interface toward the InN/Si(111) interface. The layer structure of InN on nitrided Si(111) was clearly observed. The elemental concentrations of indium and nitrogen were slightly reduced in the layer moving in the etching direction. The carbon atomic concentration decreased rapidly after the initial etching step and then remained nearly constant. Unintentional oxygen incorporation occurred because of the impurities remaining in the vacuum background, and a high concentration of oxygen was detected near the interface due to oxygen incorporation. Specht et al. [
28] reported similar results InN grown via MBE. High concentration of oxygen was also probably caused by voids between the substrate and InN nanocolumns. These results are in good agreement with the TEM images. However, no diffraction spots corresponding to crystalline oxygen-related impurities (e.g., In oxide) were observed in the selected area electron diffraction patterns; thus, the oxygen might exist as impurities or point defects in the film. Furthermore, XPS indicated a low concentration of carbon on the surfaces of the InN nanocolumns, which may have originated from the TMIn precursor and/or adsorbed carbon. Iwao et al. [
29] indicated that carbon was incorporated as a background element during the plasma-assisted metal–organic molecular beam epitaxy (PA-MOMBE) growth of InN, although the properties were improved due to they prefer to form the CO and/or CO
2 with oxygen atoms that have slightly high concentration. Oxygen contamination was previously reported in InN grown via RF-MOMBE [
30].
Figure 4b shows the In3d XPS spectra of InN nanocolumns grown at 520 °C obtained from 440 to 458 eV at various depths of the film. Overlapping peaks were deconvoluted using XPSpeak41 software. Based on the XPS spectra, the binding energies (BEs) of In3d
3/2 and In3d
5/2 were determined to be 451.2 and 443.6 eV, respectively. This indicates that the difference in the BEs of the In3d
3/2 and In3d
5/2 spin–orbit doublet was almost constant at 7.6 eV with an intensity ratio of 4:3, in agreement with the literature [
31]. The spectrum corresponding to the surface of the InN nanocolumn shows a broad peak at 444.2 eV and a shoulder peak at 444.6 eV, which are assigned to the BEs of In–N [
32] and In
2O
3 (BE = 444.6–444.9 eV) [
33], respectively. This may be related to the absorption of H
2O on the InN surface. Specht et al. [
28] reported similar results for InN grown via MBE.
The Raman spectrum of the wurtzite InN nanocolumns on a nitrided Si(111) substrate (
Figure 5) exhibits E
2(high) peaks at 489 and 590 cm
−1 for the A
1(LO) mode under 488 nm excitation. The Raman peak positions of the A
1(LO) and E
2(LO) modes of the InN nanocolumns in this study agree well with previously reported values, and the narrow linewidth is similar to that of high-quality InN films [
34,
35]. However, the strain-free Raman frequencies of the E
2(high) high and A
1(LO) modes of InN remain uncertain, and the reported values vary widely from 483 to 495 cm
−1 for the E
2(high) mode and 580–596 cm
−1 for the A
1(LO) mode [
36].