Semiconductor manufacturing relies heavily on plasma etching processes, which involve complex physics, dynamic physical reactions, and inhomogeneous chemical reactions. Maintaining consistency and quality during these operations remains a persistent challenge for the industry. The difficulty is further compounded as modern etching processes advance to require more stringent high-aspect-ratio (HAR) etch profiles, making the need to understand plasma conditions critical.

Decoding Complex Plasma Reactions with XAI

Processing plasmas are notoriously difficult to understand because of the numerous mechanical and electrical parts involved in the system, alongside the complexity of chemical and physical reactions. To investigate the physical causes behind variations in etching, the researchers fed etch profile image data directly into the XAI tool. They then analyzed how the resulting profiles changed in response to varying plasma conditions.

Through this analysis, the team discovered that the XAI’s mathematical index could function effectively as an interpretable index for plasma physics. Co-author Jihoon Park noted that this approach provided key insights, showing that process result image data—which typically receives less emphasis compared to standard sensor data—can be successfully applied to extract brand-new information regarding plasma reaction properties during manufacturing.

Framework Benefits and Future Applications

Drawing from their results, the researchers built a plasma physics-informed profile-analysis framework specifically for HAR plasma etching. This framework enables the quantification of the plasma-surface balance directly from measured etch profiles, moving beyond mere statistical attribution.

According to co-author Seolhye Park, uncovering a new source of information about plasma reactions yields an impact comparable to installing an entirely new diagnostic tool, but without requiring any physical hardware investment. The team believes these methodologies hold significant potential for other sectors of plasma physics and diagnostics. Looking ahead, the researchers intend to apply their findings toward understanding plasma process-driven phenomena within nuclear fusion.