Epitaxial AlN film with improved quality on Si (111) substrates realized by boron pretreatment via MOCVD

This study investigated the growth of AlN epitaxial films on 2-in. Si(111) via metal-organic chemical vapor deposition. By introducing trie-thylboron (TEB) during trimethylaluminum pretreatment, a nearly crack free AlN epilayer with a thickness of 500nm was acquired. The x-ray diffraction rocking curves of AlN (002) and (102) exhibited full width at half maximum values of 0.22 (cid:1) and 0.36 (cid:1) , respectively. Atomic force microscopy image analysis showed that after the introduction of TEB, larger grains appeared on the surface of Si(111) substrate, promoting the 3D growth pattern of the subsequent AlN buffer layer. Laytec reflection curves depicted the morphological transition from 3D to 2D growth mode during AlN deposition. At the same time, the curvature value was significantly reduced by 20 km (cid:3) 1 , and the Raman spectrum peak of E 2 (high) shifted from 648.7 to 652.5 cm (cid:3) 1 , indicating that the surface tensile stress was greatly reduced, effectively

2][3][4][5] Due to the high cost of the native AlN substrate, AlN is commonly grown on foreign substrates, including sapphire, SiC, silicon, etc. [6][7][8] Compared with sapphire and SiC, Si(111) emerges as a promising substrate for AlN growth owing to its cost-effectiveness, high thermal conductivity, widespread availability, large-scale production feasibility, and potential for seamless integration with electronics. 9However, the considerable $19% lattice mismatch between AlN and Si presents a formidable challenge, leading to an extremely high density of misfit dislocations (>10 10 cm À2 ) near the AlN/Si interface. 10The resulting interaction between this large lattice mismatch and a coefficient of thermal expansion (CTE) of $43% induces high biaxial tensile strains in AlN, compromising its crystal quality and causing cracks during cooling. 11rior to the growth of the AlN films on Si(111) by metal-organic chemical vapor deposition (MOCVD), it has been established that a trimethylaluminum (TMAl) pretreatment step is imperative. 12During this step, the introduction of TMAl precursor solely into the reactor is necessary to ensure the growth of high-quality AlN, primarily by safeguarding the Si surface from nitridation. 13Extensive research has been conducted on various parameters such as flow rates, duration, temperature, and pressure of the pretreatment, yielding substantial findings in this regard. 14The present study introduces TEB in the pretreatment process to investigate its influence on subsequent AlN growth, aiming to contribute valuable insights to the optimization of AlN nucleation layers on silicon for advanced electronic device applications.
AlN epitaxial films were grown on 0.5 mm thick 2-in.Si (111) substrates by Taiyo Nippon Sanso 3 Â 2 in.MOCVD with the structural diagram and process diagram depicted in Fig. 1.TMAl, TEB, and ammonia were used as precursors of Al, B, and N, respectively.Before growth, buffered oxide etchant (BOE) 7:1 treatment was used to remove the oxidation layer on the Si(111) substrates.Then, the Si(111) substrates were thermally cleaned under H 2 at 1100 C for 600 s to remove remaining oxide layer on the surface.Subsequently, the temperature was raised to 1150 C, and pretreatment processing was performed on the silicon substrates.To facilitate the investigation of different pretreatments' impacts on AlN epitaxial layer growth, samples A1 was pretreated with 32 lmol/min TMAl for 5 s, while sample B1 was pretreated with a combination of 24 lmol/min TMAl and 8 lmol/min TEB for 5 s.Subsequently, a 60 nm AlN buffer layer was deposited on the different pretreatment surfaces with a V/III ratio of 16 500 and a duration of 360 s, denoted as samples A2 and B2, respectively.Finally, AlN with thickness of 440 nm was grown on these two buffer layers for 1620 s by decreasing the V/III ratio to 110, designated as samples A and B, respectively.The pressure of reactor was kept at 50 mbar during the whole process.A 405 nm optical monitoring system of MOCVD was used to monitor reflectance and curvature of samples during the AlN growth process.The surface morphology images of AlN templates were taken by optical microscopy.The micromorphology of the AlN templates was measured by atomic force microscopy (AFM).The crystal quality of the samples was characterized by high-resolution x-ray diffractometer (HR-XRD) using rocking curve scans.Raman spectroscopy was recorded by 532 nm laser excitation in backscattering geometry at room temperature.
Figure 2(a) presents the 450 nm wavelength monitoring curve and time series of AlN growth, with both samples achieving a total thickness of approximately 500 nm.For the TMAl-pretreated sample A, the second peak of reflectance reaches its maximum value, indicating a swift transition from 3D to 2D growth mode throughout the growth process.Conversely, with the introduction of TEB in the pretreatment for sample B, the reflectance peak gradually rises over time, reaching its pinnacle at the fourth peak and stabilizing thereafter.This evolving trend suggests a progression in sample B from an initially rough surface to a progressively smoother one during growth, a transformation that aids in stress alleviation and defect mitigation.Furthermore, it is evident that the thickness of the buffer stage in sample B slightly exceeds that of sample A, hinting at a more pronounced 3D growth pattern.
In Fig. 2(b), the curvature of the samples as a function of time is illustrated.Remarkably, the curvature of both samples exhibits a consistent increase in tandem with the temperature ramp-up until the completion of AlN buffer layer.However, as the AlN growth process commences, the curvature of sample B consistently demonstrates a lower profile compared to that of sample A. By the conclusion of the growth period, a discernible difference of 20 km À1 is observed between the two samples.This noteworthy trend underscores the impact of introducing TEB during the pretreatment phase, resulting in a significant reduction in the concavity of sample B when compared to sample A. This finding suggests the considerable potential to minimize  warpage and alleviate tensile stress-induced cracking during the cooling process, emphasizing the practical implications of the TEB pretreatment technique in enhancing the structural integrity of AlN films.
Figure 3 provides a visual analysis of the surface morphology at the rounded edge, center, and flat edge of both samples A and B under a 500Â eyepiece.A clear distinction emerges upon visual examination, where sample A manifests numerous crack lines across the entire wafer.In contrast, sample B exhibits a notably improved condition, featuring a crack-free center and a few discernible crack lines at a distance of only 1 mm from the edge.This stark difference in crack occurrence highlights the efficacy of introducing TEB during the pretreatment process in effectively mitigating the cracking issues associated with AlN growth on Si substrates.
To study the effect of different pretreatment methods on AlN growth, atomic force microscopy (AFM) analysis was conducted on the pretreatment surface, buffer layer, and AlN layer.Figures 4(a of 1.21 nm, with grains' width of $30 nm and height of $8 nm.These variations in grain sizes and morphologies are likely to impact the subsequent growth of the buffer layer.Figures 4(b) and 4(e) compare the morphology of the two samples after buffer layer deposition, revealing roughness values of 1.78 and 2.48 nm, respectively.In comparison with sample A2, the buffer layer of sample B2 showcases a 3D morphology with larger dimensions and fewer defects, conducive to stress relief and defect mitigation during subsequent AlN growth processes.Figures 4(c) and 4(f) show the morphology of the two samples after growing a 500 nm AlN layer, with roughness values of 0.178 and 0.215 nm for samples A and B, respectively.It is evident that sample B exhibits fewer point defects and clearer atomic step morphology, indicating that the inclusion of TEB in the pretreatment process enhances material quality.
To investigate the influence of different pretreatments on the crystal quality of AlN films, HR-XRD patterns of samples A and B are characterized.Figures 5(a) and 5(b) illustrate the x-ray rocking curve of the symmetric (002) and asymmetric (102) orientations, respectively.For sample A, the x-ray diffraction rocking curves of AlN (002) and (102) exhibit full width at half maximum (FWHM) values of 0.3 and 0.52 , respectively.In contrast, the FWHM values of sample B on the (002) and (102) planes are 0.22 and 0.36 , respectively, which are much smaller than those of sample A and the latest reports of AlN grown on plain Si(111) by MOCVD listed in Figs.6][17][18] The FWHM values obtained from XRD are commonly used for estimating dislocations of AlN layers, which include both screw and edge dislocations.The density of screw dislocation (q s ) and edge dislocation (q e ) were calculated using the following formulas: 19 Here, b(002) and b(102) represent the FWHMs of ( 002) and (102) planes, while b c and b a correspond to the lengths of Burgers vector associated with lattice constants along c-and a-axial, respectively.The calculated values for q s and q e are 2.54 Â 10 9 and 1.96 Â 10 10 cm À2 for sample A and 1.37 Â 10 9 and 9.37 Â 10 9 cm À2 for sample B. Obviously, sample B exhibits superior quality, as indicated by the lower dislocation densities in both screw and edge dislocations.
In Fig. 6, the Raman spectra of samples A and B are presented.Both samples exhibit peaks corresponding to E 2 (high) at 648.7 and 652.5 cm À1 , A1(TO) at 616.4 and 619.2 cm À1 , and A1(LO) at 883.8 and 885.2 cm À1 , respectively.For comparison, the Raman spectra of bulk AlN crystal show E 2 (high), A1(TO), and A1(LO) peaks at 657, 610, and 890 cm À1 , respectively.The shift of the E 2 (high) peak frequency, as depicted in Fig. 6, is attributed to strain.The wavenumber of the E 2 (high) mode is 648.7 cm À1 for sample A and 652.5 cm À1 for sample B, with the FWHM of the E 2 (high) being 6.6 and 4.2 cm À1 .The observed shift suggests a nearly 8.3 and 4.5 cm À1 change in wavenumber, corresponding to the presence of tensile strain in samples A and B, respectively.Consequently, sample B is inferred to have a smaller tensile strain compared to sample A. Additionally, smaller FWHM means sample B has better crystal quality.
In conclusion, through the introduction of TEB during TMAl pretreatment, high-quality nearly crack-free AlN films were acquired, with a thickness of 500 nm.AFM revealed larger grains on the surface, which facilitated subsequent 3D island growth of the buffer layer.Laytec reflection curves demonstrated a morphological transition from 3D to 2D growth during AlN deposition.With TEB addition in the pretreatment, the curvature of AlN decreased significantly during the growth process, along with a 3.8 cm À1 blue shift decreased in the Raman spectrum peak, indicating a notable reduction in surface tensile stress.XRD rocking curves exhibited full width at half maximum values for the (002) and (102) reflections of 0.22 and 0.36 , respectively.This study not only advanced the understanding of AlN deposition processes but also paved the way for the practical realization of highquality AlN films with significant implications for advanced electronic and optoelectronic applications.

FIG. 2 .
FIG. 2. (a) The optical monitoring curve of 405 nm wavelength and (b) the curvature vs time sequence.

FIG. 3 .
FIG. 3. The surface morphology of sample A at rounded edge (a), center (b), and flat edge (c) and sample B at rounded edge (d), center (e), and flat edge (f) under the eyepiece at 500Â.