Yellow Luminescence in GaN: Predictions for Candidate Defects at Zero and 12 GPa Hydrostatic Pressure

Yellow luminescence (YL) has held great interest in GaN. The identity(ies) of the YL defect(s) has been controversial for many years. Recent experimental work by Zvanut <i>et al.</i> and by Narita <i>et al. </i>proved that carbon substituting on a nitrogen (C_N) site is at least one of the components. This identification is supported by recent theoretical results, buttressed by the photoluminescence studies of Reshchicov and coworkers, who resolved the zero phonon line (ZPL) and other phonon replicas. These theoretical studies have found other defect candidates predicted to have PL_max and ZPL close to experiment for YL. There is prior data for YL obtained at high hydrostatic pressure, showing that YL band strongly shifts and broadens, making assignment to a single defect untenable. We perform a detailed study of C_N, as well as other candidate defects (V_Ga, C_N:O_N, V_Ga:O_NH₂, V_GaH₃ ) at 0 GPa and 12 GPa using the standard PBE exchange correlation functional with the local moment counter charge (LMCC) method, an alternative to the standard <i>jellium </i>approximation for electrostatic boundary conditions. Prior published LMCC results show that the method reproduces a reasonable proxy for the band gap (3.56 eV (LMCC), vs. 3.3 eV (exp.)) bounding by the span of all defect levels arising from localized charge states for an extensive set of intrinsic and extrinsic defects. We use this “defect band gap” both to estimate the change in the experimental gap as a function of pressure, and to assign new band edges for calculation of PL_maxat 12 GPa. At 0 GPa our results agree well with Lyons <i>et al.</i>: V_Ga, V_Ga:O_NH₂, and C_N have PL_max within ~0.1 eV of experiment (2.18 eV), and also yield accurate prediction for the ZPL. We are in reasonable agreement with Lyons <i>et al. </i>that V_Ga:H₃ is within 0.2 eV of experiment ZPL. Our new results at 12 GPa provide crucial chemically differentiating information. First, we find that the prediction of the slope of E_G vs. pressure using the extremal defect levels as band edge bounds (36 meV/GPa) is in very good agreement with several experimental values (36-47), where the larger values were taken from the linear coefficient of a quadratic fit, and the lowest was taken from a purely linear fit. For V_Ga, V_Ga:O_NH₂, and C_N, we find that predicted PL_max are no longer close in energy: the values range from 2.51 eV to 2.83 eV. Almost all of the change in PL_max is in the ZPL. Even though the Frank-Condon shifts (FCS), the energy difference between PL_max and the ZPL, are quite different for each defect, ranging from 0.2 to 0.55 eV, the changes in FCS with pressure are minimal--~10-20 meV. These results point to using hydrostatic pressure, combined with computational predictions, to chemically discriminate PL.