Study on the Mechanism of Process Parameters in Pulsed Deposition Technology on the Growth and Quality of AlN Thin Films +

AlN templates on Silicon or Sapphire substrates are available for fabricating power electronics and deep ultraviolet optoelectronic devices. For more wafer specifications, please contact us at [email protected].

III-nitride materials occupy a critical position in power devices, high-frequency communications, and ultraviolet optoelectronic devices due to their excellent bandgap, high thermal conductivity, and outstanding optoelectronic properties. Among them, aluminum nitride (AlN), with the widest bandgap of 6.2eV, high UV transparency, and excellent piezoelectric properties, is not only a core material for deep ultraviolet light-emitting diodes (DUV-LEDs) and quantum cascade lasers but also an indispensable buffer layer for gallium nitride on silicon (GaN-on-Si) epitaxy.

However, the high-quality growth of AlN faces multiple challenges. First, Al atoms have low surface mobility on the substrate, requiring high growth temperatures (typically exceeding 1200°C) to achieve satisfactory crystal quality, yet high temperatures exacerbate parasitic gas-phase reactions and introduce thermal expansion issues. Second, Al precursors (e.g., TMAI) and N precursors (e.g., NH₃) readily undergo adduct reactions in the gas phase, forming TMAINH₃ adducts, which subsequently generate AlN particles, reducing growth efficiency and film quality. Furthermore, during heteroepitaxy of AlN on silicon substrates, lattice mismatch and differences in thermal expansion coefficients introduce significant tensile stress, affecting the performance of subsequent devices.

Pulsed deposition technology offers a new approach to address these issues. Unlike conventional continuous-flow MOCVD, pulsed technology alternates the supply of Group III and Group V precursors, thereby changing the dominant species on the substrate surface—from Al–N to Al—which significantly enhances the mobility of Al atoms. Research shows that a well-designed pulsed process can achieve high crystal quality AlN thin films at lower temperatures, providing an important reference for substrate engineering in III-nitride devices.

1. Parasitic Reaction Mechanism and Purge Time Optimization in Pulsed MOCVD (PMOCVD)

1.1 Formation of Adduct Reaction and the Effect of V/III Ratio

In conventional steady-state MOCVD, TMAI and NH₃ are introduced simultaneously into the reactor, where they rapidly undergo an adduct reaction in the gas phase to form the TMAINH₃ intermediate. Numerical simulation studies by Lin and Chen have revealed that as the V/III ratio increases, the generation of TMAINH₃ significantly rises, leading to a marked decrease in the average growth rate. Specifically, when the V/III ratio increases from 1 to 10,000, without a purge pulse, the average growth rate per cycle decreases substantially, due to the consumption of effective precursors by the stronger adduct reaction.

Fig. 1 Variation of average AlN growth rate with time in pulsed MOCVD under different V/III ratios

1.2 Trade-off between Benefit and Cost of Pure H₂ Purge Pulse

Inserting a pure H₂ purge pulse between the Group III and Group V pulses effectively reduces gas-phase contact between TMAI and NH₃, thereby suppressing the adduct reaction. Numerical simulation results from Lin and Chen show that at a V/III ratio of 10,000, the average growth rate continuously increases as the purge time increases from 0 to 0.1 seconds. However, longer purge times also increase the cycle time, which negatively impacts the overall growth rate.

The researchers reveal a key trade-off between the benefit and cost of purge time. When the V/III ratio is small (e.g., equal to 1), the adduct reaction is weak; the benefit of reducing TMAINH₃ formation does not compensate for the adverse effect of longer cycle times, so the growth rate decreases with increasing purge time. When the V/III ratio equals 100, increasing the purge time from 0 to 0.05 seconds raises the growth rate, but further increasing from 0.05 to 0.1 seconds causes a decrease; the optimal purge time is about 0.05 seconds, corresponding to a maximum growth rate of approximately 0.75μm/h. When the V/III ratio is increased to 10,000, the adduct reaction is extremely strong, and the benefit of inserting an H₂ purge pulse outweighs the cost of extended cycle time; thus the growth rate monotonically increases with purge time, though the maximum remains around 0.7μm/h.

Fig. 2 Relationship between average growth rate in the second cycle and purge time under different V/III ratios

The numerical simulations above reveal the nature of parasitic reactions and the basic rules of purge optimization from a mechanistic perspective. Based on this, experimental studies further explore how adjusting other parameters of the pulsed process (such as the number of cycles) can improve crystal quality.

2. Fine Control of AlN Crystal Quality by PMOCVD Cycle Number

2.1 Evolution of Surface Morphology

Hisyam et al. systematically investigated the effect of the number of PMOCVD cycles (0, 35, 70, 140) on the properties of AlN epilayers grown on Si(111) substrates. Field emission scanning electron microscopy (FESEM) surface images show significant evolution of surface morphology with increasing cycle number: at 0 cycles (nominal growth), a small number of non-uniformly distributed tiny voids are present on the surface; at 35 cycles, voids become larger but lower in density; at 70 cycles, the surface becomes denser; at 140 cycles, the entire sample surface exhibits a dense and flat morphology.

Fig. 3 FESEM surface images of AlN on Si with different PMOCVD cycle numbers: (a) 0, (b) 35, (c) 70, (d) 140

Atomic force microscopy (AFM) analysis further reveals the transition in growth mode. At 0 cycles, high-density fine grain islands indicate three-dimensional growth. As the cycle number increases, grain island size increases and density decreases; the root mean square (RMS) surface roughness significantly drops from 3.49nm at 0 cycles to 1.53nm at 35 cycles, indicating a quasi-two-dimensional growth mode. However, at 70 and 140 cycles, the RMS slightly increases to 2.13nm and 2.20nm, respectively, attributed to the suppression of smaller grain islands by larger ones.

Fig. 4 AFM topography images (2×2 μm²) of AlN surfaces with different PMOCVD cycle numbers: (a) 0; (b) 35; (c) 70; (d) 140

2.2 XRD Characterization of Crystal Quality and Strain Analysis

High-resolution X-ray diffraction (HR-XRD) rocking curve measurements show that the FWHM of AlN(0002) decreases significantly from 0 to 70 cycles, dropping from 4046 arcsec to 2756 arcsec; the FWHM of AlN(10-12) also decreases from 4783 arcsec to 3450 arcsec. However, when the cycle number further increases to 140, the FWHM of (10-12) slightly rises to 3720 arcsec, consistent with the increase in RMS observed by AFM.

Reciprocal space mapping (RSM) of the AlN(10-15) diffraction peak was used to calculate the in-plane strain state. All samples are under tensile strain, caused by the difference in thermal expansion coefficients between AlN and Si(111). As the number of PMOCVD cycles increases from 0 to 35 and 70, the in-plane tensile strain significantly decreases from 1.10% to 0.28% and 0.28%; but when the cycle number increases to 140, the strain rises back to 0.55%. A clear correlation exists between crystal quality and strain: the decrease in FWHM occurs simultaneously with the reduction of in-plane tensile stress. The condition of 70 cycles is identified as optimal, giving the best crystal quality and minimum strain.

Fig. 5 Reciprocal space maps of AlN(10-15) and in-plane strain variation with different PMOCVD cycle numbers: (a) 0, (b) 35, (c) 70, and (d) 140; (e) Variation of in-plane strain with PMOCVD cycle number

The above studies demonstrate that optimizing the number of pulse cycles can significantly improve AlN quality within the conventional MOCVD framework. However, this process still requires relatively high growth temperatures (>1000°C). To further reduce the thermal budget and expand substrate compatibility, researchers have begun exploring low-temperature pulsed deposition strategies based on novel precursors.

3. Precursor Innovation and Template Engineering for Low-Temperature Pulsed Chemical Vapor Deposition

3.1 Low-Temperature Deposition Using TDMAA and Anhydrous Hydrazine

Conventional precursor trimethylaluminum (TMA) thermally decomposes above 350°C, readily leading to carbon contamination. McLeod et al. employed a metal precursor with higher thermal stability—tris(dimethylamido)aluminum(III) (TDMAA)—combined with a highly reactive nitrogen source, anhydrous hydrazine (N₂H₄), to deposit AlN thin films by pulsed CVD at substrate temperatures of 400°C and 580°C.

X-ray photoelectron spectroscopy (XPS) analysis shows that at both deposition temperatures, the AlN films exhibit a near 1:1 stoichiometric ratio with very low carbon and oxygen impurity levels. In the C 1s spectrum, no low binding energy peak corresponding to Al–C bonds is observed, attributed to the absence of Al–C bonds in the TDMAA precursor structure.

Fig. 6 High-resolution XPS spectra of AlN thin films deposited by pulsed CVD at different substrate temperatures

3.2 Temperature-Driven Crystallinity Transition and Columnar Grain Growth

Grazing-incidence X-ray diffraction (GI-XRD) analysis shows that AlN films deposited at 580°C exhibit a clear c-axis preferred orientation on both Si(111) and Si(100) substrates, with FWHM values of the AlN(002) peak of 0.73° and 0.75°, respectively. In contrast, AlN films deposited at 400°C are nanocrystalline on Si(111) and amorphous on Si(100). X-ray reflectivity measurements further indicate that films deposited at 580°C have higher density, reaching 3.11g/cm³ on Si(111).

Transmission electron microscopy (TEM) observations reveal more detailed structural differences: films deposited at 400°C show randomly oriented nanocrystalline grains about 5nm in diameter, while films deposited at 580°C exhibit a columnar grain structure that extends through the entire film thickness, with grain diameter increasing with film thickness. This observation is highly consistent with the conclusion from the numerical simulations by Lin and Chen that Al atom mobility increases with temperature.

Fig. 7 Bright-field and dark-field TEM images of pulsed CVD AlN films at different deposition temperatures: (a, c) 400°C, nanocrystalline; (b, d) 580°C, columnar

3.3 Template Layer Effect: Epitaxial Breakthrough on SiC

McLeod et al. further explored the ability of AlN films deposited by pulsed CVD at 580°C to serve as template layers for inducing preferred orientation growth in subsequently sputtered AlN. On Si(111), both the 580°C pulsed CVD template layer (31nm) and the 400°C atomic layer annealed template layer (21nm) significantly enhanced the (002) orientation of sputtered AlN, with FWHM of 0.64° for both, whereas direct sputtering without a template gave an FWHM of 0.79° and mixed orientation.

Experiments on 4H-SiC substrates were even more groundbreaking. Due to the better lattice matching between AlN and SiC, the 20nm template layer deposited by pulsed CVD at 580°C achieved an epitaxial relationship with the SiC substrate. Selected area electron diffraction showed a single-crystal-like pattern, and high-resolution TEM revealed regularly arranged and continuously uninterrupted lattice fringes of AlN and SiC. Fast Fourier Transform (FFT) analysis further confirmed that from the SiC substrate to the CVD AlN layer and then to the sputtered AlN layer, the FFT patterns from each region exhibit the expected characteristics of 4H-SiC and the wurtzite structure of AlN, with consistent c-axis orientation. In contrast, AlN directly sputtered on SiC shows an amorphous initial layer of about 3–4nm and polycrystalline diffraction rings.

Fig. 8 (a) Cross-sectional bright-field TEM image of a 20nm AlN template layer deposited by pulsed CVD at 580°C; (b) Corresponding selected area electron diffraction (SAED) pattern; (c) High-resolution TEM (HR-TEM) image

Fig. 9 Fast Fourier Transform (FFT) analysis confirming c-axis orientation consistency: FFT patterns extracted sequentially from the 4H-SiC substrate (regions I‑V), the 580°C pulsed CVD AlN template layer (regions VI‑X), and the sputtered AlN layer (regions XI‑XV)

In summary, the quality of pulsed-deposited AlN is governed by the synergistic optimization of purge time and V/III ratio, the thermal stability of precursors, and the precise control of pulse cycle number. Introducing a pure H₂ purge under high V/III ratio conditions effectively suppresses adduct reactions, while the combination of TDMAA and anhydrous hydrazine enables columnar or epitaxial growth at 580°C, substantially reducing the deposition temperature. Meanwhile, 70 cycles of PMOCVD is identified as the optimal condition on Si(111) substrates, yielding the narrowest XRD rocking curve peak width and minimal in-plane tensile strain. Future research can further explore novel precursors and process uniformity on larger substrates to promote the practical application of pulsed-deposited AlN in power electronics and deep ultraviolet optoelectronic devices.

Whether you need AlN templates for research or for industrial applications, please contact us email at [email protected] and [email protected].

 

References:

1. Lin, W. J., & Chen, J. C. (2022). Numerical study of growth rate and purge time in the AlN pulsed MOCVD process. Crystals, 12(8), 1101.
2. Hisyam, M. I., Shuhaimi, A., Norhaniza, R., Mansor, M., Williams, A., & Mat Hussin, M. R. (2024). Study of AlN epitaxial growth on Si (111) substrate using pulsed metal–organic chemical vapour deposition. Crystals, 14(4), 371.
3. McLeod, A. J., Ueda, S. T., Lee, P. C., Spiegelman, J., Kanjolia, R., Moinpour, M., … & Kummel, A. C. (2023). Pulsed chemical vapor deposition for crystalline aluminum nitride thin films and buffer layers on silicon and silicon carbide. Thin Solid Films, 768, 139717.