[Previous] | [Session 5] | [Next]
H. Kang (Department of Statistics, Harvard University), D. van Dyk (Department of Statistics, University of California at Irvine), V. Kashyap (Harvard-Smithsonian Center for Astrophysics), A. Connors (Eureka Scientific)
The temperature distribution of Emission Measure is a powerful tool to characterize and understand the composition and physical structure of stellar coronae. Numerous methods have been proposed in the literature to compute the Differential Emission Measure (DEM) based on line fluxes measured from identifiable lines in high-resolution EUV and X-ray spectra. Here we describe a new and powerful method that we have developed to reconstruct DEMs that improves significantly on previous algorithms and further allows for incorporating atomic data errors into the calculations. Some notable features of our algorithm are: an ability to fit to either a selected subset of lines with measured fluxes or to perform a global fit to all lines over the full wavelength range of the instrument, to fully incorporate line blends, to obtain error bars to determine the significance of features seen in the reconstructed DEM, and to directly incorporate prior information such as atomic line sequences, known atomic data errors, systematic effects due to calibration uncertainties, etc.
We use highly structured models to account for the mixing of the ion/temperature specific spectra, the mixing of continuum photons with those from the multitude of spectral lines, instrumental response, the effective area of the instrument, and background contamination. We introduce the statistical framework of data augmentation (e.g., EM algorithms and MCMC samplers), in which we treat photon count in each level of the hierarchical structure as missing data. We implement a multi-scale (wavelet-like) prior distribution to smooth the DEM, which gives us the flexibility to overcome lack of information especially with low count data. In this talk we provide several simulation studies with both high-count and low-count data to evaluate the proposed method. We also provide several DEM reconstruction results of the active star alpha Aur (Capella), and validate the method by comparing our results to previous estimates of the Capella DEMs.
The authors gratefully acknowledge funding for this project partially provided by NSF grant DMS-01-04129 and by NASA Contract NAS8-39073 (Chandra X-ray Center).
The author(s) of this abstract have provided an email address for comments about the abstract: kang@stat.harvard.edu
[Previous] | [Session 5] | [Next]
Bulletin of the American Astronomical Society, 36 #3
© 2004. The American Astronomical Soceity.