Enhanced light element analysis using dual EDS detectors in advanced semiconductor structures

In advanced semiconductor devices such as 3D NAND flash memory, precise evaluation of oxygen and nitrogen distributions within SiO2/Si3N4 (ON) stacks is critical for understanding structure-property relationships and ensuring device reliability. Sensitive region in advanced semiconductor devices, where subtle variations in oxygen and nitrogen distribution can lead to measurable changes in electrical performance and long-term reliability. As device architectures continue to scale toward higher aspect ratios and tighter process margins, the ability to assess interface quality in a consistent and repeatable manner has become increasingly important for both process development and failure analysis. The oxide–nitride interface in ON stacks represents a highly process sensitive region in advanced semiconductor devices, where subtle variations in oxygen and nitrogen distribution can lead to measurable changes in electrical performance and long-term reliability. As device architectures continue to scale toward higher aspect ratios and tighter process margins, the ability to assess interface quality in a consistent and repeatable manner has become increasingly important for process development. Transmission electron microscopy (TEM)–based analytical techniques are widely used to investigate ON stack interfaces at the nanoscale. Energy dispersive X-ray spectroscopy (EDS), which detects X-ray signals generated by the interaction between an electron beam and a thin sample, is widely used for nanoscale compositional characterization due to its direct correlation with electron microscopy imaging. Although EDS provides several advantages—such as direct correlation with high-resolution imaging and the ability to detect light elements—it also presents challenges associated with long acquisition times due to limited detection of solid angle. The first issue is sample drift. During prolonged acquisition, slight movement of the specimen can distort structural information and obscure intrinsic characteristics of the interface. The second issue is electron beam induced damage caused by high accelerating voltages. Prolonged electron beam exposure may alter the specimen structure, potentially affecting the intrinsic interface characteristics. Because the O/N interface plays a critical role in semiconductor device performance, minimizing beam-induced damage is essential to accurately characterize structure–property relationships. Beam-induced damage is caused by high accelerating voltages. Extended exposure can modify the sample through mechanisms such as structural deformation or contamination. Because the SiO-induced damage is essential to accurately characterize structure–property relationships. In this paper, we present the intrinsic interface structure of ON stack using an advanced dual silicon drift detector (dual EDS) system, which provides approximately twice the detection solid angle of a single detector. The increased collection efficiency enables rapid acquisition of Xray signals before significant electron beam damage occurs. Using this enhanced EDS capability, we successfully reveal the intact structural characteristics of the SiO2/Si3N4 interface, which are essential for understanding and optimizing semiconductor device performance. EDS system, which provides approximately twice the detection solid angle of a single detector. The increased collection efficiency enables rapid acquisition of X-ray signals before significant electron-beam damage occurs. Using this enhanced EDS capability, we successfully reveal the intact structural characteristics of the SiO Cross-section lamellae of SiO2–SiN ON stacks were prepared to enable direct interrogation of the oxide–nitride interface. All analyses were conducted under 200 kV STEM conditions using a Dual-EDS system equipped with two silicon drift detectors positioned at different collection angles. STEM-EDS line scans were acquired perpendicular to the interface, allowing simultaneous detection of O Kα and N Kα signals from identical regions. Detector-specific signals (Detector A and Detector B) and their combined response (A+B) were evaluated to assess the influence of detector geometry and specimen-related effects. Rather than pursuing absolute compositional quantification, normalized O/N intensity profiles were used to derive interface-focused metrics such as transition width and gradient. To minimize beam-induced artifacts, short-dwell, multi-pass acquisition strategies were employed and subsequently summed. Single-detector STEM-EDS measurements show variability in interface metrics that can be attributed to differences in collection geometry, absorption, and local specimen conditions. In contrast, the combined Dual-EDS signal yields more consistent normalized O/N line profiles, leading to improved stability of interface metrics such as transition width. Repeated measurements across multiple line-scan positions indicate that the summed Dual-EDS approach reduces measurement scatter and enhances repeatability. In addition, multi-pass, short-dwell acquisition mitigates beam-induced broadening and contamination effects compared to single long-dwell scans, resulting in a clearer representation of interface abruptness. These observations support the use of Dual-EDS STEM-EDS as a geometry-tolerant method for comparative interface assessment rather than absolute quantification. This study demonstrates that a Dual-EDS–based STEM-EDS workflow can serve as an effective and scalable screening approach for evaluating SiO2–SiN ON stack interfaces under 200 kV STEM conditions. By focusing on normalized interface metrics instead of absolute compositional values, the proposed methodology addresses common limitations associated with conventional STEM-EDS analysis. Future work will apply these interface metrics to compare ON stacks fabricated under different deposition and thermal processing conditions, enabling rapid triage of interface quality across sample sets. Selective correlation with complementary analytical techniques may further strengthen confidence in the proposed screening strategy.