mi013.metcalf05 Posted: 24-Feb-96 Updated: 1-Sep-96 Events specified:
Thomas R. Metcalf & Hugh H. Hudson Univ. of Hawaii
David Alexander Montana State U.
Takeo Kosugi National Astronomical Observatory of Japan
We plan to compare the MEM, pixon, and Gaussian fitting reconstruction algorithms for HXT data. In particular, we wish to understand how accurate each method is, not only with respect to locating sources, but also with respect to photometry. Our method will be to examine images and light curves for simulated HXT data. This study is motivated by the important 1992 January 13 limb flare in which the coronal x-ray source is considerably fainter than the footpoint emission. The pixon method indicates a less impulsive coronal source than the MEM reconstruction and also a relatively weaker coronal/footpoint emission ratio.
In the MEM algorithm, the image entropy is maximized under the constraint that the data be reproduced adequately (i.e. within the noise in the data). Thus, the entropy term will drive all pixels to have the same value unless the data requires otherwise. While each pixel is independent, this means that there is no consideration of spatial location in the MEM algorithm. In the case of the Masuda event, there is a very bright source and a very faint source in the field of view. Although the two sources are spatially quite far apart, the entropy term will drive the two sources to have the same intensity since this will maximize the entropy of the entire image. On the other hand, the data will drive the two sources to have different intensities, in opposition to the entropy term. The final compromise between the data and the entropy terms will be determined by the level of noise in the data. Thus, to some extent, the faint source must be contaminated by the bright source, and the light curve for the faint source will mimic that for the bright source. The question we hope to address is to what extent the faint source is contaminated.
In the pixon algorithm this contamination is largely avoided since the pixon algorithm does consider the relative spatial locations of the sources. The extent to which the problem is avoided with the pixon algorithm is also a question which will be addressed with the simulated HXT data.
Update 1-Sep-96
I have recently finished a number of HXT simulations designed to determine how the MEM and Pixon reconstruction algorithms behave when a faint extended source is observed in the presence of bright, compact sources. This sheds some light on the HXT light curves for the Masuda event of 1992 January 13. Please have a look at the report I have place on the WWW:
http://diapason.space.lockheed.com/~metcalf/Papers/hxtpixons/Simulations/simulation.html