Morphological classification of galaxies by shapelet decomposition in the Sloan Digital Sky Survey. II. Multiwavelength classification

We describe the application of the shapelet linear decomposition of galaxy images to multiwavelength morphological classification using the u-, g-, r-, i-, and z- band images of 1519 galaxies from the Sloan Digital Sky Survey. This combination of morphological information in a variety of bands is unique and allows automatic separation of different classes in ways that are impossible using single-band images or simple spectrophotometric measurements such as color. We use elliptical shapelets to remove to first order the effect of inclination on morphology. After decomposing the galaxies, we perform a principal component analysis on the shapelet coefficients to reduce the dimensionality of the spectral morphological parameter space. We give a description of each of the first 10 principal components' contributions to a galaxy's spectral morphology. We find that galaxies of different broad Hubble type separate cleanly in the principal component space. We apply a mixture-of-Gaussians model to the two-dimensional space spanned by the first two principal components and use the results as a basis for classification. Using the mixture model, we separate galaxies into three classes and give a description of each class's physical and morphological properties. Galaxies were typically robustly classified, with 80% of galaxies having a probability of >= 90% of occupying their respective class. We find that the two dominant mixture-model classes correspond to early- and late-type galaxies, respectively, both in their morphology and their physical parameters ( e. g., color, velocity dispersions). The third class has, on average, a blue, extended core surrounded by a faint red halo and typically exhibits some asymmetry. The third class cannot be associated with any broad Hubble type; however, it is the most probable class for irregular galaxies. We compare our method to a simple cut on u - r color and find the shapelet method to be superior in separating galaxies. Furthermore, we find evidence that the u - r = 2.22 decision boundary may not be optimal for separation between early- and late-type galaxies and suggest that the optimal cut may be u - r similar to 2.4. We conclude with a discussion of the limitations of our method and ways in which it may be improved. Our framework provides an objective and quantitative alternative to traditional one-color visual classification, and the powerful use of both spectral and morphological information gives our method an advantage over separation techniques based on simpler calculations.