AC118, also known as Abell 2744 ( $\alpha , \delta = 00 \ 14 \ 19.5 \ -30 \ 23 \ 19, \$ J2000), is a cluster of galaxies at intermediate redshift (z=0.3, see Paper I for a summary of its properties). Its central region has been imaged in the near-infrared $K_{\rm s}$ band with SOFI at NTT (Paper I). AC118S and AC118N, the northern and southern pointings of AC118, have been observed in the $K_{\rm s}$ band in September 18 and October 31, 1999, respectively, with SOFI at NTT in the frame of an observational program aimed at deriving the Fundamental Plane at $z\sim0.3$ , as a complement to our central pointing. SOFI is equipped with a $1024\times1024$ pixel Rockwell "Hawaii" array, with a 0.292 arcsec pixel size and a $5\times5$ arcmin field of view. The two pointings are offset, with respect to the previous central pointing, by almost one SOFI field of view, to maximize the survey area while keeping enough overlap to check the consistency of the photometry of the three pointings (see Fig. 1 for the pointing layout).

**Table 1:** The data.
	AC118C	AC118S	AC118N


Exposure time (min)	265	75	90.7
Seeing (FWHM, arcsec)	0.75	1.0	1.2
Fully corrected noise^a (mag arcsec^-2)	24.0	23.0	23.3
Applied Illumination correction?	no	yes	no
Photom. zero-point RMS over the field (mag)	0.007	0.013	0.004

^a The sky noise is measured as the dispersion of adjacent pixels in the background, once the image is binned in pixels of 1 arcsec.

The data have been reduced as described in Paper I. Briefly, images are flat-fielded by means of differential dome flats and calibrated by means of Persson et al. (1998) standard stars. The background is removed by a temporal filtering of the images, using Eclipse (Devillard 1997). The combining of the individual exposures is performed by means of imcombine under IRAF, making full use of the bad pixel mask and weights and aligning images without resampling. As in Paper I, the dependence of the atmospheric absorption on airmass is computed from the science data, since the target is observed at different hour angles.

Two (minor) differences apply with respect to the data reduction described in Paper I: a) the illumination correction is found to be significant, and applied to, AC118S frames; b) a residual shallow spatial gradient is present in the frames, even after the background subtraction performed via a filtering in the time domain. Therefore, we introduce a further step in the background subtraction, by fitting, and removing a plane to the background.

The two nights were photometric, as determined from the scatter of the zero point of the standard stars and from the scatter, from frame to frame, of the instrumental magnitudes of a reference galaxy in the field of view. For objects in common between fields, that are independently calibrated, we found systematic differences less than 0.01 mag, confirming the quality of the observing nights and of the data reduction.

During the September run (AC118S), SOFI suffered a point spread function variable over the field, while in October the problem was largely solved.

Objects are detected and classified by SExtractor (Bertin & Arnouts 1996), version 2, using the exposure map for a clean detection.

Because of the shallower images, the 4.4 arcsec aperture (24 Kpc for cluster galaxies) adopted in Paper I is not an optimal aperture to measure the flux of our faint galaxies because the aperture integrates mainly noise (outer regions of faint galaxies are undetected) at a such a large radius. We adopt, therefore, a smaller (3.0 arcsec) aperture for the flux determination. By using the deep (central) AC118 pointing we verify that such an aperture misses part of the galaxy flux, even for faint galaxies, i.e. the 3.0 arcsec aperture magnitude is not a surrogate for the "total" magnitude.

The sample is complete down to $K_{\rm s}=20$ mag in the most shallow field (AC118S), and therefore the analysis is bound at $K_{\rm s}\le 20$ mag over the whole field of view.

2 Data and data reduction