In this contribution we present high resolution images of the Magellanic Clouds produced by a maximum entropy algorithm. The spatial resolution of these images is the best currently possible. At present we are in the process of scientific validation of the images in order to determine the reliability of positions and fluxes of the numerous point sources and extended structures that can be recognized from the images. To achieve this, the images are compared with data taken at other wavelengths, ranging from radio to X-ray.
The images are available for public use. Due to their size we decided not to distribute them via Internet. Requests for copies on tape or other storage mediums should be directed to the authors (e.g. MBraun@aip.de).
Readers interested in the second part of the talk dealing with `the possible interaction of high-velocity clouds with the disk of the LMC' shall be refered to Braun (1996) where this topic is discussed in detail.
It is possible to reconstruct reliable images at a resolution of nearly the diffraction limit of the IRAS telescope (Bontekoe et al. 1994). This is achieved by application of the maximum entropy package MemSys5 (Gull and Skilling 1991) which largely removes the blur of the IRAS detector response functions. The resulting spatial resolution and the dynamic range in the high-resolution image depends on the number of data samples locally in a map, and is usually not uniform over the map.
An essential ingredient for the production of high-resolution images is the consistency from scan to scan, which have different background levels. In order to bring the scan on a common baseline a first order fit to the background is subtracted. This, and other pre-processing steps are performed using the IRAS data processing routines described by Assendorp et al. (1995), followed by a series of data self-calibration steps during the maximum entropy image reconstruction process.
The scan coverage of the Magellanic Clouds is much larger than average because they are near one of the poles of the orbit of the IRAS spacecraft. The number of scans that have overlap e.g. with the LMC is roughly 600; the number of data points at 60 µm is about 6.3 million. Though the large amount of IRAS data ensures a high dynamic range and spatial resolution, the disadvantage is the sheer computation time which surpasses the capabilities of most high-end workstations. The images presented here are processed on the Convex SPP 1200 computer of the Astrophysikalisches Institut Potsdam. This is a machine with 8 cpu's with 120 MHz clock cycle and 2 GB workspace memory. Even on this machine the processing of the four images of the LMC took more than two year (single) CPU time.
Examples of images produced by the different techniques can be compared in more detail in Fig. 2 where the 60 µm emission of the south-western region of the Small Magellanic Cloud is presented. In addition to images as similarly shown in Fig. 1, the upper right panel gives the outcome of the application of MemSys5 to a combined set of IRAS survey and Pointed Observation data. A further increase in resolution is obvious. However, some artifacts are easily detactable close to bright point sources. They can be addressed to problems with the definition of the detector response functions at the lower limit. The lower right panel was constructed by scan combination from both the survey and the PO data. But in this case, only the data provided by the small detectors - 2 of the 16 60 µm detectors had smaller sizes of 1.3'×1.5' instead of 4.8'×1.5' - were used. The construction of this image was only possible due to the extremely high coverage of this area by IRAS POs. Though the signal-to-noise ratio is low in this map, it is a convincing proof of the reliability of the high resolution images provided by the MemSys5 algorithm.
First version: | 09th | August, | 1998 |
Last update: | 25th | September, | 1998 |