An optimized evaluation strategy for a comprehensive morphological soot nanoparticle aggregate characterization by electron microscopy

For a comprehensive understanding of nanoparticle formation in gas phase processes, such as soot formation, morphological parameters of fractal-like particle aggregates, like the radius of gyration, the fractal dimension and the primary particle size, have to be determined. Often transmission electron microscopy (TEM) is employed for the investigation of particle characteristics as it not only allows to investigate ensemble averages but single particle aggregates and thus also to determine statistical properties, such as the size distribution within a sample. Many different evaluation methods can be found in the literature. We investigated different approaches for the determination of morphological parameters from TEM images of soot and compared the results to derive an optimized evaluation strategy for TEM nanoparticle characterization. We compared four methods for the determination of the radius of gyration - three length- and one pixel-based methods - showing good agreement within 9% deviation for the median of the recovered lognormal size distribution. Furthermore, the fractal dimension was determined via a sample-based and various box counting methods with different limiting box sizes. Here, we could show that the upper and lower bounds (aggregate size in terms of radius of gyration and primary particle size in terms of its radius, respectively) of self-similarity of fractal-like aggregates should be accounted for by choosing corresponding upper and lower box sizes. Using box counting, we could show that for small aggregates the fractal dimension as well as its span are increased, yet with increasing aggregate size the fractal dimension converges towards 1.6. Furthermore, we could show the potential of semi-automatic aggregate detection through Trainable Weka Segmentation. However, image noise resulting in erroneous aggregate splitting often leads to smaller aggregate sizes by automatic detection compared to manual segmentation. Generalized Hough Transformation for the semi-automatic determination of primary particle sizes performs well for large soot particle aggregates as those often show spherical primary particles. For leaner combustion conditions, the primary particles of the formed clumpy soot aggregates cannot be detected well via semi-automatic detection. TEM images were taken on soot samples from premixed laminar flat flames (burner type: McKenna) under various conditions to provide comprehensive reference data.