1 Introduction

High angular resolution techniques for astronomical imaging have matured rapidly in recent years (see e.g. Beichman & Ridgway 1991) and have been applied to a variety of Galactic and extra-galactic sources. For ground-based near-IR imaging, adaptive optics techniques have been very succesful in approaching the theoretical diffraction limit of the telescope (Rigaut et al. 1998). The systems allow on-line correction for atmospheric perturbations using either a natural guide star (part of the object under study, or an unrelated star in the near vicinity) or laser guide star (see e.g. Lloyd-Hart et al. 1998). A natural extension of this technique is to achieve two-dimensional polarimetric observations in the near-infrared with the, in principle, simple provision of a polarizer in the beam. The combination of both techniques allows information on the detailed polarization of an extended source, or determination of the individual polarization of close multiple sources. It may be applied to reflection nebulae for determining the position of embedded illuminating sources, for study of the line of sight geometry of dust scattering regions and for the orientation of magnetic fields in star forming regions or quasar jets.

The extinction of interstellar dust peaks in the UV and declines to longer wavelengths (e.g. Mathis 1990), but the continuum emission from grains at typical temperatures of a few hundred K in regions heated by starlight increases strongly above 2 $\mu$m. In addition molecular emission and absorption bands are stronger above 3 $\mu$m. The $1-2~\mu$m region therefore provides an ideal window for the study of the close environment of dust embedded sources, such as regions around proto-stars or emerging young stars. For typical interstellar grains the low extinction in the near-IR enables information on the scattering properties of the grains, or the study of scattering regions, which have high optical extinction. Near-IR polarization is thus entirely analogous to optical polarization study but can be extended to more embedded environments. At longer wavelengths the grain emission dominates and any polarization of the radiation is controlled by anisotropic emission mechanisms such as aligned non-spherical grains (Davis & Greenstein 1951). Examples of IR polarimetry include: detection of extended dust disks in young stellar objects (see e.g. Piirola et al. 1992, for observational results and Berger & Ménard 1997, for theoretical work); dust structures in AGB envelopes (e.g. Sahai et al. 1997 for CRL 2688); detection of dust in interstellar jets (e.g. Hodapp 1984); magnetic field structure in star forming regions (e.g. Whittet et al. 1994) and polarization in galaxies (e.g. Jones 1997) and quasars (Sitko & Yudong 1991). Extending polarimetry to the IR also brings the potential of high spatial resolution, both through the dependence of diffraction on wavelength and the decrease in atmospheric seeing size with wavelength.

In the near-IR, the dominant contribution to polarization is therefore from scattering of radiation by grains, and their finite relative size requires that Mie theory must be used to predict the scattering properties. However the optical properties of typical interstellar grains are fairly well represented by models based on laboratory and observational data (Draine & Lee 1984), so that the scattering properties of interstellar grains in the near-IR can be predicted. Whilst polarization data naturally provides geometric information on the location of illuminating sources, the scattering efficiency with scattering angle is required to derive geometric information about the line of sight location of the scatterers (White et al. 1980). For high dust column optical depths, multiple scattering may occur and has then to be modelled using Monte Carlo methods (cf. e.g. Witt 1977; Warren-Smith 1983; Whitney & Hartmann 1992; Fischer et al. 1994; Code & Whitney 1995).

As part of a programme to study the nature of the dust in the Homunculus nebula around the massive star (or stars) $\eta$Carinae and determine information about the 3-D structure of the reflection nebula, near-IR imaging polarimetry data were obtained with the ESO ADONIS system. $\eta$ Car and the Homunculus is an ideal source for adaptive optics since the central point source is very bright and the nebula is not so extended that off-axis anisoplanicity becomes an important effect. The present paper is devoted to the details and subtleties of the data collection and removal of the instrumental signature vital to the derivation of a polarization map. A following paper will present the scientific results on the high resolution near-IR polarization of $\eta$ Car and the Homunculus. Section 2 is devoted to a brief description of the ADONIS instrument; Sect. 3 then considers the observational strategy. The fundamentals of the data reduction are described in Sect. 4 and the polarimetric calibration of the instrument in Sect. 5. Section 6 exposes the different deconvolution techniques applied to the data and their resulting effect on the polarization maps.