Abstract
By applying the informatics to the datasets of the materials simulation, we try to find the important factors and hidden relationship in the transmission spectra of molecular junction. We generated 600 molecular configurations between the electrodes randomly. Transmission spectra are simulated for each configuration by nonequilibrium Greenʼs function method based on semi-empirical extended Hückel method. We classify the simulated spectra according to the similarity around the Fermi level by hierarchical clustering. The relationship between the average spectrum shape and molecular configurations is determined for each cluster. The relationship can be understood from the coupling between the molecular orbitals and electronic states of electrodes, which is physically reasonable. The results of this study prove the usefulness of the informatics for analysis on materials simulations. Our scheme can be applied not only to the transmission spectra but also to other spectra in general.
