1 Introduction

In radio interferometry, "mosaicing'' is the practice of imaging a field much larger than the antenna primary beam response by combining data from multiple different pointings. Mosaicing is fundamentally more than a simple "cut and paste'' merging of images from different pointings: mosaicing can recover information on shorter spacings than that possible with single pointing observations. Mosaicing provides constraints on the short spacings, which allows more reliable imaging of large-scale structure. A detailed argument on the advantages of mosaicing is given by Cornwell (1988).

Mosaicing, however, is not usually sensitive to the shortest spacings. Consequently, mosaiced images are often complemented with single-dish observations to allow the Fourier plane to be fully sampled.

Although there are some examples of mosaicing polarimetric data (Wieringa et al. 1993; Dickey 1997; Beck et al. 1998; Duncan et al. 1998), these do not combine mosaic and single-dish data. This is despite many Galactic objects being both large and significantly polarized. Thus a combination of mosaiced and single-dish data is just as relevant in polarimetry as in total intensity. Additionally, the published cases do not appear to have processed the different Stokes parameters jointly and some of them do not process the individual pointings jointly.

In this note, we describe and compare two maximum entropy schemes for mosaicing and combining mosaic and single-dish data which we have used to image large-scale polarized emission. Both schemes jointly process the available pointings and single-dish data. These schemes differ in whether the four Stokes parameters are separately or jointly deconvolved. We also briefly compare these with a CLEAN-based scheme. For the data we present, we find no benefit in the joint deconvolution of the Stokes parameters over separate deconvolution, although we note some cases where this may not be the case. Issues of wide-field polarimetric purity are also briefly considered.