Previous abstract Next abstract
Session 43 - HST Instruments & Other Missions.
Display session, Tuesday, January 14
Grism or slitless spectroscopy can provide spectroscopic information of all objects in the detector field of view (FOV). Because of overlapping and object crowding, in practice, only stellar objects are exploited; most extended objects are discarded, except when some obvious emission line can be isolated. The goal of this work has been to produce an efficient code that restores the full capability of image--spectroscopy. A set of data consists of grism (or slitless) images of a given field observed at 4 different roll--angles, and for each roll-angle, an image for field identification. The IDL code is based on a predictor--corrector iterative method. The method is used to restore the true spectra of extended sources appropriate to spectroscopic observations with the new HST instruments (NICMOS, STIS). The FOV is the Hubble Deep Field. First, each pixel is given an input ``true" spectrum. Simulations of 2--DIM grism images for 4 different roll-angles are made (DATA). Then, the code calculates the first spectral data cube (3--DIM MODEL, images for each sampled wavelength) by combining the 2--DIM DATA for each point in the sky for all wavelengths. This data cube is used to simulate new 2--DIM grism images (PRED). The image ratios PRED/DATA for the 4 roll-angles are combined and used to correct the previous 3--DIM MODEL. And so on, until convergence (via chi2). Appropriate masking and normalization are made; this ensures that no false object appears. With 4 roll-angles, the true spectra are restored in about 30 iterations, i.e. 5 hours on a Sparc 20/502 running Solaris 2.4. For each iteration, error pixel maps are calculated to assess how well the DATA can be predicted. The final error maps are attached to the final 3--DIM restored cube and can be used for further data analysis. The code does not use (de)convolution. The PSFs can be measured from the final cube for each wavelength, from stars in the field. 2--DIM spatial deconvolution can be made after restoration, choosing the objects of high enough S/N. This saves considerable computer time and avoids various artefacts.
The author(s) of this abstract have provided an email address for comments about the abstract: firstname.lastname@example.org
Program listing for Tuesday