Jeremy Mettler, Ph.D. Candidate
Dr. David Ruzic, Director of Research
March 5, 2025 | 3:00pm - 5:00pm CST
This final examination will be held in 4039 CIF.
SPATIALLY RESOLVED PROBES FOR THE MEASUREMENT OF FLUORINE RADICALS
ABSTRACT: Etching is one of the fundamental steps in modern semiconductor manufacturing. As device densities grow and feature dimensions shrink, processing tolerances become increasingly strict. Meeting these tolerances will require improvements in etch process uniformity, selectivity, and pattern roughness. For high aspect ratio silicon etching, multi-step etch processes such as the Bosch process are commonly used, relying on alternating fluorinated gas chemistries to maintain high etch rates and periodically deposit inhibitor layers for sidewall protection. The importance of fluorine chemistry to silicon etching has prompted extensive study of the fundamental reaction mechanisms, especially in the presence of rate-enhancing ion bombardment. However, despite decades of study, significant disagreements remain in the literature regarding the chemical reaction rate of silicon with fluorine radicals, particularly when SF6 is used to produce those radicals. Radical densities are commonly estimated using actinometry, a convenient, non-intrusive technique which relies on light emitted from the plasma source to determine average fluorine densities. Direct comparisons between volumetrically averaged fluorine densities and localized etch rates rely on the relative uniformity of the density profile, which is often chamber and condition specific, making universal measurements such as etch probability difficult to compare across etching systems.
The primary aim of this work is to improve current fluorine radical measurement capabilities using spatially resolved radical probes. Existing radical probes for species such as O, N, and H measure densities based on the energy released as these species recombine on a catalytic surface. In plasma systems where multiple heating sources are present, the equilibrium temperature difference between a reactive catalytic probe and less reactive reference probe can be used to isolate the influence of radical heating for density measurements. Initial experiments in the Plasma-Materials Interaction Chamber (PMIC) showed the harsh chemical environment and long exposure times required to reach thermal equilibrium severely limited probe lifetime. To address these issues, the standard radical probe technique was generalized for non-equilibrium measurements and radical etching was used in place of chemical recombination as the detection heat source. Fluorine radical probes consisting of a pair of thermocouples tipped with a reactive tungsten surface and inert aluminum surface were initially tested in PMIC and a linear correlation between probe response and fluorine density was observed in both NF3 and SF6 plasmas, affirming the ability to detect fluorine radicals.
Further probe characterization was conducted in a research scale etcher provided by Tokyo Electron (TEL). Calibration for the thermal response of each probe was achieved using a two-step process consisting of inert and reactive plasma exposures, after which the probe etch rate was correlated to the probe response. Once calibrated, probe etch rate calculations remained accurate even after etching ~20% of the initial tungsten probe dimensions. High fluorine density measurements were found to be limited to ~1021 m-3 based on the transition from kinetic to diffusion limited etching, for which more modeling would be necessary to relate the reaction rate and fluorine density. At lower densities, probe measurements were compared to density results from actinometry to determine the tungsten etching kinetics in fluorine from 100ºC to 700ºC, showing good agreement with existing etch probability values above 400ºC and to existing activation energy measurements below 300ºC.
Finally, kinetic data were used in conjunction with probe measurements to measure spatial fluorine density profiles radially above the wafer stage in the TEL etcher. Independent measurements of the silicon etch rate for matching conditions found that the etch rate profile was the direct result of the spatial distribution of fluorine radicals. Actinometry measurements were shown to significantly underestimate the fluorine density near the wafer center, where etch rates are often measured for silicon kinetic experiments. Silicon etch probabilities calculated from actinometry and etch rate measurements were similar to those reported in literature, with an apparent inverse dependence on fluorine flux. Spatially resolved comparisons, however, showed no clear dependence on fluorine flux and resulted in consistent etch probabilities independent of the tested condition, demonstrating the influence of nonuniformity in actinometric measurements. The results of this work offer fresh insights into the etching mechanisms of silicon, as well as a powerful new tool for studying fluorine radical spatial distributions.