Joint plasmon and core‐loss fitting for electron energy loss spectroscopy of InGaN

Abstract

We demonstrate a method to fit electron energy‐loss spectra (EELS) of InGaN thin film samples by fitting both plasmon and core losses over the energy range of 13‐30eV. The main plasmon peak is relatively strong and broad. In our previous research, we have suppressed noise by using Lorentz fitting for the plasmon peaks of InGaN. Pure core‐loss spectra of Ga (Ga 3d transitions yield M 4,5 peaks at 23.8 and 28.5eV) and In (In 4d transitions yield N 4,5 peaks at 20.0 and 25.9eV) can be artificially constructed from EELS of binary compounds by subtracting the GaN plasmon peak (19.35eV) or the InN plasmon peak (15.5eV), see figure 1. Then the Ga 3d and In 4d reference spectra can be obtained by further smoothing the core‐loss spectra. Multiple linear regression is applied to experimental GaN and InN spectra of different relative thicknesses (t/λ), and the fitting quality is defined by adjusted R 2 , which lies higher than 0.998. In order to fit spectra from InGaN with different indium concentrations, InGaN core‐loss reference spectra were constructed by using artificial InN and GaN core‐loss spectra with linear compositional weightings, simultaneously, the chemical shift and broadening of the plasmon loss is considered in the MLLS fitting. For plasmon peak shifts we have applied our previous research results on the relationship between indium concentration and plasmon peak position [1]. The core‐loss chemical shift was assumed to follow the plasmon loss chemical shift, as depicted in figure 2. Finally, the MLLS regression can be performed to fit experimental spectra from InGaN as weighted superpositions of reference plasmon and core‐loss spectra corresponding to GaN, InN and a specific ternary InGaN alloy. InGaN samples grown at high temperatures as typically applied in metal organic chemical vapour phase deposition are prone to phase separation, which was first predicted by Ho and Stringfellow [2]. The experimental spectra were recorded in TEM mode with a conventional Schottky field‐emission transmission electron microscope (FEG‐TEM). By using the joint plasmon and core‐loss spectra fitting of GaN, InN and InGaN, we have studied InGaN spectra with nominal indium concentrations of x=0.54 and x=0.62, where energy dispersive x‐ray spectroscopy suggests the indium concentrations are close to x=0.59 and x=0.68 respectively. The fitted spectrum in figure 3 indicates a strong evidence of phase separation in the nominal x=0.62 InGaN sample.