Tyler Roberts

Tyler Roberts is Market Manager for Applied Markets at Shimadzu Scientific Instruments, where he leads technical market strategy across the Energy, Chemicals, and Materials sectors. He holds a Master’s degree in Analytical Chemistry from the University of Arizona and previously served as a GC-MS Product Specialist. Tyler collaborates with industry partners, key opinion leaders, and application scientists to translate emerging analytical and metrology challenges into practical measurement strategies that support process improvement, cost reduction, and the development of new technologies across critical industrial sectors.

Abstract

Dynamic Image Analysis and Machine Learning for Large Particle Metrology in CMP Slurries

Large-particle contamination in CMP slurries remains a near-term metrology challenge identified in the 2024 IRDS Roadmap, particularly for reliable large particle count (LPC) measurement in concentrated slurries without sample dilution. This work evaluates dynamic image analysis (DIA) as an image-based approach for characterizing coarse particles and foreign matter in silica-based slurries, with machine learning used to extend particle classification beyond conventional size metrics.

Using the Shimadzu iSpect DIA-10, particles flowing through a thin microcell were individually imaged and evaluated for size, concentration, shape, and optical morphology. In a polishing-grade colloidal silica slurry, DIA enabled measurement under undiluted conditions and revealed micron-scale foreign particles that were difficult to assess after dilution or by bulk particle-size techniques dominated by the submicron primary particle population. Particle images and morphological descriptors were then analyzed using Uniform Manifold Approximation and Projection (UMAP) for dimensionality reduction and Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) for unsupervised clustering. This approach separated distinct particle populations based on image-derived morphology, providing additional information relevant to particle origin, filtration strategy, and process control. A related high-concentration silica slurry study further demonstrated the ability of DIA to quantify coarse-particle populations before and after filtration without diluting the stock sample.

These results support DIA as a complementary method for large-particle characterization in CMP slurries. Looking forward, pairing image-based particle metrology with machine-learning classification may provide a path toward more sophisticated LPC metrology and improved discrimination of rare foreign-particle populations in increasingly demanding CMP processes.