The X-ray data (taken with the ACIS) also provide the X-ray energy distribution at each part of the image (spatially resolved spectroscopy); this is illustrated in Figure 1 by the various colors assigned to the image pixels. The lowest energy photons are colored red, intermediate energies are green, and the highest energies are blue. This representation shows how the spectrum of X-ray emission changes over the remnant. This effect can be seen in more detail in the work done by Hughes et al. (2000). As shown in Figure 1 several regions were selected from the image, and the X-ray spectrum was plotted versus an expected spectrum for a particular model of SNR ejecta. As discussed by Hughes et al. the selected spectra vary considerably, and represent material thrown out from different layers of the core of the collapsing star during the supernova explosion. Interestingly, the material from the inner-most layers of the core has traveled out to the edge of the remnant as indicated by the presence of iron in the X-ray spectrum. This effect is probably due to differences in the ejection velocity of the core layers as the collapse evolves. It provides important observational data for the theorists to match in their detailed numerical calculations for these events.
Figure 2 shows a 50ksec observation of Cas-A that was taken with the HRC. In this image, the central compact object is clearly seen. There are about 4,000,000 events from Cas-A in this image, and only about 1500 of them are from the central source. The excellent quality of the Chandra mirrors, particularly their high angular resolution and low scattering, is essential for separating the point source from the surrounding nebula. The purpose of this long HRC observation is to search for a periodic signal from the compact object. This signal, if found, would confirm the association of the point source with a neutron star that should have been formed during the supernova explosion. At the time of this writing, the data are still undergoing analysis, and results are not yet available.
These observations demonstrate the importance of coupling image processing and display systems with data analysis software. Defining regions of interest in an image to select data of interest from a non-spatially sorted data set (as for example the photon event list from Chandra) is an important tool needed for science analysis. Similarly the tools needed for detailed timing analysis present a challenge to the Chandra data system. Not only must the proper events be easily selected (as for the spectral analysis case), but there are systematic time corrections, such as accounting for orbital motion, that need to be accurately applied to the data before any temporal analysis can be performed. Searching for periods is known to be a compute intensive task. Developing efficient algorithms to implement the search (e.g., FFT, period folds, and other statistical tools) is an important need of the scientist.
Figure 4 shows the central region of the galaxy. Here the circle represents a 5 radius about the galaxy center. Where ROSAT detected a single source, Chandra resolves five individual point sources. One of these is likely to be associated with the actual center of M31, where there is a black hole (BH). The two Chandra sources closest to the BH location are just 0.5 apart.
It will take some additional work to determine which of these sources is most likely the central source. One has an unusually soft X-ray spectrum, a characteristic associated with a class of objects known as super-soft sources (SSS's). The other appears to be typical of most galactic X-ray sources. The SSS is quite variable in intensity (about a factor of 5-10 within a few months). However, none of these properties is sufficient to confidently determine an association of the black hole with the center of M31. Monitoring of the region is continuing to improve the absolute locations. HST images of the center of M31 provide matches between Chandra and HST sources so that the coordinates can be accurately aligned. If successful, this work will eventually lead to 0.2 precision in locating the center of M31 with respect to the X-ray sources.
The role of astronomical software in helping to process, analyze, and understand the data from M31 is critical to progress. In the case of M31, the entire galaxy is too large to be observed in a single detector field of view. Mosaic images are needed to give a complete view. This process requires algorithms for translating images onto a larger frame, retaining all of the information and taking into account edge effects from the detectors and telescope. Exposure maps and corrections are needed, particularly in areas of overlap, so that proper source intensities and light curves can be constructed. Stacking repeated images, as in the central region of M31, is another challenge for software developers. Easy methods for co-aligning images and calculating combined exposures are required for such observations. Matching data from Chandra and HST involves careful calculations of source centroids, transformations of coordinates, and accounting for detailed differences in the astrometry from each mission.
Figure 5 shows X-ray images obtained with the HRC and ACIS. In both images it is easy to see the strong central nucleus that is being powered by a super massive black hole (about ) and a long one sided jet of emission directed to the northeast. The jet consists of many bright knots embedded in what appears to be a well collimated diffuse emission region. The jet extends at least 6 from the nucleus.
In Figure 6, the X-ray emission is compared with radio data taken at 13 and 6cm wavelengths. The 13cm radio data emphasizes the large scale structure from Cen-A and shows what are called the inner radio lobes that extend to the northeast and southwest, as well as the jet. There is excellent correspondence between the X-ray and radio data when viewed on a large scale. Of special interest is the bright X-ray ridge of emission that corresponds to the edge of the southwest radio lobe. This ridge was barely detected in previous X-ray observations (e.g., Einstein and ROSAT) but is now seen clearly. It is very closely aligned with the edge of the radio emission. The southwest radio lobe is actually filled with diffuse X-ray emission, with the region of peak radio emission inside the lobe corresponding to the minimum X-ray emission region. Understanding the details of the physical conditions at Cen-A that give rise to these correspondences is the goal of this investigation. The ACIS can extract spatially resolved X-ray spectra at different locations in the image for studying the spectral changes as a function of location as an indicator of changes in the properties of the emitting material.
The second panel in Figure 6 shows the radio emission at 6cm plotted over the X-ray image along the jet. Again there is good general correspondence between the radio and X-ray image. However, at the detailed level, there are some striking differences. While peaks in the X-ray and radio emission correspond for the innermost knot in the jet, farther out from the nucleus the correspondence becomes less precise, and the peaks in X-ray emission lie closer to the nucleus than the corresponding radio peaks. This change in the relative locations of peak X-ray and radio emission indicates that there is likely to be a lot of shock formation, particle re-acceleration, and radiation going on along the path of the jet. The knots may well be sites of shocks from which accelerated particles propagate outward, losing energy to radiation. The higher energy particles lose their energy faster so that the spectrum softens farther from a shock. The detailed processes along the jet are much more complex at the scale of Chandra resolution than was evident from lower resolution observations.
The astronomical software impact for studying these data follows many of the points already discussed. A new process used for looking at complex images is seen in the adaptively smoothed image of Figure 6. This is an example of image processing that helps to highlight features on differing spatial scales making visual inspection easier. Various algorithms for this class of image processing are possible, and typically there are many parameters that can be set for these algorithms. Having well constructed software that can run efficiently and reliably to implement these processes is critical to their use. Being able to quickly run many cases of an adaptive smooth, for example, varying parameters to bring out desired features or suppress noise, allows a researcher to develop a sense of what is important and what is real in an image. Similarly, detecting sources in complex images such as Cen-A is an intriguing problem. Using wavelet decomposition, percolation techniques, or standard sliding box detect algorithms, each gives a particular advantage and disadvantage in the task of extracting information from an image. In practice all of these methods (as well as others not discussed here) should be tried and compared in order to obtain the most reliable results. The ease of use of these techniques, as well as their running time, often dictates what is actually done. Astronomical software developers need to work with scientists to make this task as easy and complete as possible.
Figure 8 shows a composite of images of the jet of 3C273 taken in the radio, optical and X-ray bands. It is apparent that the jet behaves differently at these three wavelengths: the X-ray image becomes fainter farther from the nucleus, the radio image gets brighter, and the optical remains largely constant - although fragmented into many knots. As for Cen-A, this comparison suggests that the physical processes responsible for emission are varying along the jet. The closest knot is well fit to a synchrotron model, while this model fits less well for the jet region farther along and fits rather poorly at the jet's end. Physical conditions along the jet are changing - not surprising since the jet is the size of our Milky Way Galaxy. More complex models must be considered if radio, optical, and X-ray emission is to be jointly understood and interpreted.
The observation of 3C273 illustrates the need to be able to simultaneously consider the multi-dimensional problem of spatially resolved spectroscopy across broad wavelength bands. The development of data-cubes, for representing the information, and then processes that can act on these large data items, appears to be the direction for the future. Chandra is only the first of several next generation observatories that will require such data types and processes.
Brinkman, B. C. et al. 2000, Proc. SPIE, 4012, 81
Garmire, G. G. 1997, BAAS, 190, 34.04
Hughes, John P., Rakowski, Cara E., Burrows, David N., & Slane, Patrick O. 2000, ApJ, 528, L109
Markert, T. H., et al. 1994, Proc SPIE, 2280, 168
Murray, S. S., et al. 1998, Proc SPIE, 3356, 974
Tananbaum, H. D. 1999, IAU Circ. No. 7246