Fig. 5: Multidimensional analysis for structure differentiation. | Nature Communications

Fig. 5: Multidimensional analysis for structure differentiation.

From: Exploiting correlations in multi-coincidence Coulomb explosion patterns for differentiating molecular structures using machine learning

Fig. 5: Multidimensional analysis for structure differentiation.

a Dimensional reduction and clustering analysis of a mixture of four isomers—cis-, trans-, twisted-1,2-DCE, and 1,1-DCE—where ball-and-stick models of these isomers are also illustrated. b Discriminative power analysis of features constructed using higher-order correlations between momentum vectors, categorized into modulus difference (green) and angle (purple) between two momentum vectors, and angle between two planes (brown) formed by four momentum vectors. Error bars show the standard error of feature importance computed over 100 Random Forest fits with different random seeds (c–e) demonstrate the effectiveness of high-dimensional data in differentiating isomers, where the separation between these structures is improved sequentially from one- to two- and three-dimensions.

Back to article page