INTEGRAL is one of the IFSs offered on the WHT, it is owned by the IAC but run jointly by the IAC and ING.
There are a few things to note if observing with or reducing data from Integral
Observing with Integral
- Integral is a flexible instrument, having access to all the ISIS/WYFFOS gratings and 3 (4 really, but one is not often used) IFUs. However, to change the gratings you need to go into the GHIRL room in the dome to do so, and this can take 15 mins or so. The echelle grating is in a different place to the others, and thus takes longer to change in and out.
- …the IFUs, however, can be changed quite quickly (and electronically).
- For the echelle grating note that (depending somewhat on the wavelength range) there will probably not be sufficient arc lines to do a wavelength calibration, so observations of the solar spectrum (i.e. twilight) will be necessary.
- The spaxels are constructed from spectra directed to the spectrograph straight though fibre-optical cables. They are therefore circular in shape and not spatially contiguous. This should be considered when deciding which IFU to select for your observations.
- If you want to check the pointing of the IFU, checking from the telescope's guide camera is usually enough. However, note, if the guiding is done in white light or a wavelength range different to yours, there will be slight shifts between the telescope's FoV and Integral's. If you have bright, and even better point sources in your FoV, you can use the pattern they make on the raw images (a pattern made up of the aperture light) to check if your FoV has changed between two observations (i.e. if the FoV does not change, neither should your aperture-light pattern). Using the aperture map you can tell in which way the FoV has drifted if your pattern has changed (e.g. if the brightest aperture has moved one to the right from aperture 57 to 58, then look at the position of these apertures on the map for IFU "SB2" will tell you which direction the FoV has moved).
Reducing Integral Data
- IRAF is an easy way to reduce Integral data, using the apall task. How this is done in general is described in data reduction; also, the IAC Integral team should give you help. Essentially, you use the task to locate, trace and extract the spectra on the raw CCD images of the target and arcs, using the flatfield to so do. You then wavelength calibrate. Work out the throughput correction using the throughput frames (sky flats) and apply. Possibly correct based on measurements of sky line fluxes. After doing this, in the header of the final fits frame should be a listing of the spectrum number and (raw CCD) coordinates for all extracted spectra. It will be easier for you if you include (i.e. do extract) even rubbish spectra, from dead spaxels, because this will make assembling the data into a cube much easier.
- Because of the non-contiguity of the spaxels it is likely that flux will "fall" between them. For point sources (e.g. flux standards) this could be important if the spaxel size is close to the seeing size (rather than at least 2 times less). Consider the spatial field to be divided into circles of light and spaces of no light (e.g. like here). Different exposures on the a point source will probably find that point source positioned in a different part of this pattern, and thus a different proportion of the flux could be lost: for example, if the star happens to be centred on a gap then more flux will be lost compared to if the star was centred right in the middle of a spaxel. This can be a problem for performing flux calibration: to calibrate a single exposure is not a problem as the flux calibration process "corrects" for this, but when transferring that calibration to another exposure of a point source (or even an extended source), if that second source was positioned differently in the spaxel/gap pattern the calibration will not be correct iff the spectrum of the flux standard was extracted by simply adding the flux from the spaxels. My work (KME: using the IFU SB2, at airmass 1.5) found a global offset of 30% in one case; this I could determine because the star I was observing the field around happened to also have a flux calibrated spectrum in the literature, to which I could compare my extracted, flux calibrated Integral spectrum. Two ways to try avoid this problem (assuming you cannot match seeing) are: dither all observations, and only extract spectra from the mosaicked-together exposures; when extracting the flux calibration source spectrum, use a PSF-extraction technique rather than simply summing the spectra from the spaxels with source light in them.
- The two sides of the CCD, the left being the red part of the spectrum and the right the blue, are better wavelength calibrated separately (at least, with the 1200l/mm and echelle gratings this is the case). Finding a stable wavelength solution for each aperture otherwise is tricky.
- Integral data reduced with IRAF does not give you an error extension, and for some analyses this will be necessary (e.g. emission line fitting). We suggest you create your own from the sqrt(counts) of the extracted but not otherwise processed astronomical frames (i.e. when first put into RSS format). The error frame will, naturally, have to be taken then through all the subsequent data reduction stages that the main image frames are.
From RSS format to cube format
To view your integral data as a image (at one wavelength) or a cube it is necessary to get hold of the position tables. These should be available on the Integral website. If you reduced the data such that all spectra, even from dead spaxels, are present, then the position tables will be as provided in this tarfile, which is from 2006. Note that is it possible an axis flip will be required, depending on whether you numbered as spectrum 1 that on the left or that on the right of the CCD. I [KME] think that for the SB1 and SB2 IFU you count from left to right on the raw CCD, for SB2 from right to left.
Turning the Integral data from RSS to cube format will require a spatial resampling. The Integral IFU heads are round and non-contiguous and this does not lend itself to conversion to cube format. We therefore direct you to the resampling page for information on how to do this. Once accomplished, you could then turn to the examples we include on the vimos extras page; as this has been written for use with VIMOS data, you will naturally need to modify the dimensions when applying to Integral data and it will be necessary also to provide a position table that is appropriate for your new, resampled Integral RSS image.
