Thick disks are not easy to determine in late-type disk galaxies compared to early-type galaxies which show much smoother surface brightness profiles compared to the patchy structure of late-type galaxies. In late-type galaxies they are hard to see by eye and one has to apply the disk models to decompose the structure and detect the thick disk component. The galaxy image requires to have a very flat background, a high S/N ratio and a high resolution, to be able to reach to very faint magnitudes, to create high quality profiles.
UGC 3186 proved to be too faint and too disrupted by bright foreground stars in front of the galaxy to obtain workable profiles in either of the bands. The images of NGC 2424 from the UKIRT set had both distorted backgrounds around the galaxy which were caused by a bright star nearby or the moonlight as the Calar Alto image showed a similar effect just outside the range of the galaxy. The contour map of both galaxies (see Appendix A) show this distorted background partially.
The CA set was our initial main sample. Structure decomposition had not been done before in the near-infrared, so we use this as our pilot sample to discover if we could find a thick disk. The UKIRT sample was added to provide a larger sample with more galaxies in the same two bands, but we knew beforehand the image frames of the galaxies were small (as they had originally been intended for study of the bulge) and that it might not be possible to use all galaxies for the profile fits. IC 2531 and NGC 973 had to be rejected from the fit.
The late-type edge-on galaxies of the CA set were selected according to the allocated observing time and for being undisturbed. Although it turned out that some contained bulges that were tilted with respect to the disk, this was not problematic. This was not the case for NGC 5348. As one can see on the left and right side on the contour map (see Appendix A), the galaxy has an U-shaped warp. This makes it very complex to create an average quadrasized galaxy, but even 1D profiles on either of the four sides will have an up- or downbending of intensity in the outer part. Fitting vertical profiles which do not follow a straight line in radius, but a 'line of nodes', is very complex to do, so it was decided not to use NGC 5348.
In the end, out of 20 data-reduced galaxy images, only 11 remained for the profile fit. An S0 galaxy, observed in the V-band, of which we already knew it had a thick disk (see Pohlen et. al. 2004) and two low-mass Sd galaxies, observed in the R-band, for comparison with Yoachim & Dalcanton (2006), as noted in Section 2.1, were added to expand the range of galaxy types and be able to test our methods and results with literature data. Obviously those did not create issues for creating profiles.