Impact of Impurities in Aluminum Nitride (AlN): From Doping Mechanisms to Optical Properties and Detection Methods +

AlN on Si templates and AlN on Sapphire templates are available for the preparation of ultraviolet light-emitting devices. For details, please consult through [email protected]

The wide bandgap and excellent ultraviolet transparency of AlN make it an ideal material for deep-ultraviolet light-emitting diodes (UVC-LEDs), laser diodes, and detectors. However, during practical growth processes, impurities—especially oxygen—inevitably incorporate into the crystal, thereby affecting device performance. Studies indicate that oxygen impurities in AlN typically exist in the form of substitutional nitrogen site (O_n), creating deep-level defects and potentially forming complex defect centers by combining with other vacancies, leading to changes in optical absorption and luminescence characteristics. Therefore, understanding the incorporation mechanisms, distribution patterns, and performance impacts of impurities is crucial for advancing AlN-based devices.

1. Source and Distribution of Oxygen Impurities: The Critical Role of Surface Roughness

Oxygen is one of the most common impurities in AlN, with concentrations reaching 10¹⁹–10²¹cm⁻³. In metalorganic chemical vapor deposition (MOCVD) growth, sources of oxygen may include reactor leaks, diffusion from gas lines, or source material contamination. Notably, research finds that the concentration of oxygen impurities is weakly correlated with growth conditions (such as V/III ratio) but closely related to surface roughness during growth.

When the AlN surface is rough, it exposes more non-c-planes (e.g., m-planes, a-planes), which exhibit stronger adsorption capacity for oxygen atoms. In contrast, the flat c-plane has a weaker capacity for oxygen adsorption. This phenomenon can be explained by atomic structure models: on the c-plane, an Al atom can bond with three N atoms, exposing only one unpaired electron; whereas on m- or a-planes, Al atoms bond with only 1–2 N atoms, exposing more unpaired electrons, thus more readily capturing environmental oxygen atoms. Consequently, surface roughness directly affects the spatial distribution uniformity of oxygen impurities in AlN.

Fig. 1 Structural differences during MOCVD growth when free Al atoms attach to different crystal planes

2. Impact of Impurities on AlN Optical Properties

2.1 Optical Absorption and Luminescence Induced by Oxygen Impurities

Oxygen impurities introduce a series of optical absorption bands in AlN. Theoretical calculations suggest that substitutional O_n can act as DX centers, causing absorption bands with peaks at 2.22eV and 3.06eV. These bands correspond to the broad absorption band observed experimentally in the visible-UV region, especially when oxygen concentration exceeds 10¹⁹cm⁻³, where the absorption extends from 2eV to the band edge.

Furthermore, cathodoluminescence (CL) spectra reveal that the luminescence peak of AlN films around 3–4eV is related to oxygen impurities. The intensity of this luminescence initially increases and then decreases with rising temperature, differing from typical thermal quenching behavior, reflecting the non-uniform distribution of oxygen impurities in AlN.

Fig. 2 Integrated intensity variation of oxygen impurity-related luminescence peaks with temperature for AlN films (samples A1, A2)

Fig. 3 Configuration coordinate diagram for substitutional oxygen impurity at the nitrogen site in AlN: optical transition paths and corresponding vertical transition energies for O_N in different charge states

2.2 Competitive Doping Effects of Carbon and Silicon

Carbon and silicon impurities also significantly affect the optical properties of AlN. Experiments show that when ([O]+[Si])/[C] < 3.3, a strong absorption peak appears at 265nm in AlN, which is associated with increased vacancy defects at high carbon concentrations. Silicon-dominated samples, however, exhibit an exponential absorption curve without distinct characteristic peaks.

Fig. 4 (a) Absorption coefficient vs. wavelength for AlN crystals with different silicon doping concentrations; (b) Absorption coefficient vs. wavelength for AlN crystals with different carbon doping concentrations

3. Rapid Detection and Prediction of Impurity Concentrations: Machine Learning-Assisted Spectral Analysis

Although traditional secondary ion mass spectrometry (SIMS) can accurately determine impurity concentrations, it is destructive, costly, and time-consuming. In recent years, rapid, non-destructive detection methods based on ultraviolet absorption spectroscopy have gained attention. Research has utilized Random Forest regression models to predict C, O, and Si concentrations in AlN using absorption spectra.

The model was trained using 37 sets of AlN samples with varying impurity concentrations, covering an absorption spectral range of 210–600nm. Results indicate that the absorption coefficient at 230nm correlates well with the overall impurity concentration (OIC), especially for samples with low carbon concentration. However, the model’s prediction accuracy heavily depends on the coverage of the training data, resulting in larger prediction errors for samples with unique impurity combinations. This suggests that future efforts should focus on expanding the sample library or introducing more complex neural network models to improve prediction robustness.

Fig. 5 Relationship between absorption coefficients at 230nm and 450nm and total impurity concentration for AlN samples with different impurity concentrations

Fig. 6 Comparison between SIMS measured values and Random Forest model predicted values for (a) undoped and oxygen-doped AlN, (b) silicon-doped AlN, (c) carbon-doped AlN

The incorporation of impurities—particularly oxygen—in AlN is significantly influenced by surface morphology, where rough surfaces exposing non-c-planes more readily adsorb oxygen atoms, leading to non-uniform impurity distribution. Oxygen impurities introduce multiple absorption bands and affect luminescence characteristics, while the presence of carbon and silicon alters optical absorption behavior through competitive doping mechanisms. Methods based on UV absorption spectroscopy combined with machine learning offer a new pathway for the rapid, non-destructive detection of impurity concentrations, but their reliability still depends on more comprehensive training data and more advanced model optimization.

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

 

References:

1. Zhang, Y., Yang, J., Liang, F., Liu, Z., Hou, Y., Liu, B., … & Zhao, D. (2024). Dependence of oxygen impurity concentration in AlN on the surface roughness during growth. Journal of Applied Physics, 135(16).
2. Yan, Q., Lyons, J. L., Gordon, L., Janotti, A., & Van de Walle, C. G. (2025). Oxygen impurities in AlN and their impact on optical absorption. Applied Physics Letters, 126(6).
3. Klump, A., Hartmann, C., Bickermann, M., & Straubinger, T. (2025). Prediction of impurity concentrations in AlN single crystals by absorption at 230 nm using random forest regression. CrystEngComm, 27(2), 184-190.