The growth of AlN samples used for this study was initiated with the deposition of the AlN rough layer (thickness ~ 0.3 µm) at low temperatures (950⁰ C), which acts as a buffer layer. On top of buffer layers, high temperature (1250⁰ C) AlN was grown to produce a 3µm thick layer with an atomically smooth surface. The AFM images of 2µm×2µm dimensions for 3.0 μm AlN layers are shown in Fig. 1(a) and Fig. 1(b) using H2 and N2 as carrier gases, respectively. Both the samples have a similar root mean square (RMS) surface roughness. The carrier gasses, N2 and H2, (growth environments) produce different morphologies of the buffer layers [8, 9]. The growth environments dependent morphologies stem from the differences in thermal conductivity of the substrates, the decomposition rates, and the diffusion coefficients of precursors due to differences in carrier gases [12]. We reported that the morphology of the buffer layer produced with the N2 carrier gas has two distinct types of islands that help reduce strain in the thick, high temperature, fully coalesced layers without any interlayer [13]. The AlN buffer layer produced with the H2 carrier gas, surface morphology is more like single-mode islands; thus, an early coalescence in subsequent high-temperature AlN layers happens, resulting in cracks around the thicknesses > 1.5 µm. A low-temperature interlayer was introduced to achieve a thickness of 3 µm for AlN layers grown with the H2 carrier gas, a proven method to mitigate the cracking [14].
Fig. 2 shows the cross-section SEM images of 3 μm samples grown with a) H2 carrier gas and b) with N2 carrier gas. We observed that the void terminates close to the sapphire substrate for samples grown with the H2 carrier gas due to early coalescence. Thus, we have to incorporate a low temperature 0.1 µm interlayer at the thickness of 1.5µm to achieve a 3 µm AlN thick layer. We have shown that the low-temperature buffer layer grown in the N2 environment has two distinct islands that impede early coalescence; thus, Fig. 2 (b) shows the termination of voids at the larger thickness parallel to the growth directions, which helps to grow thick crack-free AlN samples without the low-temperature interlayer.
The 3 µm AlN samples were thoroughly analyzed with x-ray diffraction; the details of the method were reported in the reference [8, 9]. The full-width-at-half-maximum (FWHM) of the asymmetric (10ī2) scan of III-nitrides (AlN, GaN, and AlGaN) can indicate the total dislocation density of the material. The FWHM of (10ī2) (off-axis) ω scan for 3 µm samples grown with H2 carrier gas was ~330 arcsec whereas it was 307 arcsec for 3µm sample grown with N2 carrier gas. The calculated screw dislocations were in the order of 107 cm-2, and total dislocation density was in the order of lower 109 cm-2 (using the Williamson and Hall method) [9, 15, 16]. The slightly lower off-axis value can be delayed coalescence or the absence of low-temperature interlayer in the AlN samples grown in N2 gas environment. The uncoalesced layers can act as a secondary buffer layer to grow slightly reduced dislocation high-quality thick AlN; however, the differences in dislocation densities in both cases are within the margins of error in the calculations.
The XPS survey scans and high-resolution (HR) scans were performed at two locations – the surface and 5 nm depth from the surface of the samples. We will analyze the scan at a depth of 5 nm only because of ambiguities in the surface scan due to surface contamination associated with sample handling. The full spectra of two AlN thin films and all HR spectra were calibrated to the (carbon 1 s) C1s peak at 284.6 eV. Survey scan shows that XPS spectra are composed of aluminum, nitrogen, carbon, oxygen, and maybe hydrogen. There is a chance that both samples might contain some hydrogen, but hydrogen cannot be detected as it has only one electron and a small photoionization cross-section. Fig. 4 (a) and 4(b) show the deconvoluted C1s peak for the AlN samples grown with N2 and H2 carrier gases. The main peak of the spectra of C1s states measured at the surface of AlN is around 284.6 eV has high intensity, which indicates C. We observed that the intensity of C1s drops significantly at 5 nm depth, leading to the actual amount of C in the sample. The C1s graph can be deconvoluted into four different chemical states such as C=C at 284.6 eV, C-C/C-N close to 286.6 eV, C-O close to 288.4 eV, and O-C=O bond close to 288.6 eV [10, 11]. The XPS high-resolution scan of (aluminum 2 p) Al2p chemical state at 5 nm depth is shown in Fig. 4(c) and 4 (d) for AlN samples grown in N2 and H2 with carrier gasses, respectively. The atomic concentration of Al is higher in the case of Fig. 4 (d). Note that the deconvoluted peak positions are the same for both deconvoluted peaks confirming the same type of ionic bonding states in both samples [17, 18].
Table 1. Atomic concentration (AC) of AlN samples at the depth of 5 nm grown in N2 and H2 environments.
AC of
AlN
|
Grown in N2 environment (%)
|
Grown in H2 environment (%)
|
Al
|
24.8
|
29.4
|
C
|
12.3
|
12.9
|
N
|
43.6
|
42.3
|
O
|
19.3
|
15.4
|
Table 1 shows the XPS detected elements after 5 nm Ar ion etching where components are expressed in terms of atomic concentration (AC) in percentage. We observe that both samples exhibit the presence of C and O. In the case of sample grown in the H2 gas environment, the amount of oxygen is slightly less than the sample grown in the N2 gas environment. On the contrary, the amount of Al is much higher in the case of the sample grown in H2 gas environment, which can be expressed as the Al-N stoichiometric ratio that is closer to the ideal case. The source of the C and O containing impurities can be following - a) oxygen/carbon incorporation during the growth process from the a) growth precursors and carrier gases, b) impurities absorbed when sample exposed to air, c) Ar+ ion sputtering to check the elements at 5 nm depth, d) from the XPS chamber [19].
To further investigate the presence of impurities and at different depts from the sample surface in the AlN films, the secondary-ion mass spectrometry (SIMS) was utilized in time-of-flight (ToF) mode. The SIMS measurements were performed at the Georgia Tech. Material Characterization Facility. The ToF SIMS machine uses two ion guns: one for sputtering only and another one for analysis purposes. For sputtering, Cesium ions were used with an energy of 2 keV for creating a crater and bismuth ions with 25 keV for analyzing the sputtered area along the way. To measure these samples, they were mounted on a sample holder using Cu tape, and the sputtering created a crater area of 300 μm2; the area of analysis by the bismuth ions was 150 μm2. Fig. 5 shows the average relative SIMS detected counts for Al2+, Al+, H+, N+, O+ (with respect to Al+ normalized to 1) for both AlN samples grown with N2 carrier and H2 carrier gases. No significant change was observed in the composition of various compounds up to 0.8 µm (and beyond data is not shown here). Compared to N2 grown samples, H2 grown samples have 33-39% less C, 33% less H, 41% less O; a similar trend was observed in XPS analysis of these samples.
To further understand the presence of impurities, the samples were studied by cathodoluminescence (CL). In III-nitride semiconductors near band edge emissions, such as free exciton or bound exciton recombination are characteristics of high-quality materials with lower impurity concentration and lower defect density [20, 21]. Fig. 6 shows CL spectra of the AlN sample grown with the N2 carrier gas and H2 carrier gas and the HVPE grown free-standing AlN. The peak close to 6.0 eV dominates in all the samples. Davis et al. simulated the CL spectra for aluminum gallium nitride, AlxGa1-xN, samples, where x is the composition of the aluminum [22]. For x=1, in those calculations, the pure AlN near band edge emission is close to 6.0 eV, which matches our measurements. If we closely observe these peaks, we can see that there is hump-like peaks are present in all the samples close to 3.8 eV, which is well known for oxygen presence [20]. We have grown our samples on sapphire substrates, so to exclude the idea that this peak can come from sapphire substrate oxygen signal, we have measured the free-standing AlN grown by hydride vapor phase epitaxy (HVPE). We observed a peak close to the 3.8 eV for the free-standing AlN sample, confirming that the source of the oxygen-related peak is the AlN, not the sapphire substrate. In addition, the electron penetration depth is only a few nm in the AlN samples in the case of 10kV accelerating voltage. So, there is less probability of incident electrons reaching the sapphire substrate. Thus, we can conclude with certainty that the source of oxygen signal is from the oxygen impurities in the AlN samples. According to Youngman et al., the luminescence peak position from oxygen-related defects in AlN shifts from 4 eV to 3.3 eV as the oxygen increases from less than 0.1 to about 0.8% [23]. We have assigned the peak at 3.8 eV, which agrees with the previously reported data [20]. The near band edge emission (NBE) peak (~ 6.0 eV) peak arises from the free excitons or excitons bound to the shallow donor or acceptor impurities. The NBE peak of the samples grown with N2 and H2 carrier shows a blue shift with respect to HVPE grown free-standing AlN sample, which is related to stress in the MOCVD grown samples. In the next section, we evaluate the stress in these films by Raman spectroscopy measurement.
The impurities, point defects, and extended defects alter the stress values in the epitaxially grown films. Raman spectroscopy has been previously demonstrated as an effective tool to calculate stress in thin films [24]. In the case of AlN, high-resolution phonon linewidth can provide the crystalline quality of the thin films. It has been previously reported that for single-crystal bulk AlN, FWHM of E2 (high) phonon mode is 3 cm-1 [25]. For a 3 μm AlN sample grown with N2 carrier gas, the E2 (high) phonon linewidth was 3.8 cm-1; on the other hand, a 3 μm AlN sample grown with of H2 carrier, it was 5 cm-1. We employed the biaxial stress coefficient to calculate stress based on the Raman shift. Biaxial stress coefficient () depends on the growth technique [24]. We considered 657.04 cm-1, the bulk sample E2(high) as the stress-free frequency, and calculated the stress based on ± 0.3 cm-1/GPa [24]. Figure 7 shows the normalized Raman spectra of E2(high) phonon mode where the blue shift towards the right for the grown samples indicates compressive stress. The compressive stress in AlN film results from the lattice mismatch between sapphire substrate, thermal expansion coefficient mismatch, and stress generated during the coalescence process and impurities [26]. We have previously demonstrated that, due to the delayed coalescence, the N2 carrier grown samples show less compressive stress [8, 9].