Impact of electron beam propagation on high-resolution quantitative chemical analysis of 1-nm-wide GaN/AlGaN quantum wells

Recent advancements in high-resolution spectroscopy analyses within the scanning transmission electron microscope (STEM) have paved the way for measuring the concentration of chemical species in crystalline materials at the atomic scale. However, several artifacts complicate the direct interpretation of experimental data. For instance, in the case of energy-dispersive X-ray (EDX) spectroscopy, the linear dependency of local X-ray emission on composition is disrupted by channeling effects and cross-talk during electron beam propagation. To address these challenges, it becomes necessary to adopt an approach that combines experimental data with inelastic scattering simulations. This method aims to account for the effects of electron beam propagation on X-ray emission, essentially determining the quantity and the spatial origin of the collected signal. In this publication, we propose to assess the precision and sensitivity limits of this approach in a practical case study involving a focused ion beam (FIB)-prepared III-N multilayers device. The device features nominally pure ∼1.5-nm-wide GaN quantum wells surrounded by AlGaN barriers containing a low concentration of aluminum (∼5 at%). By employing atomic-scale EDX acquisitions based on the averaging of more than several thousand frames, calibrated $\zeta$ factors combined with a multilayer X-ray absorption correction model for quantification, and by comparing the X-ray radiation obtained from the quantum well with a reference 10-nm-wide structure, we demonstrate that the quantitative impact of beam propagation on chemical composition can be precisely accounted for, resulting in a composition sensitivity at the atomic scale as low as $\pm$0.25 at%. Finally, practical aspects to achieve this high precision level are discussed, particularly in terms of inelastic multislice simulation, uncertainty determination, and sample quality.