Monday, April 17, 2017 3:00PM (110 INSCC)
Title: In Search of Stellar Sisters
The study of star clusters have been of fundamental importance to the study of stars and the Milky Way. Until recently our ability to identify stars born together was limited to spatially and kinetically coherent clusters. This represents only a relatively small fraction of Milky Way stars. With the advent of modern large scale chemical surveys such as APOGEE the prospect of using stellar composition alone to identify potential birth cluster associations (chemical tagging) for large numbers of stars has become feasible. The task of determining cluster memberships and confidences from chemical composition poses a significant challenge. Standard clustering algorithms such as k-means assign a cluster identity to each and every data point and are designed under the assumption that the underlying number of clusters from which our samples have been drawn are significantly fewer in number than the number of samples at our disposal. Chemical tagging places us in the opposite regime where we have a sample of hundreds of thousands of stars drawn from potentially millions of birth clusters. I will discuss statistical methods for working in this peculiar situation and present an analysis of abundances of APOGEE survey stars. Although I do not detect individual clusters at high significance I will present evidence that chemical space is clumpier than would be expected when drawing from a smooth distribution.
This excess clumpiness is suggestive of the signature of birth cluster association hinting that the reality of obtaining cluster memberships through chemical tagging is just on the horizon.