Fig. 1: Schematic summary of our method for unsupervised atomic-scale analysis of high-resolution GLC-EM density maps (see Materials and methods for details). | Communications Chemistry

Fig. 1: Schematic summary of our method for unsupervised atomic-scale analysis of high-resolution GLC-EM density maps (see Materials and methods for details).

From: Small, solubilized platinum nanocrystals consist of an ordered core surrounded by mobile surface atoms

Fig. 1

a–c Atomic density detection and identification of atomic positions. a Denoising of the experimental map (raw vs. denoised), binarization (binary), and map simulation (simulated). b Determination of the optimal map binarization threshold through maximization of the correlation between simulated and denoised maps. c Atom positions are assigned as the centers of mass of the grayscale voxel values within the connected components. d Validation of atomic positions. Atomic densities that are still connected due to limited local resolution are split. Outlier atoms are rejected based on atomic density quality (validation correlation). e Unsupervised per-atom statistics calculations include coordination number (CN) and atomic displacement parameters. Thermal ellipsoids representing fit anisotropic displacement parameters. Example central plane of atoms in NP4 shows that atoms close to the surface move more (isotropic displacement) and with a greater degree of isotropy (DOI). Examples of atomic densities with different DOI in NP4 illustrate the heterogeneity of atomic thermal motion (bottom right).

Back to article page