Charge-state control of carbon-related optical absorption in AlN

Technical Summary: Charge-state control of carbon-related optical absorption in AlN

Problem Statement
Aluminum nitride (AlN), with its 6.2 eV ultrawide bandgap, is a critical material for ultraviolet photonics and high-power electronics. However, its performance is often limited by sub-bandgap optical absorption (OA) caused by unintentional impurities and intrinsic defects. While a specific absorption band near 4.7 eV has been convincingly attributed to carbon on the nitrogen site ($C_N$), the microscopic origins of the broad absorption features spanning 2 eV to 4 eV remain contested. Previous studies have attributed peaks in this range to various defects, including nitrogen vacancies ($V_N$), aluminum vacancies ($V_{Al}$), and their complexes, often relying on decomposition of broad spectra into Gaussian components without independent structural or chemical identification of the responsible defects.

Methodology
To resolve these ambiguities, the authors employed a multi-modal approach combining experimental spectroscopy with first-principles calculations on bulk AlN substrates grown via physical vapor transport.

1. Experimental Techniques: The study utilized photo-induced electron paramagnetic resonance (photo-EPR) and optical absorption spectroscopy (OA) on the same samples. Samples were subjected to pre-illumination using light-emitting diodes (LEDs) at specific wavelengths (ranging from 1200 nm to 265 nm) to manipulate the charge state of defects.
   • Set 1 Samples: Initially contained $C_N$ in the negative charge state ($C_N^-$), which is EPR silent. Illumination at 4.7 eV (265 nm) photoionized $C_N^-$ to the neutral state ($C_N^0$), making it EPR active. Subsequent illumination at 2.3 eV (530 nm) returned the system to the dark state.
   • Set 2 Samples: Contained $C_N^0$ in the dark, allowing observation of minimal perturbation under similar LED exposure.
2. Theoretical Calculations: The authors performed hybrid functional (HSE) density functional theory (DFT) calculations to simulate the optical absorption line shape for transitions between the valence band and the $C_N$ acceptor level. A key methodological distinction was the explicit calculation of the energy dependence of optical matrix elements using a dense k-point grid, rather than assuming they were constant across the valence band. The line shape was modeled using a one-dimensional configuration coordinate approach to account for electron-phonon coupling.

Key Results

• Correlation of Charge State and Absorption: The study demonstrated a direct correlation between the neutral charge state of substitutional carbon ($C_N^0$) and the enhancement of optical absorption between 2 eV and 4 eV. In Set 1 samples, pre-illumination at 4.7 eV increased the OA in this range and generated the $C_N^0$ EPR signal. Illumination at 2.3 eV quenched both the EPR signal and the enhanced OA. In Set 2 samples, where $C_N^0$ was present in the dark, the OA was already high and minimally affected by the LEDs.
• Identification of the 3.4 eV Peak: Difference spectra (illuminated minus dark) revealed distinct peaks. While a 4.7 eV feature was attributed to the photoionization of $C_N^-$, the authors focused on lower-energy peaks. A peak near 3.4 eV was identified as the primary feature linked to the $C_N^0$ charge state.
• Theoretical Validation: Standard Franck-Condon approximations, which assume energy-independent optical matrix elements, predicted a peak at 2.5 eV. However, the authors' calculations showed that optical matrix elements are strongly suppressed near the valence band maximum (VBM) and peak approximately 1 eV below the VBM due to the growing Aluminum character of the valence band states. Incorporating this energy dependence shifted the predicted absorption peak to 3.3 eV, aligning closely with the experimentally observed 3.4 eV peak.
• Exclusion of Alternative Defects: The authors ruled out aluminum vacancies ($V_{Al}$) and $V_{Al}$-oxygen complexes as the origin of the 3.4 eV peak in their samples. The stability of the $C_N^0$ state requires a Fermi level between 1.1 eV and 1.9 eV above the VBM, a range where $V_{Al}$ defects would be in positive or neutral charge states, rendering them incapable of producing the observed transition.

Significance and Claims
The paper claims to provide the microscopic origin of a specific sub-bandgap absorption feature in AlN (the 3.4 eV peak) as a transition between the valence band and the neutral $C_N$ acceptor level. The authors emphasize that this assignment is only possible through the combination of two specific factors:

1. Independent Charge State Identification: Using photo-EPR to unambiguously determine the charge state of the carbon defect prior to optical measurement.
2. Advanced Theoretical Treatment: Using first-principles calculations that account for the energy dependence of optical matrix elements and the dispersion of the valence band, rather than relying on simplified models with constant matrix elements.

The work concludes that reliable assignment of sub-bandgap absorption in wide-bandgap semiconductors requires this dual approach of experimental charge-state control and rigorous theoretical modeling of optical transitions.