Multi-dimensional astronomical datasets
Together with the development of new instruments also the data recording techniques are changing quite drastically. The times are gone when one always obtained either an image or a spectrum or a photometric light curve of an object. These days techniques exist to leave that up to the choice of the astronomer while analysing the recorded data. I will present one example of this below.
X-ray observatories have detectors that register the arrival of individual photons. And, instead of registering them on a photographic plate or another device organising them into a specific strucutre, these photon arrivals are written into so-called "event tables". In these, all salient parameters of each event are recorded, such as the position where the photon hit the detector, when it came in and, in the case of an energy-sensitive detector, the energy the photon carried (see example below).
And so forth... Let us consider having a long table of tens of thousands of events, with pixel coordinates from 1 to 1024 in both directions, X and Y (assuming that we have a 1024x1024 pixel detector) and energies ranging from 100 eV to 10000 eV. The photons might have been received over a timespan of 40000 seconds. This would be quite a typical table coming from one of the current generation X-ray observatories, such as XMM-Newton or Chandra (external link).
Now what does an astronomer with such a table? In itself it is obviously not a very suggestive presentation of the observations. One would obtain an image by using software to select (for instance) all photons and regroup them onto a 1024x1024 X,Y grid. In this case it is very simple. There are other applications that involve more work. Here one selected example, namely spatially resolved spectroscopy of one object at the centre of the field of view, with background correction.
One would instruct the appropriate program to select only photons in an area where (one will know this by creating a test image beforehand) the target of the investigation is seen, then sort the photon events by energy and plot the number of photons per energy interval against energy, thereby creating a spectrum. To correct for instrumental effects and possible low-level sky background emission, one would tell the software to select photons from a second, empty sky region, create a spectrum for that area and then subtract that from the spectrum of the target of interest. Then one will know the spectral characteristics of the target itself, corrected for all other influences by comparing the source spectrum with one of an area on the sky where no obvious sources are seen.
Similary, one could select photons by arrival times to create a photometric series or one could select photons of a certain energy interval only to image the sky at that particular energy (which might contain valuable diagnostic information). The concept allows for maximum flexibility, thus allowing the astronomer to use the data in every conceivable way and thus making optimal use of them.
Note that there is a peculiarity going hand in hand with the recording of such event tables, namely that, in order to preserve the correct energy and arrival time information of each photon, the detector (which stores an electrical charge proportional to the incoming photon's energy) must be read out fast enough that no second photon hits the same pixel prior to readout! Readout times for X-ray detectors are therefore extremely short, of order seconds, and down to milliseconds for studies of bright objects (from which one receives lots of photons in a short amount of time).