1 Introduction

Deep near-IR imaging has an important role in the studies of very distant objects, whose redshift moves the rest frame visible light into the near-IR. In fact, from $z \geq 1$ the wavelength range where normal stars emit most of their energy and have their prominent spectral features (4000 $\leq \lambda \leq$ 10000 Å) is shifted in the near-IR, between 1 and 2.5 $\mu$m. In addition, the stellar light of $z \ge 1$ galaxies observed in the near-IR is less likely to be affected by extinction and by an active nuclear component (i.e. a dust scattered AGN) than in the optical and is not contaminated by thermal dust emission. Therefore the near-IR is of fundamental importance in order to investigate the stellar component of high redshift galaxies.

This paper describes the results of IR imaging of sources selected from the Westerbork Northern Sky Survey (WENSS, Rengelink et al. 1997), a new large-scale low-frequency radio survey that covers the whole sky for $\delta \ge 30$ degree at 325 MHz, and about a quarter of this region at 609 MHz, to a limiting flux density of 18 and 15 mJy respectively. Contrary to previous IR works concentrated mostly on powerful radio sources (e.g. Lilly & Longair 1984; Rigler et al. 1992; Dunlop & Peacock 1993), the radio catalogue from which we selected our sources extends to low flux densities and therefore to galaxies with a much lower level of nuclear activity.

The use of IR imaging allows one to address three important questions: first, the identification of the sources in these subsamples with their optical/IR counterparts. At high redshift ($z \ge 1$) the peak of the spectral energy distribution of galaxies is shifted in the near IR where their flux density is therefore larger than in the optical by orders of magnitude. So it is easier to identify sources, which correspond to distant galaxies, in the IR than in the optical.

Secondly, IR imaging allows the study of the IR "alignment effect'', i.e. the alignment of the IR morphology with the radio axis, as a function of the redshift and radio power. The alignment effect (see McCarthy 1993 for a review) has been discovered in the optical, and demonstrates a strong influence of the AGN on the optical morphology and luminosity. However, its origin is still matter of debate and its extension to the IR is important to understand it. Dunlop and Peacock (1993) have detected a clear IR alignment effect in a 3CR sample of radio galaxies in the redshift range $0.8 \leq z \leq 1.3$ (they show the IR-radio and the optical-radio alignment histograms for this subset, claiming that the alignment effect is just as clear at K as in the optical) in contrast with the conclusions of Rigler et al. (1992) for the same subsample of objects. Therefore it is necessary to study the IR alignment in different samples to shed some light on the origin of this discrepancy

(probably, much of the apparent discrepancy arises from different methods of analysis) and on the possible presence of different components which dominate in the IR and in the optical, respectively.

Finally, the IR observations enable the study of the stellar populations of distant galaxies through the analysis of the light emitted by normal stars, in order to determine, modelling their spectral energy distributions by means of stellar population synthesis models, the age of the oldest stars and therefore provide us with information about the epoch of galaxy formation. We emphasize that the starlight analysis is easier and more productive in the samples at lower radio power, as are the ones selected from the WENSS survey, since the properties of the stellar populations are less contaminated by the AGN component.

In the next paragraph we discuss the selection criteria of our samples. Then we report on our observations and immediate results for individual sources. We then analyse the radio-IR alignment and discuss the results obtained for a gravitational lens system.