StarCAT is a catalog of high resolution ultraviolet spectra of objects classified as "stars," recorded by Space Telescope Imaging Spectrograph (STIS) during its initial seven years of operations (1997-2004). StarCAT is based on 3184 echelle mode observations of 545 distinct targets, with a total exposure duration of 5.2 Ms. For many of the objects, broad ultraviolet coverage has been achieved by splicing together echellegrams taken in two or more FUV (1150-1700 Å) and/or NUV (1600-3200 Å) settings. In cases of multiple pointings on conspicuously variable sources, spectra were separated into independent epochs. In cases of nonvariable, or lightly variable, objects, different epochs were combined to enhance signal-to-noise (S/N).

The StarCAT sample was assembled by two crosscutting searches through the HSTonline catalog hosted at MAST. The initial search focussed on objects identified by the Guest Observer as "STAR-∗" for the first broad category keyword. The second search focussed on the second choices of GOs when the first choice was "STAR-∗." (Sometimes, a GO would use these less preferred secondary descriptors as the leading broad category keyword.)

A naming convention was adopted to unify the somewhat diverse target IDs assigned by the GOs, taking in order of priority: HD number, variable star name (including supernovae: SN), WD number (from McCook et al. 2006), and assigning "STARHHMM±DDMM" to all others (equinox/epoch 2000). In one case, "NOIDHHMM±DDMM" was tagged to an observation lacking a SIMBAD object within 20″.

Post-processing began at the level of the calstis pipeline "x1d" file; somewhat a misnomer because it actually is a 2D tabulation of wavelengths, flux densities, photometric errors, and data quality flags for the up to several dozen orders of the particular grating setting (e.g., E140M-1425, where the first part is the mode and the second is central wavelength in Å). The x1d file contained at least one — and sometimes several — subexposures, which were treated as separate observations. The initial processing included a post facto correction to compensate for subtle wavelength distortions identified in a recent study of the STIS dispersion relations (the "Deep Lamp Project"); a recalculation of the photometric error (replacing N^½ with [N+1]^½); and more aggressive edge trimming of E230 modes to avoid unflagged "dropouts" that sometimes occurred at the beginning of the low orders. The x1d orders then were merged, averaging the overlap regions weighted by the individual sensitivity functions s _λ, but accounting for bad pixels and wavelength gaps.

Next, a series of different layers of coaddition and splicing were applied to the sets of order-merged 1D spectra of each object.

STAGE ZERO — the subexposures, if any, of an observation were combined. The individual spectra were aligned in velocity by cross-correlation against the first observation. The initial exposure should have the most reliable wavelength scale because it normally would have been taken closest to an acquisition and peak-up (the zero point is dependent on the accuracy of the target centering, which is affected over time by jitter and/or drifts). The subexposures then were interpolated onto the wavelength scale of the initial spectrum and coadded, weighting by exposure time, taking into account bad pixels and gaps. The resulting files are called "o-type" and have the same rootname as the original STIS observation, e.g., o61s01010.

STAGE ONE — the independent exposures of like mode and aperture, all taken in a well defined group within a single "visit," were combined. As in Stage Zero, the spectra were aligned by cross-correlation against the first of the sequence, interpolated onto that scale, then coadded, again weighting by exposure time. The resulting files are called "E-type," e.g., "E140M-1425_020X020_51613." The initial part is the particular mode/tilt (here, "E140M-1425" for the medium-res FUV echelle mode with central wavelength 1425 Å), followed by an aperture code (here, "020X020" for the 0.20″×0.20″ photometric slit), and ending with the start time of the first exposure expressed in integer days as a Modified Julian Date (MJD: J.D. - 2,400,000).

STAGE TWO — like-mode exposures of an object taken in different visits and/or using different apertures were combined. Now, the cross-correlation alignments were done relative to the highest S/N exposure of the group (perhaps a Stage 0 or 1 coadd), and the average shift was subtracted. Also, the coaddition weighted the constituent spectra by a factor closely related to the total net counts at each wavelength, to compensate for the different transmission factors of, say, a narrow versus broad slit. The multi-epoch coadditions were undertaken only if the object displayed minimal variability during its timeline. The resulting files, like Stage 1, are E-type, but with extended aperture codes and/or MJDs to reflect the diversity of the constituent exposures, e.g., "E230M-2707_020X020_020X006_51163-51202" for one extreme example.

The top-level object list (StarCAT Portal), lists brief stellar characteristics abstracted from SIMBAD, and links to several layers of data tables. It looks like this:

The first layer down (linked through object name [Column 1]) represents the final spectrum (or spectra) for each object, which might be only a single o-type, say from a GO "SNAPSHOT" program; or a single E-type if only one mode/setting had been utilized, but, say, with several independent o-type exposures; or a full-blown U-type covering all or part of the 1150-3200 Å range, perhaps as a function of epoch for variable targets.

This "final spectra" layer then links down to the constituent spectra in the splice, if any; and each of these in turn links down to whatever grouping, if any, of exposures constituted it. Thus, the lowest layers of a tree always are the o-type exposures; the next one or two layers up are the E-types; and finally the top layer holds the U-types (at best; an o-type at worst; and E-types in intermediate cases).

At the same time, the author cautions users of StarCAT that — as with any large scale project like this — there are inevitable errors and omissions that have escaped the albeit numerous checks that the author has conducted. The author would appreciate any feedback along these lines, or any other comments for that matter, so that corrections can be made and/or enhancements undertaken to improve the catalog for future users.